Impact of auditor independence on earnings management : the role of internal control

(1)

Amsterdam Business School

Master Thesis 2014:

Impact of auditor independence on

earnings management:

The role of internal control

Name

Frank de Vries

Student number

10297626

Concept version

June 19

th

2014

Supervisor

Dr. G. Georgakopoulos

Second examiner

Dr. Ir. S.P. van Triest

Master thesis Accountancy & Control, variant Accountancy, 2013/2014

Faculty of Economics and Business, University of Amsterdam

(2)

Abstract

Listed companies in the US are, since the implementation of the Sarbanes Oxley Act (SOX) in 2002, obliged to disclose the magnitude of the fees they pay to their external auditor, divided in audit fees and non-audit fees. This research investigates the relationship between auditor independence (i.e. fee ratio) and earnings management (i.e. real earnings management) as well as the influence of internal control (i.e. SOX 302 and SOX 404) upon this relationship. Firstly, auditor independence is measured by the fee ratio. Secondly, I focus on real earnings management and this will be measured by sales manipulation, reduction in discretionary expenditures, and overproduction. Finally, I will make assumptions for internal control as I use SOX 302 and SOX 404 disclosures as an indicator for the quality of company’s internal control systems. The research sample consists of 25.647 firm-year observations of US listed companies only, with available financial data of fiscal years 2006-2013. I find that companies with ICD are more likely to change from auditor, are smaller, less profitable, and are less likely to be audited by a Big 4 accounting firm. Especially for these companies it is a challenge to develop/maintain a strong internal control system. Furthermore, more independent auditors are more likely to disclose an ICD compared to auditors who are less independent.

Keywords: Auditor independence, fee ratio, earnings management, real activities manipulation, sales manipulation, reduction in discretionary expenditures, overproduction,

(3)

1. Introduction

One of the requirements for US listed companies, by the SEC (hereafter; Securities and Exchange Commission), is to compile an annual report. An annual report can be seen as a summary of a company’s business for the year and is essential in terms of explaining how the company has performed during the fiscal year. Annual reports are designed to comply with the information needs of many different parties, like for example; (potential) shareholders and creditors, economists, financial analysts, suppliers and customers (Friedlob and Welton, 2008, p. 3). One of the features of an annual report is to reduce the information asymmetry that could arise between the principle (i.e. the shareholder) and the agent (i.e. the manager), also known as the agency theory. Managers perform daily operations on behalf of the shareholders, however they do not always act in the best interest of the shareholders and could attempt to maximize their self-interests (Wallace, 2004). The agency theory is concerned with resolving these conflicts and subsequently align the interests of the principal and the agent (Jensen and Meckling, 1976). Furthermore, annual statements should provide users of annual statements with useful information about the financial position and performance of the company, in order to help them making appropriate economic decisions. Hence, the importance that annual reports reflect a true and fair view of the company’s financial position. Therefore, financial statements are obliged to be prepared in accordance with accounting and reporting standards established by the Financial Accounting Standards Board (hereafter; FASB), the SEC, and various committees of the American Institute of Certified Public Accountants (hereafter; AICPA), who attempt to prevent scandals. Users of financial statements can rely on the fair presentation of the information in the annual reports due to the involvement of independent Certified Public Accountants (hereafter; CPAs).

The Sarbanes-Oxley Act (hereafter; SOX) supplements the requirements of the SEC, since the annual reports must contain management’s objective assessment of the effectiveness of the company’s system of internal control over financial reporting (Friedlob and Welton, 2008, p. 6).

The objective of this study is to provide empirical evidence regarding the influence of internal control on the potential relationship between auditor independence and earnings management. Resulting in the following main research question:

What is the influence of internal control on the potential relationship between a lack of auditor independence and earnings management?

(5)

My motivation for this study is the growing interest of internal control. In addition, a number of major accounting scandals (for example: Enron, WorldCom, and Parmalat) resulted in a loss of confidence in the accounting profession and placed the accounting profession in the spotlight. Especially after these accounting scandals, there have been several studies on auditor independence and earnings quality. However, it is still claimed to be controversial (Alexander and Hay, 2013), this created the drive to investigate auditor independence. In addition, prior literature shows mixed results regarding audit quality, like Ashbaugh et al. (2003) and Chung & Kallapur (2003) that both measure earnings quality as a proxy for audit quality. Furthermore, the implementation of SOX in 2002 is crucial for this investigation since two of its requirements enable this study; (1) SOX requires companies to disclose weaknesses in their internal control systems, and (2) SOX requires companies to disclose the magnitude of fees paid to an external auditor, with a distinction between audit fees and non-audit fees. The former enables to investigate internal control in depth, since disclosures of weaknesses in internal control systems are used as a proxy to capture the quality of company’s internal controls. The latter enables to investigate auditor independence in depth, since disclosures of fees paid for services provided by auditors could indicate the independency of the auditor to their clients.

To investigate internal control I use section 302 and section 404 of SOX (hereafter SOX 302 and SOX 404). These two sections of SOX provide information regarding company’s internal control systems. SOX 302 requires management to evaluate the effectiveness of their internal control systems and disclose their conclusions about the effectiveness in the financial report (Ashbaugh-Skaife et al., 2008). So, management is required to report any significant change in their internal controls that could have negative impact on the internal control systems (Zhang et al., 2007). SOX 404 on the other hand complements SOX 302 since the auditor of their financial reports is obliged to provide an opinion about management’s assessment regarding the effectiveness of the internal control system. Furthermore, the auditors opinion needs to be disclosed in the same financial report.

Another requirement of SOX is that companies are obliged to disclose the fees they pay for the services the auditor provide, divided in audit fees and non-audit fees. This information is crucial in order to determine the independency of the auditor. Since, non-audit fees are more lucrative they could affect auditors opinion during the audit in order to please the client. Therefore, I calculate the fee ratio since the fee ratio is frequently used in prior literature and indicates whether an auditor is independent from a particular client or that the auditor depends too much on that particular client, which could affect the independency for audit services.

Finally, earnings management plays an important role for management to beat earnings benchmarks. My motivation to use real earnings management is based on prior literature. According to Cohen et al. (2008) the use of discretionary accruals shows a significant decline after the

(6)

implementation of SOX. However, the use of real earnings management shows a significant increase after SOX’s implementation (Cohen et al., 2008). Therefore, this study focused on real activities manipulation instead of discretionary accruals.

This study attempt to fill the gap in the existing literature, since Ge and McVay (2005) stated that “future studies could explore links between disclosure of material weaknesses and fraud, earnings management, or restatements. These studies could help provide insights to the benefits of Sarbanes-Oxley, and Section 404 in particular (p. 155).

The results of this research show that companies with ICD are more likely to change from auditor, are smaller, less profitable, and are less likely to be audited by a Big 4 accounting firm. Especially for these companies it is a challenge to develop/maintain a strong internal control system. These findings are consistent with the results of Doyle et al. (2007a). Furthermore, more independent auditors are more likely to disclose an ICD compared to auditors who are less independent, which is consistent with Zhang et al. (2007).

The remainder of this paper proceeds as follows. In the next section I discuss relevant prior literature, which subsequently will lead to the development of my hypotheses. I describe the empirical methodology in Section 3, while I present the tests and results in Section 4. Finally, a summary of the findings appear in Section 5.

(7)

2. Literature review and hypotheses

In this chapter prior literature is reviewed regarding auditor independence, earnings management, and internal control systems respectively. This chapter clarifies definitions and examines results from prior literature. Eventually, this will lead to the development of my hypotheses.

