Amsterdam Business School
MSc Thesis Accountancy & Control
The effect of auditor independence on audit quality and the moderating role of audit committee effectiveness
Name: Reimer Stoeten
Student number: 11147237
Date: 19-06-2016
Word count: 12.910
Thesis supervisor: Dr. Georgios Georgakopoulos MSc Accountancy & Control, specialization Accountancy Amsterdam Business School
2 Statement of Originality
This document is written by student Reimer Stoeten who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3
Abstract
This study is complementary to the auditor independence literature, because it tests the relation between auditor independence and audit quality in a recent timeframe (2010-2014) and investigates whether this possible relation is influenced by audit committee effectiveness. Using data of 1620 firm-year observations from US firms located in the Standard’s & Poor’s 500 index, the results of the OLS regressions do no provide evidence of a moderating influence of audit committee effectiveness. Furthermore, the results show that audit quality is not reduced by non-audit service (NAS) provision. In contrast, this study even found little evidence that audit quality improves as a result of NAS. This implies a support of the knowledge spillover theory at the expense of the economic bonding theory, which may be of interest to regulators.
4
Tabel of contents
1 Introduction ... 6
2 Literature ... 9
2.1 Literature review ... 9
2.1.1 Agency theory and the demand for auditing ... 9
2.1.2 Auditor independence ... 10
2.1.3 Audit quality ... 12
2.1.4 Auditor independence and audit quality ... 13
2.1.5 Audit committee ... 14
2.2 Hypotheses development ... 16
3 Research methodology ... 18
3.1 Research method and data collection ... 18
3.2 Sample selection ... 18 3.3 Dependent variable... 20 3.4 Independent variable ... 22 3.5 Moderating variable ... 23 3.6 Control variables ... 23 3.7 Empirical models ... 25 3.8 Expectations ... 27 4 Results ... 29 4.1 Descriptive statistics ... 29 4.2 Regression results ... 32
5 Discussion and conclusion ... 39
5.1 Limitations and recommendations for future research ... 41
5.2 Theoretical implications ... 42
5.3 Practical implications ... 43
5.4 Conclusion ... 43
5 7 Appendices ... 51 7.1 Appendix A ... 51 7.2 Appendix B ... 52 7.3 Appendix C ... 53 7.4 Appendix D ... 55
6
1 Introduction
“The independence requirement serves two related, but distinct, public policy goals. One goal is to foster high quality audits by minimizing the possibility that any external factors will influence an auditor's judgments … The other related goal is to promote investor confidence in the financial statements of public companies ... Investors are more likely to invest, and pricing is more likely to be efficient, the greater the assurance that the financial information disclosed by issuers is reliable. The federal securities laws contemplate that assurance will flow from knowledge that the financial information has been subjected to rigorous examination by competent and objective auditors.”
- Securities and Exchange Commission (SEC) (2001)
Over the years, a big debate is going on about auditor independence. Since an auditor is the counselor of civil life and has a public trust, it is important that he is objective while producing opinions about financial statements of companies. Worldwide, many laws and regulations are developed to limit the dependency between auditors and clients, to in the end improve the quality of audits. The introduction of many of those laws and regulations was a consequence of financial reporting scandals, like Enron and WorldCom. An example is the development of the Sarbanes-Oxley Act (SOX)1 in 2002.
Most of the financial reporting scandals are a consequence of unethical behavior by top management of organizations and their auditors. Corporate values and behavior and the capitalistic culture of today’s society are important causes of these scandals (Low et al., 2008). Like every profit organization, the intention of audit firms is to make profit. De Angelo (1981) argues that an economic bond between an auditor and its client, because of audit fees and non-audit service2
(NAS) fees, can impair auditor independence, what can result in a reduced audit quality (economic bonding theory). She also mentions that audit quality can be defined as the probability that an auditor will report a breach when he or she discovers one. Audit quality is related to auditor independence and it reflects the role of auditors in diminishing estimation errors in accruals (De Angelo, 1981). Totally independent auditors will ask management to make changes in estimations and accounting methods to improve the quality of accruals. In contrast to the economic bonding theory, the knowledge spillover theory states that audit quality increases as a result of NAS purchase. According to this theory, audit firms are more integrated with a client due to NAS
1 The Sarbanes-Oxley Act is a regulation introduced by the US Congress to strengthen the regulation regarding
financial reporting, resulting in a reduction of the risk of accounting errors and fraudulent activities. Additional disclosure requirements, corporate governance mandates and greater penalties in case of violation are important components of this Act (Romano, 2005).
2 Non-audit services are services provided by an audit firm to its client, which deviate from services belonging to an audit.
7 purchase and are therefore more able to recognize potential financial reporting problems (Simunic, 1984).
Audit committees play an important role in corporate governance (Crişan & Fülöp, 2014).
Previous studies found that the existence of an audit committee is associated with less financial reporting problems (Defond & Jiambalvo, 1991; McMullen, 1996). Highly effective audit committees (reflected by its size, the degree of independence, the expertise of its members and the meeting frequency) are negatively related to internal control problems (Krishnan, 2005). Audit committees play an important role in internal control and the corporate governance within a company. Since SOX, audit committees are responsible for the selection of auditors and the approval NAS and NAS fees. It is possible that effective audit committees, based on their role as representatives of the corporate governance in a company, can mitigate a possible relation between low auditor independence and low audit quality. Therefore, the following research question will be investigated in this study:
RQ: Is auditor independence related to audit quality, and does the effectiveness of an audit committee affect this relation?
Prior research has tested the effect of auditor independence on audit quality. Some find a positive relation (Frankel et al., 2002). Others find no relation between auditor independence and audit quality (Chung & Kallapur, 2003), and some even find a negative relation (Simunic, 1984). Although the huge amount of existing studies regarding the relation between auditor independence and audit quality (De Angelo, 1981; Simunic, 1984; Beck et al., 1988; Beck & Wu, 2006; Chung & Kallapur, 2003; Frankel et al., 2002; Ferguson et al., 2004; Srinidhi & Gul, 2007; Ashbaugh et al., 2003; Gul et al., 2007), this study contributes primarily to existing literature by using more recent data (2010-2014). In a recent, post-financial crisis setting in which laws and regulations concerning NAS are stricter than ever, it provides evidence regarding the ongoing trade-off between the economic bonding theory and the knowledge spillover theory. This is helpful to regulators, because it could be used to evaluate the effectiveness of changes in regulation.
Furthermore, a literature research for the purpose of this study showed a literature gap regarding the moderating influence of the effectiveness of an audit committee on the relation between auditor independence and audit quality. Due to its important monitoring role as corporate governance mechanism, the audit committee and in particular its effectiveness could be an explaining factor regarding the aforementioned contrast in results of studies. This study provides
8 useful information to regulators in two different ways. Firstly, within the boundaries of current laws and regulations, NAS does not lead to more accrual-based earnings management. This is neither influenced by the effectiveness of an audit committee. Secondly, regulators should not focus on increasing the financial expertise of an audit committee, but instead they should focus on increasing the audit committee itself.
