The effect of audit firm rotation on audit fee & audit fee disclosure in the Dutch context

(1)

The effect of audit firm rotation on audit fee & audit fee disclosure in

the Dutch context

Tristan Brouwer

1

MSc. Accountancy & Controlling, Track Accountancy

Supervisor: drs. W. (Wilfred) Kevelam

Co-assessor: prof. dr. R.L. (Ralph) ter Hoeven

Abstract

This paper tries to find out what the effect of audit firm rotation is on the audit fee and audit fee disclosure of companies. The sample of the paper consist out of Dutch listed firms, which were listed on the AEX, AMX or AScX index on March 1st_{, 2017. The data is handpicked from the annual reports}

of the companies on the period 2012-2016. The audit fee is obtained from the audit fee disclosure and the quality of audit fee disclosure is measured using a disclosure index. Using pooled OLS models and fixed effects models, I find that the most important variable of audit fee is the size of the auditee. Also the audit fee is lower after an audit firm rotation. The audit firm rotation has no influence on the quality of the audit fee disclosure of companies, but the inherent risk of a company has a significant negative influence on the audit fee disclosure. Lastly the size of the company is positively related to the quality of the audit fee disclosure.

Word count: 12,309 (excluding tables)

Groningen, June 9

th

, 2017

(2)

1. Introduction

Introduction

After the financial crisis, auditors have appeared negatively in the news. Many people argued that the auditors were not independent of their clients and therefore they are also partly responsible for the financial crisis (independent, 2011). Committees, established by parliaments, which researched the role of auditors during the financial crisis concluded that the way audit firms operate and the familiarity with the client were a cause of the financial crisis (Financial Times, 2011). In response, the European Union made new regulation (Regulation (EU) No 537/20142_{) in which is specified that organizations of public interest need to appoint a new audit firm at least}

every ten years. This measure was taken to prevent a breach in the independence of the audit firm because of a familiarity threat. The Dutch regulator already made his own regulation in 2012, in which it is stated that audit firms should rotate at least every eight years. The conflicting regulations caused confusion and therefore the Dutch regulator adopted the regulation made by the EU (Wijzigingswet Financiële Markten 20163_{). The}

regulation from the EU will be effective from 2016 onwards. Since most audit firms have long standing relationships with their clients and the Dutch regulation was effective from 2014 onwards, a large rotation of audit firms in 2014 till 2016 is the result of the new regulation.

The newly assigned audit firm should be more independent due to the new regulation and hence the audit quality should improve (DeAngelo, 1981). The audit fee is a good proxy of the audit quality according to Hoitash, Markelevich & Barragato (2007). So when studying the audit fee, by induction, we can measure the audit quality. There are however multiple factors that determine the audit fee. This study attempts to find what the effect of audit firm rotation is on the audit fee and which other factors determine the audit fee. Besides the audit fee, the audit fee disclosure is researched to find out what the effect of audit firm rotation is on audit fee disclosure and which factors influence the disclosure of audit fees. The disclosure of audit fees is researched because the audit fee disclosure of companies should motivate a change in audit firm. Best practice indicates that the audit fee disclosure should comprise why a different audit firm performs the audit and why the audit fee differs from previous year. This paper attempts find the underlying variables that determine the quality of the audit fee disclosure.

Scientific contribution & practical relevance

The research on audit fee is relatively extensive and almost all thinkable determinants have been researched. This paper will extend the research done so far by researching in a Dutch context, which has barely been researched

2_See

http://publications.europa.eu/en/publication-detail/-/publication/567809be-e656-11e3-8cd4-01aa75ed71a1/language-en

(4)

https://www.rijksoverheid.nl/documenten/kamerstukken/2015/05/07/wetsvoorstel-wijzigingswet-financiele-since most research focusses on the US. Therefore this paper tries to find if the research done so far is also generalizable to the Dutch context and other jurisdictions similar to the Dutch’s. There is a lot of evidence on determinants of audit fee but most studies test only one or a couple of determinants. This paper will test all determinants in conjunction with each other to find evidence which is more robust in reality, since the determinants influence each other and this can alter the results. Besides audit fee, this paper will research the audit fee disclosure which has barely been researched because parts of this disclosure are not mandatory in many jurisdictions. There is for instance no literature connecting the change of auditors to the audit fee disclosure. In addition, there is almost no evidence on the determinants of the audit fee disclosure. This paper tries to fill this gap in the literature, connecting determinants of auditees and audit firms to audit fee disclosure including the audit firm rotation. Furthermore this paper will cover the five most recent years, which complements the existing research with new data and new elements since the audit market has changed quite a bit since the first studies were published. Most importantly this research is performed in a unique environment where there are many auditees changing from audit firm due to new regulation, and this incorporates the first mandatory change in audit firms in the Netherlands. This is a unique opportunity that enables a thorough investigation of the effect of an audit firm rotation on the audit fee and audit fee disclosure. This study might serve a practical use as well. In defining the determinants of the audit fee and audit fee disclosure, the shareholders and directors of companies might use this study in explaining the underlying factors that define the audit fee and audit fee disclosure. If shareholders or directors want to influence the costs of an audit, they know which firm characteristics or auditor characteristics influence the audit fee after reading this paper.

Development of the central question

The research about audit fee is relatively extensive. One of the first researchers that researched audit fee is Simunic (1980). He performed a study of the pricing of audit services using a mathematical model and a questionnaire in 1977 in the US. He finds that the pricing of audits depend on the possible litigation risk that may arise from the audit. Also the interests of the auditor and the external world are aligned in providing reliable information and the competitiveness on the audit market is irrelevant for the price of an audit. This research is replicated in other geographical regions. For instance Cobbin (2002) conducts a research that incorporates 17 different countries. He finds similar results as Simunic (1980) and further finds that the size of the auditee is the single and most dominant determinant of the audit fee. After these papers the research extended fast and also individual determinants of audit fee were researched. An example of this in a French context is Gonthier-Besacier & Schatt (2007). They research which determinants of an auditee have an influence on the audit fee. They find that the size and firm risk of an auditee increases the audit fee. Also if the two auditors of the auditee (in France every mandatory audit is performed by two audit firms) are both Big-4 auditors the audit fee is lower. In an Italian context Corbella et al. (2015) research the effect of auditor rotation on audit fee. They find that audit fee was lower after the rotation if the new auditor is a Big-4 firm and unchanged for a non-Big-4 firm. They also find that audit quality improves after rotation.

(5)

The literature about audit fee disclosure is relatively scarce. Because the audit fee disclosure is a partially voluntary and partially mandatory disclosure. EU-IFRS mandates that large companies disclose the overall audit fee, other assurance fees, tax fees, the change of audit firm and comparative figures (ICAEW, 2011). The hypotheses concerning the quality of the audit fee disclosure are built on the literature of voluntary disclosure, since most content of the audit fee disclosure is disclosed voluntary and there is little evidence on the audit fee disclosure. Myers, Myers & Omer (2003) call for a mandatory auditor rotation because they find that the quality of the disclosures (earnings quality) decreases with auditor maturity. Dunn & Mayhew (2004) research if there is a connection between auditor industry specialization and disclosure. They find that if a firm is audited by the industry specialist, the disclosure and the disclosure quality are enhanced.