2.1. Auditor independence

Auditor independence is crucial for credible financial reports, since auditing should reduce the information asymmetry between the principles (i.e. shareholders) and agents (i.e. management) by verifying the validity of financial statements by an independent external auditor. Hence, The Securities and Exchange Commission’s (hereafter SEC) aim for independent auditors in order to protect investors and other users of financial statements (SEC, 2001). DeAngelo (1981a) clarifies the concept of auditor independence, she describes that auditor independence depends on the perceived ability to (1) discover errors or breaches in the accounting system, and (2) withstand client pressure to disclose selectively in the event a breach is discovered. The level of auditor independence depends on the probability that the auditor report a breach after one is discovered (DeAngelo, 1981a, pp. 115-116).

One of the most important conditions for independency is related to financial interest, and to be independent the auditor is expected to have no financial interest in the client. The objective of an audit is to enhance credibility of financial statements by providing reasonable assurance that these financial statements represent a true and fair view about the company’s financial position. However, if users of financial statements believe that the auditor is not independent then the objective of the audit is at stake. Therefore, all auditors are required to be independent in appearance and in fact (SEC, 2000b). These two distinct dimensions are explained by the SEC in the Final Rule S7-13-00, ‘Revision of the Commission’s Auditor Independence Requirements’. Firstly, the SEC defines independence in fact as the auditor’s “mental state of objectivity and lack of bias” (SEC, 2000a). In other words, that the auditor is actually unbiased, however, according to Frankel et al. (2002) and Schneider et al. (2006) this is not observable and impossible to prove. Independence in appearance is assessed and the SEC will not consider an auditor to be independent with respect to a particular client “if a reasonable investor, with knowledge of all relevant facts and circumstances, would conclude that the auditor is not capable of exercising objective and impartial judgment” (SEC, 2000a, Section I, quoted in Frankel et al., 2002, p, 73). So, independence in appearance is entirely based on perceptions by a reasonable observer (Beattie et al., 1999) and this type of independence can harm companies’ results as well, by for example stock market reactions (Ghosh et al., 2009), even when independence in fact is not compromised. So, auditors need to be able to carry out their work freely

(8)

and in an objective manner (Krishnan et al., 2005) and to establish objectivity and integrity for certifying financial statements, both dimensions (i.e. independence in appearance and independence in fact) are a necessity.

Auditor independence can be affected in several ways. According to Holland and Lane (2012) a decline in independence is conceptualized in two ways; (1) the role of the auditor could change resulting in a modified perspective from independent outsider to internal advisor whereby their judgment is affected, (2) through the economic bond between the organization and its external auditor. This research focus on the latter where audit fees and non-audit fees create an economic bond between auditor and client, this economic bond could harm auditor independence. When an auditor has a strong bond with a client, potential problems could be ignored in order to issue a clean opinion. Such a strong bond refers to the fees involved during the audit engagement. Auditors have incentives to retain their clients because these clients pay fees for the services provided by the external auditor. Fees can be divided in audit fees and non-audit fees. The higher the fees the more income the auditor receives. However, this can lead to a decrease in auditor independence. Especially the non-audit fees, since non-audit fees have higher margins and have therefore more impact on auditor independence (Chung and Kallapur, 2003). The magnitude of audit fees are affected by audit quality and auditor tenure (Hay et al. 2006).

Ashbaugh et al. (2003), Ghosh et al. (2009), Zhang et al. (2007) [and others] use the fee ratio to capture auditor independence. Since, non-audit fees are more lucrative the fee ratio will be calculated in order to determine the independency of the auditor. The fee ratio can be calculated in different ways (for example; the ratio of audit fees to total fees or non-audit fees to total fees). This ratio is frequently used in prior literature and considerably gives a representative indication of auditor independence. Prior literature often used fees to indicate a relation between auditor independence and audit quality. However, the findings are not consistent and prior literature shows that audit fees and non-audit fees can have different influences on audit quality. Ghosh et al. (2009) mentioned that there are little studies about independence in appearance that provide evidence if the independency is affected by audit fees and non-audit fees. Frankel et al. (2002) argue that an auditor is not fully independent when a client is able to manage earnings without the auditor reporting it or demanding restatements, which leads to a lower quality financial statements. Therefore, I use earnings management as a proxy for audit quality, to test for a relationship between auditor independence and earnings management.

(9)

2.2. Earnings management

Earnings and other accounting information are crucial for creditors to assess companies’ health, credibility, and viability (Ge and Kim, 2014). However, management has discretion in these accounting figures and can even manage earnings for the companies and/or private benefit. Healy and Wahlen (1999) stated that “earnings management occurs when managers use judgment in financial report 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” (p. 368). In other words, earnings management is applied in order to meet or beat earnings targets. Earnings management is extensively studied in prior literature and frequently used as a proxy to measure audit quality (Fields et al., 2001). Francis (2011) indicate that earnings quality potentially capture the entire continuum of audit quality, from low- to high-quality audits. A company has low earnings quality when they have a high amount of earnings management. This indicates that earnings management is not discovered during the audit process or deliberately not reported, either way it indicates that earnings quality is low.

Earnings management can be divided in two different categories namely, (1) accruals-based earnings management and (2) real earnings management (Zang, 2012). Manipulated earnings are no longer reliable after applying one of the earnings management types, since it is not a representative reflection of companies’ performance anymore. The majority of prior literature focuses on accrual-based earnings management. Accrual-accrual-based earnings management is caused by accrual accounting, whereby expenses are matched with revenues. This should help investors to assess companies’ performance. However, there are some concerns because the existence of management discretion could result in smoothing reported earnings. A commonly used approach to measure accrual-bases earnings management is the modified Jones (1991) model, or a variant thereof, whereby discretionary accruals are estimated by identifying total accruals to explain the normal level of accruals and the regression model will subsequently show the discretionary accruals (Beattie and Fearnley, 2002).

Unlike accrual-based earnings management, management can use real earnings management by managing earnings through real economic actions (Ge and Kim, 2014). Real earnings management is studied in prior literature as well, but in a much smaller stream. Real earnings management is a mechanism whereby management manipulate real transactions to distort earnings (Zang, 2012). Real earnings management jeopardizes a firm’s competitive advantage in the long-term (Cohen and Zarowin, 2010; Zang, 2012). In addition, according to Roychowdhury (2006) and Chi et al. (2011) real earnings management could have negative consequences for future cash flows as well.

(10)

Despite these negative consequences, real earnings management it is not forbidden, and therefore it differs from accrual-based earnings management, since real earnings management cannot influence auditors’ opinions or regulators’ actions as long as it is properly disclosed in the financial statements (Chi et al., 2011). However, according to Cohen et al. (2008) the implementation of SOX should “restore the integrity of financial statements by curbing earnings management and accounting fraud. Therefore, the extent of earnings management prior to SOX and the effect of SOX on earnings management is an important research topic” (Cohen et al, 2008, pp. 759-760). Earnings management plays an important role for management to beat earnings benchmarks. According to Cohen et al. (2008) the use of discretionary accruals shows a significant decline after the implementation of SOX. However, the use of real earnings management shows a significant increase after SOX’s implementation (Cohen et al., 2008). Therefore, this study focused on real activities manipulation instead of discretionary accruals.