This study uses a sample of 1620 firm-year observations from 324 different Standard’s & Poor’s 500 (S&P 500) firms to test the four hypotheses that are set up to answer the research question. The absolute value of discretionary accruals, calculated with the modified Jones model (Dechow et al., 1995), is used as a measure of audit quality. NAS purchase is used as a proxy for auditor independence and is measured by two different variables. Furthermore, audit committee effectiveness is reflected by three of its characteristics, namely the degree of financial expertise, the size of the committee and the number of meetings. The results of this study differ from the on literature based expectations. NAS purchase by clients shows not to be related to a reduction of audit quality. In fact, little evidence supporting the later described knowledge spillover theory is found. Furthermore, it turns out that the effectiveness of an audit committee does not affect the relation between auditor independence and audit quality.
The remainder of this paper is structured in the following way. In chapter 2 a literature review is given and the hypotheses are formulated. Chapter 3 covers the research methodology, including the research method, sample selection, operationalization of variables, empirical models and the expectations regarding the associations between variables. The descriptive results and regression results are presented in chapter 4. Finally, chapter 5 shows the discussion and conclusion.
9
2 Literature
This literature section is divided into two parts. Part one is a literature review, which will provide a theoretical background to show what already is known about the different topics covered in this paper. In part two, the hypotheses will be developed based on this theoretical background.
2.1 Literature review
2.1.1 Agency theory and the demand for auditing
The roots of corporate governance lie in the agency theory (Jensen & Meckling, 1976; Watts & Zimmerman, 1978). As a result of a conflict of interests between the principal (investor, outsider) and agent (manager, insider), firms do not act in line with the interests of principals. The existence of information asymmetry, divergent objectives and different risk preferences between principal and agent can all be labeled as agency problems (Eisenhardt, 1989). Agency problems arise in organizations where ownership and control are separated, because in those organizations the agent controls the capital invested by the principal. This means that the principal cannot control its invested money. Someone else, the agent, makes decisions what to do with the invested money of the principal.
Corporate governance, in the form of internal control mechanisms, can reduce agency problems (Shleifer & Vishny, 1997). Internal corporate governance mechanisms are mechanisms inside an organization that should ensure the achievement of organizational objectives, by aligning the interests of shareholders and managers and by monitoring actions taken to achieve those organizational objectives (Walsh & Seward, 1990).
External auditing, due to its efficient contracting role (Ball, 1989) and its valuation role (Jensen & Meckling, 1976), supports internal control mechanisms in reducing agency problems. The IAASB (2013) says the following about an audit: “The purpose of an audit is to enhance the degree of confidence of intended users in the financial statements. This is achieved by auditors gathering sufficient appropriate audit evidence in order to express an opinion on whether the financial statements are prepared, in all material respects, in accordance with the applicable financial reporting framework. Usually, that opinion is on whether the financial statements “present fairly, in all material respects” or give “a true and fair view” of the entity’s financial position as at the period end and of its results and cash flows for the period.” This means that users of financial statements benefit from an audit, because an audit gives reliability to information provided by managers, resulting in less information asymmetry. Therefore, auditing plays an important role in investor protection (Newman et al., 2005). Furthermore, auditing helps with monitoring managers by
10 checking their work (Jensen & Meckling, 1976). This implies that the existence of the audit profession can be explained by the agency theory.
Moreover, the agency theory is not the only theory explaining the demand for external auditing. Three other well-known theories exist, namely the policeman theory, the lending credibility theory and the theory of inspired confidence (Hayes et al., 2014, p.43-46). The policeman theory sees the auditor as a preventer and detector of fraud (Hayes et al., 2014). The lending credibility theory describes adding credibility to financial statements as the main function of auditing. With audited financial statements, management can strengthen the faith of stakeholders that the financial statements give a true and fair view of the economic situation of a company. In line with this theory, auditing has a positive effect on stakeholders’ faith in management’s stewardship. According to the theory of inspired confidence, which has similarities with the agency theory, the demand for auditing can be explained by the involvement of external stakeholders in a company. In return for their contribution to a company, stakeholders ask for accountability of management. This accountability is expressed by financial statements. However, information provided by management might be biased. Therefore, stakeholders demand for auditing of this information.
The American Accounting Association defines auditing as: “a systematic process of objectively obtaining and evaluating evidence regarding assertions about economic actions and events to ascertain the degree of correspondence between those assertions and established criteria and communicating the results to interested users” (Silvoso, 1972, p.18). Auditors help to align the interests of agents and principals by ensuring that information provided by agents to shareholders is reliable. Investors make decisions based on audited financial statements. It is important that they can trust auditors about the reliability of audited information, by which they can make good decisions and risk is minimalized.
2.1.2 Auditor independence
As stated in Section 100.5 of International Ethics Standards Board for Accountants’ (IESBA) Code of Ethics for Professional Accountants (IESBA, 2013), objectivity is one of the five fundamental principles of the accounting profession. In this Code, objectivity is described as: “not to allow biases, conflict of interest or undue influence of others to override professional or business judgments”. Impaired auditor independence could form a self-interest threat to objectivity (Moore et al., 2006). Therefore, auditor independence is fundamental for the existence of the auditor profession (Mautz & Sharaf, 1961). Independence means that there is no conflict of interest between an auditor and users of audited financial information (Bollom, 1988). This should result in a bias-free opinion. The American Institute of Certified Public Accountants (AICPA) (1988) makes a distinction between
11 independence in mind and independence in appearance. Independence in mind can be defined as the state of mind that allows to draw conclusions that are unaffected by influences that compromise professional judgment. Independence in appearance is about the avoidance of circumstances from which a well-informed third party might conclude that professional judgment is compromised. Independence in mind can be seen as real independence and independence in appearance is the perceived independence by other parties.
Auditor independence could be impaired by an economic bond and a social bond between an auditor and its client (De Angelo, 1981). An economic bond arises from audit fees and NAS fees. A social bond as a result of a longer auditor tenure or personal relationships are also important in relation with auditor independence. The economic bond threat and social bond threat can be divided into four main threats to auditor independence, namely client importance, NAS, auditor tenure and client affiliation with audit firms (Tepalagul & Lin, 2015). Auditors get paid by their clients. If a payment is higher, the client plays a larger role in the portfolio of the auditor and will be more important to the auditor. Therefore, the auditor has more intention to retain that client, which could be harmful for the independence of the auditor. In case of a longer auditor tenure it is more likely that auditors are going to act in line with management, due to the social bond between auditor and management. Client affiliation with audit firms is about a close (working) relationship between the client and its auditor. The last of the four main threats is NAS. According to Moore et al. (2006), all provision of NAS should be forbidden. Because NAS is such an immense threat, auditors should only perform audit services and no NAS. However, some NAS is still permitted nowadays. The question can be asked whether NAS forms indeed such a huge threat as it is assumed.
Therefore, the focus of this study will be on the effects of NAS. Providing NAS is often more profitable for audit firms than providing audit services. Audit firms prefer NAS over audit services, because of the economic advantages of NAS (Dopuch & King, 1991). Audit firms benefit enormously from the provision of NAS. The benefits are the fees that audit firms earn with the provision of NAS and knowledge spillovers as a result of that provision (Simunic, 1984). Knowledge spillovers are information advantages as a result of NAS and impaired auditor independence, which are helpful during the audit process. An auditor is able to detect material weaknesses at an earlier stage if his audit firm provides NAS to a client (De Simone et al., 2015). There also are costs for auditors when they provide NAS. These costs are the loss of reputation (Watts & Zimmerman, 1983) and risk of clientele adjustment (Shu, 2000). Loss of reputation could arise when it becomes clear for the society that the independence of an auditor is impaired as a result of NAS. The faith in his objectivity could be impaired. This could be extremely harmful,
12 because the reputation of an auditor is essential for its existence (De Angelo, 1981). Clientele adjustment could occur when an audit firm only wants to provide services to clients that also purchase NAS of the audit firm (Shu, 2000). Clients that do not purchase NAS of the audit firm should possibly pay a higher premium for the audit to compensate. If such clients do not want to pay that premium, it is possible that the contract with the auditor will be terminated. However, for audit firms the benefits outweigh the cost of NAS, otherwise NAS would not be provided to clients.