The change of audit firm is central in this research but we also attempt to find out what the relation is between the audit fee and auditee’s risk, governance effectiveness, and the auditor effectiveness. Also the audit fee disclosure is investigated for the auditee’s risk, governance effectiveness and auditor effectiveness. All the variables are examined before and after the change of audit firm to find if there are differences in audit fee and audit fee disclosure. Therefore the research question is:

What is the effect of a change of audit firm by auditees on the audit fee and the quality of the audit fee disclosures of the auditees in the Netherlands?

Methodology & results

This research is conducted in the Dutch context. The sample consists of the large, mid, and small cap stocks, in total 75 companies, which are investigated for five consecutive years from 2012 till 2016. Due to IPOs and delistings, an unbalanced panel of data is obtained. These data are hand-picked from the annual reports of the concerned companies. The audit fee can be obtained from the audit fee disclosure but for the quality of the audit fee disclosure, a disclosure index is developed to transform the disclosure into an interval variable. A fixed effects model with cross-sectional fixed effects and pooled OLS models are used to investigate this panel data for the audit fee and audit fee disclosure. The results show that the most important determinant of audit fee is the size of the auditee. The rotation of audit firm has a negative influence on the audit fee, but for companies composed in the AMX the audit fee increased after the rotation of audit firm. This suggest a heterogeneous reaction to the rotation of audit firm. The rotation of audit firm has no influence on the audit fee disclosure. The only stable determinants of the audit fee disclosure are the size of the auditee and the inherent risk of the auditee.

Structure of the paper

The second section of this paper will first introduce the underlying theory that is applicable to an audit and thereafter the empirical results found by other researchers on audit fee and audit fee disclosure are presented.

(6)

out how the data is gathered, how the determinants are operationalized and which methodology is used in this paper to answer the central question. The fourth section presents the empirical results found per determinant on the basis of the data and thereafter robustness test are conducted to ensure results of high quality. The fifth section concludes and discusses the limitations of this research and will come up with ideas for further research.

(7)

2. Theory

,

empirical findings and hypotheses development

This section provides the theoretical basis of this paper. First the theory on audits underlying the hypotheses is developed. Next the empirical findings are presented and in conjunction with the theory, the hypotheses that are tested in this paper are constructed.

Underlying theory

There are several theories that are used to develop the hypotheses in the next section. The most prominent one is the “Agency Theory” which is described in the first subsection. Another view on the relation between the agent and the principal is given in the second subsection with the “Stewardship Theory.” Thereafter a more economical view is taken on the audit market and the “Theory of Market Failure” is described in the third subsection. Lastly the “Voluntary Disclosure Theory” is set out in the fourth subsection.

2.1.1. Agency Theory

The agency theory is founded and described by Jensen & Meckling (1976). This theory describes the relation between the shareholder of a corporation and the manager of the same corporation. In this relation the shareholder is the principal and the manager is the agent. The principal hires the agent to perform work on behalf of the principal, in return the agent gets remunerated. Both the agent and the principal are maximizing their own utility without taking into account the utility of the other party. According to Eisenhardt (1989), agency theory is concerned with resolving two problems, namely the misalignment of utility between the principal and the agent and the verification of what the agent is actually doing. The second problem arises because the shareholder is not involved in the company and therefore he cannot observe the actions that the manager takes. This gives rise to information asymmetry because the agent has much more information about the corporation than the principal does. Since the agent is only concerned with his own utility, he would use his information advantage to maximize his own utility at the cost of the utility of the principal, which is called moral hazard. Beaver (1981) states that this moral hazard can be solved in two ways. The first way is designing incentive contracts for the agent so that the incentives of the agent and principal are aligned. The second way is to provide complete public disclosure of the firm’s information in order to remove the superior information position of the agent. In order to eliminate the information asymmetry and to resolve the agency problems, agency costs need to be made to align the utility of the agent and the principal. An example of agency costs is the hiring of an audit firm which will perform an audit on the corporation (Williams, 1988). The principal might hire an audit firm because this is an independent monitor and can assure the fairness of the financial statement disclosure. This should resolve agency problems since the audit firm verifies what the agent is actually doing and the information asymmetry is (partly) resolved. The auditor reports his findings to the principal using an audit report in the financial statements of the annual report of the company. Francis & Wilson (1988) and Defond (1992) state that the remuneration of the auditor should be

(8)

disclosure and therefore the principal can judge if remuneration of the audit firm is sufficient to prevent the moral hazard. A long standing relation between the client and the audit firm can cause familiarity between the agent and the auditor which increases the risk of moral hazard. This familiarity threat can be eliminated by changing from audit firm.

2.1.2. Stewardship Theory

The stewardship theory takes a different view on the relation between the agent and the principal than the agency theory. The stewardship theory states that agents are not motivated by individual goals, but rather the agent’s objectives are aligned to the objectives of the principal (Davis, Schoorman & Donaldson, 1997). Bernstein, Buse & Bilimoria (2016) state that the stewardship theory may be viewed in two ways. The first way is that the agent will act in the best interest of the principal even if their interests diverge, because in doing so they will accomplish higher personal outcomes of achievement, affiliation, and self-actualization. The second way is that the principal's and agent's goals are in fact perfectly aligned because of commonality of interests. Williams (1988) argues that under the stewardship hypothesis, an agent would seek an audit firm who satisfies the principal’s need for assurance. In return, the principal expects that the agent will seek an audit firm that can provide quality. The agent would like to hire an audit firm, so that the audit firm can reflect the favorable image of the agent as a good steward of the principal’s investment. A way for the audit firm to reflect the favorable image of the agent as a good steward of the principal’s investment is by issuing an unqualified opinion. A long standing relation with the audit firm and a low remuneration of the audit firm might affect the image of a good steward due to the familiarity threat. Therefore the agent demands a new audit firm every so often and discloses the audit fee in the audit fee disclosure to improve the image of a good steward of the principal and to prevent the appearance of familiarity.

2.1.3. Theory of Market Failure

The agency theory already made it clear that there exist information asymmetry between the agent and the principal. Since there exist information asymmetry it can be stated that the market of holding shares (current and future principals) of corporations is not efficient and hence there is a Pareto improvement possible. This problem is described by the theory of market failure (Bator, 1958). Leftwich (1980) describes that accounting information can make the market more efficient in the case of a market failure. The accounting information is compiled by the agent. To make the accounting information reliable, and achieve a Pareto improvement, an audit firm is hired on behalf of the principal. The audit firm will audit the accounting information compiled by the agent. The hiring of an audit firm should therefore limit the market failure and produce reliable accounting information for the principal. The hiring of the audit firm should therefore reduce the information asymmetry and create more efficient working markets. DeFond, Wong & Li (1999) find that after the implementation of the rotation of audit firms the perceived audit quality increases. This should create more efficient working markets since investors can rely more on the opinion of the audit firm.