Roychowdhury (2006) uses three different measures to capture real earnings management. The results of Roychowdhury (2006) suggests that managers avoid reporting annual losses or missing analyst forecasts by manipulating sales, reducing discretionary expenditures, and/or overproduction to decrease the cost of goods sold, all of which are deviations from otherwise optimal operational decisions, with the intention of biasing earnings upward. Although, there is little evidence on real activities manipulation (Roychowdhury, 2006), most of the evidence from prior literature show that reported expenses are reduced by a reduction of R&D expenditures. Dechow and Sloan (1991) documents that CEOs increase (short term) earnings by reducing R&D expenditures towards the end of their tenure. Consistent with Dechow and Sloan (1991), Bens et al. (2002) and Bushee (1998) documents that R&D expenditures are reduced to meet earnings benchmarks. So, evidence shows that R&D expenditures are reduced to increase earnings. However, prior literature is not consistent in the evidence they provide regarding the relationship between auditor independence and earnings management. Frankel et al. (2002) argue that auditor independence is negatively associated with earnings management, measured by non-audit services provision and earnings quality. Supported by the findings of Antle et al. (2006), since they find a negative association between audit fees and audit quality, where higher audit fees lead to lower audit quality. Non-audit fees on the other hand are positively related to audit quality, where higher non-audit fees lead to higher audit quality (Antle et al., 2006). Contrary to these findings, Ashbaugh et al. (2003), Chung and Kallapur (2003) and Frankel et al. (2002) argue that non-audit fees decrease audit quality. Craswell et al. (2002) investigates the influence of the magnitude of fees on auditor independence. Their results did not support the view that a higher magnitude of fees harm auditor independence, however they find evidence that mandatory disclosure of fees result in an increase in audit quality (Craswell et al., 2002).

(11)

Beattie and Fearnley (2002) argue that results from prior literature are inconsistent and unclear. Given the lack of clarity and inconsistencies in prior literature regarding the effect of auditor independence on earnings management, resulted in the following hypothesis:

H1: Lack of auditor independence has a positive effect on earnings management.

This hypothesis analyse the potential relationship between auditor independence and earnings management, where I expect to find evidence for. This means that I posit that earnings management will decrease when the auditor become more independent.

2.3. Internal control

A number of major accounting scandals (e.g. Enron, WorldCom, and Parmalat) resulted in a loss of confidence in the accounting profession, and subsequently increasing concern of investors arose. In 2002, SOX was implemented to restore confidence in the accounting profession and to address the concerns of investors (Zhang et al., 2007). Besides, its implementation facilitates to investigate internal control, because SOX requires companies to establish, maintain, and evaluate internal control. Internal control is best described by The Committee of Sponsoring Organizations of the Treadway Commission (hereafter COSO) since this definition would serve the needs of organizations, independent auditors, legislators, and regulatory agencies (COSO, 2013). According to the COSO framework internal control is described as “a process, effected by an entity’s board of directors, management, and other personnel, designed to provide reasonable assurance regarding the achievement of objectives relating to operations, reporting, and compliance” (COSO, 2013, p. 3). The objectives can be subdivided into three different categories; (1) operations, which refers to effectiveness and efficiency of operations, (2) reporting, which refers to reliability, timeliness, and transparency of internal and external financial and non-financial reporting, and (3) compliance, which refers to laws and regulations to which the entity is subject (COSO, 2013).

In order to ensure accurate financial statements, it is required that companies’ internal control systems are of high quality (Jiang et al. 2010). Since, internal control systems of low quality increase the risk of misstatement, which lead to less reliable financial reports (Doyle et al., 2007b; Ashbaugh-Skaife et al., 2008). Besides, it could also lead to e.g. higher audit fees (Hoitash et al., 2008), and auditor resignations (Elder et al., 2009). Companies with recent auditor changes are more likely to have internal control weaknesses (Zhang et al., 2007), and therefore SOX requires companies to disclose the quality and changes in their internal control system in both quarterly and annual financial reports (Hoitash et al., 2012) in order to protect investors and other users of financial

(12)

statements. Several studies argue that internal control weaknesses are more pronounced by companies with audit committees that have less accounting financial expertise (Krishnan, 2005; Zhang et al., 2007). However, Zhang et al. (2007) argues that there is a link with auditor independence as well, since companies are more likely to be identified with an internal control weakness if their auditor is more independent (p. 315).

In order to evaluate the effectiveness of companies’ internal control systems, COSO developed a broad framework of criteria and investigators of internal control in pre-SOX period were assigned to mandated file like, 8-K [indicated a change in the auditor or an organizations’ financial statements]. One important aspect of SOX is that there are two specific sections devoted to internal control issues related to financial reporting: (1) SOX 302, and (2) SOX 404. So, the mandated 8-K file is complemented by 10-K [disclosed information about an organizations’ yearly performance], and 10-Q [disclosed information about an organizations’ quarterly performance].

2.3.1. SOX 302

August 29, 2002 is the date that section 302 of SOX (hereafter SOX 302) went into effect. Under SOX 302 a firm’s CEO and CFO are mandated to evaluate the effectiveness of their internal control systems, subsequently they are mandated to present their conclusions about the effectiveness of their internal controls based on their evaluation in the financial report (Ashbaugh-Skaife et al., 2008). So, under SOX 302 management is obliged to disclose significant internal control deficiencies when certifying financial statements, on a periodic, annual and quarterly basis. Personnel responsible for internal controls are required to evaluate the internal control and compose a list containing all deficiencies in the internal controls and fraudulent activities of personnel who is involved with internal control activities. Furthermore, management is required to report any significant change in internal controls or related factors that could have a negative impact on the internal control systems (Zhang et al., 2007).

2.3.2. SOX 404

Another specific section devoted to internal control issues is SOX 404. Both sections are implemented in order to inform users of financial statements about the quality of organization’s internal control systems. SOX 404 is an extension of SOX 302 and requires management to assess the effectiveness of the company’s internal control systems as of the fiscal year-end (Ashbaugh-Skaife et al., 2008) and require that this assessment be audited by the auditor of its financial statements (SEC,

(13)

2008). The effectiveness of the internal control structure and procedures, and the disclosure of such information in the annual financial reports are crucial in SOX 404, together with the external auditor’s opinion on the assessment made by the management (Zhang et al., 2007). So, the financial reports must include; a statement of management’s responsibilities for establishing and maintaining adequate internal control systems at the company’s fiscal year-end, the framework used by management to evaluate the effectiveness of the internal control systems of the company. Furthermore, it is required that an attestation report is included in the financial report, composed by the registered public accounting firm that audited the company’s financial statement, about management’s assessment of the company’s internal control systems (SEC, 2008). These two measures (i.e. SOX 302 and SOX 404) are frequently used in prior literature and therefore I combine these two measures in order to give a representative indication of the quality of a company’s internal control system.

Prior literature argue that auditor independence is an important determinant of internal control weaknesses, since auditor who are more independent from their client are more likely to identify an internal control weakness (Zhang et al., 2007). When an auditor issued a clean opinion despite potential problems, it will affect audit quality. Ashbaugh-Skaife et al. (2008) argue that companies with internal control weaknesses are more likely to commit earnings management. The results in Doyle et al. (2007b) confirms the findings of Ashbaugh-Skaife et al. (2008) as they find that earnings are of lower quality when a company disclosed internal control weaknesses. In order to test the influence of internal control on the relationship between auditor independence and earnings management, I test the following hypothesis:

H2: The [expected] negative effect of a lack of auditor independence on earnings management is

more pronounced by companies with weak internal control systems.