The regulation regarding auditor independence is tightened in the past fifteen years. Many laws and regulations were developed after big scandals, where auditors were involved and the independence was insufficient. The development of the SOX is the most important and well-known example of this emerging regulation in the US. The SEC gives in its revision of the Commission’s auditor independence requirements (SEC, 2001) a by SOX supported list of prohibited NAS that impair auditor independence: bookkeeping and related services, financial information system design and implementation, appraisal or valuation services and fairness opinions, actuarial services, internal audit services, management activities, human resources activities, broker-dealer services, legal services and expert services which are unrelated to the audit. All allowed activities need to be pre-approved by the audit committee of an organization before an auditor can provide those activities.
2.1.3 Audit quality
As mentioned before in paragraph 2.1.1, investors benefit from an audit (Newman et al., 2005). This is only the case when the audit is of high quality. An audit forms no value-added service to the reliability of information when the quality is low. Audit quality is defined by the Public Company Accounting Oversight Board (PCAOB) (2013) as: “meeting investors’ needs for independent and reliable audits and robust audit committee communications on: (1) financial statements, including related disclosures; (2) assurance about internal control; and 3) going concern warnings” (p.3-4). De Angelo (1981) describes audit quality as the probability that an auditor discovers a breach and that he or she reports that breach. Audit quality has two important determinants: auditor reputation and the ability to withstand managerial pressure (Deis & Giroux, 1992). Studies regarding the relation between auditor independence and audit quality mostly measure audit quality with accrual-based earnings management (Frankel et al., 2002; Ferguson et al., 2004; Chung & Kallapur, 2003; Ashbaugh et al., 2003), restatements (Kinney et al., 2004) or going-concern opinions (Sharma & Sidhu, 2001). This study also uses accrual-based earnings management as a proxy for audit quality. Earnings management occurs when a manager use judgment in financial reporting and organize
13 transactions in such a way, that the true economic performance of a firm is not reflected by the financial statements, in order to mislead stakeholders or to influence contractual outcomes resulting in benefits for the manager (Healy & Wahlen, 1999). Earnings management can be divided into two important forms: accrual-based earnings management and real earnings management. (Badertscher, 2011) In the first form, managers adjust financial reporting by using revenue or expense accruals. Managers that use the second form of earnings management change the timing or structuring of operating, investing or financing decisions to adjust financial reporting.
2.1.4 Auditor independence and audit quality
Over the years many studies have investigated the influence of auditor independence on audit quality. An important article is that of De Angelo (1981). She argues that audit quality can be reduced by impaired auditor independence. Later on more studies examined this relation. The results of these type of studies are very conflicting. Some find a positive relation, as a result of an economic bond between auditor and client (Frankel et al., 2002; Ferguson et al., 2004; Srinidhi & Gul, 2007). Others find no significant relation (Ashbaugh et al., 2003; Chung & Kallapur, 2003). In contrast, the third group of studies even find a higher audit quality as a result of more provided NAS and hence impaired auditor independence, due to knowledge spillovers (Simunic, 1984; Beck et al., 1988). Tax NAS is positively related to correctly issuing a going-concern opinion by an audit firm (Robinson, 2008). In contrast, Basioudis et al. (2008) find a negative relation between high NAS fees and the probability of a going-concern modified opinion for financially stressed organizations in the UK. Moreover, Whisenant et al. (2003) doubt the existence of knowledge spillovers, because they do not find a relation between audit fees and NAS fees. This implies that the joint provision of audit services and NAS does not lead to knowledge spillovers. However, the results of a similar and more recent study show the opposite (Krishnan & Yu, 2011). They find that audit fees are negatively related to non-audit fees and vice versa, indicating the existence of knowledge spillovers. In conclusion, in the literature a trade-off exists between the economic bonding theory and the knowledge spillover theory. For both theories there is supporting evidence gathered by previous studies, but the findings regarding the existence of knowledge spillovers are mixed. Regulators try to limit the provision of NAS, what is reflected by the prohibition of some NAS. This indicates that regulators prefer the economic bonding theory at the cost of the knowledge spillover theory. Independence and objectivity seem to outweigh the possible benefits from knowledge spillovers.
14
2.1.5 Audit committee
The audit process by external auditors is an external control mechanism of corporate governance. An important internal control mechanism is the audit committee. In Section 205 of the SOX, an audit committee is defined as: “a committee or equivalent body established by and amongst the board of directors of an issuer for the purpose of overseeing the accounting and financial reporting processes of the issuer and audits of the financial statements of the issuer”. According to the SEC (2003), “the audit committees plays a critical role in providing oversight over and serving as a check and balance on a company's financial reporting system. The audit committee provides independent review and oversight of a company's financial reporting processes, internal controls and independent auditors. It provides a forum separate from management in which auditors and other interested parties can candidly discuss concerns. By effectively carrying out its functions and responsibilities, the audit committee helps to ensure that management properly develops and adheres to a sound system of internal controls, that procedures are in place to objectively assess management's practices and internal controls, and that the outside auditors, through their own review, objectively assess the company's financial reporting practices.” This means that audit committees have an important oversight function regarding the financial reporting process and audit process. Hiring and dismissal of auditors and the approval of NAS are two important tasks related to this oversight function. The Cadbury Committee (1992), the committee that took the first steps in the development of the UK Corporate Governance Code3 (formerly Combined Code),
describes that audit committees are powerful resources to improve corporate governance and may improve the independence of an audit and audit quality. Wild (1994) shows that managers are easier held to account by shareholders due to the formation of audit committees.
The effectiveness of an audit committee is related to its characteristics (De Zoort et al., 2002). The Blue Ribbon Committee (1999) and the SOX had both a big influence on the composition and authority of audit committees. Audit committee composition is associated with the quality of the oversight activities of the committee (Raghunandan et al., 2001). The four most important characteristics that reflect the effectiveness of an audit committee are the meeting frequency of a committee, the independence of members, the size of a committee and financial expertise of the members (Kent et al., 2010; Sultana et al., 2015; Goh, 2009). These four characteristics will be discussed below.
The first characteristic explaining audit committee effectiveness is its meeting frequency. Via an experiment in Australia, Stewart and Munro (2007) find that the existence of an audit committee and subsequently the number of meetings are associated with a lower perceived audit risk. Diligence in the form of the number of meetings shows the effort of an audit committee in fulfilling its
3 The UK Corporate Governance Code includes the best practices regarding corporate governance for firms in the
15 oversight function (Menon & Williams, 1994). The Blue Ribbon Committee (1999) suggests that audit committees should meet at least four times a year to be effective. Sultana (2015) argues that audit committees are better monitoring mechanisms when they have a high number of meetings. Abbott et al. (2000) find that at least two audit committee meetings per year are needed to have less financial misstatements, both fraudulent and non-fraudulent. The frequency of audit committee meetings is also negatively related to the degree of restatements (Abbott et al., 2004).