(9)

2.1.4. Voluntary Disclosure Theory

In the absence of market imperfections or externalities, firms have incentives to optimally trade off the costs and benefits of voluntary disclosure according to Healy & Palepu (2001). Also firms, in the absence of market imperfections, have an incentive to produce an efficient level of information for investors because there is no information asymmetry. However in reality, the markets are imperfect. The theory of market failure describes the need for accounting data in markets that are not perfect, according to Leftwich (1980). A reason for this is that information is produced for the current shareholders but future shareholders also benefit for the availability of the information. This will lead to a potential underproduction of information by the firm. A second explanation by the same author is that disclosure regulation, which may limit the production of voluntary information, is motivated by other reasons than market failure. For instance the redistribution of wealth can be a reason to implement new disclosure regulation instead of market efficiency. So to overcome the market inefficiencies, companies tend to disclose extra information voluntary. This extra information also includes the audit fee disclosure. Ball, Jayaraman & Shivakumar (2012) note that since shareholders want credible information, the voluntary disclosure must be audited by an audit firm and this can be done at a relatively low cost. They also note that there is no difference between an audit on voluntary information containing good news or bad news. Myers, Myers, & Omer (2003) note that the earnings quality improves after an audit firm rotation. This means that after a rotation of audit firm the disclosures, voluntary and mandatory, are of higher quality.

Empirical findings & Hypotheses development

In this subsection the empirical literature per determinant will be presented. The relevant empirical literature about audit fee will first be presented for auditee’s risk, governance effectiveness, audit firm effectiveness and audit firm rotation. Thereafter the relevant literature about audit fee disclosure will be presented for the same determinants. Once the empirical background is laid, the hypotheses will be developed in conjunction with the theory mentioned in the previous subsection.

2.2.1. Empirical literature & hypotheses development about audit fee

As mentioned in the introduction, the literature surrounding audit fee is extensive and is frequently researched. Hay (2013) performs a meta-analysis on the research done so far in the field of audit fees. He finds several significant relations with respect to client attributes, auditor attributes, and engagement attributes. The client attributes and auditor attributes will also be used in this paper to explain the underlying determinants of the audit fee and audit fee disclosure because they are easily observable in the annual report. In contrast, the engagement attributes are difficult to observe since this usually contains confidential information. Because of the limited availability of data, this attribute will not be used in this paper. The client attributes and auditor attributes are subdivided into two smaller determinants. The client attributes are divided into auditee’s risk and governance effectiveness. The auditor attributes are divided into audit firm effectiveness and audit firm rotation. The

(10)

empirical results found so far on each determinant will be stated below. The research framework on audit fee is graphically depicted in figure 1 on page 11.

The first empirical study that connects audit fee to the risk arising from the auditee is Simunic (1980). He tests if possible litigation risk influences the audit fee. He finds that auditors take the possible litigation risk into account when setting the audit fee. This means that auditors ask a risk premium on top of the normal audit fee if there is a higher than normal risk for an auditee. This is in line with the Agency Theory and the Theory of Market Failure which state that if there is more information asymmetry or risk, a larger audit effort is needed which results in a higher audit fee. Several studies confirm the existence of a risk premium (Bell, Landsman & Shackelford, 2001; Bell, Doogar & Solomon, 2008; Seetharaman, Gul & Lynn, 2002). In addition, these studies find that the hourly wage of the auditors is not increased buy only the amount of hours worked on the auditee. O'Keefe, Simunic & Stein (1994) search for determinants of risk that effect the audit fee. In their research they use the following variables: Client’s total assets at the end of the fiscal year, client’s percentage of foreign assets to total assets, the leverage of the auditee, and the assessed riskiness (a dummy variable) of the client compared to an average client. They find that all these variables have a significant relation on the audit fee. Basioudis, Papakonstantinou & Geiger (2008) research the relation between audit fee and going concern reporting decisions. They find that the audit fee is higher if there is a going concern issued by the audit firm about the auditee. This means that a going concern reporting decision may be an indicator of increased audit risk. Hay, Knechel & Wong (2006) perform a meta-analysis on the determinant of audit fee and confirm the results mentioned above. Building on previous research we therefore expect that there is a positive relation between the risk of an auditee and the audit fee. The following hypothesis is:

H1: There is a positive relation between the risk of an auditee and the audit fee.

DeZoort et al. (2002) state that if the governance effectiveness is studied, the only observable determinant is the composition of the board. Therefore the governance effectiveness is measured by the composition of the board in this study. Most empirical research done so far is executed in an American context, where a one-tier board is standard. This paper will research Dutch listed companies where a two-tier board is standard. In the American context the board, which consist out of the executive board and supervisory board, will choose the audit firm and its accompanying audit fee. In the Dutch context the audit firm and its accompanying audit fee is selected by the supervisory board. This difference in governance set up may causes some generalization issues, but this paper will research if the results found in other papers is also generalize to the Dutch context. Abbott et al. (2003) find that the audit fee is positively related to the audit committee independence and the presence of financial expertise in the audit committee. In the Netherlands audit committees are mandatory and consists of only outside directors, therefore all audit committees in the Netherlands are independent. The tenure of audit committee members is researched by Chan, Liu & Sun (2013). They state that the tenure of the audit committee members can have a positive and a negative influence on the audit fees. The positive influence is due to the protection of the reputation

(11)

of the audit committee members, an increased audit effort is mandated. The negative influence is due to the familiarity of the organization and therefore a decreased audit effort is requested. Chan, Liu & Sun (2013) find that audit fee is negatively related to tenure of the audit committee members. Kalelkar (2017) researches the relation between audit fee and the size of the audit committee and finds that larger audit committees have lower audit fees. This finding may be due to a greater ability to oversee the company. This finding is in line with the Agency Theory, the Stewardship Theory and the Theory of Market Failure which state that if the governance is more effective, less information asymmetry exist which will result in a lower audit effort and lower audit fee. The effect of a female chair of the audit committee is researched by Ittonen, Miettinen & Vähämaa (2011). They find that the audit fee is reduced if the chair of the audit committee is hold by a female. This may be due to a lower assessed audit risk by the external auditor. Based on the previous research we therefore expect that there is a negative relation between the governance effectiveness and the audit fee. The hypothesis is formulated the following:

H2: There is a negative relation between the governance effectiveness of an auditee and the audit fee.

The audit fee is related to the characteristics of the audit firm. The characteristics of an audit firm, in turn, are related to the effectiveness of an audit firm. One of the first empirical studies that relate auditor characteristics to audit fees is Craswell, Francis & Taylor (1995). They research the effect of Big-8 audit firms and the industry specialisation on the audit fee. They find that industry specialist Big-8 audit firms earn a premium of 34% over non-specialist Big-8 audit firms. In addition they find a Big-8 brand premium of 30% over non-Big-8 audit firms. Casterella et al. (2004) confirm these results. Also the meta-analysis of Hay, Knechel & Wong (2006) confirm the existence of a premium if the audit firm is an industry specialist and/or a Big-4 firm. The premium for the industry specialist is explained by Casterella et al. (2004) by using the strategic framework of Porter (1985). Since the industry specialist follows a differentiation strategy, the industry specialist has a competitive advantage over non-specialist audit firms. This will lead to better audit quality and higher audit fees. The premium for the Big-4 audit firms over non-Big-4 audit firms is related by the supply of audit quality. DeAngelo (1981) hypothesizes that larger audit firms supply more audit quality because they are more independent of their clients. So based on the research mentioned above we can state that the characteristics of an audit firm can increase the audit quality and in return the audit fee. The hypothesis is formulated as follows:

H3: There is a positive relation between the audit firm effectiveness and the audit fee.