To test this hypothesis I am forced to make some assumptions and therefore I assume that companies that disclose an internal control weakness under either SOX 302 and/or under SOX 404 are categorized as a company with a weak internal control system. Subsequently, companies without any internal control weakness disclosure is categorized as a company with a strong internal control system. This assumption is based on a statement of the Public Company Accounting Oversight Board (hereafter PCAOB), that a material weakness in internal control systems results in “more than a remote likelihood that a material misstatement of the annual or interim financial statements will not be prevented of detected” (PCAOB, 2004). I predict that companies with strong internal control systems have a positive influence on the potential relationship between a lack of auditor independence and earnings management. This means that more independent auditors result in a decrease in earnings management and internal control will enhance this relationship.

(14)

Furthermore, the financial crisis has major impacts on financial reporting and provides incentives for earnings management. Since, the financial results deteriorates and managers avoid reporting annual losses or missing analyst forecasts by conducting earnings management with the intention of biasing earnings upward (Roychowdhury, 2006). Hence, my motivation to test the influence of the financial crisis on earnings management. This resulted in the following hypothesis:

H3: The financial crisis has a positive effect on earnings management.

I predict to find evidence that the financial crisis resulted in an increase in the use of earnings management by management. As I expect that the incentives for managers, to manage earnings, increase because Brown and Caylor (2005) find that managers try to avoid negative earnings surprises, which could arise due to the financial crisis.

(15)

2.4. Overview

This paragraph provides an overview of this thesis, depicted in figure 1. The first hypothesis of this thesis investigates the potential relationship between auditor independence and earnings management. Auditor independence is captured by the fee ratio and I expect that a lack of auditor independence will have a positive association with earnings management. Thus, the higher the independency of the auditor the lower the likelihood that a company will use earnings management to manipulate earnings. I focus on real earnings management and specifically on sales manipulation, reduction in discretionary expenditures, and overproduction.

The second hypothesis of this thesis investigates the potential influence of internal control on the potential relationship between auditor independence and earnings management. To capture the quality of internal control I focus specifically on the disclosure of weaknesses under either SOX 302 or SOX 404. It depends on the quality of the internal control system to determine whether internal control has a positive or negative influence on the potential relationship between auditor independence and earnings management.

Finally, I compare two distinct periods with each other; prior to the crisis and during the crisis, in order to verify if the financial crisis have influence on the quantity of earnings management.

(16)

3. Research design (methodology)

This study investigates the relationship between auditor independence and earnings quality in a quantitative manner, with data from US listed organizations only during the period 2006-2013, which enables me to investigate the influence of the crisis.

3.1. Measures to capture auditor independence

To measure the independence of auditors I will focus on the fee ratio. The fee ratio capture the relative monetary value of the audit versus non-audit services (Ashbaugh et al., 2003). So, this proxy form a good measure to capture auditor independence. Since, it enables me to investigate whether the audit fee is more important or the non-audit fee. The fee ratio received support from the SEC as well as several prior academic studies and is calculated as the ratio of audit fee to total fee or as the ratio of non-audit fee to total fee from a client. Auditor independence may be at risk when the amount of non-audit fees become higher relative to audit fees (Reynolds et al., 2004), because non-audit fees are more lucrative (Chung and Kallapur, 2003).

Revised auditor independence rules, implemented by the SEC in November 2000, requires firms to disclose the amounts of all audit and non-audit fees paid to the auditor and therefore enables me to investigate auditor independence with the use of the fee ratio, which will give me a reliable representation of auditor independence.

3.2. Measures to capture real earnings management

Prior literature focused mainly on three different measures to capture real earnings management (Cohen et al., 2008; Cohen and Zarowin, 2010; Ge and Kim, 2014; Kim et al., 2012; Roychowdhury, 2006; Zang, 2012). I have chosen to use all three measures as this will enhance the reliability and comparability of my research. In particular, I follow Roychowdhury (2006) to support my research method.

The first measure is sales manipulation. This measure of real earnings management focus on the sales that deviate from normal business practices, in other words, the abnormal level of cash flow from operations (CFO). Roychowdhury (2006) explains sales manipulation as an attempt made by managers to a tentative increase in sales by offering price discounts or more generous credit terms.

(17)

• Price discounts will probably result in increased sales levels but on the other hand it will lead to abnormally high production costs since the margins decrease due to a discount in sales. • Another kind of discount is more generous credit terms, what basically means that an

organization lowers its interest rates, which leads to lower cash inflows during the life of the sale but will probably result in an increase in sales levels.

So, a boost in sales volumes is preferred by managers as it lead to an increase in earnings, making meeting or beating certain earnings thresholds more and/or easier feasible, otherwise the organization probably missed their earnings thresholds. Roychowdhury (2006) expect that sales manipulation lead to lower CFO in current period and higher production costs compared to the normal sales levels. Equation (1) shows the relevant regression to measure the abnormal levels of cash flows:

CFOt/At-1 = α0 + α1(1/At-1) + β1(St/At-1) + β2(ΔSt/At-1) + εt, (1) Where:

CFOt = cash from operations at the end of period t; At-1 = total assets at the end of last year’s period; St = sales during period t;

ΔSt = difference between current year’s sales and last year’s sales.

The second measure is the reduction of discretionary expenditures. This measure of real earnings management focusses on the abnormal level of discretionary expenses. Organizations can increase earnings by reducing discretionary expenditures like; Research & Development (R&D), advertising expenses, and Selling, General & Administrative expenses (SG&A). Generally these costs are expensed in the period that they occur, and lowering such expenses will increase earnings as the outflow of money is reduced. Roychowdhury (2006) expects that the reduction of discretionary expenditures “has a positive effect on abnormal CFO in the current period, possibly at the risk of lower cash flows in the future” (p. 340). Equation (2) shows the relevant regression to measure the abnormal levels of discretionary expenditures:

DISEXPt/At-1 = α0 + α1(1/At-1) + β(St-1/At-1) + εt, (2) Where:

DISEXPt = discretionary expenses in period t;

At-1 = total assets at the end of last year’s period; St-1 = sales at the end of last year’s period.

(18)

The third measure is overproduction. This measure of real earnings management focus on the production levels of manufacturing firms, in other words, the abnormal level of production costs. Managers organize the production in order to meet the expected demand, generally managers tend to do this as efficiently as possible. However, there are circumstances whereby managers chose to produce more goods than necessary, since this can increase earnings and enables managers to meet or beat earnings benchmarks. Overproduction basically means that managers increase the production levels in order to spread the fixed costs over a larger number of products which leads to a lower fixed cost per product. Cohen et al. (2008) mention that “as long as the reduction in fixed costs per unit is not offset by any increase in marginal cost per unit, total cost per unit declines” (p.765). Roychowdhury (2006) mentioned that “reported Cost of Goods Sold (COGS) is lower, and the firm reports better operating margins” (p. 340). So, managers reporting lower costs of goods sold through increased production. However, there are some consequences related to this approach to manage earnings since it leads to an increase in holding costs to store the over-produced products that not have been sold. Equation (3) shows the relevant regression to measure the abnormal levels production costs:

PRODt/At-1 = α0 + α1(1/At-1) + β1(St/At-1) + β2(ΔSt/At-1) + β3(ΔSt-1/At-1) + εt, (3) Where:

PRODt = COGSt + ΔINVt;

At-1 = total assets at the end of last year’s period; St = sales during period t;

St-1 = sales at the end of last year’s period.

Although, some prior literature (e.g. Kim et al., 2012; Ge and Kim, 2014) combine these three measures in one combined measure to capture earnings management. However, I will use each measure separately, this enables me to find out which measure is frequently used and the impact of each measure.