The audit committee requirements of the SOX show that audit committee independence and financial expertise are important in exhibiting corporate governance within an organization (SEC, 2003). Klein (2002) finds in a research under S&P 500 firms that audit committee independence is associated with less accrual-based earnings management. Bédard et al. (2004) and Dhaliwal et al. (2010) come to a similar conclusion with comparable studies. Independent audit committees are also associated with less misstatements (Abbott et al., 2000) and less restatements (Abbott et al., 2004). As mentioned in the Final Rule of the SEC (2003), all audit committee members of a firm need to be independent according to specified criteria. Because all US listed firms are required to have independent audit committee members, audit committee independence will not fall under the scope of this study. There should not be a firm in the later described sample that does not have independent audit committee members.
The size of an audit committee is also an important determinant of its effectiveness (Kent et al., 2010; Sultana et al., 2015). Although small committees are better manageable compared to large committees, a committee that consists of more members is likely to contain more knowledge, experience and expertise (Bédard et al., 2004; Blue Ribbon Committee, 1999). The results of previous studies regarding the influence of audit committee size on audit quality are mixed. Kent et al. (2010) find an association between audit committee size and the quality of accruals. Large audit committees are more able to remediate material weaknesses in a timely manner than small audit committees (Goh, 2009), probably because larger audit committees are more able to produce extensive discussions and have more monitoring power (De Zoort et al., 2002). However, Xie et al. (2003) do not find any relationship between audit committee size and financial reporting quality. The Blue Ribbon Committee (1999) argues that every audit committee should exist of at least three members in order to fulfil their monitoring function in a sufficient manner.
The fourth important determinant of audit committee effectiveness is financial expertise. The SEC (2003) describes an audit committee financial expert as someone who has: “(1) an understanding of generally accepted accounting principles and financial statements; (2) The ability to assess the general application of such principles in connection with the accounting for estimates, accruals and reserves; (3) Experience preparing, auditing, analyzing or evaluating financial statements that present a breadth and level of complexity of
16 accounting issues that are generally comparable to the breadth and complexity of issues that can reasonably be expected to be raised by the registrant's financial statements, or experience actively supervising one or more persons engaged in such activities; (4) An understanding of internal controls and procedures for financial reporting; and (5) An understanding of audit committee functions.” These attributes can be acquired by one or more of the following: “(1) Education and experience as a principal financial officer, principal accounting officer, controller, public accountant or auditor or experience in one or more positions that involve the performance of similar functions; (2) Experience actively supervising a principal financial officer, principal accounting officer, controller, public accountant, auditor or person performing similar functions; or (3) Experience overseeing or assessing the performance of companies or public accountants with respect to the preparation, auditing or evaluation of financial statements.” Furthermore, according to the SEC (2003), firms are obliged to disclose that at least one financial expert serves on its audit committee, with the name of that expert included in the disclosure. Firms that do not have financial experts in their audit committee needs to disclose this. In the latter situation, firms must also explain why the audit committee does not contain a financial expert. This means that disclosure about having a financial expert in the audit committee is an important aspect, but it is not mandatory to have one. The presence of a financial expert in an audit committee is negatively associated with the degree of accrual-based earnings management within a firm (Bédard et al., 2004), negatively associated with misappropriation of assets (Mustafa & Ben Youssef, 2010) and positively related to audit quality (Ghafran & O’Sullivan, 2013). Furthermore, audit committee financial expertise and audit committee independence are negatively associated with NAS fees (Zaman et al., 2011). High quality audit committees want to enhance auditor independence, resulting in a lower purchase of NAS by firms.
2.2 Hypotheses development
In this paragraph the hypotheses are formulated, based on the theory that is explained in the literature review section. First, the direct influence of auditor independence on audit quality will be tested. NAS purchase will be used as a proxy for auditor independence. There will be more elaborated about this proxy in the research methodology chapter. As mentioned before, a trade off exists between the economic bonding theory and the knowledge spillover theory. Literature shows that supporting evidence is gathered for both theories (Frankel et al., 2002; Fergurson et al., 2004; Srinidhi & Gul, 2007; Simunic, 1984; Beck et al., 1988; Robinson, 2008; Krishnan & Yu, 2011). However, some studies question the existence of knowledge spillovers (Basioudis et al., 2008; Whisenant et al., 2003). Because evidence regarding knowledge spillovers is mixed and because regulations, like the SEC, prefer the economic bonding theory, what is reflected by the prohibition of some NAS, and argue that NAS purchase by clients reduce auditor independence, which can
17 induce accrual-based earnings management and hence reduce audit quality, the following hypothesis is formulated:
Hypothesis 1 (H1): NAS purchase is associated with a lower audit quality.
Based on the role of the audit committee as an important internal corporate governance mechanism, it can be expected that effective audit committees, measured with the characteristics financial expertise, size and diligence, can mitigate a potential influence of impaired auditor independence on audit quality. NAS needs to be pre-approved by the audit committee of a firm (SEC, 2001). As a result of the monitoring function of an audit committee, it can be expected that an effective audit committee only accepts the purchase of NAS in cases that it does not impair audit quality. Previous studies show that financial expertise, size and diligence are important characteristics of effective audit committees (Kent et al., 2010; Sultana et al., 2015; Goh, 2009). Therefore, based on this literature and pre-described expectations, the following three hypotheses can be formulated to test the moderating role of audit committee effectiveness:
Hypothesis 2a (H2a): the association between NAS purchase and audit quality is weaker when the percentage of financial experts on an audit committee is higher.
Hypothesis 2b (H2b): the association between NAS purchase and audit quality is weaker when an audit committee is larger.
Hypothesis 2c (H2c): the association between NAS purchase and audit quality is weaker when the number of audit committee meetings is higher.
18
3 Research methodology
In this section, firstly the research method and data collection are described. After this, the process of sample selection is shown. Furthermore, the proxies and measures of the dependent variable, the independent variable, the moderating variable and the control variables are described. Finally, the empirical model is presented.
3.1 Research method and data collection
This study comprises a quantitative research. The multiple hypotheses are tested via OLS regressions. The focus of this study is on US listed firms, represented in the S&P 500 for the period 2010-2014. This timeframe is used, because it includes the most recent period of available data. Furthermore, this period excludes years during the financial crisis, because in those years NAS fees differed from NAS fees in the years after the financial crisis (Alexeyeva & Svanström, 2015). To obtain all required data, a combination of data from multiple databases, provided by Wharton Research Data Services (WRDS), and hand collected data is used. Data obtained from AuditAnalytics, Compustat and Institutional Shareholder Services (ISS) (formerly RiskMetrics) is combined with hand collected data into one file to get a single dataset that includes all required variables. Data regarding NAS fees are available with the AuditAnalytics database. This database covers all SEC registrants and therefore a focus on US listed firms makes sense. All financial data is conducted with the Compustat database. The ISS database is used to obtain data regarding financial expertise of audit committee members. Finally, data regarding the frequency of audit committee meetings were not available in one of the databases provided by WRDS. Therefore, these data are hand collected from proxy statements. According to Item 407 of Regulation S-K of the SEC, registrants are required to disclose whether audit committee members are independent, the number of meetings held and whether or not the registrant has at least one audit committee financial expert serving on its audit committee (SEC, 2003). Therefore, the number of audit committee meetings should be presented in proxy statements. These proxy statements are found with the EDGAR database of the SEC under filings type ‘DEF 14a’.
3.2 Sample selection
Table 1 summarizes the process of sample selection. The initial sample contained all firms that were included in the S&P 500 for the period 2010-2014. For all these years the total number of firms exceed 500, because during a year the list of firms included in the S&P 500 is changing regularly. Firms may take the place of other firms in the S&P 500.
19 Table 1 Sample selection 2010 2011 2012 2013 2014 Total firm-year observations Number of different firms Firms in S&P 500 518 519 518 519 515 2589 585 Financial institutions (SIC codes 6000-6999) (86) (87) (86) (88) (87) (434) (92) 432 432 432 431 428 2155 493 Missing data (43) (40) (35) (32) (33) (183) (41) 389 392 397 399 395 1972 452
Firms not for whole period (2010-2014) in S&P 500 or with deleted firm-year observations because of missing data (55) (58) (63) (65) (61) (302) (118) 334 334 334 334 334 1670 334
Firms not audited by a Big 4 auditor (1) (1) (1) (1) (1) (5) (1) 333 333 333 333 333 1665 333 Under-represented industries (9) (9) (9) (9) (9) (45) (9) Final sample 324 324 324 324 324 1620 324
All firms that have been in the S&P 500 for a part of at least one of the years of the aforementioned period are included in the initial sample. This resulted in an initial sample of 2589 firm-year observations derived from 585 different firms. In accordance with Frankel et al. (2002), financial institutions (SIC codes 6000-6999) are excluded from the initial sample, because a different way of
20 calculating discretionary accruals is required for these type of firms. Furthermore, firm-year observations with missing data regarding the variables used in this study were excluded. All firms that were not in the S&P 500 for the whole period or firms with excluded firm-year observations due to missing data were excluded. After this step all included firms have the same amount of firm-year observations (5 per firm, with a total of 1670 firm-firm-year observations). Only one of the firms was not audited by a Big 4 auditor (Deloitte, EY, KPMG or PwC). A Big 4 auditor is more likely to report a breach when he discovers one, than a non-big4 auditor (Francis et al., 1999). According to Khurana and Raman (2004), Big 4 auditors provide higher quality audits than non-Big 4 auditors, because of their reputation. The single firm that was not audited by a Big 4 auditor is excluded to end up with a sample that is as consistent as possible. The later explained discretionary accruals are calculated separately for every industry. Per industry a minimum number of observations is required to obtain reasonable parameter estimates (Jones, 1991). Some industries where poorly represented and therefore not usable. These industries were Agriculture, Forestry and Fishing, Construction, Public Administration and Non-Classifiable. The firms in these industries are deleted from the sample to be able to calculate the discretionary accruals in an accurate way. Therefore, the final sample exists of 324 firms that all have observations of the same five years. This gives a total of 1620 firm-year observations.
3.3 Dependent variable
Most studies regarding the influence of NAS purchase on audit quality assume that accrual-based earnings management, measured with discretionary accruals, is an indicator of low-quality audit (Frankel et al., 2002; Ashbaugh et al., 2003; Chung & Kallapur, 2003). Therefore, discretionary accruals are used as a proxy for the dependent variable audit quality. Discretionary accruals are the part of total accruals that are not due to changes in the firms’ economic condition or normal operating activities. In accordance with Frankel et al. (2002), the absolute level of discretionary accruals is used. Discretionary accruals can be calculated with the use of several models, namely the Healy model (Healy, 1985), the De Angelo model (De Angelo, 1986), the Jones model (Jones, 1991), the modified Jones model (Dechow et al., 1995) and the Industry model (Dechow & Sloan, 1991). In this study, the modified Jones model (Dechow et al., 1995) is used, because this is the most powerful in testing accrual-based earnings management (Dechow et al., 1995). The model is a modified version of the original model by Jones (1991). The original Jones model uses the difference in revenue between year t and year t-1, the difference in plant, property and equipment (PPE) between year t and year t-1 and the total assets at year t-1 to calculate the non-discretionary
21 accruals. With respect to the original Jones model, the only amendment in the modified Jones model is the inclusion of the difference in net receivables between year t and year t-1.
According to the modified Jones model, first the total accruals need to be calculated to compute the discretionary accruals. In accordance with Chung and Kallapur (2003), this is done separately for every industry, based on the 2-digit SIC code (see Table A1 in Appendix B). After that, the non-discretionary accruals need to be estimated and subtracted from the total accruals to calculate discretionary accruals. In accordance with Dechow et al. (1995), total accruals are calculated with the following formula:
TAt = (ΔCAt – ΔCLt – ΔCasht + ΔSTDt - Dept) / (At-1)
Where:
TAt = Total accruals in current year
ΔCAt = Change in current assets
ΔCLt = Change in current liabilities
ΔCasht = Change in cash and cash equivalents
ΔSTDt = Change in debt included in current liabilities
Dept = Depreciation and amortization expense in current year
At-1 = Total assets in previous year
The non-discretionary accruals are estimated using the following model: NDAt = α1 (1 / At-1) + α2 [(ΔREVt – ΔRECt) / At-1] + α3 (PPEt / At-1)
Where:
NDAt = Non-discretionary accruals in current year
At-1 = Total assets in previous year
ΔREVt = Revenues in current year less revenues in previous year
22 PPEt = Gross property, plant and equipment in current year, scaled by total assets in
previous year
α1, α2, α3 = Firm-specific parameters
All data used to calculate the discretionary accruals is found in the Compustat database.
After calculating the total accruals and estimating the non-discretionary accruals, the quantity of discretionary accruals can be calculated with the following formula:
DAt = TAt - NDAt
Where:
DAt = Discretionary accruals in current year
TAt = Total accruals in current year
NDAr = Non-discretionary accruals in current year
3.4 Independent variable
In this study the level of NAS purchase is the proxy for auditor independence. Most of the existing literature regarding auditor independence also use the level of NAS purchase as a proxy for auditor independence (e.g. Chen et al., 2005; Gul et al., 2006; Zhang et al., 2007). This in combination with the theory about economic bonding, as a result of NAS fees paid by clients to auditors, ensures that it makes sense to use NAS purchase as a proxy for auditor independence. NAS purchase is measured in two different ways. In accordance with Frankel et al. (2002) and Ferguson et al. (2004), NAS purchase is first measured as the ratio of NAS fees to total fees (NASFTotFR). This is in accordance with the SEC’s vision regarding the required presentation of audit fees and NAS fees in annual reports (SEC, 2003). Investors are able to assess the independence of an auditor on the basis of the proportion of NAS fees to total fees. A disadvantage of this measure is that it only reflects the ratio of NAS fees to total fees, but it ignores the absolute amount of NAS fees (Frankel et al., 2002; Ferguson et al., 2004). Through this, the financial importance of the client and the economic bonding with a client would be ignored. In accordance with Ferguson et al. (2004), Ashbaugh et al. (2003) and Srinidhi and Gul (2007), the natural logarithm of NAS fees (LnNASF) is used as a second measure of NAS purchase to inhibit this problem.
23 3.5 Moderating variable
To measure audit committee effectiveness, this study uses three different audit committee characteristics, namely financial expertise, size and the diligence of an audit committee. Financial expertise reflects the presence of financial experts in an audit committee. In previous studies (e.g. Bédard et al., 2004; Abbott et al., 2004), financial expertise is often tested as a dummy variable (1 if audit committee contains a financial expert; 0 otherwise). Nowadays, all audit committees are required to contain at least one financial expert. Therefore, a similar dummy variable would not make any sense, hence financial expertise is, in accordance with Sultana (2015) and Badolato et al. (2014), measured as the percentage of audit committee members with financial expertise. Size is measured as the number of directors in an audit committee and diligence is measured as the number of meetings of an audit committee (Kent et al., 2010). Based on the argumentation of the Blue Ribbon Committee (1999) that an audit committee should meet at least four times a year, existing literature about audit committee diligence use this number of meetings as a threshold for a dummy variable (Abbott et al., 2004; Sultana et al., 2015). Nowadays almost all audit committees meet four or more times a year. Because of this, a dummy variable with the aforementioned threshold would not be usable. Therefore, the meeting frequency of an audit committee is not measured with a dummy variable, but in accordance with Kent et al. (2010) and Othman et al. (2014), with the absolute number of meetings.
3.6 Control variables
To control for factors that have an influence on discretionary accruals, the following control variables will be used in this study: firm size (Size), auditor tenure (AudTenShort and AudTenLong), leverage (LEV), firm performance (ROA), loss (Loss) and growth (Growth).
This study controls for firm size, because there is some evidence that smaller firms generally operate without being investigated as much as larger firms (Xie et al., 2003). Therefore, smaller firms are more able to manage their earnings. Firm size also is an important factor relating to the demand for NAS (Frankel et al., 2002). In accordance with Frankel et al. (2002), size is measured as the log of the market value of equity.
Auditor tenure is added as control variable for audit tenure, because Gul et al. (2007) find that the positive relation between NAS fees and the level of discretionary accruals is dissimilar in situations of different auditor tenure. Myers et al. (2003) find a lower accruals quality when the auditor tenure is lower. In accordance with Chi and Huang (2005), a distinction is made between short (<4 years), medium (>3 years and <9 years) and long (>8 years). AudTenShort is used as a
24 dummy variable for short auditor tenure (1 if auditor tenure is less than 4 years; 0 otherwise) and AudTenLong is used as a dummy variable for long auditor tenure (1 if auditor tenure is more than 8 years; 0 otherwise). In case of medium auditor tenure, both dummy variables AudTenShort and AudTenLong have a value of 0.
LEV is used as control variable, because high-leverage firms are more likely to be involved in earnings management (Iatridis & Kadorinis, 2009). LEV is measured as total liabilities divided by total assets.
In accordance with Frankel et al. (2002), firm performance is added as control variable. Firm performance is measured as the return on assets (ROA). ROA is calculated as net income divided by the average total assets. Furthermore, firms that make a loss are less likely to use discretionary accruals to report positive earnings surprises, but use more often discretionary accruals to increase the loss in favor of the profit of next year (Brown, 2001). Therefore, Loss is added as control variable. The dummy variable is 1 if a firm makes a loss and 0 if a firm makes a profit.
The last control variable is growth. Matsumoto (2002) finds that growth-firms have an incentive to manage earnings, because those firms want to avoid negative earnings surprises. In accordance with Skinner and Sloan (2002), growth is measured with the market-to-book ratio. For a complete overview of the variables used in this study, see Table 2.
25 Table 2
Variable descriptions
Variable Proxy Measure Acronym
Dependent variable
Audit quality Accrual-based earnings management
Absolute discretionary accruals DACabs Independent variable
Auditor
independence NAS purchase NAS fee to total fee ratio NASFTotFR Natural logarithm of NAS fee LnNASF Moderating variable
Audit committee
effectiveness Audit committee financial expertise Percentage of financial experts in audit committee ACFExp Audit committee
diligence Number of audit committee meetings ACMeet Audit committee size Number of members in audit
committee ACSize
Control variables
Firm size Firm size Logarithm of market value of
equity Size
Auditor tenure Short auditor tenure Dummy variable (1 if tenure <4 years; 0 otherwise)
AudTenShort Long auditor tenure Dummy variable (1 if tenure
>7 years; 0 otherwise) AudTenLong Leverage Leverage Total liabilities to total assets LEV
Firm performance Return on assets Net income divided by average
total assets ROA
Loss Dummy variable (1 if firm
makes a loss; 0 otherwise) Loss
Growth Growth Market-to-book ratio Growth
3.7 Empirical models
The following empirical models will be used to test the four hypotheses. This will be done using OLS regression analysis. In the first model the association between the control variables and absolute discretionary accruals will be tested. Model 2a and model 2b (H1) examine the association between NAS purchase and DACabs, in which NAS purchase is measured as NASFTotFR. Model 3a and model 3b (H1) will test the association between NAS purchase and DACabs, where NAS purchase is measured as LnNASF. Model 4a, 4b and 4c (H2a, H2b and H2c) examine the
26 moderating influence of respectively ACFExp, ACMeet and ACSize on the association between NAS purchase and DACabs, where NAS purchase is measured as NASFTotFR. According to Brambor et al. (2006), interaction terms need to be used to test the influence of the moderating variables. Finally, model 5a, 5b and 5c (H2a, H2b and H2c) will test the moderating influence of respectively ACFExp, ACMeet and ACSize on the association between NAS purchase and DACabs, in which NAS purchase is measured as LnNASF. Centered values are used in the models in which interaction terms are used, to diminish the effects of potential multicollinearity problems.
Model 1: DACabs = ß0 + ß1* Size + ß2*AudTenShort + ß3*AudTenLong + ß4*LEV + ß5*
ROA + β6* Loss + ß7* Growth + ϵ
Model 2a: DACabs = ß0 + ß1* NASFTotFR + ϵ
Model 2b: DACabs = ß0 + ß1*NASFTotFR + ß2* Size + ß3*AudTenShort + ß4*AudTenLong
+ ß5*LEV + ß6* ROA + β7* Loss + ß8* Growth + ϵ
Model 3a: DACabs = ß0 + ß1* LnNASF + ϵ
Model 3b: DACabs = ß0 + ß1*LnNASF + ß2* Size + ß3*AudTenShort + ß4*AudTenLong +
ß5*LEV + ß6* ROA + β7* Loss + ß8* Growth + ϵ
Model 4a: DACabs = ß0 + ß1*NASFTotFR + ß2*ACFExp + ß3*NASFTotFR*ACFExp +
ß4*Size + ß5*AudTenShort + ß6*AudTenLong + ß7*LEV + ß8* ROA + β9* Loss + ß10* Growth
+ ϵ
Model 4b: DACabs = ß0 + ß1*NASFTotFR + ß2*ACMeet + ß3*NASFTotFR*ACMeet +
ß4*Size + ß5*AudTenShort + ß6*AudTenLong + ß7*LEV + ß8* ROA + β9* Loss + ß10* Growth
+ ϵ
Model 4c: DACabs = ß0 + ß1*NASFTotFR + ß2*ACSize + ß3*NASFTotFR*ACSize +
ß4*Size + ß5*AudTenShort + ß6*AudTenLong + ß7*LEV + ß8* ROA + β9* Loss + ß10* Growth
+ ϵ
Model 5a: DACabs = ß0 + ß1*LnNASF + ß2*ACFExp + ß3*LnNASF*ACFExp + ß4*Size +
27 Model 5b: DACabs = ß0 + ß1*LnNASF + ß2*ACMeet + ß3*LnNASF*ACMeet + ß4*Size +
ß5*AudTenShort + ß6*AudTenLong + ß7*LEV + ß8* ROA + β9* Loss + ß10* Growth + ϵ
Model 5c: DACabs = ß0 + ß1*LnNASF + ß2*ACSize + ß3*LnNASF*ACSize + ß4*Size +
ß5*AudTenShort + ß6*AudTenLong + ß7*LEV + ß8* ROA + β9* Loss + ß10* Growth + ϵ
The four assumptions regarding linear regression (Miles & Shevlin, 2001, p. 58 – p. 112) are tested (see Appendix D). All assumptions are probably infringed. The implications of these tests are presented in the limitations of this study.