One of the first empirical studies that researched the effect of a change in audit firm on the audit fee is Simon & Francis (1988). They find that the audit fee is decreased after a new auditor is appointed. Ettredge & Greenberg (1990) and the meta-analysis of Hay, Knechel & Wong (2006) confirm the findings of Simon & Francis (1988). Ettredge & Greenberg (1990) also research why the audit fee is reduced. They find that if the new audit firm is a Big-4 firm the new audit fee will be lower than last year. If an audit firm is an industry specialist the audit fee will also be lower than previous year. The same holds for an auditee with bankruptcy risk, however this finding

(12)

is not significant. The finding that Big-4 audit firms have a negative impact on audit fees when an audit firm change takes place may be explained by the lowballing argument to win new clients. Based on these findings a negative relation between the change of an audit firm and the audit fee is suggested. To my knowledge there is no empirical theory connecting audit fee to auditee’s risk or governance effectiveness after a change in auditor. Since the auditee’s risk has a positive influence on audit fee and the change in audit firm has a negative influence, the relation between the auditee’s risk on audit fee after a change of audit firm is ambiguous. The effect of governance effectiveness and audit firm effectiveness on audit fee is negative just as the change of audit firm is. Therefore it is proposed that the combined effect will also be negative. The hypotheses based on the previous research can be formulated as follows:

H4: There is a negative relation in the first year after the rotation of an audit firm and the audit fee. H5: There is no relation between the risk of an auditee and audit fee in the first year after the rotation

of an audit firm when it is added as moderator.

H6: There is a negative relation between the governance effectiveness of an auditee and audit fee in the first year after the rotation of an audit firm when it is added as moderator.

H7: There is a negative relation between the audit firm effectiveness and audit fee in the first year after the rotation of an audit firm when it is added as moderator.

Figure 1: A graphical overview of the research framework concerning audit fee.

2.2.2. Empirical literature & hypotheses development about audit fee disclosure

The literature about audit fee disclosure is scarce and the topic is not often researched. This may be because audit fee disclosures are not mandatory in all jurisdictions. This in turn means that there are still some gaps in the literature about audit fee disclosure that need to be filled. But since the literature is so scarce, also the literature of voluntary disclosure will be employed because the audit fee disclosure also consists partly out of voluntary disclosures. The same attributes used in the previous model will also be used in the model for audit fee disclosure. This means that the audit fee disclosure will be researched for auditee’s risk, governance effectiveness, audit firm effectiveness and audit firm rotation. The research framework on audit fee disclosure is graphically depicted in figure 2 on page 14.

(13)

One of the first researchers that researched audit fee disclosure is Dye (1991). He states that auditor independence is improved if the audit fee is disclosed because there are no quasi-rents that can influence the opinion of the auditor and the public can assess the independence of the audit firm. The study that is most prominent in the field of audit fee disclosure is that of Lai (2009). Lai (2009) researches if auditors are more likely to issue modified audit opinions, which is a measure of auditor independence, after the disclosure of audit fees became mandatory in the US. He finds that auditors are more likely to issue going concern opinions after the audit fee is disclosed than when there is no audit fee disclosure. This confirms the theory developed by Dye (1991) that auditor independence is increased after the disclosure of the audit fee and hence the audit fee disclosure is relevant. Also large and well-known (Big-4) audit firms may induce firms to disclose more information (Firth, 1979). The research of Firth (1979) also finds that the size of an auditee is an important determinant to voluntary disclosure. This may be the case because the information production is less costly for large firms. This finding is confirmed by Raffournier (1995). He also suggest that the industry in which a company is active and the leverage can be determinants of voluntary disclosure and are positively associated with the amount of voluntary disclosure. These relations are hypothesized because if companies are active in more risky industries and have more leverage, companies are more risky and shareholders demand more information. The Theory of Market Failure and the Voluntary Disclosure Theory also predict that more risky companies disclosure more information because in risky companies there is more information asymmetry and to overcome this, more information must be disclosed. Cooke (1992) and Meek, Roberts & Gray (1995) find that the industry in which a company operates is a determinant of voluntary disclosure. The research if leverage of a company influences the voluntary disclosure, is mixed. Some studies find that leverage has a positive relation on the voluntary disclosure (Meek, Roberts & Gray, 1995) and other studies find that there is no relation between the leverage and the voluntary disclosure (Raffournier, 1995). Based on the previous research it is hypothesized that there is a relation between auditee’s risk and voluntary disclosure. Therefore the hypothesis is:

H8: There is a positive relation between the risk of an auditee and audit fee disclosure.

The board of a company discloses information voluntary to increase the transparency according to the Voluntary Disclosure Theory (Ho & Wong, 2001). In the literature many determinants are proposed that influence the voluntary disclosure of information. These determinants are independent board directors (Chen & Jaggi, 2001; Eng & Mak, 2003; Ho & Wong, 2001), the appointment of a nonexecutive director as chairman (Forker, 1992; Ho & Wong, 2001), and the existence of an audit committee (Forker, 1992; Ho & Wong, 2001). The audit committee is mandatory in Dutch listed companies and they contract the audit firm. Mangena & Pike (2005) state that the audit committee has an influence on the information disclosed. It is expected that the audit committee also has an influence on the audit fee disclosure since they are the body within the company that oversees the audit. Li, Mangena & Pike (2012) find that the size of the audit committee is positively related to the disclosure of a company. To my knowledge there is no research connecting other specific elements of the audit committee

(14)

to voluntary disclosure and/or audit fee disclosure. Based on the literature mentioned above, a positive relation between the audit committee and the audit fee disclosure can be proposed. The hypothesis is:

H9: There is a positive relation between the governance effectiveness of an auditee and audit fee disclosure.

Several studies find that more information is disclosed if the audit firm is one of the Big-4 audit firms (Firth, 1979; Raffournier, 1995; Braam & Borghans, 2014). Clarkson, Ferguson & Hall (2003) state that Big-4 audit firms have a positive influence on the disclosure of voluntary information because Big-4 audit firms are conservative and therefore avoid litigation risk and demand more disclosure of information. Dunn & Mayhew (2004) research if there is a connection between auditor industry specialization and the overall disclosure strategy. They find that industry specialist auditors have a positive influence on the amount of information disclosed. This may be caused by the choice of hiring an industry specialist audit firm. This signals the shareholder that the company is enhancing disclosures which is in line with the Stewardship Theory. To my knowledge there is no further research connecting characteristics of audit firms to voluntary disclosure or disclosure of audit fees. Based on the previous research a positive relationship is proposed between the audit firm effectiveness and audit fee disclosure. The hypothesis is:

H10: There is a positive relation between the audit firm effectiveness and the audit fee disclosure.

The main topics of research after a change of audit firm are the change in audit quality and the change in earnings quality. Myers, Myers & Omer (2003) find that after a change of audit firm, the earnings quality improves. The earnings quality improves because long auditor tenure places greater constraints on management decision about reporting and long auditor tenure is related to conservatism. The audit quality should also improve according to Bowlin, Hobson & Piercey (2015). They state that the audit quality should improve after the appointment of a new audit firm because a new audit requires a higher audit effort. To my knowledge there is no research connecting the change of audit firm to voluntary disclosure or audit fee disclosure, but Kothari et al. (1988) state that the disclosures of companies increase if the legal liability of the audit firm decreases. Venkataraman, Weber & Willenborg (2008), in turn, state that the legal liability decreases if the audit quality is improved. Therefore since the audit quality & earnings quality is increased after a change in audit firm, it can be proposed that the audit fee disclosure should increase in quality. Because a positive relation between audit fee disclosure and auditor change is proposed, and the relations of the other determinants to audit fee disclosure are also positive, a combined positive effect is expected. This gives rise to the following hypotheses:

H11: There is a positive relation in the first year after the rotation of an audit firm and the audit fee disclosure.

H12: There is a positive relation between the risk of an auditee and audit fee disclosure in the first year after the rotation of an audit firm when it is added as moderator.

(15)

H13: There is a positive relation between the governance effectiveness of an auditee and audit fee disclosure in the first year after the rotation of an audit firm when it is added as moderator. H14: There is a positive relation between the audit firm effectiveness and the audit fee disclosure in the

first year after the rotation of an audit firm when it is added as moderator.

(16)

3. Methodology

This section develops the methodology that is used in this paper. First the data that is used, will be described. The second subsection discusses how the dependent variables of both models are operationalized. The third subsection describes how the independent and control variables that are used in the models are measured. Lastly the regression equations that are employed to research the central question are developed.

Data gathering

This research is performed in the Dutch context, this means that the data gathered originates from Dutch listed companies. The data consists of large (AEX), mid (AMX) and small (AScX) cap companies, that are listed on the Euronext Amsterdam on March 1st_{, 2017. In total this yields 75 companies that are researched for five}

consecutive years in the period 2012 – 2016. This means that 375 observations are researched. All companies that comprise the indices will be researched, so no companies or industries are excluded. The composition of the indices changes every year because the composition is dependent on the trading revenue (share price multiplied by the volume of shares traded) of a company, also IPOs and delistings change the composition of an index. Therefore an unbalanced panel of data is obtained. There are 14 companies where not all data is available for. Therefore 37 observations cannot be included in the sample. The data is obtained by extracting it from the annual reports of the companies over the particular years. Below it is set out which data is extracted from the annual reports and how this is used to research the topics of interest of this paper.

The dependent variables

This paper comprises two subjects, namely audit fee and audit fee disclosure. Audit fee and audit fee disclosure are both a mandatory part of the annual report. Companies have however the liberty to disclose more information about the audit fee. The audit fee can be obtained from the audit fee disclosure and the audit fee disclosure is included in the notes of the annual report. The depended variables of the two models are the audit fee in a particular year for a particular company and the quality of the audit fee disclosure in a particular year for a particular company. The audit fee can easily be observed and is an objective measure. The quality of the audit fee disclosure however is a subjective measure and therefore an audit fee disclosure index is developed. The audit fee disclosure index answers 12 questions about the content and quality of the disclosure and if the content and quality is present, a point added, so the questions have equal weights. The disclosure index asks questions about the presence of the audit fee, other assurance fees, tax fees, non-assurance fees and comparable figures of last year. The disclosure index will also assess the quality of the disclosure by asking questions about the presentation of the figures, the breakdown of the figures and the qualitative explanation on the different fees. This will result in a value of the audit fee disclosure between 0% and 100%. The quality or internal consistency of the disclosure index is measured using the Cronbach’s Alpha. The disclosure index has a Cronbach’s Alpha of 0.7814. The

(17)

critical values of the Cronbach’s Alpha are stated in Table 1 and are based on the critical values stated by Gliem & Gliem (2003) and Smith (2015). The value of the Cronbach’s Alpha is acceptable and therefore reliable. The questions of the disclosure index are stated in the Appendix A.

Table 1: Critical values of Cronbach’s Alpha. Cronbach's Alpha Internal consistency

α ≥ 0.9 Excellent 0.9 > α ≥ 0.8 Good 0.8 > α ≥ 0.7 Acceptable 0.7 > α ≥ 0.6 Questionable 0.6 > α ≥ 0.5 Poor 0.5 > α Unacceptable

The independent and control variables

The two models that are used in this paper have both four determinants. The determinants are auditee’s risk, governance effectiveness, auditor effectiveness and audit firm rotation. These determinants can be divided into several variables that measure the determinants. Below for each determinant is described which proxies are used to measure the determinant. In addition the control variable will also be described. In Table 2, a summary of all variables, their names that are used in the regressions and their proposed effect on the dependent variables are listed.

Table 2: Overview of the variables used in the regressions.

Variable _Name _{Kind of}

variable

Expected sign for audit

fee

Expected sign for audit fee disclosure

Audit Fee AUF Interval

Audit Fee Disclosure AFD Interval

Leverage LEV Interval + +

Inherent Risk INR Dummy + +

Going Concern COC Dummy + +

Size AC SIZ Interval - +

Female Chair AC FEC Dummy - +

Big-4 Audit Firm BI4 Dummy + +

Market Share Audit Firm MAS Dummy + +

Audit Firm Rotation AFR Dummy - +

Size of Auditee SIA Interval + +

3.3.1. Auditee’s risk

The auditee’s risk is a broad concept and can comprise many variables. The variables that are researched the most and that have a significant relation with audit fee are the leverage of the auditee, the inherent risk, and the issuance of a going concern by the auditor (O'Keefe, Simunic & Stein, 1994; Basioudis, Papakonstantinou & Geiger, 2008;

(18)

Hay, Knechel & Wong, 2006). The empirical literature surrounding audit fee disclosure also finds significant relations for the inherent risk and the issuance of a going concern by the auditor (Lai, 2009; Raffournier, 1995; Cooke, 1992; Meek, Roberts & Gray, 1995). The evidence on the effect of leverage on disclosures is mixed, but the leverage is included in this paper because some papers find that leverage has a significant influence on the audit fee disclosure. Based on the empirical results found by other studies, this paper defines the auditee’s risk by the leverage of the auditee (LEV), the inherent risk (INR), and the issuance of a going concern by the auditor (COC). LEV is defined as the total debt divided by the total shareholder’s capital. INR is measured by a dummy variable for companies that are active in more risky or regulated business sectors. The sectors that will receive a value of 1 are the financial services sector and the utilities sector, because they are identified as more risky in the meta study of Hay, Knechel & Wong (2006). The remaining sectors receive a value of 0. COC is also measured using a dummy variable. This dummy receives a value of 1 if there is a going concern for a company in a particular year and a 0 otherwise.

3.3.2. Governance effectiveness

The governance effectiveness can be described by how the board is organized, as is done in many papers. DeZoort et al. (2002) state that if the governance effectiveness or audit committee effectiveness is studied, the only observable determinant is the composition of the board or audit committee. Therefore in this paper the composition of the audit committee, as part of the board, is researched because they are the body within the board that selects the audit firm. The variables identified by other studies that have a significant influence on the audit fee and audit fee disclosure of an auditee is the size of the audit committee (Li, Mangena & Pike, 2012; Kalelkar, 2017) and the gender of the chairmen, based on the notion that this will increase the effectiveness of the audit committee (Ittonen, Miettinen & Vähämaa, 2011). The variables that are used in this paper for governance effectiveness are the size of the audit committee (SIZ) and if the chairmen of the audit committee is a female (FEC). SIZ is measured by the amount of people that have a seat on the audit committee. FEC is measured using a dummy variable. If the chairman of the audit committee is a female, a value of 1 is assigned to this dummy variable and if the chairman is a male, a value of 0 is assigned to this dummy variable.

3.3.3. Audit firm effectiveness

Several studies find that the characteristics of audit firms are of influence on audit fee and audit fee disclosure. The characteristics of audit firms that influence the audit fee and audit fee disclosure are if the audit firm is a Big-4 audit firm (Firth, 1979) and if the audit firm is an industry specialist (Dunn & Mayhew, 200Big-4; Craswell, Francis & Taylor, 1995).The characteristics of audit firms that influence the audit fee and audit fee disclosure that are used in this paper are the size of an audit firm (BI4) and if the audit firm is an industry specialist (MAS). BI4 is measured using a dummy variable. This dummy variable receives a value of 1 if the audit firm is a Big-4 audit firm and receives a value of 0 if the audit firm is not a Big-4 audit firm. MAS is measured using the market share of an audit firm in a particular industry for a particular year. To calculate the market share in an industry, the

(19)

approach of Mayhew & Wilkins (2003) is used. They use the following formula to calculate the market share in an industry:

𝑀𝐴𝑆

𝑖,𝑘

=

∑

𝐽𝑖,𝑘

_{√𝑆𝐼𝐴𝑖,𝑗,𝑘}

𝑗=1

∑

𝐼𝑘 𝑖=1

∑

√𝑆𝐼𝐴𝑖,𝑗,𝑘

𝐽𝑖,𝑘 𝑗=1

Where: i = an index of audit firms; j = an index of client firms; k = an index of client industries; 𝐼𝑗 = number of audit firms in industry k;

𝐽𝑖,𝑘 = the number of clients served by audit firm i in industry k;

𝑆𝐼𝐴𝑖,𝑗,𝑘 = total client assets for audit by auditor i of client j in industry k; 𝑀𝐴𝑆𝑖,𝑘 = market share of auditor i in industry k.

To qualify as a specialist in a particular industry and a particular year, two rules must be satisfied. The first rule is that the audit firm has at least a market share of 40% (Mayhew & Wilkins, 2003). The second rule is that the audit firm with the largest market share, has a market share which is at least 15% higher than the audit firm with the second largest market share (Bae, Choi & Rho, 2016). When an audit firm satisfies both rules in a particular industry and in a particular year, it receives a value of 1 and all other audit firms receive a value of 0.

3.3.4. Audit firm rotation

The rotation of an audit firm may also contribute to the effectiveness of the audit firm, but because the audit firm rotation is central in this paper it is considered to be an independent determinant. Several studies identified that the change of an audit firm has a significant relation on the audit fee and the disclosure that the auditee makes (Simon & Francis, 1988; Myers, Myers & Omer, 2003). The rotation of an audit firm (AFR) is defined by a dummy variable. If the audit firm that audits the financial statements this year, is different from the audit firm that audited the financial statements previous year, a value of 1 is assigned to that year for the auditee. A value of 0 is assigned to the years that do not satisfy this rule.

3.3.5. Control variables

To control for the relations between the independent variables and the dependent variables, a control variable is added to the regression equations which also has an influence on the dependent variables. The regression controls for the size of the auditee (SIA). SIA is measured by the total assets of the auditee. The size of auditees is added because the empirical literature shows that this is an important variable (Simunic, 1980), but not of interest in this research. Lastly, the models are also controlled for variations throughout the years (e.g. mergers & acquisitions) and variations between auditees.

(20)

Model development

This research deals with panel data and therefore panel data methods should be employed. The three main methods for panel data are pooled OLS, fixed effects and random effects. The F-test of the least squares dummy variable estimator confirms that pooled OLS is not preferable and fixed effects or random effects should be employed. Because of collinearity issues between the fixed effects model and the INR and COC variables, these variables have to be omitted in a fixed effects or random effects model. Therefore pooled OLS regressions are also employed to include the variables specified earlier. This assumes that the data is homogeneous which will lower the explanatory power of the regression. To control for the heterogeneity in the pooled OLS models, White cross-sectional standard errors are employed.

To know if a fixed effects model or a random effects model should be used, the Durbin-Wu-Hausman test is conducted. The test concludes that the fixed effects model should be employed. The redundant fixed effects test shows that only cross-sectional fixed effects must be included and time fixed effects must not be included. There is still not a solution for multicollinearity in panel data, but the model can be estimated if the standard errors are ‘reasonable’ (Statalist, 2015). The fixed effects and pooled OLS models that research the effect of the determinants on the audit fee and audit fee disclosure will look as follows:

𝐴𝑈𝐹𝑖,𝑡 = 𝛽1,𝑖+ 𝛽2𝐿𝐸𝑉𝑖,𝑡+ 𝛽3𝐼𝑁𝑅𝑖,𝑡+ 𝛽4𝐶𝑂𝐶𝑖,𝑡+ 𝛽5𝑆𝐼𝑍𝑖,𝑡+ 𝛽6𝐹𝐸𝐶𝑖,𝑡+ 𝛽7𝐵𝐼4𝑖,𝑡+ 𝛽8𝑀𝐴𝑆𝑖,𝑡+ 𝛽9𝐴𝐹𝑅𝑖,𝑡 + 𝛽10 ln(𝑆𝐼𝐴𝑖,𝑡) 𝐴𝑈𝐹𝑖,𝑡= 𝛽1+ 𝛽2𝐿𝐸𝑉𝑖,𝑡+ 𝛽3𝐼𝑁𝑅𝑖,𝑡+ 𝛽4𝐶𝑂𝐶𝑖,𝑡+ 𝛽5𝑆𝐼𝑍𝑖,𝑡+ 𝛽6𝐹𝐸𝐶𝑖,𝑡+ 𝛽7𝐵𝐼4𝑖,𝑡+ 𝛽8𝑀𝐴𝑆𝑖,𝑡+ 𝛽9𝐴𝐹𝑅𝑖,𝑡 + 𝛽10 ln(𝑆𝐼𝐴𝑖,𝑡) 𝐴𝐹𝐷𝑖,𝑡 = 𝛽1,𝑖+ 𝛽2𝐿𝐸𝑉𝑖,𝑡+ 𝛽3𝐼𝑁𝑅𝑖,𝑡+ 𝛽4𝐶𝑂𝐶𝑖,𝑡+ 𝛽5𝑆𝐼𝑍𝑖,𝑡+ 𝛽6𝐹𝐸𝐶𝑖,𝑡+ 𝛽7𝐵𝐼4𝑖,𝑡+ 𝛽8𝑀𝐴𝑆𝑖,𝑡+ 𝛽9𝐴𝐹𝑅𝑖,𝑡 + 𝛽10ln(𝑆𝐼𝐴𝑖,𝑡) 𝐴𝐹𝐷𝑖,𝑡 = 𝛽1+ 𝛽2𝐿𝐸𝑉𝑖,𝑡+ 𝛽3𝐼𝑁𝑅𝑖,𝑡+ 𝛽4𝐶𝑂𝐶𝑖,𝑡+ 𝛽5𝑆𝐼𝑍𝑖,𝑡+ 𝛽6𝐹𝐸𝐶𝑖,𝑡+ 𝛽7𝐵𝐼4𝑖,𝑡+ 𝛽8𝑀𝐴𝑆𝑖,𝑡+ 𝛽9𝐴𝐹𝑅𝑖,𝑡 + 𝛽10ln(𝑆𝐼𝐴𝑖,𝑡)

(21)

4. Results

This section deals with the results from the regression specified in the previous section. First the descriptive statistics of the data will be discussed. Next the results on audit fee is discussed for each determinant and thereafter the results on audit fee disclosure is discussed. Lastly the robustness of the results will be discussed.

Descriptive statistics

This study uses the sample of the 75 companies with the highest trading revenue in the Netherlands on March 1st_,

2017. This constitutes the companies in the three Dutch stock market indices, namely the AEX, AMX and AScX. These companies are studied for 5 consecutive years, from 2012 till 2016. Due to IPO’s and delistings only 338 of the 375 observations are obtained. In Table 3 the descriptive statistics of the dependent, independent and control variables are listed. The variables going concern and Big-4 audit firm deserve special attention. The mean of the going concern variable is very low, which means that there are very few observations with a going concern. Due to the low amount of observations it is hard to obtain inferences from this variable. The variable Big-4 audit firm has a very high mean, which means that almost all observations are a Big-4 audit firm. This also means that it is hard to obtain inferences from this variable. The other variables seem to have reasonable distributions based on the means, maximums, minimums and the standard deviations.

Table 3: Descriptive statistics of the variables.

Observations Mean Median Maximum Minimum Std. Dev.

Dependent variables

Audit Fee 338 4.528 1.200 52.000 0 8.956

Audit Fee Disclosure 338 0,6857 0,7273 1,0000 0,0000 0,2281 Independent variables Leverage 338 4,3554 1,2905 172,6680 -5.593.951 15,2861 Inherent Risk 338 0,1923 0,0000 1,0000 0,0000 0,3947 Going Concern 338 0,0089 0,0000 1,0000 0,0000 0,0939 Size AC 338 3,3314 3,0000 7,0000 2,0000 1,1229 Female Chair AC 338 0,1036 0,0000 1,0000 0,0000 0,3051

Big-4 Audit Firm 338 0,9941 1,0000 1,0000 0,0000 0,0768

Market Share Audit

Firm 338 0,2041 0,0000 1,0000 0,0000 0,4037

Audit Firm Rotation 338 0,1361 0,0000 1,0000 0,0000 0,3434 Control variable

Size of Auditee 338 39.504.539 2.831.713 1,17E+09 6.281 1,38E+08

To detect multicollinearity in panel data the correlation between variables must be inspected. In Table 4 the Pearson correlation matrix is stated. Variables are collinear if the correlation between the variables is either very high or very low. In this study we define variables to be collinear if the correlation coefficient is higher than 0.8 or lower than -0.8, which is a rule of thumb (Brooks, 2014). Almost all correlation coefficients are low, which

(22)

Table 4: Pearson correlation matrix with the asterisks indicating p-values of 0.10, 0.5 and 0.01 respectively.

AUF AFD LEV INR COC SIZ FEC BI4 MAS AFR SIA

AUF 1,000 AFD 0,188*** 1,000 LEV 0,155*** -0,010 1,000 INR 0,262*** -0,168*** 0,273*** 1,000 COC -0,047 -0,019 -0,008 -0,046 1,000 SIZ 0,381*** -0,069 0,032 0,170*** -0,112** 1,000 FEC -0,100* -0,042 0,017 0,056 -0,032 0,029 1,000 BI4 0,037 0,090* 0,016 0,038 0,007 -0,046 0,026 1,000 MAS 0,404*** 0,090* 0,121** 0,293*** -0,048 0,138** -0,124** 0,039 1,000 AFR -0,043 0,060 0,077 0,025 -0,038 -0,010 0,063 0,031 -0,030 1,000 SIA 0,757*** 0,094* 0,324*** 0,384*** -0,027 0,324*** -0,083 0,022 0,298*** 0,012 1,000

(23)

indicates no collinearity issues. Only the correlation coefficient between SIA and AUF is high, with a value of 0.7571. When used in the regressions, this does not cause collinearity issues. This is however evidence that the audit fee is largely dependent on the size of the auditee.

The pattern of the dependent variables over the years is different for both dependent variables. In Figure 1 the average audit fee and the average audit fee disclosure over the years is plotted. The total audit fee and audit fee disclosure are decomposed in the three stock indices to detect any heterogeneity. Despite that these plots cannot be seen as evidence, they give a good representation of the development of the audit fee and audit fee disclosure over time. The average audit fee is fairly stable over time, but a small decrease in audit fee is visible in 2015 which is the year where many auditees changed from audit firm. The pattern between de indices seem to differ. The companies in the AEX seem to have a lower audit fee after rotation, companies in the AMX seem to have a higher audit fee after rotation and companies in the AScX have the same audit fee after rotation. Contrary to the audit fee, the quality of the audit fee disclosure seems to increase over time. Especially the auditees in the AMX and AScX increase the quality of their audit fee disclosure. The auditees in the AMX and AScX have on average a lower quality of the audit fee disclosure than auditees in the AEX, which is indicated by the lower scores. In my opinion the quality of the audit fee disclosure is high. There are however some items that are consistently missing in most audit fee disclosures. These items are the breakdown of the fee into the part for the Dutch audit firm and the foreign audit firm, which method is used for the disclosure, the fee for a subsidiary if this is a material part of the fee and lastly the qualitative explanation on the audit fee. In Appendix B, two snapshots are included from two companies to illustrate the audit fee disclosure. The snapshots are from the annual report of Wolters Kluwer from 2015 and the annual report of TomTom from 2016. In the audit fee disclosure of Wolters Kluwer it can be seen that they give a good explanation of the audit fee and incorporate the best practice into the disclosure to facilitate a good transfer of information from the company to the investor. In the audit fee disclosure

4.600 4.628 4.683 _4.315 _4.445 11.019 _10.875 11.281 10.457 10.315 1.859 1.905 1.953 1.810 2.309 602 509 516 520 555 2012 2013 2014 2015 2016 A u d it f ee (in € '0 0 0 ) Year

Total AEX AMX AScX

67% _66% 68% 69% 72% 74% 75% 77% 76% 76% 60% 60% 62% 62% 66% 66% 63% 65% 70% 74% 2012 2013 2014 2015 2016 A u d it f ee d isc lo su re Year

Total AEX AMX AScX

(24)

from TomTom it can be seen that they only incorporate the mandatory parts and the disclosure is poorly structured which complicates the transfer of information. TomTom can take an example of Wolters Kluwer to structure their information in the same manner and provide additional notes to the disclosure. The audit fee disclosure of Wolters Kluwer received a score of 100% while the audit fee disclosure of TomTom only received a score of 58%. The next sections attempt to find if the observations in figure 1 are statistically significant and what the determinants of the audit fee and audit fee disclosure are.

Results on audit fee

This section attempts to find the determinants of the audit fee. The regressions specified in section 3.4 are used. First the variables of each determinant are individually inspected for a statistically significant result and thereafter the regression with all the variables included is inspected for statistically significant results. The results of the relation between the determinants and the audit fee are listed in Table 5. In this table Model A till Model G test the hypotheses that are constructed in section 2. Model H and I test the relation of audit fee and all the determinants using two different statistical approaches.

The first hypothesis is examined using Model A. In this model all variables are statistically significant except for the leverage variable. The adjusted R-squared is high, which means that the variables explain a large part in the variability of the audit fee. The inherent risk variable has the opposite sign than expected but the going concern variable has the expected sign. Due to the contradicting results the first hypothesis is rejected. An explanation for the opposite sign of the inherent risk is given by Pearson & Trompeter (1994). They state that while the industry is more difficult to audit, the assets are larger than other industries and therefore require less audit effort. The lower amount of audit work translates into a lower audit fee. This explanation can also explain the result that is obtained in Model A.

In Model B the second hypothesis is examined. In this model all variables are statistically significant and the model has a high adjusted R-squared. The size of the audit committee variable has to opposite sign than what is expected. The other variable has the sign that is expected. Because the two variables that measure the governance effectiveness of the auditee contradict each other, the second hypothesis is rejected. The opposite sign of the size of the audit committee is explained by Carcello et al. (2002). They state that the size of the audit committee is a substitute for the size on an auditee and therefore the sign should be the same. This explanation can also be applied here because the size of the audit committee is reasonably correlated with the size of the auditee.

The third hypothesis is examined using Model C. All variables are highly statistical significant and the adjusted R-squared is high. The variable Big-4 auditor has the opposite sign as hypothesized. This variable has only two observations where the audit firm is not a Big-4 audit firm. This causes that the inference of this variable is limited, as mentioned under the descriptive statistics. Therefore the findings of this variable are discarded. The remaining variable has the expected sign. Since the audit firm effectiveness is measured using the Big-4 audit

(25)

firm variable and the market share of the audit firm variable and these variables have the opposite sign, the third hypothesis must also be rejected.

In model D the fourth hypothesis is examined. In this model all variables are statistically significant and the adjusted R-squared is high. The sign of the audit firm rotation variable is as expected and also economically significant. This means that the fourth hypothesis cannot be rejected and that the audit fee is lower after the rotation of an audit firm.

The fifth, sixth and seventh hypothesis tests the relations of Model A, B and C after the audit firm rotation is added as a moderator. The results in Model E, F and G are similar to the results in Model A till Model D, but the coefficients are closer to 0, and therefore less pronounced than in the original models. The differences are tested if they are statically significant from one another using a Wald-test. The Wald-test shows that the differences in coefficients are not statistically significant for Model E and G but the differences are statistically significant for Model F. Therefore there is no the relation between the risk of an auditee and audit fee after the audit firm rotation is added as moderator. The relation between the governance effectiveness and the audit fee is lower after audit firm rotation is added as moderator, because the coefficients are closer to zero and therefore less pronounced. Lastly there is no relation between the audit firm effectiveness and audit fee after the audit firm rotation is added as a moderator. This, in turn, means that the fifth and sixth hypothesis are not rejected, but the seventh hypothesis is rejected.

Model I shows the same results as the individual models. Model H however has only significant coefficients for the size of the audit committee variable and the size of the auditee variable. The adjusted R-squared of the fixed effects models is very high, which states that this model explains the variation in the audit fee really good. The signs for the different variables change as well. This in turn means that the hypothesized relations are not present when controlled for the individual heterogeneity in the sample. The most important finding on audit fee is that the size of an auditee is the most important determinant for the audit fee and that the audit fee is lower after the rotation of an audit firm.

Results on audit fee disclosure

This section attempts to find the determinants of the quality of the audit fee disclosure. The regressions specified in section 3.4 are used. In Table 6 the results of the regressions are presented. First the determinants are examined individually on their statistical significance and thereafter two models with all the variables included, are tested. Model J till P test the hypotheses developed for the audit fee disclosure.

In Model J hypothesis eight is examined. In this model only the inherent risk variable and the size variable are statistically significant, where the inherent risk has a negative impact on the audit fee disclosure and the size has

(26)

Table 5: Results of the models related to Audit Fee with the asterisks indicating p-values of 0.10, 0.5 and 0.01 respectively.

Model A Model B Model C Model D Model E Model F Model G Model H Model I

Related to Hypothesis: H1 H2 H3 H4 H5 H6 H7

Variable Expected Sign Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient Intercept -40478,89*** -37713,99*** -32279,00*** -38155,30*** -40290,25*** -37649,11*** -32404,72*** -14660,38*** -35214,82*** LEV + -2,82 -0,14 -12,85 2,36 INR + -834,13*** -832,45*** -1510,49*** COC + 9687,97*** 9480,44*** 9260,26*** SIZ - 649,49** 632,78** 372,18*** 652,26*** FEC - -1455,03*** -1335,98*** 26,25 -501,69 BI4 + -3007,5*** -2820,66*** 12,49 -2232,6*** MAS + 4020,63*** 3962,96*** 168,93 4159,84*** AFR - -1668,391** -1574,00** -1542,04** -1462,79** -363,36 -1283,426** SIA + 3004,04*** 2679,97*** 2596,49*** 2858,60*** 3005,07*** 2692,52*** 2606,54*** 1199,41*** 2622,39*** N 338 338 338 338 338 338 338 338 338 Adjusted R² 0,487 0,485 0,507 0,483 0,489 0,487 0,509 0,961 0,519 F-statistic 80,91*** 106,64*** 116,54*** 158,51*** 65,47*** 80,86*** 88,24*** 104,73*** 41,46*** DW Statistic 0,101 0,102 0,168 0,112 0,113 0,117 0,177 1,706 0,189

Type of model Pooled OLS Pooled OLS Pooled OLS Pooled OLS Pooled OLS Pooled OLS Pooled OLS

Cross-sectional fixed effects

The effect of audit firm rotation on audit fee & audit fee disclosure in the Dutch context