(19)

3.3. Measures to capture the quality of internal control

Two proxies will help me to capture the quality of internal control systems. The introduction of SOX enables me to measure the quality of organizations internal control systems, since US listed organizations are obliged to reports about their internal control weaknesses using SOX 302 and SOX 404. I refer to a weak internal control system when organizations disclosed a weakness under either SOX 302 or SOX 404 and vice versa.

The assessment of SOX 302 took place each quarter, this leads to four outcomes each year. The quality of an internal control system is not optimal when a material weakness is disclosed in either SOX 302 or SOX 404. Furthermore, when the opinion of the external auditor about management’s assessment about the effectiveness of the internal control systems is negative, I posit that the quality of the internal control systems is weak. I collect both SOX 302 management disclosures of ICDs and SOX 404 auditor opinions from “Audit Analytics” to identify the quality of companies internal control systems. Therefore, I include all types of internal control deficiencies – material weaknesses, significant deficiencies, and control deficiencies. Companies are assigned to the ICMW sample if there is either a weakness disclosure in SOX 302 or SOX 404.

3.4. Control variables

In this study, I used six control variables. Firstly, I included a dummy variable to control for the influence of a Big 4 auditor, since DeAngelo (1981b) mentioned that larger audit firms have more to lose, resulting in better audit quality. The dummy variable takes the value 1 for clients audited by a Big 4 auditor and 0 for clients audited by a non-Big 4 auditor.

Secondly, I control for size because the results of Doyle et al. (2007a) and Ge and McVay (2005) indicate that smaller firms are more likely to disclose material weaknesses.

Thirdly, I included a growth indicator since Doyle et al. (2007a) and Ge and McVay (2005) find that rapidly growing organizations are more likely to disclose material weaknesses in their internal control system compared to organizations with less growth potential. Furthermore, companies with high growth opportunities are more willing to meet earnings forecasts since missing forecasts will affect companies value which could lead to earnings management (Chung and Kallapur, 2003; Roychowdhury, 2006). Market-to-book equity ratio is used as a proxy for growth opportunities which is measured as market value of equity divided by book value of equity.

Fourthly, I control for return on assets (ROA) as Ge and McVay (2005) argue that disclosing material weaknesses is negatively associated with company’s performance. Since, the profitability of companies with material weaknesses is lower than company’s without material weaknesses.

(20)

Fifthly, I included a dummy variable to control for companies who switched from auditor, since the results of Zhang et al. (2007) indicate that companies with recent auditor changes are more likely to have weaknesses in their internal control systems. The dummy variable (AUD_change) takes the value 1 if the company switched from auditor and 0 if the company stay with the incumbent auditor.

Finally, I control for companies undergoing a restructuring, because Doyle et al. (2007a) show that companies undergoing a restructuring are more likely to disclose material weaknesses. This is partly due to a loss of experienced and valuable employees during a restructuring. The dummy variable (Restr) takes the value 1 if the company is undergoing a restructuring and 0 otherwise.

3.5. Empirical models

The following empirical models are estimated to capture the relation between real earnings management and auditor independence, whereby I use three different dependent variables since I capture earnings management with three different measures:

Res_DISEXP = β0 + β1(Fee_Ratio) + β2(Mag_Fees) + β3(Size) +

β4(ROA) + β5(BIG4) + β6(Aud_change) + β7(Restr); (4)

Res_SALESM = β0 + β1(Fee_Ratio) + β2(Mag_Fees) + β3(Size) +

β4(ROA) + β5(BIG4) + β6(Aud_change) + β7(Restr); (5)

Res_PROD = β0 + β1(Fee_Ratio) + β2(Mag_Fees) + β3(Size) +

β4(ROA) + β5(BIG4) + β6(Aud_change) + β7(Restr); (6) Where:

Res_DISEXP = reduction in discretionary expenditures as indicator of real earnings

management, measured by the level of abnormal discretionary expenses, where discretionary expenses are the sum of R&D expenses, advertising expenses, and SG&A expenses;

Res_SALESM = sales manipulation as indicator of real earnings management,

measured by the level of abnormal cash flows from operations;

Res_PROD = overproduction as indicator of real earnings management, measured

by the level of abnormal production costs, where production costs are defined as the sum of cost of goods sold and the change in inventories;

(21)

Fee_Ratio = ratio total audit fee and non-audit fee;

Mag_Fees = total fees scaled by total assets;

Size = firm size, measured as the natural logarithm of total assets;

ROA = return on assets, measured by net income scaled by total assets;

BIG4 = dummy variable, takes the value 0 if audited by non-Big 4 and 1 if

audited by Big 4 (Big 4 = Deloitte, EY, KPMG and PwC);

Aud_change = dummy variable, takes the value 0 if the company did not change

from auditor and 1 if the company did change from auditor;

Restr = dummy variable, takes the value 0 if the company does not

undergoing a restructuring, and 1 if the company does undergoing a restructuring.

The main relation of my research investigates the relation between auditor independence and earnings management. Specifically, the relation between each of the real earnings management measures (i.e. sales manipulation, reduction is discretionary expenditures, and overproduction) and auditor independence. Furthermore, I have include several control variables that potentially influence this relationship as explained in 3.4.

3.5.1. Influence of internal control

To test the influence of internal control on the potential relationship between auditor independence and earnings management I used exactly the same model as depicted above. However, I divided my dataset in two parts; (1) companies with internal control deficiencies (hereafter ICDs), and (2) companies without ICDs, whereby ICDs functions as an indicator for the quality of the companies’ internal control systems. To determine whether a company has ICD(s) I used SOX 302 and SOX 404 disclosures. When a company has either a material weakness disclosure under SOX 302 or under SOX 404 then the company is placed in the category ‘companies with ICDs’. The other group functions as a control group as this group consists of companies with no material weakness disclosure under SOX 302 or under SOX 404. I assume that companies without any disclosure under either SOX 302 or under SOX 404 as companies with high quality internal control systems. However, on the other hand, I assume that companies with one or more disclosures under SOX 302 or SOX 404 as companies with low quality internal control system. So, to capture the influence of internal control I use exactly the same empirical models as mentioned above, thus with three different dependent variables. However, I used the model for both sample in order to compare

(22)

the outcomes of the regressions with each other and to compare the three dependent variables (i.e. sales manipulation, reduction in discretionary expenditures, and overproduction).

3.5.2. Influence of the Financial crisis

Finally, I test whether the financial crisis have influence on the use of earnings management. The financial crisis resulted in a deterioration of companies financial results. Probably leading to an increase in earnings management, since incentives to use earnings management increase because management try to avoid reporting annual losses or missing analyst forecasts. In order to determine whether the financial crisis stimulated the use of earnings management I divided my dataset in two parts; (1) companies’ results during the financial crisis (i.e. 2008-2013), and (2) companies’ results before the financial crisis (2006-2007). Once again, the same empirical models are used for both separate datasets. In order to see which form of earnings management is more pronounced during the crisis, I use again three different dependent variables (i.e. sales manipulation, reduction in discretionary expenditures, and overproduction). Additionally, I controlled for fixed year effects in all the regressions I undertake.

(23)

4. Evidence

In this chapter the results of multiple regressions are presented. This chapter starts with my sample distribution, followed by descriptive statistics and finally I will deal with auditor independence and earnings management, internal control and earnings management, and the financial crisis, internal control and earnings management, in section 4.3, 4.4 and 4.5 respectively.