3.8 Expectations
Table 3a summarizes the on literature based expectations regarding the associations between the independent variables (NASFTotFR, LnNASF) and the dependent variable (DACabs) and the associations between the control variables and the dependent variables. Table 3b gives an overview of the expectations regarding the moderating influence of audit committee characteristics on the association between NAS purchase and audit quality.
Table 3a Expectations of associations DACabs NASFTotFR + LnNASF + Size +/- AudTenShort + AudTenLong - LEV + ROA + Loss +/- Growth +
28 Table 3b
Expectations of influence of moderating variables
Association between NAS purchase and audit quality
ACFExp -
ACSize -
29
4 Results
In this chapter all results will be presented. First, the descriptive statistics and the correlation between the variables used in the models will be presented. Furthermore, the results of the regression analyses will be shown and explained.
4.1 Descriptive statistics
The descriptive statistics for the dependent variable and all independent, moderating and control variables are summarized in Table 4. All variables, except of ACFinExp and ACSize (no extreme outliers), and AudTenShot, AudTenLong and Loss (dummy variables), are winsorized at the 1st and
99th percentile to diminish the effect of potential outliers. This means that for these variables all
observations in the 1st percentile are replaced by the smallest value of the 2nd percentile and all
observations in the 100th percentile are replaced by the largest value of the 99th percentile.
The absolute value of discretionary accruals (DACabs) is on average 0,0342 with a minimum value of 0,0005 and a maximum value of 0,2281. The proportion of NAS fees to total fees (NASFTotFR) is on average 18,99%. The other 81,01% are audit fees. Of all audit committee members of the firms included in the sample, 58,27% is categorized as an audit committee financial expert (ACFExp). All audit committees met at least four times a year and the maximum of meetings was fifteen times (ACMeet). The smallest audit committee consisted of two members and the largest ten (ACSize). On average an audit committee consisted of around four members. The auditor tenure was in 3,12% of the firm-year observation less than four years (AudTenShort) and in 89,13% more than seven years (AudTenLong). This means that the auditor tenure was between four and seven years in the remaining 7,75% of the observations. Firms made a loss (Loss) in only 4,20% of the observations. Finally, the lowest ROA is negative (-5,77%), because some firms reported a negative net income.
Table 5 provides the Pearson correlations between the dependent variable and all independent, moderating and control variables. A perfect negative correlation has a value of -1 and a perfect positive correlation has a value of +1. The closer a value is to 0, the smaller the correlation. In general the correlations between the variables used in this study are weak. A small but significant correlation exist between LnNASF and DACabs (r=0,109). This means that when the natural logarithm of a NAS fee is higher, the level of absolute discretionary accruals also is higher. However, in contrast to the results of Frankel et al. (2002), NASFTotFR is not correlated with DACabs at all. NASFTotFR and LnNASF are highly correlated (r=0,734), because both measures
30 Table 4
Descriptive statistics
Mean Q1 Median Q3 Minimum Maximum SD
DACabs 0,0342 0,0104 0,0216 0,0426 0,0005 0,2281 0,0387 NASFTotFR 0,1899 0,0881 0,1698 0,2694 0,0023 0,5460 0,1280 LnNASF 13,7750 12,8886 13,8250 14,7987 9,5774 16,8359 1,4413 ACFExp 0,5827 0,2857 0,5000 1,0000 0,0000 1,0000 0,3023 ACMeet 8,4748 7,0000 8,0000 10,0000 4,0000 15,0000 2,5018 ACSize 4,2523 4,0000 4,0000 5,0000 2,0000 10,0000 1,0389 Size 10,2145 9,9069 10,1431 10,4653 9,4564 11,4089 0,4380 AudTenShort 0,0312 0,0000 0,0000 0,0000 0,0000 1,0000 0,1740 AudTenLong 0,8913 1,0000 1,0000 1,0000 0,0000 1,0000 0,3114 LEV 0,5906 0,4675 0,5906 0,7099 0,1477 1,2137 0,1862 ROA 0,0763 0,0381 0,0678 0,1074 -0,0577 0,2451 0,0558 Loss 0,0420 0,0000 0,0000 0,0000 0,0000 1,0000 0,2007 Growth 3,6454 1,6882 2,7197 4,3314 -11,2465 31,8997 4,5670
are at least partly based on NAS fees. A small but significant correlation exist between ROA and DACabs (r=0,129). In contrast, ACSize (r=-0,152) and LEV (r=-0,101) are negatively correlated with DACabs. A very small negative significant correlation exist between ACSize and ACFExp (r=-0,143) Furthermore, a significant correlation exist between LnNASF and Size (r=0,445). ROA is positively correlated with Size and negatively with LEV. Size and Growth are correlated (r=0,154) probably because both variables use the market value of equity as a basis.
The variance inflation factor (VIF) values of all independent, moderating and control variables are calculated to recognize potential multicollinearity problems (see Table A2a and Table A2b in Appendix C). The VIF values of all variables are below 2. According to the boundary formulated by Miles and Shevlin (2001, p.131), this indicates that multicollinearity does not play a role.
Table 5
Pearson correlation matrix
1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 1. DACabs 2. NASFTotFR 0,000 3. LnNASF -0,109*** 0,734*** 4. ACFExp 0,080** 0,050* 0,081*** 5. ACMeet 0,003 0,043 0,150*** 0,096*** 6. ACSize -0,152*** -0,109*** 0,050* -0,143*** -0,069** 7. Size -0,068** 0,107*** 0,445*** -0,038 0,160*** 0,116*** 8. AudTenShort 0,037 0,001 -0,024 0,062* -0,035 0,003 -0,046 9. AudTenLong -0,039 0,024 0,083*** -0,129*** 0,083*** 0,031 0,105*** -0,514*** 10. LEV -0,101*** -0,015 0,130*** 0,050* -0,077** 0,196*** -0,056* 0,046 -0,036 11. ROA 0,129*** -0,011 -0,108*** 0,015 0,013 -0,049* 0,201*** -0,008 0,004 -0,261*** 12. Loss 0,032 -0,043 -0,021 0,030 0,018 -0,037 -0,127*** -0,020 0,025 0,084*** -0,396*** 13. Growth 0,055* -0,021 -0,035 -0,043 -0,022 0,044 0,154*** -0,016 0,025 0,096*** 0,282*** -0,020
4.2 Regression results
To answer the research question, four hypotheses are tested in this study. The first hypothesis examines the association between NAS purchase and audit quality. The results of the OLS regressions are presented in Table 6. The association between the control variables and the absolute value of discretionary accruals is tested in model 1. Larger firms (b=0,009; p<0,001), firms with a higher ROA (b=0,106; p<0,001) and firms that make a loss (b=0,017; p<0,001) are associated with a higher value of absolute discretionary accruals. This implies that accrual-based earnings management occurs more often at large firms, at better performing firms and at firms that make a loss. Further, a negative relation exist between LEV and DACabs (b=-0,017; p<0,01). This indicates that the higher the level of leverage in a firm, the more often accrual-based earnings management occurs. In contrast to what was expected based on literature (Gul et al., 2007; Myers et al., 2003; Matsumoto, 2002), auditor tenure and growth do not significantly affect absolute discretionary accruals.
In model 2a and model 2b the first hypothesis is tested with NASFTotFR as a measure of NAS purchase. Model 2a test the single association between NASFTotFR and DACabs. In model 2b the control variables are added. Both models show that in contradiction to what is expected, NASFTotFR is not associated with DACabs. The control variables that were associated with DACabs in the first model are still significantly associated in model 2b.
Model 3a and model 3b test the first hypothesis with LnNASF as a measure of NAS purchase. A positive association between LnNASF and DACabs is expected, but instead a negative association is found in model 3a (b=-0,003; p<0,001). This implies that the level of absolute discretionary accruals decrease for firms with a higher absolute level of NAS fees. Only a small fraction of the variance in absolute discretionary accruals is explained by LnNASF (R2=0,012).
After adding the control variables in model 3b, LnNASF is not significant anymore. However, the control variables Size, LEV, ROA and Loss are still significantly associated with DACabs.
The five models show that R2 is the lowest in model 2a. Here, 0,000 of the variance in
DACabs is explained by NASFTotFR. In this model, the F-value also is 0,000 and insignificant. This is a sign that NASFTotFR does not explain the variance in DACabs.
33 Table 6
OLS regression: the association between NAS purchase and absolute discretionary accruals
Model 1: DACabs = ß0 + ß1* Size + ß2*AudTenShort + ß3*AudTenLong + ß4*LEV + ß5* ROA + β6* Loss + ß7* Growth + ϵ
Model 2a: DACabs = ß0 + ß1* NASFTotFR + ϵ
Model 2b: DACabs = ß0 + ß1*NASFTotFR + ß2* Size + ß3*AudTenShort + ß4*AudTenLong + ß5*LEV + ß6* ROA + β7* Loss + ß8* Growth + ϵ
Model 3a: DACabs = ß0 + ß1* LnNASF + ϵ
Model 3b: DACabs = ß0 + ß1*LnNASF + ß2* Size + ß3*AudTenShort + ß4*AudTenLong + ß5*LEV + ß6* ROA + β7* Loss + ß8* Growth + ϵ
Variable Model 1 Model 2a Model 2b Model 3a Model 3b
Intercept 0,127 (5,44)*** 0,034 (19,85)*** 0,123 (5,46)*** 0,075 (8,05)*** 0,124 (5,29) NASFTotFR 0,000 (0,02) 0,001 (0,71) LnNASF -0,003 (-4,39)*** -0,001 (-1,63) Size 0,009 (-3,91)*** -0,009 (-3,96)*** -0,007 (-2,70)** AudTenShort 0,006 (1,02) 0,006 (1,01) 0,006 (0,96) AudTenLong -0,003 (-0,76) -0,003 (-0,77) -0,002 (-0,69) LEV -0,017 (-3,12)** -0,017 (-3,11)** -0,015 (-2,87)** ROA 0,106 (5,24)*** 0,107 (5,27)*** 0,101 (4,90)*** Loss 0,017 (3,33)*** 0,017 (3,36)*** 0,017 (3,25)*** Growth 0,000 (1,45) 0,000 (1,47) 0,000 (1,31) R2 0,042 0,000 0,042 0,012 0,043 Adj. R2 0,038 -0,001 0,037 0,011 0,038 F 9,97*** 0,000 8,79*** 19,28*** 9,02***
34
N 1620 1620 1620 1609 1609
***, ** and * denote significance level at respectively p<0,001, p<0,01 and p<0,05. See 3.7 Empirical models for definitions of variables.
The models in which NAS purchase is measured by LnNASF show all a smaller sample size compared to the models in which NAS purchase is measured by NASFTotFR. In 11 firm-year observations, firms did not purchase NAS. A natural logarithm of zero does not exist. Therefore, those 11 firm-year observations are deleted from the total sample when NAS purchase is measured by LnNASF.
Hypotheses 2a (ACFExp), 2b (ACMeet) and 2c (ACSize) examine the moderating influence of audit committee effectiveness on the association between NAS purchase and discretionary accruals. These hypotheses are tested with respectively model 4a, model 4b and model 4c. Table 7 presents the results of these models. Model 2a and model 2b already showed that no evidence is found regarding a significant association between NASFTotFR and DACabs. Therefore, significant interaction terms of one of the three moderating variables would be highly unlikely. However, the three models do not show a significant effect of the interaction terms on the association between NASFTotFR and DACabs.
Although the three interaction terms that test the moderating influence of audit committee effectiveness do not show a significant relation, a significant association is found between ACFExp and DACabs (b=0,009; p<0,01) in model 4a. This indicates that a higher percentage of financial experts in an audit committee is associated with more absolute discretionary accruals. In model 4c a significant negative association is found between ACSize and DACabs (b=-0,005; p<0,001). This indicates that a larger audit committee is associated with less absolute discretionary accruals.
Table 8 shows the influences of the moderating variables ACFExp, ACMeet and ACSize on the association between LnNASF and DACabs. This means that the models in this table again test hypotheses 2a, 2b and 2c, but with another measure of NAS purchase. In model 5a, the moderation of ACFExp is presented. Model 5b uses ACMeet as a moderator. ACSize is used as a moderator in model 5c.
The results of Model 5a show, in accordance with model 3a, a negative association between LnNAS and DACabs (b=-0,002; p<0,05). This indicates that a higher value of NAS fees is related to less absolute discretionary accruals. However, the interaction term
35 Table 7
OLS regression: the moderating influence of audit committee effectiveness on association between NAS purchase and absolute discretionary accruals
Model 4a: DACabs = ß0 + ß1*NASFTotFR + ß2*ACFExp + ß3*NASFTotFR*ACFExp + ß4*Size + ß5*AudTenShort + ß6*AudTenLong + ß7*LEV + ß8* ROA + β9* Loss + ß10* Growth + ϵ Model 4b: DACabs = ß0 + ß1*NASFTotFR + ß2*ACMeet + ß3*NASFTotFR*ACMeet + ß4*Size + ß5*AudTenShort + ß6*AudTenLong + ß7*LEV + ß8* ROA + β9* Loss + ß10* Growth + ϵ Model 4c: DACabs = ß0 + ß1*NASFTotFR + ß2*ACSize + ß3*NASFTotFR*ACSize + ß4*Size + ß5*AudTenShort + ß6*AudTenLong + ß7*LEV + ß8* ROA + β9* Loss + ß10* Growth + ϵ
Variable Model 4a Model 4b Model 4c
Intercept 0,126 (5,37)*** 0,130 (5,49)*** 0,109 (4,61)*** NASFTotFRa 0,003 (0,42) 0,005 (0,70) -0,000 (-0,03) ACFExpa 0,009 (3,02)** NASFTotFR*ACFExpa 0,047 (1,89) ACMeeta 0,000 (0,46) NASFTotFR*ACMeeta 0,000 (0,05) ACSizea -0,005 (-4,97)*** NASFTotFR*ACSizea -0,007 (-0,98) Size -0,009 (-3,84)*** -0,009 (-3,98)*** -0,007 (-3,24)*** AudTenShort 0,006 (0,97) 0,006 (1,01) 0,007 (1,06) AudTenLong -0,002 (-0,57) -0,003 (-0,79) -0,002 (-0,66) LEV -0,018 -0,016 -0,012
