The relation between non-audit services and the quality of financial reporting in the Netherlands

The Relation between Non-Audit Services and the Quality of Financial Reporting in the Netherlands

Daan Wildschut University of Amsterdam

June, 19, 2016 S.W. Bissessur

Verklaring eigen werk

Hierbij verklaar ik, Daan Wildschut, dat ik deze scriptie zelf geschreven heb en dat ik de volledige verantwoordelijkheid op me neem voor de inhoud ervan.

Ik bevestig dat de tekst en het werk dat in deze scriptie gepresenteerd wordt origineel is en dat ik geen gebruik heb gemaakt van andere bronnen dan die welke in de tekst en in de referenties worden genoemd.

De Faculteit Economie en Bedrijfskunde is alleen verantwoordelijk voor de begeleiding tot het inleveren van de scriptie, niet voor de inhoud.

Abstract

This paper examines whether the cross selling of non-audit services to audit clients, results in lower quality of financial reporting in the Netherlands. Previous empirical literature that investigated the relationship between non-audit fees and financial reporting quality showed mixed results. Using abnormal working capital accruals as a surrogate for earnings management, we found positive relations between total non-audit fees and absolute abnormal working capital accruals, and between total non-audit fees and positive abnormal working capital accruals. Next, we found negative relations between total non-audit fees and negative abnormal working capital accruals, and positive relations between audit-related fees and positive working capital accruals. Finally, we found positive relations between tax fees and abnormal working capital accruals, and negative relations between tax fees and negative abnormal working capital accruals. Our evidence suggests that cross selling total non-audit services, audit-related services and tax services to audit clients is associated with higher levels of earnings management. The results of this study are distinctive as most prior empirical research finds no or less conclusive evidence that cross selling of non-audit services impairs financial reporting quality.

Contents

H1 Introduction ... 5

1.1 Background ... 5

1.2 Research Question ... 6

1.3 Research Approach, Results and Contribution ... 8

1.4 Structure ... 9

H2 Literature Review ... 9

2.1 Introduction ... 9

2.2 Auditing and Agency Conflict ... 9

2.3 Audit Quality ... 10

2.4 Earnings Management ... 11

2.5 Non-Audit Fees ... 14

2.6 Relation Non-Audit Fees and Financial Reporting Quality ... 16

H3 Development of Hypotheses and Research Methodology ... 19

3.1 Development of Hypotheses ... 19 3.2 Research Method ... 21 H4 Empirical Results ... 27 H5 Conclusion ... 42 References ... 45 APPENDIX A ... 52

The Relation between Non-Audit Services and the Quality of Financial Reporting in the Netherlands

H1 Introduction

1.1 Background

In October 2010, in the aftermath of the global financial crisis of 2008 that rocked the foundations of the world’s economies, the European Commission has published its final Green Paper on Audit Policy: Lessons from the Crisis. According to Michel Barnier, European Commissioner for Internal Market and Services, the financial crisis highlighted failings in the audit sector that needed to be addressed (EC, 2010). The main objective of the Green Paper was to initiate a debate on the role and governance of auditors (Humphrey et al., 2011). One of the key issues addressed in the Green Paper are the issues around the

independence of audit firms in relation to the audited company. In the Green Paper radical reforms were proposed that would change the shape of the audit market, including a ban on the provision of non-audit services to audit clients. This was believed to enhance the quality of audits, and as a result the quality of financial statements of public interest companies. After a public consultation of the participants, including representatives of the audit profession, of companies, of investors and of business associations, the commission started preparing the new legislation. Late 2011 the European commission adopted proposals for a regulation on the quality of audits of public-interest entities and for a directive to enhance the market for audits. By the end of 2013 European Union Member States and the European Parliament reached an agreement on a revised Directive and a new Regulation on audit. The new rules have entered into force in 2014 (EC, 2014). An important part of the Regulation are the severe restrictions that have been imposed on cross selling of non-audit and audit services to public interest companies.

In anticipation of the new European legislation the Dutch government promulgated a number of amendments to the Audit Firms Supervision Act (“Wet toezicht

accountantsorganisaties” (2012). One of these amendments, Section 24b, prohibits audit firms to cross sell non-audit services to public interest entities (PIE) in the Netherlands as per January 2013. This prohibition pertains to audit firms conducting statutory audits of PIE in the

Netherlands. The effective date is January 2013, although there is a transitional provision for activities for which the audit firm was already engaged prior to 2013.

The regulatory changes aiming to promote auditor independence that have been

introduced in the Netherlands and the rest of the European Union, followed the legislation that was already enacted a decade earlier in the United States of America (U.S.) after a series of large bookkeeping scandals in the early stages of the twentieth century. These bookkeeping scandals caused some of the largest bankruptcies in corporate history with Enron being the most prominent example. The Enron case also led to the collapse of the firm’s auditor Arthur Andersen. In conjunction with audit services, Arthur Andersen also provided a significant portion of non-audit services to Enron. In 2001, the year before Enron went bankrupt, the fees Arthur Andersen received relating to non-audit services were even higher than the audit fees: $ 27 million vs $ 25 million (Financial Times 2002).

The perception among regulators and the general public is that this cross selling of audit and non-audit services by audit firms strengthens the economic bond between the auditor and his client. This increases the incentive for the auditor to acquiesce to client pressure to allow earnings management (Simunic 1984) and as a consequence results in impaired quality of financial reporting. The Sarbanes-Oxly act, which was enacted in the U.S. after the Enron scandal, therefore imposes severe restrictions on cross selling of non-audit and audit services to listed companies.

1.2 Research Question

The provision of non-audit services by audit firms to their audit clients has been one of the most hotly debated audit independence issues in the past three decades with regular

features in the mainstream media (e.g. Financial Times 2002; The Economist 2010; The Economist 2011; Financial Times 2015a; Financial Times 2015b). This has resulted in extensive research in this area (e.g. Frankel et al.; 2002; Ashbaugh et al. 2003; Kinney et al. 2004; Schneider et al., 2006; Paterson et al, 2011; Habib, 2012; Causholli et al., 2014). And although the legislation surrounding the practice of cross selling non-audit services to audit clients has become more and more stringent, the analytical theory that covers this issue is less conclusive.

Strong arguments have been made that support the theory that auditor independence is impaired by the provision of non-audit services due to the economic bond between the auditor and his client (Simunic, 1984). Economic bonding is also referred to as economic

dependence. Economic bonding is the result of cost savings (economies of scope) which arise because of the joint provision of audit and non-audit services by an auditor. There are two types of economies of scope: contractual economies of scope and knowledge spillovers (Simunic, 1984; Beck et al. 1988a; DeBerg et al.1991; Arruñada, 1999). The fact that these cost savings are partly retained by the auditor causes a threat to his independence. Auditor independence is one of the key attributes of audit quality. Although the concept of auditor independence is difficult to define precisely, most definitions of auditor independence that can be found in literature reflect objectivity and integrity as its two key aspects (Antle 1984; Knapp 1985; Magill and Previts 1991; Dunmore and Falk, 2001; Baker 2005). Economic bonding can be illustrated by an auditor who is concerned about the possible loss of revenues from non-audit services. Under those circumstances it is less likely that he will object to earnings management since this could result in a deteriorated relationship with the client.

In contrast, also arguments have been made that the cross selling of non-audit services increases the auditor’s knowledge of the client’s business (“knowledge spillover”) resulting in higher competency and therefore higher audit quality (Goldman and Barlev, 1974; Simunic 1984; Wallman 1996)).