4.1. Sample

The sample consists of US listed companies for the years 2006 through 2013. I focus specifically on the period after the introduction of SOX, since the use of discretionary accruals shows a significant decline after its implementation. Besides, the use of real earnings management shows a significant increase after SOX’s implementation (Cohen et al., 2008). Furthermore, I have chosen this particularly timeframe since it allows me to compare the results before the financial crisis with the results during the financial crisis. I have collected my data for auditor independence (i.e. audit and non-audit fees) and internal control (i.e. SOX 302 and SOX 404) from the database ‘Audit Analytics’. Additionally, I use data from the database ‘Compustat’ in order to calculate real earnings management (i.e. sales manipulation, reduction in discretionary expenditures, and overproduction). Subsequently, I merged both datasets, yielded in 33.933 firm-year observations. I excluded 7.691 firm-year observations in financial services industries (SIC codes 6000-6900) because real earnings activities are likely to be different in comparison with firms in other industries. Also, I excluded 595 firm-year observations because of data availability regarding real earnings management. Above that I winsorized the bottom and top 1% to mitigate for the effects of extreme observations (i.e. outliers). Resulting in a total sample of 25.647 firm-year observations of US listed companies. My sample selection is summarized in table 1.

Table 1 is an overview of the sample distribution by the two-digit SIC code industry. The two most heavily represented industries are; (1) Chemicals and Allied Products (12.57 percent), and (2) Business Services (12.45 percent), followed by Electronic and Other Electric Equipment (9.30 percent), and Instruments and Related Products (7.57 percent).

(24)

Industry Two-Digit

SIC # of Obs. % of Sample

Metal Mining, Ores 10 324 1.26

Oil and Gas 13 1,318 5.14

Food, Beverage 20 619 2.41

Apparel and Other Textile Products 23 203 0.79

Lumber and Wood Products 24 118 0.46

Furniture and Fixtures 25 151 0.59

Paper and Allied Products 26 222 0.87

Printing and Publishing 27 230 0.90

Chemicals and Allied Products 28 3,225 12.57

Petroleum 29 182 0.71

Rubber 30 207 0.81

Primary Metal Industries 33 319 1.24

Fabricated Metal Products 34 263 1.03

Industrial Machinery and Computer Equipment 35 1,449 5.65

Electronic and Other Electric Equipment 36 2,386 9.30

Transportation Equipment 37 618 2.41

Instruments and Related Products 38 1,942 7.57

Miscellaneous Manufacturing 39 238 0.93

Trucking and Warehousing 42 176 0.69

Water Transportation 44 141 0.55

Air Transportation 45 208 0.81

Communication 48 886 3.45

Electric, Gas, Sanitary Services 49 1,572 6.13

Wholesale-Durable Goods 50 576 2.25

Wholesale-Non-Durable Goods 51 385 1.50

General Merchandise Store 53 166 0.65

Auto Dealers, Gas Stations 55 139 0.54

Apparel and Accessory Stores 56 328 1.28

Eating and Drinking 58 395 1.54

Miscellaneous Retail 59 535 2.09

Business Services 73 3,193 12.45

Amusement and Recreation Services 79 329 1.28

Health Services 80 543 2.12

Educational Serives 82 153 0.60

Engineering and Management Services 87 539 2.10

Other 1,369 5.34

Total 25,647 100

TABLE 1

Sample Description:

(25)

4.2. Descriptive statistics

The descriptive statistics are reported in Table 2 and are divided over three different panels. Panel A present the descriptive statistics of the full sample, panel B present the differences in descriptive statistics of companies with Internal Control Deficiencies (hereafter ICD) and companies without ICD, Table 2 concludes with panel C where differences between descriptive statistics before the crisis with the descriptive statistics in the crisis are presented.

Panel A of Table 2 show a mean value of 0.0054, -0.0022, and 0.0001 for the absolute value of Res_DISEXP, Res_SALESM, and Res_PROD, respectively. These mean values suggest that, on average, companies are more likely to engage real earnings management through a reduction in discretionary expenditures, and is similar to findings in prior literature (Kim et al., 2012). The mean value of the Fee_Ratio indicate that 14.12 percent of the companies had higher non-audit fees in relation to audit fees. Furthermore, in my sample 13.63 percent of the companies have internal control deficiencies (hereafter ICD) and almost 75 percent of the sample is during the financial crisis. The control variables indicate that 65.71 percent of the companies are audited by a Big 4 accounting firm, 46.92 percent of the sample switched from auditor in the period from 2006 through 2013, and 27.71 percent of the firms are undergoing a restructuring.

Observations Mean Median Std. Dev. Min Max Dependent Variables Res_DISEXP 25647 0.0054 - 0.0570 0.8160 - 8.0240 9.0073 Res_SALESM 25647 - 0.0022 0.0297 0.4116 - 4.9019 4.2434 Res_PROD 25647 0.0001 - 0.0091 0.3956 - 3.6476 4.8360 Variables of Interest Fee_Ratio 25647 0.1412 0.1009 0.1493 0 1 Magnitude_Fees 25647 0.0138 0.0026 0.0541 0.0001 0.4727 Control Variables Size 25647 4.6953 5.0201 2.5542 - 3.8167 9.7080 ROA 25647 - 1.1140 0.0561 5,6898 - 46.3268 1.2262 Big4 25647 0.6571 1 0.4747 0 1 Auditor_Change 25647 0.4692 0 0.4991 0 1 Restructure 25647 0.2771 0 0.4476 0 1

(Continued on next page)

TABLE 2 Descriptive Statistics Panel A: Full Sample

(26)

The descriptive statistics in panel B indicate that companies with ICD have higher Res_DISEXP and Res_PROD, indicating that companies with ICD are more likely to engage real activities manipulations through a reduction in their discretionary expenditures and overproduction relative to companies without ICD.

Observations Mean Median Observations Mean Median t-test Dependent Variables Res_DISEXP 3496 0.2402 - 0.0246 22151 - 0.0316 - 0.0601 - 10.3343*** - 9.694 *** Res_SALESM 3496 - 0.1435 - 0.0148 22151 0.0201 0.0348 13.2505*** 19.977 *** Res_PROD 3496 0.0390 - 0.0001 22151 - 0.0060 - 0.0105 - 4.0929 *** - 3.523 *** Variables of Interest Fee_Ratio 3496 0.1235 0.0584 22151 0.1440 0.1067 6.6915 *** 17.131 *** Mag_Fees 3496 0.04645 0.0081 22151 0.0086 0.0022 - 20.6407*** - 46.575 *** Control Variables Size 3496 2.5764 2.9635 22151 5.0298 5.2363 45.0470*** 44.454 *** ROA 3496 - 5.0888 - 0.2845 22151 - 0.4867 0.0740 22.8397*** 41.113 *** Big4 3496 0.3538 0 22151 0.7050 1 40.6045*** 40.655 *** Aud_Change 3496 0.7303 1 22151 0.4280 0 - 36.8176*** - 33.284 *** Restr 3496 0.2031 0 22151 0.2887 0 11.4895*** 10.517 ***

Observations Mean Median Observations Mean Median t-test

Dependent Variables Res_DISEXP 19165 0.0056 - 0.0638 6482 0.0050 - 0.0389 - 0.0642 8.084 *** Res_SALESM 19165 - 0.0027 0.0318 6482 - 0.0008 0.0234 0.3497 - 4.689 *** Res_PROD 19165 0.0005 - 0.0101 6482 - 0.0009 - 0.0059 - 0.2530 2.027 ** Variables of Interest Fee_Ratio 19165 0.1398 0.0991 6482 0.1455 0.1053 2.6155 *** 4.473 *** Magnitude_Fees 19165 0.0157 0.0026 6482 0.0081 0.0025 - 12.9401*** - 3.205 *** Control Variables Size 19165 4.5506 4.9077 6482 5.1234 5.2727 17.5421*** 13.426 *** ROA 19165 - 1.3221 0.0435 6482 - 0.4990 0.0884 12.7038*** 14.751 *** Big4 19165 0.6263 1 6482 0.7484 1 19.0040*** 17.900 *** Auditor_Change 19165 0.4851 0 6482 0.4221 0 - 8.8519 *** - 8.787 *** Restructure 19165 0.2783 0 6482 0.2734 0 - 0.7714 - 0.769

*, **, *** Indicate statistical significance at the 0.10, 0.05, and 0.01 levels, respectively, based on two-tailed tests. Variables are defined in Appendix A.