In addition litigation risk and reputational risk could deter auditors from succumb to earnings management practices. Possible litigation claims can be substantial which gives the auditor an incentive to reduce the risk of material misstatement (Simunic, 1980; Frankel et al. 2002, DeFond and Zhang, 2014). Furthermore, the fear of losing clients due to the

reputational damage that occurs after an audit failure creates a reputation incentive for the auditor to deliver high quality audits (Arruñada, 1999; Weber et al. 2008; Skinner and Srinivasan, 2012; DeFond and Zhang, 2014).

The extensive empirical research in this field provides mixed evidence. A few studies find evidence which suggests that cross selling of non-audit services impairs real auditor independence due to the economic bonding (e.g. Frankel et al., 2002; Causholli et al., 2014). In contrast, most studies indicate that knowledge spillover cancels out economic bonding (e.g. Ashbaugh et al. 2003; Chung and Kallapur, 2003; Larcker and Richardson, 2004; Reynolds, et al., 2004; Habib, 2012).

Other researchers make a distinction between the various types of non-audit services in order to examine differences in their relationship with auditor independence. Their findings indicate that certain non-audit services actually strengthen auditor independence (Kinney et al. 2004; Huang et al., 2007; Paterson et al., 2011; Krishnan and Visvanathan, 2011), whereas

other types of non-audit services show a negative association (Kinney et al. 2004; Paterson et al., 2011).

Given this long standing debate whether cross selling non-audit services to audit clients impairs audit quality, which in turn affects the quality of financial reporting, this remains an interesting area for research. As from 2013 there have been regulatory changes in the Netherlands which further prohibit cross selling of non-audit services to audit clients classified as a public interest entity. The introduction of this new legislation in the Netherlands provides a good occasion to obtain further evidence. This paper therefore investigates the following research question:

- Does cross selling of non-audit services to audit clients result in lower quality of financial reporting in the Netherlands?

1.3 Research Approach, Results and Contribution

In the present study we investigate whether the cross selling of non-audit services to audit clients, results in lower quality of financial reporting in the Netherlands.

We use the level of abnormal accruals as a proxy for earnings management and we use the abnormal working capital accrual model of DeFond and Park (2001) to determine the level of abnormal accruals.

We collected fee data from companies listed on the Amsterdam stock exchange in 2012 and 2013. Besides the relationship between total non-audit fees and earnings

management, we also examined the relationship between the three different types of non-audit services (audit-related services, tax services, and other services) and earnings management separately.

We found evidence of a positive association between total non-audit fees and

abnormal accruals and tax fees and abnormal accruals. Furthermore, we found some evidence for a positive association between audit-related fees and abnormal accrual. Our results thus indicate that cross selling total non-audit services, audit-related services and tax services to audit clients is associated with higher levels of earnings management in the Netherlands.

This study is distinctive and contributes to prior research as it investigates data from the Netherlands that has become available after the introduction of legislation in 2013 which further prohibits cross selling. Moreover, we found evidence of an association between the

cross selling of non-audit services and impaired financial reporting quality, whereas most prior empirical research finds no, or less conclusive evidence.

1.4 Structure

The structure of this paper is as follows. First, an in depth evaluation of previous research concerning the cross selling of non-audit services and financial reporting quality is provided. Second, hypothesis are developed using abnormal working capital accruals as a surrogate for financial reporting quality and also taking into account the fee disclosure requirements in the Netherlands. Third, the research method is presented, including sample description, the method of collection, the chosen non-audit fee measures, the abnormal working capital accruals model and the control variables. Fourth, the empirical results are provided. Finally, the conclusion regarding the hypothesis is provided together with a reflective analysis of the research, its limitations and also suggestions for further research.

H2 Literature Review

2.1 Introduction

In order to provide more theoretical background with respect to the research question, the concepts of financial reporting, audit, audit quality, auditor independence, earnings management and the provision of non-audit services and how these concepts are intertwined will be discussed below.

2.2 Auditing and Agency Conflict

Financial reporting plays a vital role in the world’s economies. It provides financial information to the various stakeholders of a company, including its shareholders. For larger companies there is typically a division between ownership and management: directors are appointed on behalf of the shareholders. The financial reporting which is being prepared under the responsibility of management, enables the investors to monitor the performance of the company.

However, conflicts of interest can arise between management of the company and its shareholders. Agency theory suggests that an agency problem exists as managers will seek to maximize their own utility at the expense of the shareholders. Management has the ability to operate in its own self-interest rather than in the best interest of the shareholders because of uncertainty and asymmetric information. The nature of the agency theory has been described by Eisenhardt (1989).

The agency problem might lead to concern about the reliability of financial reporting as it is being prepared under the responsibility of management. There are various mechanisms that may be used to try to align the interests of management with the shareholders and to allow shareholders to measure and control the behavior of management and reinforce trust.

One of these mechanisms is an audit (DeFond and Zhang, 2014). Audits are conducted by an external auditor and provide verification of the financial statements prepared by

management, which helps to maintain confidence and trust. In essence the auditor provides assurance to the shareholders on the truth and fairness of the financial statements (Quick, 2012).

According to Jensen and Meckling (1976) an audit serves both as a bonding mechanism and a monitoring device. The bonding mechanism limits the discretion of

management to manage earnings (e.g. it will prevent aggressive revenue recognition policies). From the perspective of the shareholder its serves as a monitoring device as it enables them to monitor management.

2.3 Audit Quality

An essential element for enriching the credibility of financial statements by external auditors is audit quality (DeFond and Zhang, 2014). The concept of audit quality is often defined in literature as a combination of two characteristics associated with the external auditor: the technical ability to identify misstatements, and independence required for the correction of misstatements (DeAngelo 1981). Because of this, independence is considered to be an important attribute of external auditors. Regulators around the world have rules that require auditors to maintain their independence. However, there is no concise definition of the concept and throughout the years it has been subject to constant reinterpretation (Antle 1984; Baker 2005). A number of definitions of auditor independence which can be found in

- ”The ability to resist client pressure” (Knapp 1985)

- “A function of character, with the attributes of integrity and trustworthiness being key (Magill and Previts 1991)

- “Independence is taken to be the willingness of the auditor to reflect in the audit report all problems and defects he has detected in the financial statements” (Arruñada 1999)

It is generally agreed that a distinction can be made between auditor independence in fact (also referred to as real independence) and auditor independence in appearance (also referred to as perceived independence) (Beattie and Fearnley 2002). Independence "in fact" is defined as the absence of mental bias in the conduct of an audit. Independence in appearance is the perception of auditor independence that the users of financial information have (Ghosh et al. 2009). Both are essential for the audit mechanism to be effective (Beattie et al. 1999). As such auditing standards require both.

Literature suggests that there are various factors that can affect auditor independence consciously or unconsciously. Beattie et al. (1999) identify “four principal factors to affect auditor independence: the economic dependence of the auditor on the auditee, competition within the external audit market, the provision of non-audit services by the auditor, and the degree of laxity of the regulatory framework”. The cross selling of non-audit services by the auditor has been debated most intensively the past decades and is therefore the factor which is further investigated in this paper.