All test statistics and significance levels are calculated based on the standard errors adjusted by a two-dimensional cluster at the firm and year levels.

Wilcoxon Test

Panel C: Descriptive Statistics by Crisis versus Non-Crisis

Crisis No-Crisis

Difference Tests: Panel B: Descriptive Statistics by ICD Firms versus Non-ICD Firms

(27)

Furthermore, companies without ICD show a higher mean value of Res_SALESM, indicating that the companies with ICD are, on average, more likely to conduct sales manipulations compared to the sample with ICD. However, the amount is negative, indicating that the sales manipulation resulted in a lower sales instead of an increase in sales. Companies without ICDs have a higher Fee_Ratio, indicating, ceteris paribus, that either the non-audit fees increased or the total fees decreased compared to companies with ICD. This is an indication that auditors are less independent in the sample of companies without ICD and more independent in the sample of companies with ICD. So, when the auditor is more independent the likelihood increase that the company is identified with ICD, this is consistent with results of prior literature (e.g. Zhang et al., 2007). However, the magnitude of fees scaled by total assets is higher for companies with ICD, suggesting that companies with ICD pay more total fees relative to their total assets than companies without ICD. The control variables in panel B indicate that companies with ICD are relatively smaller, which is consistent with prior literature (Doyle et al., 2007a; Ge and McVay, 2005), are less profitable, also consistent with prior literature (Ge and McVay, 2005), are less likely to be audited by a Big 4 audit firm, and are more likely to change from auditor, which is also consistent with findings in prior literature (e.g. Zhang et al., 2007). Although, these companies have a higher likelihood to face a restructuring, which is inconsistent with prior literature since Doyle et al. (2007a) find evidence that companies with ICD are more likely to face a restructuring.

The descriptive statistics in panel C are divided in two categories, before the crisis (i.e. 2006 and 2007) and during the crisis (i.e. 2008-2013). The results indicate that during the crisis more companies manage their earnings through real activities manipulations like reducing their discretionary expenditures and overproduction. For overproduction the amount is negative, which can be a result of higher storage costs. Also, Res_SALESM is higher during the crisis, indicating that there is more earnings manipulation during the crisis compared with the results before the crisis. However, sales manipulation resulted in a reduction in sales since the amount of sales manipulation is negative1_{. Furthermore, companies have a higher Fee_Ratio before the crisis, indicating that the} auditor is more independent during the crisis because, ceteris paribus, either the non-audit fees decreased relative to the total fees or the total fees increased. This is an indication that auditors are more independent because they earn less non-audit fees and especially these fees could affect auditors independency because these fees are more lucrative and therefore affect the quality of the services rendered for the audit fees. However, the magnitude of fees scaled by total assets is higher during the crisis, suggesting that companies pay more total fees relative to their total assets than

1_{Due to the simplicity of the T-test I also conducted an additional test to control. The T-test did not find significant}

differences in real activities manipulations, however the Wilcoxon Test did, because this test is more extensive and rank the sample where the T-test does not rank the data.

(28)

Re s_ DI S E XP Re s_ S A L E S M Re s_ P RO D F ee _ R at io M ag ni tu d e_ F ee s Q ua lit y_302 Q ua lit y_404 S iz e RO A BI G 4 A udi tor _ ch an ge R es tr uc tu re R es _ DI S E XP -0.6279 -0.2219 -0.0043 0.0839 -0.0894 -0.1128 -0.1706 -0.1931 -0.1189 0.0669 -0.0682 0.0000 0.0000 0.4885 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 R es _ S A L E S M -0.3827 -0.2945 0.0132 -0.1116 0.1100 0.1245 0.2064 0.2174 0.1495 -0.0929 0.0613 0.0000 0.0000 0.0328 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 Re s_ P RO D -0.3703 -0.3339 -0.0149 0.0096 -0.0339 -0.0260 -0.0449 -0.1104 -0.0395 0.0266 -0.0147 0.0000 0.0000 0.0157 0.1202 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0170 F ee _ R at io -0.0278 0.0531 -0.0081 -0.0676 0.0480 0.0478 0.1400 0.0728 0.0920 -0.0554 0.0531 0.0000 0.0000 0.1957 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 M ag ni tu d e_ F ee s 0.1347 -0.1770 -0.0568 -0.1297 -0.2245 -0.2607 -0.5280 -0.6428 -0.2616 0.1474 -0.1093 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 Q ua lit y_302 -0.0470 0.1048 -0.0179 0.1000 -0.2543 0.6647 0.2771 0.2415 0.2172 -0.1847 0.0508 0.0000 0.0000 0.0041 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 Q ua lit y_404 -0.0604 0.1205 -0.0144 0.1050 -0.2851 0.6639 0.3409 0.2920 0.2615 -0.2029 0.0766 0.0000 0.0000 0.0212 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 S iz e -0.1459 0.2008 0.0395 0.2316 -0.7769 0.2303 0.2819 0.5237 0.6576 -0.3615 0.3186 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 RO A -0.1081 0.3418 -0.1352 0.1500 -0.5431 0.2185 0.2516 0.4918 0.2615 -0.1555 0.1041 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 BI G 4 -0.0792 0.1462 0.0080 0.1704 -0.4870 0.2136 0.2572 0.6671 0.3395 -0.4582 0.2530 0.0000 0.0000 0.2020 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 A ud ito r_ ch an ge 0.0483 -0.1159 0.0113 -0.1311 0.2836 -0.1826 -0.2003 -0.3666 -0.2445 -0.4560 -0.1046 0.0000 0.0000 0.0707 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 R es tr uc tu re -0.0879 0.0566 0.0042 0.0940 -0.1189 0.0495 0.0746 0.3291 0.0049 0.2535 -0.1053 0.0000 0.0000 0.5063 0.0000 0.0000 0.0000 0.0000 0.0000 0.4366 0.0000 0.0000 TA BL E 3 P ear son C or re lat ions ( top) and Spe ar m an C or re lat ions ( bot tom )

(29)

before the crisis. The control variables in panel C indicate that during the crisis the companies are relatively smaller, less profitable, and less companies are audited by a Big 4 accounting firm compared to the results before the financial crisis. However, there are no significant differences in restructuring before and during the financial crisis.

Table 3 presents the pair-wise correlations, where the lower-left-hand portion displays the Spearman correlations and the upper-right-hand portion displays the Person correlations for all variables, although the correlations are generally consistent with each other. The Fee_Ratio is negatively correlated with Res_DISEXP and Res_PROD, however, the Fee_Ratio is positively correlated with Res_SALESM. In addition, both measures of internal quality (i.e. Quality302 and Quality404) have a highly significant correlation with all three real activities manipulations (i.e.