2.4 Earnings Management

Agency theory suggests that conflicts of interest can arise between management of the company and its shareholders (Eisenhardt, 1989). This can lead to earnings management which has been defined by Healy and Wahlen (1999) as the alteration of firms’ reported economic performance by insiders to either mislead some stakeholders or to influence contractual outcomes. In previous literature earnings management is commonly used as a proxy for financial reporting quality (e.g. Frankel et al., 2002; Kinney et al. 2004; Paterson et al. 2011). Earnings management affects financial reporting quality as it devaluates the quality of earnings, and as a result the quality of financial reporting (DeFond and Zhang, 2014). Other definitions of earnings management that can be found in literature are:

- “Purposeful intervention in the external financial reporting process, with the intent of obtaining some private gain.” (Schipper 1989)

- “The active manipulation of earnings towards a predetermined target.” (Mulford and Comiskey 2002)

There can be numerous motives for earnings management. Literature suggest that these motives can be divided in the following three categories (Walker, 2013):

- To achieve targets or contractual terms which are related to reported earnings (e.g. equity compensation and bonus schemes)

- To influence the decisions of investors

- To influence the decisions of other stakeholders (e.g. customers; competitors, suppliers, etc.)

Reported earnings are based on accrual accounting. The difference between free cash flows and the operating income are the accounting accruals (DeFond and Zhang, 2014). Certain accruals are judgmental in nature, which implies managerial discretion over these accruals. Examples of such accruals are inventory provisions, bad debt provisions and tax provisions. Management can exercise control over these accruals to a certain extent and it can use its discretion to ‘stash accruals in cookie jars during good times and reach into them when needed in the bad times’ (Levitt, 1998).

Typically management, that has the discretion to make accounting decisions that affect earnings management, needs to balance the three motives for earnings management (Walker, 2013). Depending on the specific motives of management this can lead either to upwards earnings management or to downwards earnings management.

In order to determine the level of earnings management a surrogate measure of earnings management is required. The most common approach in previous studies is to capture a firm’s accrual accounting choices in a single number. Collectively these models are referred to as models of discretionary accruals (DAC’s). Models of discretionary accruals try to estimate a hypothetical level of accruals that would have occurred in absence of earnings management. This level of accruals is referred to as ‘normal’ (Walker, 2013).

The most prominent model of dictionary accruals is the Jones (1991) model. Later studies have further developed the model of Jones in order to improve the statistical

performance. The modified Jones model (Dechow et al., 1995) was the most frequently used model of discretionary accruals in literature up till 2005 (Jones et al. 2008).

The approach of Jones (1991) and Dechow et al. (1995) was further developed by Kothari et al. (2005). Kothari et al. added performance matching on return on assets to control for the effect of performance on measured discretionary accruals. They argued that this would enhance the reliability of the discretionary accruals models.

The discretionary accrual models have been widely used in earnings management studies over the past 20 years, however they have also been criticized for being unreliable due to noisy or biased estimates of discretion (e.g., Dechow et al., Bernard and Skinner, 1996; Kothari et al. 2005, Jones et al., 2008).

Stubben (2010) used a different measure of earnings management – discretionary revenues – by means of an unsophisticated accrual model. His results indicate that

discretionary revenue models are more likely to detect earnings management than the more commonly used accrual models. This suggests that the commonly used accrual models need to be further developed in order to improve their ability to detect earnings management.

Apart from discretionary accrual models, also other measures for earnings

management can be found in literature. Kinney et al. (2004) and Paterson et al. (2011) used accounting restatements as a proxy. An advantage of using restatements is that it is typically a strong indicator for low audit quality. However, restatements are rare events which makes the measure less suitable for smaller sample sizes. Another limitation is that the absence of a restatement does not prove high audit quality (DeFond and Zhang, 2014).

Researchers also used the propensity of auditors to render a going concern audit opinion as a proxy for earnings management is (e.g. DeFond et al., 2002; Geiger and Rama (2003). Going concern opinions typically have substantial adverse effects for companies, which gives management an incentive to pressure the auditor into rendering an unqualified opinion. The limitation of using this proxy is the fact that it does not capture small variations in quality. Furthermore, going concern opinions are fairly rare and only applicable for companies in financial distress. This reduces its statistical power for smaller sample sizes or for samples that include a large number of healthy firms (DeFond and Zhang, 2014).

In other studies the propensity to meet or beat earnings targets was used as a proxy for financial reporting quality (e.g. Huang et al., 2007; Davis and Trompeter, 2009). The

assumption is that a higher likelihood to meet analysts’ forecast indicates increased earnings management. Similar to discretionary accrual models these type of studies are prone to potential bias and measurement errors (DeFond and Zhang, 2014).

Finally, literature suggests that all measures of earnings management are sensitive to excessive conservatism of auditor’s that reduce litigation risks. Excessive conservatism could be misinterpreted as increased audit quality (DeFond and Zhang, 2014). However,

management has incentives to dismiss auditors that are overly conservative, which reduces the risk of excessive auditor conservatism (DeFond and Subramanyam, 1998).

2.5 Non-Audit Fees

Besides audit services, audit firms typically provide non-audit services to their clients, such as audit related services, tax advice, management consultancy, human resources

consultancy and so on. In case these services are provided to an audit client this could result in a real or perceived threat to the independence of the auditor. Threats to independence arising from the provision of non-audit services might include the following principal threats (IESBA1 Code of Ethics, 2010):

- The self-review threat: auditing one’s own work

- The familiarity or trust threat: becoming over-influenced by management and as a result to sympathetic to their interests, or becoming too trusting of management representations

- The advocacy threat: becoming an advocate for the client’s position in adversarial circumstances

- The self-interest threat: becoming inappropriately influenced by a financial or other interest

In various countries over the world legal or professional frameworks lay down the regulations for auditors. These regulations can be either principle or rule based depending on the country specific environment. The regulatory frameworks provide rules or guidance on provision of non-audit services. Cross selling of non-audit services is prohibited when the type of service is considered to be a threat to independence for which no safeguards are available to mitigate the risk.

1 The International Ethics Standards Board for Accountants (IESBA) is an independent body that sets international ethics standards under the auspices of International Federation of Accountants (IFAC).

The overall trend in the past two decades is that regulators, legislators and the general public are becoming more and more skeptic with regards to the joint provision of services (e.g. EC 2010; Audit Firms Supervision Act 2012; Financial Times 2002; The Economist 2010; The Economist 2011; Financial Times 2015a; Financial Times 2015b; Financial Times 2016). As a consequence this has led to further restrictions on the cross selling of non-audit and audit services, especially for public companies.

However, theoretical studies that analyze the relationship between the joint provision of non-audit and audit services, and audit quality, are ambiguous. Economic bonding between the auditor and his client is considered to be the main argument against cross selling of non-audit services as it is perceived to impair non-auditor independence (DeFond and Zhang, 2014). The economic dependence between the auditor and his client is a consequence of contractual economies of scope and knowledge spillovers (Simunic, 1984; Beck et al. 1988a; DeBerg et al.1991; Arruñada, 1999).

In contrast, strong arguments have been made that support the cross selling of non-audit services. Non-non-audit services may create knowledge spillovers that improve non-auditor competency, which in turn result in higher audit quality (Goldman and Barlev, 1974; Simunic 1984; Wallman 1996). Banning the joint provision of services might therefore have an

adverse effect on audit quality (Lu and Sapra, 2009).

Further litigation and reputation concerns may offset the threat to an auditors

independence caused by cross selling of non-audit services. The litigation and reputation risk provide strong incentives to auditors to be concerned with audit quality (Simunic, 1980; Arruñada, 1999; Weber et al. 2008; Skinner and Srinivasan, 2012; DeFond and Zhang, 2014).

Joint provision of services increases the economic bond between the auditor and the auditees, but it may be efficient due to knowledge spillovers and is thus generally favored by both parties (Simunic 1984; Beck et al. 1988a).

The widespread perception among regulators and the general public is that knowledge spillovers and litigation and reputation incentives do not fully offset the threat to

independence that is caused by economic bonding (e.g. The Economist 2010; The Economist 2011; Financial Times 2015b). Recent legislation in Europe therefore imposes further

restrictions on cross selling of non-audit and audit services (EC 2014; Audit Firms Supervision Act 2012).

2.6 Relation Non-Audit Fees and Financial Reporting Quality

Prior to 2002 the number of empirical studies that examined the relationship between non-audit fees and financial reporting quality was limited, due to the fact that fee data was not publicly available. However as of 2001 public companies in the United States are required by the Securities and Exchange Commission (SEC) to disclosure of both audit fees and non-audit fees, which has led to a surge of research in this field.

An early study by Frankel et al. (2002), using the modified Jones model, reported a significant positive relationship between the ratio of non-audit fees to total fees, and abnormal accruals2 . Furthermore, Frankel et al. found that the likelihood to meet analysts’ forecasts seemed to be higher as non-audit fees increased.

Subsequent studies questioned the findings of Frankel et al. (2002). Ashbaugh et al. (2003) attempted to further develop the model of Frankel et al., however they did not find evidence that a positive relationship between non-audit fees and abnormal accruals exists. They argued that the results of Frankel el al. were distorted by research design choices. Also other researchers (e.g. Chung and Kallapur, 2003; Larcker and Richardson, 2004; Reynolds, et al., 2004) that performed similar studies as Frankel et al. failed to find evidence of impaired financial reporting quality due to the provision of non-audit services.

DeFond et al. (2002) and Geiger and Rama (2003) studied the relationship between cross selling of services and propensity of auditors to issue a going concern opinion. They did not find evidence which indicates that financial reporting quality is impaired.

Regulators such as the SEC and the European Commission require disclosure of non-audit fees per type, which suggest that this is useful information (Paterson et al., 2011). However earlier studies only used total non-audit fees in their research models. In contrast, Kinney et al. (2004), which used restatements as a proxy for earnings management, broke down the non-audit fees into five different types: internal audit, financial information systems design and implementation, tax, audit-related, and unspecified fees. Their findings showed a positive association between restatements and unspecified fees, and a negative association between restatements and tax fees. This is consistent with the theory that tax services improve audit quality due to knowledge spillovers (Simunic 1984; Arruñada, 1999; Paterson et al., 2011). Further, Kinney et al. (2004) found no evidence that aggregate non-audit fees and

2 _{In literature accrual estimates are interchangeably referred to as either “abnormal” (DeFond and Jiambalvo}

restatements are associated. The study of Kinney et al. is distinctive as it indicates that various types of non-audit services have different relationships with financial reporting quality.

Huang et al. (2007) also made a distinction between the various types of non-audit fees. In contrast to Frankel et al. (2002), Huang et al. (2007) found some evidence that that non-audit services are actually negatively related to earnings management, suggesting that banning cross selling of non-audit services lowers audit quality because knowledge spillover effects outweigh economic bonding (Simunic, 1984; Lu and Sapra, 2009).

Paterson et al. (2011) extended the research of Kinney et al. (2004). They identified three types of non-audit services: tax, audit-related and other services. Further, following suggestions of Schneider et al. (2006), Paterson et al. also made a distinction between recurring and nonrecurring non-audit services in their model. They found a positive association between nonrecurring tax fees and restatements and a negative association between recurring tax fees and restatements. It should be noted that tax services include different types of services, such as the preparation of annual tax returns, tax compliance services and tax consulting services. As of 2006 the joint provision of certain tax consulting services is prohibited by the Public Company Accounting Oversight Board (PCAOB). Paterson et al. (2011) considered nonrecurring tax fees to be a proxy for tax consulting and recurring tax fees to be a proxy for other tax services. As such they conclude that their findings support the decision of the PCOAB to only prohibit the joint provision of tax consulting services.

In addition, Paterson et al. (2011) found positive associations between audit-related fees and restatements as well as other service fees (both for recurring and non-recurring). However, the difference between the recurring and nonrecurring audit-related fees is

statistically significant, which indicates that nonrecurring audit-related engagements are more likely to threaten independence.

A study from Krishnan and Visvanathan (2011) investigated auditor-provided tax services. Their results were consistent with Paterson et al., suggesting that tax consulting services are associated with increased earnings management.

Causholli et al. (2014), used the discretionary accruals model of Kothari et al. (2005) and incorporated future non-audit service fees into the model. Consistent with the findings of Frankel et al. (2002), they found evidence that anticipated future cross selling is negatively associated with financial reporting quality.

Other empirical studies examined the relationship between non-audit services and the perception of independence (e.g. Frankel et al. 2002; Francis and Ke, 2006; Raghunandan,

2003). Proxies for the perception of independence include earnings response coefficients (ERC’s), the cost of debt and capital, and the reaction of stock markets to audit related events such as the issuance of a going concern opinion (Hackenbrack and Hogan, 2002; Francis and Ke, 2006; DeFond and Zhang, 2014).

Frankel et al. (2002) investigated the relationship between non-audit fees and stock market reactions. They found evidence that non-audit fees are associated with stock discounts, indicating that the perception of auditor independence is impaired by cross selling of non-audit services.

DeFond and Park (2001) argued that abnormal accruals have little net effect on

lifetime earnings of a company, since they are reversing in nature. As such, assuming efficient capital markets, abnormal accruals should have little or no impact on stock prices. They used a firm specific working capital accruals model to measure the level of abnormal accruals. Furthermore, they used ERC’s to measure whether the market responds differently to earnings surprises that include abnormal accruals. Consistent with their hypothesis they found higher ERC’s when abnormal accruals are negative and lower ERC’s when abnormal accruals are positive. This provides evidence that earnings management is at least partly captured by the response of stock markets to earnings surprises.

Francis and Ke (2006) used ERC’s and found that non-audit fees are positively related to the market valuation of earnings surprises. This implies a negative relationship between non-audit services and the perception of independence. Dhaliwal et al. (2008) and Mitra and Hossain (2007) also investigated this relationship and they had similar findings as Frankel et al. (2002) and Francis and Ke (2006). In contrast, other studies that investigated the

relationship between non-audit services and auditor independence in appearance (e.g. Ashbaugh et al., 2003; Raghunandan, 2003) did not find evidence that non-audit services impairs independence.

Concluding, previous empirical literature that examined the relationship between non-audit fees and financial reporting quality showed mixed results. There is little empirical evidence that cross selling of non-audit services in general impairs financial reporting quality. However, there is some evidence that it affects the perception of independence (e.g. Francis and Ke (2006), Dhaliwal et al. (2008), Mitra and Hossain (2007). Other empirical researchers found evidence that the effect of non-audit services on earnings management depends on the type of non-audit service which is provided (e.g. Kinney et al. 2004; Huang et al., 2007; Paterson et al., 2011; Krishnan and Visvanathan, 2011). Their research showed that some types of non-audit services are associated with improved financial reporting quality (e.g.

Kinney et al. 2004; Huang et al., 2007; Paterson et al, 2011), whereas other types of non-audit services lead to increased earnings management (e.g. Kinney et al., 2004; Paterson, et al., 2011; Krishnan and Visvanathan, 2011).

Considering the inconclusive evidence of prior studies the relationship between non-audit services and financial reporting quality remains an interesting area for research. The recent prohibition to cross sell non-audit services to PIE in the Netherlands provides a good occasion to obtain further evidence from the Netherlands.

H3 Development of Hypotheses and Research Methodology

3.1 Development of Hypotheses

In this study we examined the relationship between non-audit fees and financial reporting quality in the Netherlands. For this particular research, restatements (see Kinney et al.; 2004; Paterson et al., 2011)) and the propensity of auditors to render a going concern audit opinions (see DeFond et al., 2002; Geiger and Rama, 2003), are not suitable as surrogate for financial reporting quality, due to the relatively small sample size (DeFond and Zhang, 2014). Further, the propensity to meet or beat earnings targets as a proxy for financial reporting quality (see Huang et al., 2007; Davis and Trompeter, 2009) is also not suitable, due to the fact that it is very difficult to obtain earnings targets data. Therefore we chose a model of discretionary accruals as a proxy for earnings management. The accruals model we used in this study to determine the level of abnormal accruals is the working capital accrual model of DeFond and Park (2001). We considered other discretionary accrual models such as the modified Jones model (Dechow et al., 1995) and the model of Kothari et al. (2005), however these models are cross sectional models and are therefore not suitable for our research due to the relatively small sample size.

The theory on economic bonding (Simunic, 1984) suggests that cross selling non-audit services to non-audit clients, impairs non-audit quality. This in turn leads to more managerial opportunism (Jensen and Meckling, 1976), resulting in higher levels of abnormal accruals and thus more earnings management (i.e. lower financial reporting quality) (Frankel et al., 2002; Kinney et al. 2004; Paterson et al. 2011). In contrast, analytical theories on knowledge spillover (Goldman and Barlev, 1974; Simunic 1984; Wallman 1996), litigation and

reputation risk (Simunic, 1980; Arruñada, 1999; Weber et al., 2008; Skinner and Srinivasan, 2012; DeFond and Zhang, 2014) suggests a negative relationship between the cross selling of

non-audit services and audit quality. This study aims to obtain further evidence regarding this relationship, which resulted in the following null hypothesis:

- Hypothesis 1: Non-audit service fees are not associated with abnormal accruals

Audit firms are typically organized in different branches such as audit, tax, consulting, mergers and acquisitions and so on. Most of the times these branches are all separate legal entities and together they form a network of members. The legislation on cross selling non-audit services to non-audit clients as included under section 24b of the Audit Firms Supervision Act, pertains to audit firms (including its Dutch network members) conducting statutory audits of PIE in the Netherlands. It does not cover non-Dutch branches of the audit firm’s network.

Following the 8th Directive of the European Union, Dutch legislation on the

disclosure of fees paid to the auditor is applicable since 2008. The Dutch Civil Code article 2:382a BW stipulates that PIE must disclose the total fees paid to the auditor and split these into the following four categories: audit, audit related, tax and other non-audit services.

However, disclosure of fees paid to non-audit and non-Dutch branches is not mandatory by Dutch law since these entities are not allowed to conduct audits under Dutch legislation. The Dutch Institute of Certified Public Accountants (NBA) finds this undesirable as this might withhold valuable information from the users of the financial statements. The NBA has therefore developed a voluntary fee disclosure template. PIE are advised to make use of this template in their annual reports.

X Auditors Other X network Total X network

Audit Audit related Tax

Other non-audit Total

The distinction that is being made between fees from audit, audit related, tax and other audit services, is relevant as prior empirical research indicates that various types of non-audit services differ in their relationship with non-audit quality (Kinney et al. 2004; Huang et al., 2007; Paterson et al, 2011; Krishnan and Visvanathan, 2011). Their findings suggest that

certain non-audit services actually strengthen auditor independence, while other types of services show a positive association with earnings management.

Further, section 24b of the Audit Firms Supervision Act the auditor still allows auditors to perform certain audit related services, such as assurance services as well as fact-finding services. Tax services and other non-audit services are however totally banned. This shows that the Dutch legislator is more concerned with tax fees and other fees, than with audit-related fees. It would therefore also be interesting to investigate relationship between the three different non audit fee types and earnings management separately.

The availability of voluntary fee disclosure data in the Netherlands, based on the NBA guidance, provides a good opportunity to further investigate the relationship between the different fee categories and financial reporting quality in the Netherlands. As such the following null hypotheses were added:

- Hypothesis 2: Audit-related fees are not associated with abnormal accruals - Hypothesis 3: Tax fees are not associated with abnormal accruals

- Hypothesis 4: Other fees are not associated with abnormal accruals

3.2 Research Method

The number of observations are the total number of companies listed on the

Amsterdam stock exchange in 2012 and 2013, since these are all public interest entities. Only firms with full fee disclosure are taken into account, so this includes disclosure of fees paid to non-audit and non-Dutch branches. Further, financial institutions and real estate firms were removed from the sample since these are not comparable because specific accounting rules (Frankel et al. 2002). This resulted in a total sample of 118 observations:

Table 1. Sample description.

Selection Criteria Observations

Total no. firms listed in 20123 ₁₃₇

Financial institutions and real estate firms4 _-28

Firms with missing data5 _-50

Number of samples 2012 59

Total no. firms listed in 20133 ₁₃₈

Financial institutions and real estate firms4 _-27

Firms with missing data5 _-52

Number of samples 2013 59

Total number of samples 118

Note: No financial and real estate firms are included in the study

DataStream has been used as a source to collect the data as it includes most of the data that is required to calculate the abnormal working capital accruals. Further, DataStream also includes most of the control variables data. The fee data was hand collected as it is not

available in any database. In addition, the auditor type data (Big 4 or no Big 4), was also hand collected.

Like Frankel et al. (2002), Kinney et al. (2004), Larcker and Richardson (2004) and Francis and Ke (2006), we used the proportion of the different non-audit fee categories to total fees as fee variables. However, these proportions are linked to the scale of the fees. In order to capture the relative client importance to the auditor also the natural logarithm of the fees is used. This is also a relatively common method (e.g. Ashbaugh et al. 2003; Geiger and Rama, Reynolds et al., 2004, Paterson et al. 2011). This results in the following fee variables:

- ATO/TF (Audit related + Tax + Other)/Total fees - A/TF Audit related/Total fees

- T/TF Tax/Total fees - O/TF Other/Total fees

- LN(ATO) Natural log (Audit related + Tax + Other)

3 _{Obtained from www.euronext.com} 4 _{ICB code 8500/8600/8700}

- LN(A) Natural log (Audit related) - LN(T) Natural log (Tax)

- LN(O) Natural log (Other)

The descriptive statistics regarding the fees from the 59 firms that have full disclosure in 2012 as well as 2013 can be found in Table 2.

Table 2. Descriptive statistics of non-audit fees.

Panel A Non-audit fees in 2012 (n = 58)

Variable* Mean Standard Deviation First Quartile Median Third Quartile Minimum Maximum Total fees 5.801 10.631 336 1.623 5.550 41 49.000 Audit fees 4.825 9.826 283 1.198 4.322 32 47.000 Audit related 386 908 6 82 400 0 4.500 Tax fees 400 588 0 100 700 0 3.200 Other fees 189 388 0 19 56 0 1.400 ATO/TF 0.24 0.16 0.10 0.22 0.34 0.00 0.61 A/TF 0.07 0.09 0.00 0.05 0.10 0.00 0.50 T/TF 0.10 0.12 0.00 0.05 0.17 0.00 0.55 O/TF 0.07 0.10 0.00 0.03 0.12 0.00 0.58

Panel B Non-audit fees in 2013 (n = 57)

Variable* Mean Standard Deviation First Quartile Median Third

Quartile Minimum Maximum

Total fees 5.535 10.302 336 1.410 5.550 41 49.000 Audit fees 4.563 9.525 283 1.000 4.322 32 47.000 Audit related 426 860 6 60 400 0 4.500 Tax fees 404 646 0 100 700 0 3.200 Other fees 142 343 0 0 56 0 1.400 ATO/TF 0.22 0.15 0.10 0.22 0.32 0.00 0.67 A/TF 0.08 0.10 0.01 0.05 0.10 0.00 0.40 T/TF 0.10 0.12 0.00 0.04 0.17 0.00 0.53 O/TF 0.05 0.09 0.00 0.00 0.06 0.00 0.47

Panel C Non-audit fees in 2012 and 2013 (n = 115) Variable* Mean Standard Deviation First Quartile Median Third

Quartile Minimum Maximum

Total fees 5.669 10.424 346 1.600 5.800 37 50.000 Audit fees 4.696 9.637 274 1.195 4.625 31 48.000 Audit related 406 881 5 66 400 0 5.600 Tax fees 402 615 0 100 700 0 3.200 Other fees 165 366 0 0 161 0 1.900 ATO/TF 0.23 0.16 0.10 0.22 0.32 0.00 0.67 A/TF 0.07 0.10 0.01 0.05 0.10 0.00 0.50 T/TF 0.10 0.12 0.00 0.05 0.17 0.00 0.55 O/TF 0.08 0.10 0.00 0.00 0.09 0.00 0.58

Note: *All the data is stated in thousands of euro. The audit fee data was hand collected.

In the working capital accrual model of DeFond and Park (2001) the expected accruals for each company are based on the adjusted ratio of working capital to sales. Abnormal

accruals are estimated by calculating the difference between the realized working capital, and an estimate of the working capital which is based on the current sales level. This difference is the portion of working capital accruals which is not likely to be sustainable and is therefore expected to reverse against future earnings. This results in the following model:

AWCA t = WC t – [(WC t-1/S t-1) x S t] Where:

AWCA t = abnormal working capital accruals in current year;

WC t = non-cash working capital in current year calculated as: (current assets6 - cash and cash equivalent7_{) - (current liabilities}8 _{- short-term debt}9_);

WCt-1 = working capital in previous year; S t = sales10 in current year; and

S t-1 = sales10 in previous year

6 _{DataStream code WC02201} 7 _{DataStream code WC02001} 8 _{DataStream code WC03101} 9 _{DataStream code WC03051} 10 _{DataStream code WC01001}

The abnormal working capital accruals are corrected for the size of a company, by means of dividing the abnormal accruals by the total assets11 _{of the firm’s current year. In} addition, the absolute, positive, and negative abnormal working capital accruals are calculated for each firm in 2012, and 2013. The descriptive statistics for the resulting abnormal accruals can be found in Table 3.

Table 3. Abnormal working capital accruals (n = 115***). Mean Standard Deviation First Quartile Median Third

Quartile Minimum Maximum

AWCA* -0,005 0,067 -0,030 -0,004 0,020 -0,289 0,285

AWCA_ABS 0,040 0,054 0,008 0,022 0,046 0,000 0,289

AWCA_POS** 0,017 0,044 0,000 0,000 0,020 0,000 0,285

AWCA_NEG** -0,023 0,043 -0,030 -0,043 0,000 -0.289 -0,000

Note: *AWCA is corrected for the size of a firm by dividing it by total assets. The mean of AWCA is close to zero leading to unreliable analyses, therefore three adjusted measures are used in addition to AWCA. **The positive and negative abnormal working capital accruals are divided using zero as a cutting point, furthermore, the resulting missing values are capped by zero. ***Three outliers are removed from the sample (above and below 3SD). For 2012 Corbion (0.44) is left out, and for 2013 Corbion (-1.24) and Fugro (-0.42) are left out.

Besides the type and level of fees, and the year of the abnormal accruals of a firm, there are several other variables that could affect the abnormal accruals. These control

variables are taken into account as well, descriptive can be found in Table 3. The variables are further described below:

- LN(Ass) - Company size: We control for larger companies as these are less likely to indulge in earnings management due to the fact that they already have a higher change to meet their earnings targets when compared to smaller companies. We use the natural logarithm of total assets11 _{as a proxy for company size (Frankel et}

al., 2002).

11 _{DataStream code WC02999}

- OCF - Operating Cash Flow12_{: Literature suggest that companies with high}

operating cash flow are more likely to beat their earnings targets, giving them less incentive for earnings management (Frankel et al., 2002).

- LEV - Leverage: We also control for leverage, which is calculated by taking the total liabilities13 divided by total assetsError! Bookmark not defined._{, as prior research}

suggest that leverage is related to abnormal accruals, since companies with high leverage might have an incentive to manage earnings in order to refinance their debt (Becker et al. 1998; Frankel et al. 2002)

- B4 - Big 4 auditor14: Big 4 auditors have a lager client base and thus they have more incentive to protect their reputation (Defond and Zhang, 2014). Further they have more industry specific knowledge which enables them to detect fraud more easily (DeAngelo, 1981; Krishnan, 1999). As such we control for this variable.

This results in the following model:

AWCAt = β0 + β1Fee + β2(LN(Ass) + β3OCF +β4LEV + β5B4 + fixed effects + e

- AWCA: Abnormal working capital accruals corrected for size (total assets) for the regarding company and year. In addition to AWCA, we used absolute

(AWCA_ABS), positive (AWCA_POS), and negative (AWCA_NEG) forms of AWCA.

- FEE: See Appendix A for the definition of the fee measures used in the model. - LN(ASS): Assets are measured in thousands of euros at the end of the fiscal year,

and are corrected by taking the natural logarithm.

- OCF: Operating cash flow is corrected for the size of a firm by dividing it by total assets.

- LEV: Leverage is calculated by taking the total liabilities divided by total assets at the end of the fiscal year.

- B4: For each firm is specified if it was audited by one of the Big 4 auditors (1 = ‘Big 4 auditor’, 0 = ‘other auditor’).

12 _{DataStream code WC04860} 13 _{DataStream code WC03351}

- Fixed effects: We included the variable Time to control for the effect of fiscal year (0 = ‘2012’, 1 = ‘2013’).

- e: Error term

Table 4. Descriptive statistics for the control variables (n = 115).

Variable Mean Standard Deviation First Quartile Median Third

Quartile Minimum Maximum

Assets* 16.028.129 49.052.660 223.624 1.134.842 6.881.000 3.199 270.060.240

OCF* 1.661.917 5.950.177 4.207 49.600 703.478 -323.700 34.791.047

Leverage 0.562 0.205 0.456 0.576 0.692 0.061 1.284

B4 No: 10 Yes: 105

Note: This table presents descriptive statistics for the control variables that are included in the abnormal working capital accruals model. Assets and operating cash flow are in thousands of euro. Please refer to Appendix A for variable definitions

H4 Empirical Results

Relations of the Model Variables

Correlations between all variables of the model were calculated to get a first overview of the relationship between the abnormal working capital accruals and the non-audit fees, and the control variables (see Table 5). First, we looked at the relations between the four types of abnormal working capital accruals (raw AWCA corrected for size, absolute, positive, and negative) and the non-audit fees, then we looked at the relations between the control and fee variables.

First, the relations between the non-audit fees and absolute abnormal working capital accruals showed that ATO/TF (r = 0.24, p = 0.011) and T/TF (r = 0.23, p = 0.014) are positively related to absolute abnormal working capital accruals. These results indicate that companies with higher proportions of non-audit and tax fees, have higher levels of absolute abnormal working capital accruals. Furthermore, ATO/TF (r = -0.20, p = 0.030) and T/TF (r = -0.27, p = 0.004) were negatively related with the negative abnormal working capital accruals, indicating that firms with higher proportions of non-audit and tax fees have more negative abnormal working capital accruals. We found no significant relations between non-audit fees and the positive abnormal working capital accruals.

Next, the relations between the control variables (i.e., size, leverage, cash flow, Big 4) and the three different abnormal working capital working capital accruals showed a negative relation between LN(Ass) and absolute abnormal working capital accruals (r = -0.24, p = 0.010), indicating that larger companies have a lower level of absolute abnormal working capital accruals. In addition, LN(Ass) was marginally related to negative abnormal working capital accruals (r = 0.19, p = 0.045), such that larger companies have less negative abnormal working capital accruals. Furthermore, OCF was positively related with absolute abnormal working capital accruals (r = 0.46, p < 0.001), and positive abnormal working capital accruals (r = 0.47, p < 0.001), indicating that companies with a larger amount of cash flow have higher absolute accruals, and more positive abnormal working capital accruals.

Finally, we looked at the relations between the control variables (i.e., assets, leverage, cash flow, Big 4) and the non-audit fees. The results showed that LN(Ass) correlates

positively with all LN-fees (all P’s < 0.05), and negatively with TO/TF and O/TF (P’s < 0.05). These result suggest that larger companies have higher levels of non-audit fees, but lower proportions of tax and other, and other fees. In addition, LEV (leverage) was positively correlated with A/TF (p < 0.01), and with all LN-fees (P’s < 0.05; except LN(TO) and LN(O)), which indicates that companies with higher leverage have a higher proportion of audit related fees, and higher levels of non-audit, audit related, and tax fees. In contrast, LEV(leverage) was negatively related with O/TF (p < 0.05), showing that companies with more leverage have a lower proportion of other fees. Furthermore, B4 also predicted the LN-fees (all P’s < 0.05), the positive b-values indicates that firms which are audited by one of the Big 4 auditors have higher levels of non-audit fees than firms which are audited by other auditors.

In sum, the results of the correlations show that higher proportions of several non-audit fee variables are related to higher levels of absolute accruals, and more negative abnormal working capital accruals. In addition, larger companies have lower absolute

accruals, and less negative abnormal working capital accruals. Companies with higher levels of cash flow have higher absolute accruals, and more positive abnormal working capital accruals. Finally, larger companies, companies with higher leverage, and companies audited by a Big 4 auditor, have higher levels of non-audit fees. Larger companies have lower

proportion of tax and other fees, and companies with more leverage lower proportion of other fees, but higher proportion of audit related fees.

Table 5. Correlations of the abnormal working capital accruals model. AWCA AWCA _ABS AWCA _POS AWCA _NEG ATO/ TF A/TF T/TF O/TF LN (ATO) LN(A) LN(T) LN(O) LN (Ass) OCF LEV B4 AWCA_ABS 0.018 - AWCA_POS 0.783*** 0.636*** - AWCA_NEG 0.774*** -0.619*** 0.212* - ATO/TF -0.068 0.236* 0.095 -0.202* - A/TF 0.118 -0.025 0.075 0.108 0.422*** - T/TF -0.156 0.230* 0.022 -0.268** 0.588*** -0.142 - O/TF -0.031 0.116 0.048 -0.098 0.447*** -0.132 -0.147 - LN(ATO) 0.056 -0.052 0.011 0.076 0.443*** 0.341*** 0.288** 0.018 - LN(A) 0.116 -0.111 0.020 0.162 0.128 0.558*** -0.075 -0.246** 0.794*** - LN(T) 0.033 -0.008 0.021 0.031 0.325*** 0.044 0.570*** -0.220* 0.807*** 0.563*** - LN(O) 0.025 0.066 0.060 -0.022 0.4381*** -0.026 0.053 0.556*** 0.455*** 0.155† _{0.286*** -} LN(Ass) 0.046 -0.239* -0.113 0.187† _{-0.110 0.116 -0.037 -0.238* 0.725*** 0.753*** 0.642*** 0.225* -} OCF 0.236* 0.463*** 0.470*** -0.107 -0.047 -0.135 0.056 -0.011 -0.156† -0.130 -0.044 -0.029 -0.160† - LEV -0.035 -0.104 -0.092 0.038 0.107 0.272** 0.082 -0.194* 0.405*** 0.369*** 0.239* 0.139 0.340*** 0.212* B4 -0.024 -0.141 -0.107 0.070 0.118 0.129 -0.042 0.110 0.390*** 0.305** 0.200* 0.214* 0.484*** -0.321*** 0.157† - Note: †p < 0.10, *p < 0.05 (two-tailed), **p < 0.01 (two tailed), ***p < 0.001 (two tailed). The table reports the correlations of the abnormal working capital accruals model: AWCAt = β0 + β1Fee + β2(LN(Assets) + β3OCF + β4LEV + β5B4 + fixed effects + e.. The variables are explained in Appendix A.

Difference in Earnings Management Between High- versus Low Non-Audit Fees

In order to explore the differences between high versus low non-audit fees in accruals, we conducted independent sample t-tests (see Table 6). More specifically, based on a median split we separated a high fee group from a low fee group, for the total non-audit, audit related, tax and other fee variables. Next, we tested whether the high fee group has significantly different abnormal working capital working capital accruals (AWCA and AWCA_ABS) than the low fee group. Positive and negative working capital accruals are not taken into account, given that they analyses would be unreliable due to the large number of zeros.

The results for the abnormal working capital accruals (AWCA) showed that companies with lower audit related fees differed from companies with higher audit related fees (p = 0.051). The means of AWCA show that companies with low audit related fees have a negative average of AWCA (M = -0.172), however companies with high audit related fees have a positive average of AWCA (M = 0.007). The low audit-related fee group has a relatively larger deviance from zero, indicating that the earnings management is higher for companies with lower audit related fees than for companies with higher audit related fees. In addition, the results for the absolute abnormal working capital accruals showed that

companies with lower non-audit fee proportions (ATO/T; M = 0.026), had lower abnormal working capital accruals than companies with higher non-audit fee proportions, (M = 0.054, p = 0.006). A similar pattern was found for the comparison between high and low tax fee proportions (T/TF), the low fee group (M = 0.030) had lower absolute abnormal working capital accruals than the high fee group (M = 0.067, p = 0.060). These last results suggest that companies with higher total non-audit, or tax fee proportions have more earnings management than companies with lower total non-audit, or tax fee proportions. However, it should be noted that these results might be unreliable, given that the effects are not corrected for the control variables such as leverage or the size of cash flow. In addition, splitting data based on median split might lead to biased results. Next, we will test the full model, using linear/tobit regression, and will also control for the in control variables as outlined in section 3.

Table 6. Independent sample t-tests comparing differences in abnormal working capital accruals between low and high (median split) non-audit fees (n = 115).

Panel A: Total non-audit fees (audit related + tax + other)

Below Median Above Median Difference t-value p-value

AWCA -0.010 (0.072) 0.000 (0.062) -0.011 -0.87 0.389

AWCA_AB 0.042 (0.059) 0.038 (0.049) 0.004 0.41 0.686

N 58 57

Panel B: Audit related fees

Below Median Above Median Difference t-value p-value

AWCA -0.172 (0.073) 0.007 (0.059) -0.024 -1.97 0.051

AWCA_AB 0.045 (0.060) 0.035 (0.048) 0.010 0.98 0.329

N 58 57

Panel C: Tax fees

Below Median Above Median Difference t-value p-value

AWCA -0.009 (0.062) -0.001 (0.073) -0.007 -0.58 0.560

AWCA_AB 0.038 (0.049) 0.042 (0.059) -0.004 -0.38 0.705

N 60 55

Panel D: Other fees **

Equal to zero Above zero Difference t-value p-value

AWCA 0.007 (0.056) 0.078 (0.010) -0.003 -0.27 0.788

AWCA_AB 0.033 (0.045) 0.047 (0.062) -0.014 -1.37* 0.172

N 60 55

Panel E: ATO/TF

Below Median Above Median Difference t-value p-value

AWCA -0.009 (0.036) -0.001 (0.088) -0.0077 -0.62* 0.539

AWCA_AB 0.026 (0.027) 0.054 (0.069) -0.0276 -2.83* 0.006

N 58 57

Panel F: A/TF

Below Median Above Median Difference t-value p-value

AWCA -0.010 (0.063) -0.000 (0.071) -0.010 -0.80 0.428

AWCA_AB 0.037 (0.052) 0.043 (0.057) -0.006 -0.58 0.563

Panel G: T/TF

Below Median Above Median Difference t-value p-value

AWCA -0.010 (0.045) -0.001 (0.083) -0.009 -0.70 0.486

AWCA_AB 0.030 (0.005) 0.067 (0.009) -0.019 -1.90* 0.060

N 56 59

Panel H: O/TF**

Equal to zero Above zero Difference t-value p-value

AWCA -0.007 (0.060) -0.003 (0.078) -0.003 -0.72 0.788

AWCA_AB 0.033 (0.045) 0.047 (0.062) 0.014 -1.37* 0.172

N 60 55

Note: Means and standard deviations (M(SD)), and the mean differences, are reported for the below and above median groups. *Reported t-values and p-values are adjusted for inequality of group variances, given that Levene’s test for equality of variances was significant. **O/TF was split into a group with values equal to zero and a group with values above zero, given that the O/TF values of sixty companies equalled zero. AWCA and AWCA_AB are explained in Appendix A.

Testing the Abnormal Working Capital Accruals Model

The hypotheses of the model are tested by estimating the abnormal working capital accruals model with regression analyses. First, the absolute abnormal working capital working capital accruals model is estimated using a linear regression (ordinary least squares, OLS). Next, both the positive abnormal working capital accrual model and the negative abnormal working capital accrual model are estimated by means of a tobit regression. For all three abnormal working capital accruals outcome measures (absolute, positive, and negative), we subsequently estimated ten models, such that the predicting effect of the fee variables were tested separately. More specifically, we tested the predicting value of 1) the proportion, and the level of audit related, tax and other fees (ATO/TF; LN(ATO)), 2) the proportion, and level of audit related fees (A/TF; LN(A)), 3) the proportion, and level of tax fees (T/TF; LN(T)), 4) the proportion, and level of other fees (O/TF; LN(O)), 5) the unique predicting value of the separate proportions, and levels of audit related, tax and other fees (A/TF, T/TF, O/TF; LN(A), LN(T), LN(O)). In addition, in order to obtain a reliable and unbiased estimate of earnings management the total amount of assets (LN(Ass)), size of cash flow (OCF), leverage (LEV), and auditor type (B4, yes/no) were added as control variables, as well as the year of the relevant data (Time, 2012/2013). Tables 7.a, 7.b, 8.a, 8.b, 9.a and 9.b display an overview of the estimated model and the relevant results of the estimated models.

Absolute Abnormal Working Capital Accruals Model

The absolute abnormal working capital working capital accruals models explain a moderate amount of 21-26% of the variance in the data (Adjusted R2_{). First, looking at the} control variables shows that the company size (LN(Ass)), and leverage (LEV) both

significantly predict the size of absolute accruals. More specifically, larger companies have a lower level of absolute abnormal working capital accruals, which is in accordance with the correlations reported above. Also, the model shows that a higher level of operating cash flow is related to a higher level of absolute abnormal working capital accruals. Furthermore, leverage, and whether or not the firm was audited by one of the Big 4 audit firms, did not significantly predict the level of absolute accruals.

Next, inspecting the predicting values of the audit fee variables shows evidence that there is a positive relationship between absolute abnormal working capital accruals and several fee measures. First, companies with a higher proportion of non-audit fees (ATO/TF; p = 0.008), or a higher level of non-audit fees (LN(ATO); p = 0.023) have higher levels of absolute accruals. In addition, companies with a higher proportion of tax fees (T/TF; p = 0.018), or a higher level of tax (LN(T); p = 0.037), have higher levels of absolute accruals. The proportion, or level, of other (O) fees did not significantly predict the level of absolute abnormal working capital accruals. Finally, when all three fee categories are separately included in the model, companies with a higher proportion of tax fee (T/TF; p = 0.008), or a higher level of tax fee (LN(T); p = 0.090 (marginal)), have higher levels of absolute accruals. The proportion, or level, of audit related (A) and other (O) fees did not significantly predict the level of absolute abnormal working capital accruals. These results show that controlling for relevant variables, reveals that not only proportions of non-audit and tax fees are

positively related to absolute abnormal working capital accruals (as found in the correlational analyses), but also the level of non-audit and tax fees are positively related to absolute

abnormal working capital accruals.

Positive Abnormal Working Capital Accruals Model

The positive abnormal working capital working capital accruals models have a good fit, which is reflected by the p values of the Wald statistic (all P’s < 0.001). Inspecting the estimates of the control variables shows that larger companies (LN(Ass)) have less positive accruals, when the level of audit related (LN(A)), or the level of the three separate fees are included in the model (LN(A), LN(T), LN(O)). In addition, the results showed that companies

with a larger operating cash flow (OCF) have more positive accruals, regardless of which fee variable is included, which is in accordance with the previous reported correlation.

Furthermore, the regression analyses showed that companies with a higher proportion of non-audit fees (ATO/TF; p = 0.033), or a higher level of non-audit fees (LN(ATO); p = 0.017), have more positive abnormal working capital accruals. In addition, companies with higher proportion of audit related fees (A/TF; p = 0.001/ p < 0.001), or higher level of audit related fees (LN(A); p = 0.004/ p = 0.004), have more positive abnormal working capital accruals as well, regardless of whether the other fee variables (T and O) were separately added to the model or not. The proportion, or level, of tax (T) and other (O) fees did not significantly predict the level of positive abnormal working capital accruals. Although we did not find correlations between the fee variables and positive accruals, these results show that when we control for other relevant variables, non-audit, and audit related fees are positively related to positive abnormal working capital accruals.

Negative Abnormal Working Capital Accruals Model

The negative abnormal working capital accruals models have a good fit as well, again reflected by the p values of the Wald statistic (all P’s < 0.003). The results show that larger companies (LN(Ass)) have less negative abnormal working capital accruals. Although this relation was only marginal significant in the correlational analyses, the relation was now significant in all estimated models, which indicates that controlling for other relevant

variables increased the effect size of the relation between company size and negative accruals. We also found a negative relation between operating cash flow (OCF) and negative abnormal working capital accruals in all models, indicating that companies with larger cash flows have more negative abnormal working capital accruals.

Inspecting the estimates of the fee variables shows that companies with higher

proportions of non-audit fees (ATO/TF; p = 0.075), have marginally more negative abnormal working capital accruals, given their negative relationship. In addition,f companies with a higher proportion (T/TF; p = 0.004/p = 0.003), or a higher level (LN(T); p = 0.071/p = 0.073) of tax fee, also have more negative abnormal working capital accruals. The other fee variables did not significantly predict negative accruals.

Summary Abnormal Working Capital Accruals Model Analyses

In sum, the correlations, t-tests, and regression analyses show converging evidence that several control variables, and fees are related to abnormal working capital working