Res_DISEXP, Res_SALESM, and Res_PROD). This evidence suggests that companies with strong

internal control systems are less likely to engage in real earnings management compared to companies with a weak internal control system.

4.3. Auditor independence and earnings management

The results of multivariate regression analyses are reported in Table 4. Together with the test statistics I also shows the significance levels based on the standard errors adjusted by a two-dimensional cluster at the company and year levels, because the residuals can be correlated across company and/or over time (Kim et al., 2012).

The results in Table 4 are significant for DISEXP and SALESM, this indicates that the independent variables explain a significant portion of companies real earnings management. The

Fee_Ratio is positive and significant related with DISEXP, this means that when the Fee_Ratio

increases it will lead to higher magnitude of real earnings manipulation through a reduction of discretionary expenditures. This result indicate that when the auditor become more dependent on the client (since the audit firm receives more non-audit fees) it will lead to a higher magnitude of earnings manipulation by reducing discretionary expenditures. However, for SALESM the results are significant and negative, meaning that when the auditor become more independent (since the audit firm receive less non-audit fees) the company is more likely to participate in sales manipulation. This is not according to my expectations. The results for overproduction are not significant. So, H1 is only partly supported by the results since I find evidence for the relationship between a lack of auditor independence and real earnings management, in particular for DISEXP. However, the variable

Mag_Fees indicates that when the total fees in relation to total assets decrease, the magnitude of

(30)

Furthermore, I find evidence that smaller companies are more likely to commit earnings management by reducing their discretionary expenditures. Although, this does not apply for sales manipulation, since the results indicate that larger and more profitable companies are more likely to manipulate earnings by sales manipulation. Additionally, I find evidence for earnings manipulation through a reduction in discretionary expenditures and overproduction are more likely committed by less profitable companies and it is more likely that they are audited by non-Big 4 audit firms.

4.4. Internal control and earnings management

To investigate the influence of internal control I divided my sample in two parts. The first sample consists of companies with weak internal control systems and the other sample is a control sample consisting of companies with strong internal control systems. Table 5 present the results of multiple regression of real activities manipulation for companies with and without internal control deficiencies. DISEXP Coefficient (t-stat) SALESM Coefficient (t-stat) PROD Coefficient (t-stat) Fee_Ratio 0.109 -0.044 -0.22 (2.39) ** (-1.91) * (-0.93) Mag_Fees -1.311 0.488 -0.708 (-2.30) ** (1.73) * (-2.98) *** Size -0.026 0.016 0.003 (-4.33) *** (5.44) *** (1.01) ROA -0.028 0.014 -0.012 (-6.21) *** (6.47) *** (-6.14) *** Big4 -0.073 0.048 -0.023 (-3.99) *** (5.22) *** (-2.18) ** Aud_change -0.012 -0.018 -0.006 (0.74) (-2.32) ** (0.76) Restr -0.048 0.006 -0.005 (-5.57) *** (1.27) (-0.99)

Variables are defined in Appendix A.

*, **, *** Indicate statistical significance at the 0.10, 0.05, and 0.01 levels, respectively, based on two-tailed tests.

TABLE 4

Multiple Regression of Real Activities Manipulation (n = 25.647)

(31)

The results indicate that companies with ICD are more likely to engage in sales manipulation when the Fee_Ratio declines and vice versa, however, again the results are negative meaning that it lead to lower sales instead of higher sales. The sample of companies without internal control deficiencies show other results, as in this sample significant results are shown for DISEXP (positively) and PROD (negatively), meaning that when the Fee_Ratio increase, earnings manipulation by reducing of discretionary expenditures also increase, and earnings manipulation through overproduction decrease. Furthermore, in the sample of companies with no ICD I find significant and negative relation between Mag_Fees and DISEXP and PROD, meaning that when the total fees in relation to total assets decrease earnings management through DISEXP and PROD increases and vice versa. This indicates that when companies paying less fees the amount of earnings management through DISEXP and PROD increases. So, the results indicate that hypothesis two is only partly supported. Additionally, reducing discretionary expenditures and overproduction are more likely to be used by smaller companies. The results regarding ROA are for both samples negative and highly significant which allows me to compare both samples and since the coefficients are exactly the same they both show that less profitable companies conduct more earnings management through a

DISEXP Coefficient (t-stat) SALESM Coefficient (t-stat) PROD Coefficient (t-stat) DISEXP Coefficient (t-stat) SALESM Coefficient (t-stat) PROD Coefficient (t-stat) Fee_Ratio 0.259 -0.149 0.039 0.081 -0.020 -0.038 (1.43) (-1.81) * (0.51) (2.18) ** (-1.03) (-1.76) * Mag_Fees -0.466 -0.080 -0.381 -2.092 1.016 -1.022 (-0.66) (-0.26) (-1.17) (-2.38) ** (2.17) ** (-3.44) *** Size 0.040 -0.013 0.009 -0.034 0.018 0.004 (1.56) (-1.08) (1.07) (-6.83) *** (7.01) *** (1.23) ROA -0.029 0.010 -0.006 -0.029 0.022 -0.021 (-5.02) *** (3.77) *** (-2.48) ** (-3.97) *** (6.08) *** (-6.03) *** Big4 -0.334 0.186 -0.061 -0.022 0.020 -0.016 (-4.58) *** (5.37) *** (-2.24) ** (-1.38) (2.45) ** (-1.41) Aud_Change -0.117 0.030 0.029 0.022 -0.020 0.001 (-1.51) (0.81) (1.04) (1.68) * (-2.92) *** (0.18) Restr -0.169 0.048 -0.029 -0.033 0.002 -0.005 (-1.28) *** (2.17) ** (-1.15) (-4.10) *** (0.51) (-0.93) TABLE 5

*, **, *** Indicate statistical significance at the 0.10, 0.05, and 0.01 levels, respectively, based on two-tailed tests.

Variables are defined in Appendix A.

Multiple Regression of Real Activities Manipulation

ICD Firms (n = 3.496) No-ICD Firms (n = 22.151)

𝑅𝑒𝑠_𝐷𝐼𝑆𝐸𝑋𝑃 / Res_SALESM / Res_PROD =

(32)

reduction in discretionary expenditures, however, the t-statistic for ICD companies is higher which indicates that companies with ICD are more likely to reduce discretionary expenditures in order to manipulate earnings. According to sales manipulation and overproduction the results correlate with

ROA, however they are stronger for companies without ICD. Finally, companies with ICD are more

likely to manipulate earnings through a reduction in discretionary expenditures compared to companies without ICD when facing a restructure.

4.5. Earnings management and the financial crisis

The final test investigates whether companies are more likely to manipulating earnings during crisis by comparing two samples, one before the crisis and the other during the crisis. According to the results I can conclude that during the crisis there are significant relations between the Fee_Ratio and DISEXP and SALESM. Since, the results do not show significant results between earnings management and the Fee_Ratio I can conclude that hypothesis three is supported by the results in table six. Furthermore, the results regarding Mag_Fees indicate that before the crisis there was a stronger negative relation with DISEXP and PROD. These results suggests that the lower the total fees in relation to total assets, the more earnings management is committed. Additionally, the results show that smaller, less profitable companies are more likely to engage in earnings management. Finally, the results show that there is a stronger relationship between Big4 and DISEXP during the crisis than prior to the crisis. This indicates that it is more likely that during the crisis, the companies that conducted earnings management were not audited by a Big 4 accounting firm.

Impact of auditor independence on earnings management : the role of internal control

Amsterdam Business School

Master Thesis 2014: