The effect of firm ownership type on audit quality in relation to auditor tenure and the provision of non-audit services : evidence from the Netherlands

(1)

The effect of firm ownership type on audit quality in relation to

auditor tenure and the provision of non-audit services: Evidence

from the Netherlands.

Student name: Jan Goedhart Student number: 6074995

Date: 22 July 2014

Program: Msc Accountancy and Control, track Accountancy University: Universiteit van Amsterdam

Faculty: Economics and Business Supervisor: dr. P. (Patrick) Klijnsmit RA

(2)

1. Introduction

The last two decades are turbulent in the accountancy world. Auditor independency stands negatively in the spotlights and governments and regulators are trying to improve the independence of the auditor to safeguard the quality of the audit. This phenomenon started with the large accounting scandals in the years after 2000. Enron, U.S. (2001), WorldCom, U.S. (2002), Xerox, U.S. (2002), Ahold, Netherlands (2003) and Parmalat, Italy (2003) are a few examples of accounting scandals which were the basis of an intense discussion about the quality of the audit (Low et al., 2008).

Audit quality has not only been a major topic on political level. It is also extensively studied by researchers. All kinds of relationships within the audit quality, like auditor tenure, audit fees and auditor type, were tested in the last decades (Simunic, 1984; Frankel et al., 2002; Geiger & Raghunandan, 2002; Johnson et al., 2002; Myers et al., 2003; Ferguson et al., 2004; Larcker & Richardson 2004; Ghosh & Moon, 2005; Antle et al., 2006; Basioudis et al., 2008; Lim & Tan, 2010; Krishnan et al., 2011; Parkash et al., 2012; Svanström, 2012, Hope et al., 2013). Most research to these relationships is done for public firms. Private firms are still not sufficient studied in these relationships to audit quality.

The purpose of this research is to investigate the differences between private and public firms in the relation to audit quality for the Netherlands. The audit quality will be examined by two different mechanisms, respectively, auditor tenure, the length of the auditor-client relationship, and the provision of non-audit services, the amount of non-audit fees. Both mechanisms could have a positive or a negative effect on audit quality.

1.1 Background

Despite of the shortage in the literature there are researchers who looked on the possible effects of firm ownership type on audit quality. For some countries, like the Belgium (Sercu et al., 2002; Bauwhede & Willekens, 2004; Knechel & Vanstraelen, 2007), Norway (Hope & Langli, 2009), Sweden (Svanström, 2012), United Kingdom (Ball & Shivakumar, 2005), United States (Hope et al., 2013), and Europe overall (Burgstahler et al., 2006; Van Tendeloo & Vanstraelen 2005) this typical relation is examined, but for most other countries, like the Netherlands, research in this field is new and unexamined.

Prior literature suggests a difference in audit quality between both types of ownership. 3

(4)

This assumption is studied under different conditions, like common/code law and high/low litigation risk. Most of this research indicates a higher audit quality for public firms than for private firms. Ball & Shivakumar (2005) examined the earnings quality of U.K. private firms and discovered that the financial reporting quality of privately held firms is of a lower quality compared to the quality of public firms. This conclusion is remarkable, because the auditing regulation in the U.K. for private firms and public firms is equivalent, thus both firms have to follow the same requirements (Ball & Shivakumar, 2005).

More prior studies indicate a difference in audit quality between both types of firm ownership. A recent research, Hope et al. (2013), indicates the same pattern as Ball & Shivakumar (2005). A large sample of U.S. private and public firms shows a higher audit quality for public firms. Underlying differences in behavior between both types of firms could have influence on this difference in audit quality. In comparison, Burgstahler et al. (2006) conclude the opposite. They show that earnings management, which results in a lower audit quality, is more pervasive for privately held firms than for public firms in Europe based on a relative large sample (287,354 firm-year observations). Variations between countries have a strong influence on the differences in the results (Burgstahler et al., 2006). Thus, it is not straightforward how the type of firm ownership affects the audit quality.

This research narrows the view to audit quality into two major mechanisms; auditor tenure and the provision of non-audit services. Both mechanisms could have a positive as well as a negative effect on audit quality. Many studies investigated the effects of auditor tenure and the provision of non-audit services. Like the research to audit quality, research to auditor tenure and the provision of non-audit services is also extensively examined for public firms,

but studies to private firms are rare.

Two Scandinavian studies (Hope & Langli, 2009; Svanström, 2012) examined the effects of the provision non-audit services on audit quality for private firms as one of the first. These pioneers looked on the effects which one typical type of ownership, private firms, has on the audit quality. Hope & Langli (2009) and Svanström (2012) show the same results. They provide evidence from the perspective of private firms that the provision of non-audit services does not affect the audit quality negatively. Svanström (2012) goes one step further and concludes that the provision of non-audit services associates positively with the quality of the audit. In comparison with prior research for public firms are these results differently. A meta-analysis about the effect of non-audit services on financial reporting quality of public firms proves a negative relation (Habib, 2012).

As mentioned, auditor tenure is like the provision of non-audit services also one of the 4

(5)

variables which is intensively tested in relation to audit quality for public firms. A wide range of research underpins an effect of that auditor tenure on audit quality. Even as the effect of the provision of non-audit services is the effect of auditor tenure also dividable into two forms. Some researchers, like Geiger and Raghunandan (2002) and Myers et al. (2003), indicate a positive effect and other studies, like Al-Thuneibat et al. (2011), point out a negative effect of auditor tenure. As the research to the provision of non-audit services is the research to auditor tenure also mostly done for public firms and is the research to privately held firms rare. Just one scientific study investigated the relation between auditor tenure and the audit quality for private firms (Knechel & Vanstraelen, 2007). This Belgium research concludes that auditor tenure associates negatively with audit quality for private firms. Overall, there is a lack of literature to audit quality in relation with auditor tenure and the provision of non-audit services for private firms.

1.2 Motivation

The brief overview of the prior research in section 1.1 gave an insight in the effects of ownership type on audit quality. However, most research to this relationship is done with the use of public firms. This is, despite of the difference between both kinds of firms, a bit curious, because private firms represent a large part of the worldwide economy. Berzins et al. (2008) suggest that private firms have four times more employees than public firms and the sales of these firms are also four times higher in comparison with public firms. Even the amount of assets is twice as high as for public firms. These numbers apply on average for every country in the world (Berzins et al., 2008). Thus, private firms are important for most economies in the world. However, the research to the behavior of private firms is in its infancy. This could be the result of data unavailability through no mandatory pressure to publish annual reports for private firms. In most countries only public firms are forced to publish their annual reports by law. For private firms these requirements do not exist, but there are some exceptions like in Norway (Berzins et al., 2008). This lack of publishing requirements leads to a data scarcity of private firms, which results in less available information for researchers.

This shortcoming in the literature and especially the shortcoming in the relationships between auditor tenure and audit quality and between the provision of non-audit services and audit quality is interesting from an academic point of view. This study will provide new

(6)

insights in the literature regarding auditor tenure and the provision of non-audit services, especially from perspective of the private firms. In addition, this research will be the first research to the effects of audit tenure and the provision of non-audit services1 on the audit quality of private firms for the Netherlands, and also the first research which compares these relationships for private and public firms in the Netherlands.

Even from a societal point of view is this research useful and interesting. The Dutch government implemented a new regulation regarding the independence of the auditor in 2013. This new regulation means that auditors were not longer allowed to provide audit services and non-audit services to the same clients if those clients were public interest entities (PIEs), as first new provision (Artikel 24b, Wta). PIEs are listed firms, banks and insurers. The second new provision of this regulation is a mandatory audit firm rotation, which means that an audit firm may not conduct an audit for a PIE for more than eight years in succession. The

segregation of audit and non-audit services has taken effect on 1 January 2013. The mandatory firm rotation will take effect on 1 January 2016 (Artikel 23c, Wta).

Both provisions of the new regulation have as objective to safeguard the independence of the auditor, but as mentioned above this will only apply for PIEs. Thus the Dutch

government makes a separation between public firms (PIEs) and private firms in the regulation regarding auditor independence. This separation is remarkable. It could indicate that standard setters and regulators have some unknown presumptions about the auditor independence for the two firm ownership types. To evaluate this assumption by using this research, it is possible to give feedback on this separation in regulation and it is possible to give recommendation for other standard setters and regulators regarding auditor independence for private firms relative to public firms. But that is not the aim of this research. The aim is to provide better insight in the relationships between auditor tenure and audit quality for both private and public firms and to investigate the relationship between the provision of non-audit services and audit quality for public firms. So I will examine how the type of firm ownership affects the audit quality.

The paper is organized as follows. First I discuss the audit quality, the main

differences between private and public firms, the auditing environment in the Netherlands and the relationships between auditor tenure and audit quality and between the provision of non-audit services and non-audit quality. In the third section I explain the sample, the research design and the variables measures. Section four contains the empirical results. Finally, the paper ends

1

Due data unavailability is the provision of non-audit services not testable for private firms.

6

(7)

up in section five with the conclusions and the limitations plus some recommendations for future research.

2. Literature review and hypotheses

2.1 Audit quality

The main problem with auditor tenure and also with the provision of non-audit services is the possible negative effect on the audit quality. DeAngelo (1981) defines audit quality as the market-assed joint probability of two mechanisms; (1) auditor will discover a breach in the client’s accounting system, and (2) the auditor will report that breach. The chance that the auditor will discover a breach in the client’s accounting system depends on his or her capabilities. The chance that the auditor will report that breach depends on the degree of independency of the auditor (DeAngelo, 1981). The general assumption is that the probability or the capability that the auditor will discover a breach is constant across different auditors. But the audit quality, which is the end product of the auditor, can differ. One of the reasons of this difference is the discrepancy in the independence of the auditor. Auditor tenure and the provision of non-audit services affect this independence. The negative assumption is that the longer the auditor-client relationship is the less independent the auditor will be (Svanström, 2012). The same applies for the provision of non-audit services. The larger the amount of non-audit fees the less independent the auditor will be (Frankel et al., 2002). But some researchers show an opposite effect of auditor tenure and the provision of non-audit services (Myers et al., 2003; Antle et al., 2006). Sections 2.4 and 2.5 give an in-depth explanation about both mechanisms and their effects.

2.2 The auditing environment in the Netherlands

This study is focused on the Netherlands. There is a lack of literature regarding the audit quality for this country. Especially for the relationships within the audit quality and the effect of firm ownership type. Notwithstanding, the government of the Netherlands decided to implement stricter rules regarding auditor tenure and the provision of non-audit services.

In the Netherlands applies a mandatory audit for every firm which meets at least two 7

(8)

of the next criteria in two consecutive years; (1) total assets greater than €4.4 million, (2) sales greater than €8.8 million, and (3) number of employees greater than 50 (Artikel 393 BW2). It does not matter if the firm is privately held or publicly held. In general, these requirements imply that every medium and large sized firm needs an audit in the Netherlands. These requirements are the same for every country in Europe, but some countries use stricter rules and lower values for these requirements2 (Van Almelo, 2011). In this way the Netherlands handle a less strict regulation, but still comply with the European laws. Despite of the differences in institutional factors between the European countries, European accounting regulation takes a straight-line between private and public firms. Both types of firm ownership face the same accounting standards in Europe (Burgstahler et al., 2006).

In 2011 the AFM, the Dutch Authority for the Financial Markets, conducted a study on the independence of the auditor (AFM, 2011). In the research report of this study the AFM announced his concerns regarding the danger of auditors’ contradictory interests for the auditor independence and the AFM appealed for a clear, unambiguous and more restrictive regulation regarding the auditor independence. The AFM believes that auditors in the Dutch audit market are exposed to contradictory interests and that these interests could lead to less independence. On the one hand auditors have to be objective and critical towards the clients in order to be able to fulfill the public duty of giving assurance regarding the financial statements, but on the other hand auditors have to satisfy their clients in order to retain the client. These contradictory interests arise because audit firms are commercial organizations and will be paid by their own clients, according to the AFM (2011). The Dutch government shared these concerns and adopted in 2013 a new Accountancy Profession Act (in Dutch: Wet op het accountantsberoep, (Wab)) (AFM, 2013). In addition to this new law, the Dutch

government also introduced two new provisions which had to improve the safeguards of the auditor independence. These provisions are developed as restrictions on auditor tenure and the provision of non-audit services for PIEs. Since 1 January 2013 it is not allowed for an auditor to provide audit services and non-audit services to the same client, if this client is a public interest entity (PIE) (Artikel 24b, Wta). And from 1 January 2016 the regulation regarding auditor tenure will also be stricter. An auditor(audit firm)-client relationship for PIEs cannot be longer than eight consecutive years from then. This implies a mandatory audit firm rotation (Artikel 23c, Wta).

2

Belgium (assets: 3.6; sales: 7.3; employees: 50), France (1.5; 3.0; 10), Sweden (0.2; 0.4; 50)

8

(9)

2.3 Differences between private and public firms

Although this research field is in its infancy, some assumptions can be made regarding the differences in audit quality of private and public firms. First of all, private and public firms differ in ownership. Public firms are listed firms and got range of other intentions and mechanisms compared to private firms. These differences have effects on the audit quality between both types of firms. Incentives to report financial statements are one of the

differences. This is because the ability to raise capital is unequal for both (Ball &

Shivakumar, 2005). Public firms have the ability to offer shares and receive money back for their shares. For private firms this kind of fundraising does not exist. Also other manners to raise capital differ between private and public firms. Private firms are mostly dependent on banks for debt financing, because it is hard to attract large investors as a private firm

(Svanström, 2012). Public firms are in that way better able to raise capital in public markets (Burgstahler et al., 2006). This distinction results in other needed audit quality between both types of firms. Namely, external providers of capital, capital providers for public firms, want more certainty about the financial condition of the firms. This implies that they are more likely to rely on published financial statements if they invest in public firms. Thus public firms have a greater incentive to report high quality financial statements for this reason (Hope et al., 2013).

In addition, it is less important for private firms to report financial statements publicly. Ownership structures are relatively concentrated and that makes it is easier to communicate via private manners (Burgstahler et al., 2006). Ball & Shivakumar (2005) suggest that private firms communicate with their debt financiers (banks) on a more private basis. This can result in a lower demanded audit quality of private firms (Ball & Shivakumar, 2005). Plus the disclosure standards for publicly held firms, as a result of listing rules, are more rigorous than for private firms which will imply that the financial statements of public firms are more complete and thus of a higher quality in this context (Coffee, 2005).

Another significant difference between private and public firms is the pressure to meet earnings benchmarks and targets. In public firms this pressure is more relevant. There are several reasons for this mechanism. For example, investors in stock markets rely more on earnings targets than banks in general (Beatty et al., 2002). This implies a stronger incentive for public firms to meet the earnings benchmarks if they want to raise external capital. Also the equity-based incentive schemes, which are a more common tool in public firms, result in extra motivations for managers for short-term financial manipulation to meet earnings

(10)

benchmarks and targets (Coffee, 2005; Hope et al., 2013). These two incentives make public firms more likely to manage earnings. Earnings management results in lower financial reporting quality, which implies that private firms could have higher audit quality based on the assumptions regarding the pressure to meet earnings benchmarks and targets (Beatty et al., 2002; Coffee, 2005; Hope et al., 2013).

From the perspective of the auditor are there also differences between private and public firms. Prior literature (St. Pierre & Anderson, 1984) shows that the public firms are more risky for audit firms in comparison with private firms. Most lawsuits against auditors are from public firms. Plus the costs of lawsuits are higher for public firms compared to private firms. This suggests that auditors will work more carefully with public firms, which could result in a higher audit quality. In addition, the likelihood that an audit failure will be discovered is lower for private firms than for public firms. Public firms will be analyzed by analysts and stock-market regulators. So the opportunity to find out an audit failure is greater for public firms. The costs which are involved in that kind of findings will also be higher for public firms compared to private firms (Sercu et al., 2002).

Based on the prior literature regarding the differences in private and public firms which could affect the audit quality I formulize the first hypothesis. The majority of the arguments pointing to a higher audit quality for public firms based on the multiple

characteristics of both ownership types. Following the motivations of Beatty et al. (2002), Ball & Shivakumar (2005), Coffee (2005) and Hope et al. (2013) I form the first hypothesis; H1: Audit quality for public firms is higher than for private firms

2.4 Auditor tenure and audit quality

Auditor tenure is the time that an auditor provides audit services to a same client. Thus, auditor tenure is about the auditor-client relationship. This relationship is possible in two forms. Auditor tenure on the level of the partner and auditor tenure on the level of the audit firm (Lim & Tan, 2010). Both forms are sufficient studied in prior literature for public firms. The focus in this research is on the audit firm tenure, mainly because the new audit rotation regulation in the Netherlands that will take effect on 1 January 2016 is focused on the audit firm level (Artikel 23c, Wta).

The relation between audit tenure and the quality of an audit can be explained in two 10

(11)

effects; a positive and a negative effect. To begin with the positive side of auditor tenure. This side argues that the audit quality will improve over time, because the auditor will get more knowledge about the operational and financial organization of his client. This knowledge will help to conduct an audit of a higher quality. The theory behind this positive effect of a longer auditor-client relationship is characterized by learning effects (Svanström, 2012). These learning effects are also one of the main arguments in the discussion regarding auditor rotation. James E. Copeland, former CEO of Deloitte and Touche, formulated that the learning effect which have to restart by every auditor rotation again will reduce the audit quality. Plus this repetitive learning curve increases the start-up costs for auditors (Daniels & Booker 2011). These arguments speak for a negative effect of auditor rotation, which is interpretable as a positive point of audit tenure.

Besides these positive points of audit tenure are there also other explanations regarding the effect of auditor tenure on audit quality. The negative effects of audit tenure indicate that the objectivity of the auditor can be affected by a relative long auditor-client relationship. Thus the independence of the auditor can decrease over time and this can result in auditors’ support for a more aggressive accounting policy that is looking for the limits of the financial accounting standards which implies a lower audit quality, because it could result in a failure to discover and/or report an breach in the financial statements (Myers et al., 2003). This effect is the biggest concern regarding a long auditor-client relationship.

In contrast, prior literature shows more support for positive side of auditor tenure than for the negative effects of it (Lim & Tan, 2010). Several researchers conclude that audit tenure affects the audit quality positive, thus the longer the auditor-client relationship is the higher the quality of the audit will be (Geiger & Raghunandan, 2002; Myers et al., 2003; Ghosh & Moon, 2005). Beside of this evidence, there is a smaller voice which concludes that auditor tenure has a negative effect on audit quality (Al-Thuneibat et al., 2011). More

specific, Johnson et al. (2002) found that auditor-client relationships of 2 to 8 years are associated with lower audit quality and that auditor-client relationships of more than 9 years will result in a higher audit quality.

Overall, auditor tenure is explainable from a positive as a negative point of view, but it is likely that audit tenure influences the quality of the audit. In formulizing the second

hypothesis I follow the arguments of the negative side of auditor tenure. The new regulation regarding audit firm rotation also hints on this possible negative side. This results in the following hypothesis;

(12)

H2: Auditor tenure is negatively associated with audit quality.

Hypothesis 2 is formulized to investigate a possible negative effect of auditor tenure on the quality of the audit in general. The next step is to specify the effect of auditor tenure for both types of firms to investigate a possible difference between private and public firms in the relationship between auditor tenure and audit quality. In doing this, I formulize an additional hypothesis;

H3: Auditor tenure has a larger negative effect on audit quality for public firms than for private firms.

2.5 The provision of non-audit services and audit quality

Comparable to auditor tenure has the relationship between the provision of non-audit services and audit quality also a positive and a negative effect. The positive form is characterized by ‘knowledge spillovers’. These are positive knowledge effects of doing different types of work for the same client. Thus by doing non-audit services for a client, you will have more

knowledge about the financial condition of that client. This extra knowledge is helpful by doing an audit for that client and can result in an audit of a higher quality. This positive effect of non-audit services has been demonstrated in many prior researches (Frankel et al., 2002,

Antle et al., 2006; Krishnan et al., 2011; Parkash et al., 2012).

The negative side of the relationship between the provision of non-audit services and audit quality is most of the time explained by the term ‘economic bonding’. The economic bonding can decrease the independence of the auditor. Auditors or audit firms could have high valued clients in their portfolio. This could be a client to which many services are provided, like audit and non-audit services. In the extreme context it is even possible that the auditor or audit firm becomes dependent on a client, because a client could represent a high value for the auditor or audit firm. Losing this client will have significant impacts on the revenues (Frankel et al., 2002). This could have influence on the objectivity of the auditor and could result in a reduced independency in advantage of the client. In that case the auditor is probably more willing to give an unqualified audit opinion. This negative side of the

provision of non-audit services is also extensively examined (Frankel et al., 2002; Ferguson et al., 2004; Larcker & Richardson 2004; Basioudis et al., 2008).

(13)

The two effects of the provision of non-audit services are comparable with the effects of auditor tenure. In the new regulation the Dutch government assumes that the positive side of the provision of non-audit services is overwhelmed by the negative side. This matches with the assumption about the effect of auditor tenure. To test the assumption of the Dutch

government I formulize the fourth hypothesis in the same way as the second hypothesis3; H4: The provision of non-audit services is negatively associated with audit quality

3. Sample, research design and variables measures

3.1 Sample

This study covers the fiscal year 2012. The main reason for this is that the first part of the new provisions for PIEs has taken effect on 1 January 2013. So the fiscal year 2012 is the last year without the new regulation. The auditor tenure period of 8 years goes back to the fiscal year 2005 for this reason. Non-audit services will also be available for the fiscal year 2012, because from 2005 it is mandatory for European firms to disclose their audit and non-audit fees (Directive 2006/43/EC, 2006). This research looks on the period before the new regulation, thus data of the year 2013 and later are excluded.

In the dataset the distinction which the new regulation in the Netherlands made between public (PIEs) and private firms is followed (AFM, 2013). So banks and insurers are included in the dataset ‘public firms’ instead of excluding them as is done in prior studies. Most private firms in the Netherlands are small. It was likely that information about financials and/or auditor tenure and non-audit services was unavailable for these small private firms. Burgstahler et al. (2006) is followed in excluding these firms which implies that the criteria for a mandatory audit in The Netherlands are used. Thus the private firms had to meet at least two of the next criteria in two consecutive years; (1) total assets greater than €4.4 million, (2) sales greater than €8.8 million, and (3) number of employees greater than 50 (Artikel 393 BW2.). This requirement results in a bigger chance of available data, because audits are mandatory for these firms. In addition to this requirement, I excluded firms which switched from private ownership to public ownership and vice versa.

3

The fourth hypothesis relates only to public firms due data unavailability.

13

(14)

The final sample consists of 86 public firms and 125 private firms for the year 20124. All these firms are active within the Netherlands. The financial information has been obtained from the Amadeus database of Bureau van Dijk. This database consists European private and public firm information. But not all financial information was available in the Amadeus database. Some financial information was missing which forced me to look for other ways to capture this information. For public firms this information was available in the annual reports, which can be found on the internet5. In comparison with public firms are private firms not required to publish their annual reports. This was one of the major difficulties in this study, but I resolved this problem by buying online6 several annual reports of private firms via the Kamer van Koophandel (KvK), which is a Dutch interest group who supports entrepreneurs in the Netherlands. This organization owns annual reports of most private firms in the

Netherlands.

Besides the missing financial information in the database a much larger data problem pops up. Information about auditor tenure and the provision of non-audit services was not available in Bureau van Dijk. The database didn’t include any data about the length of the auditor-client relationship or the amount of non-audit fees. The solution for public firms in this case was the same as for the financial information problem. The publicly available annual reports included this specific data for public firms. Gathering this data for private firms was harder, and in the case of the provision of non-audit services even impossible. There was no possibility to compute the auditor tenure, because I didn’t have access to all annual reports of all years (2005-2012) of the private firms and Bureau van Dijk was also not sufficient. This forced me to direct mailing and/or filling in contact forms on the websites of the private firms. In total 396 private firms has been contacted. 155 firms responded on the request to share private information about their auditor tenure. 125 of these responses were useful for the research. This leads to a response rate of 39,1% and a useful response rate of 31,6% (table 1).

4_{Three outliers are detected and removed from the original sample}

5_{Official websites of the firms, but also on several other websites which offer annual reports, like}

http://xbrl.rienks.biz/jaarverslagen and http://www.beursgorilla.nl/jaarverslagen

6

http://www.kvk.nl/producten-bestellen/

14

(15)

# Firms % Firms

396 100 %

Responses 155 39,1 %

Useful responses 125 31,6 % Non-useful responses 30 7,5 %

Table 1: Distribution responses from private firms

These rates a quite high and could be the result of a very simple request with only one question; ‘Is the current auditor (accounting firm) still the same auditor (accounting firm) as the auditor (accounting firm) in 2005?’. The appendix contains the email with the request to the private firms. Answers like ‘YES’ or ‘NO’ were already sufficient. In the case of the provision of non-audit services this method did not help. Most private firms do not distinguish total audit fees in audit fees and audit fees. This makes it impossible to capture the non-audit part of the total non-audit fees. This discovery had major consequences for the research, because this data unavailability leads to changes in the research method and design of the paper.

In completing the data sample an outlier analysis is performed. To exclude outliers from the sample the outliers are determined by the 68-98-99,7 rule. In this case the outliers are determined as the observations of the absolute value of discretionary accruals which lie outside three standard deviations of the mean. Performing this analysis results in three outliers. The three outliers are public firms. These outliers are removed from the original sample (214), which leads to the final sample size of 211 firms.

Despite of the data issues, the approach of database information combined with the purchased annual reports and the direct mailing method results in a unique data sample. This sample enables me to do a unique research and it gives the opportunity to examine the difference in firm ownership types for the two relationships, especially the auditor tenure relationship, with audit quality based on the new regulation in a context that not has been done before for the Netherlands.

3.2 Research design

Due the data unavailability of non-audit fees for private firms it is not possible to perfectly compare the relationships between the provision of non-audit services and audit quality for

(16)

private and public firms. Fortunately, the auditor tenure data is complete, so the new auditor tenure provision can be tested extensively for both types of firm ownership. The lack of information about the private firms leads to some major consequences for this paper. Two different methods are used to test the possible effects of auditor tenure and/or the provision of non-audit services on the two types of firm ownership. The first method, the auditor tenure test, follows the Baron and Kenny (1986) model in testing for a mediating effect of auditor tenure for both firm ownership types. The second model, the non-audit services tests, are two regression models which investigate the effect of the provision of non-audit services on the audit quality of public firms.

3.2.1 Auditor tenure test

All needed data for the auditor tenure test part is available, so the best fitting test to determine the difference in the effect of auditor tenure on audit quality between public and private firms is feasible. As discussed above, the audit quality between public and private firms differs (Ball & Shivakumar, 2005; Burgstahler et al., 2006; Hope et al., 2013). The assumptions that the audit quality differs between both types of firm ownership and that auditor tenure could play a role in this relationship lead to a model with a mediator. In this case auditor tenure is the mediating variable. To do so, I follow the model of Baron and Kenny (1986). The central idea in the mediating variable model of Baron and Kenny (1986) is that ‘the effects of stimuli on behavior are mediated by various transformation processes internal to the organism’ (p. 1176). In other words, the mediation model identifies a mechanism, the mediating variable or intervening variable, which is able to influence the relationship between an independent variable and a dependent variable (Baron and Kenny, 1986; Kim et al., 2001).

Figure 1: Mediating effect

(17)

AUDITOR TENURE is the possible mediating variable in the test for a difference between private and public firms in the effect of auditor tenure. The independent variable is SIZE, which means private or public firms, and AUDIT QUALITY is the dependent variable.

Following Baron and Kenny (1986) we can distinct four steps in the process to determine a possible mediating effect of auditor tenure;

Mediating variable step 1

Regression of SIZE (independent variable), equals 1 for public firms and 0 otherwise, on AUDIT QUALITY (dependent variable), the absolute value of the discretionary accruals, to look if there is a (significant) relation in the first place between the independent variable en the dependent variable.

Independent Variable Dependent Variable

There is only a chance of a mediating variable when there is a significant relation between SIZE (independent variable) and AUDIT QUALITY (dependent variable).

This mediating step has an additional advantage, because the regression of SIZE on AUDIT QUALITY will also confirm or reject hypothesis 1; audit quality for public firms is higher than for private firms

Regression of SIZE (independent variable), equals 1 for public firms and 0 otherwise, on AUDITOR TENURE (mediating variable), the length of the auditor-client relationship and equals 1 when the length of the auditor–client relationship is 8 years or more and 0 otherwise.

Independent Variable Mediator

There is only a chance of a mediating variable when there is a significant relation between SIZE (independent variable) and AUDITOR TENURE (mediating variable).

(18)

Regression of AUDITOR TENURE (mediating variable), the length of the auditor-client relationship and equals 1 when the length of the auditor–client relationship is 8 years or more and 0 otherwise, on AUDIT QUALITY (dependent variable), the absolute value of the discretionary accruals.

Mediator Dependent Variable

There is only a chance of a mediating variable when there is a significant relation between AUDITOR TENURE (mediating variable) and AUDIT QUALITY (dependent variable).

Regression of SIZE (independent variable), 1 for public firms and 0 otherwise, on AUDIT QUALITY (dependent variable), the absolute value of the discretionary accruals, with the mediating variable AUDITOR TENURE (mediating variable), the length of the auditor-client relationship and equals 1 when the length of the auditor–client relationship is 8 years or more and 0 otherwise.

Independent Variable Mediator Dependent Variable

This step in determining a mediating effect of auditor tenure between the private and public firms as independent variable and the absolute value of discretionary accruals as dependent variable is the final step in the Baron and Kenny (1986) model and gives the answer on the question if a mediating variable exists. This step also leads to a clear view about a possible difference between private and public firms in the effect of auditor tenure on audit quality. Control variables will be used to search for any effect of these control variables on the possible mediating relation.

3.2.2 Non-audit services tests

It is not possible to do the same mediating test for the provision of non-audit services as discussed. For this reason I investigate the association between audit quality and the provision

(19)

of non-audit services by the following regression models with the same control variables as in the test for auditor tenure;

Audit Quality (DA) = b0 + b1 x NASRATIOi;t + b2 x TENUREi;t + b3 x BIG4i;t + b4 x TOT.ASSETi;t + b5 x LEVERAGEi;t + b6 x ROAi;t + ɛi;t.

Audit Quality (DA) = b0 + b1 x LNNASi;t + b2 x TENUREi;t + b3 x BIG4i;t + b4 x TOT.ASSETi;t + b5 x LEVERAGEi;t + b6 x ROAi;t + ɛi;t.

As explained, the provision of non-audit services has to two possible effects on audit quality. It could be positive, thus improving the quality of the audit, but the effects could also be negative and reducing the audit quality. Following Ferguson et al. (2004) and Svanström (2012) I take both effects into account by measuring the provision of non-audit services; NASRATIO (positive effect) and LNNAS (negative effect). These two measures give a better picture of the total effect of the provision of non-audit services on audit quality.

3.3 Variables measures

3.3.1 Dependent variable measure – Audit quality proxy

The research to audit quality is dividable into multiple manners. A main part of the researchers made use of ‘accruals’ as the proxy for audit quality. Other proxies for audit quality based on prior literature are ‘restatements’ and ‘going concern opinions’. Some other less used proxies are ‘dispersion of analysts’ forecasts’, ‘audit failure’ and ‘public criticism or ‘regulatory investigation of firm accounting practices’. The measurement of audit quality is a difficult decision, because it is not definitely clear which measurement option is the best one. Although, a significant part of the research to audit quality used discretionary accruals as the measure of audit quality. Furthermore, this manner is used for both audit tenure (Johnson et al., 2002; Lim & Tan, 2010; Al-Thuneibat et al., 2011) and non-audit services (Frankel et al., 2002; Ashbaugh et al., 2003; Chung & Kallapur 2003; Larcker & Richardson 2004; Antle et al., 2006; Krishnan et al., 2011) and also for private and public firms (Ball & Shivakumar, 2005; Burgstahler et al., 2006; Svanström, 2012; Hope et al., 2013). Based on these prior

(20)

studies and the possibility to test the different relationships with one single audit quality measure I decided to use discretionary accruals as the measure of audit quality.

Besides the explanation of the common use of accruals, another explanation indicates the advantage of the use of accruals over other manners. If firms want to manage earnings they will try to do it in a manner that is hard noticeable. Accruals are in that way an option. Total accruals (TA) are derivable in two forms; non-discretionary accruals (NDA) and discretionary accruals (DA).

TAit = DAit + NDAit

Non-discretionary accruals are expected accruals. Discretionary accruals are accruals which can be controlled by the management, thus this type of accruals is depended on the

subjectivity of the management. Firms or managements will use accruals to change reported earnings towards outcomes which are more profitable for the firm (Jones, 1991). This

opportunistic use of discretionary accruals is a manner to achieve earnings benchmarks and/or targets with the use of accounting measures. Earnings management by the use of accruals is in that way an indicator of low-quality financial reporting. Auditors should constrain this type earnings management. This makes discretionary accruals a good measure of audit quality (Svanström, 2012).

The model which is used to calculate the discretionary accruals is the Modified Jones Model (Dechow et al., 1995). Modified Jones Models are often used in several studies (Becker et al., 1998; Frankel et al., 2002; Gul et al., 2003; Krishnan, 2003; Ferguson et al., 2004; Svanström, 2012) and are based on the original Jones Model for discretionary accruals (Jones, 1991). The discretionary accruals in the Modified Jones Model of Dechow et al. (1995) are calculated as the difference between the total accruals (TA) and the non-discretionary accruals (NDA). This calculation consists of two major steps.

The Modified Jones Model has to be estimated as first. This means that the all variables have to be calculated with the use of the information which is collected from the database and the annual reports. The model is as followed;

it it it it it it it it jt e assets total ppe assets total s receivable revenues assets total assets total accruals total + + ∆ − ∆ + = − − − − 1 2 1 1 1 1 . . . 1 . . β β α 20

(21)

it

accruals

total. = Total accruals for sample firm i for the year 20127

1

.assets_it₋

total = Total assets for sample firm i for the year 2011

it

revenues

∆ = Change in net revenues for sample firm i in for the year 2012

it

s receivable

∆ = Change in net receivables for sample firm i for the year 2012

it

ppe = Gross property, plant and equipment for sample firm i for the year 2012

it

e = Error term for sample firm i in for the year 2012 i = 1, …, N for firm index (N = 211)

The next step in the process to calculate the discretionary accruals is to determine the

parameters of the Modified Jones Model by performing the regression. These parameters are used to calculate the non-discretionary accruals (NDA) for every firm in the sample.

Deducting this amount of the total accruals (TA) results in the discretionary accruals (DA) for every firm. The absolute value of the discretionary accruals is the audit quality variable.

3.3.2 Independent variables measures – Public/Private proxy and non-audit service proxy

The data unavailability forced me to create different tests for auditor tenure and the provision of non-audit services. This also leads to different independent variables. In the test for auditor tenure the independent variable is formed by SIZE. SIZE equals 1 for public firms and 0 otherwise. This proxy indicates the type of firm ownership.

In the other test, the test of non-audit services, two independent variables are used. The first proxy for non-audit services is the proportion of non-audit fees of the total audit fees paid to the auditor (ratio of non-audit fees to total audit fees), in this research called

NASRATIO. This is the most often used measure for non-audit services (Frankel et al., 2002; Ashbaugh et al., 2003; Larcker & Richardson 2004; Svanström, 2012). In addition to

NASRATIO a second measure for non-audit services is also determined, because

NASRATIO will not capture the financial importance of a client to the auditor (Frankel et al., 2002). LNNAS, the natural logarithm of non-audit fees, will do this. Both measures are consistent with prior literature and capture both effects of the provision of non-audit services. The positive side, the knowledge spillovers, is captured by NASRATIO, because NASRATIO 7_{Total accruals = Net Income - Cash Flow from Operating Activities (CFO). CFO = EBIT + Depreciation –}

Taxes +/- Change in Working Capital (Frankel et al., 2002)

21

(22)

captures the relative level of knowledge that the auditor gains from the provision of non-audit services (Ferguson et al., 2004; Svanström, 2012). The negative side, the economic bonding, is captured by the second measurement of non-audit services, LNNAS (Ferguson et al., 2004; Svanström, 2012).

3.3.3 Mediating variable measure - Auditor tenure proxy

In the mediating variable model of Baron and Kenny (1986) the possible mediating effect of auditor tenure is tested. The measure of auditor tenure is the length of the auditor-client relationship and equals 1 when the length of the auditor–client relationship is 8 years or more and 0 otherwise. The new provision of the government will directly be measured by

AUDITOR TENURE, because the mandatory auditor rotation provision has been determined on 8 years (Artikel 23c, Wta).

3.3.4 Control variables measures

Apart from auditor tenure and the provision of non-audit services there might be other variables which could have an influence on the audit quality. Auditor type could be one of them, so auditor type will be included as control variable. Prior literature suggest that big four auditors (KPMG, E&Y, PWC and Deloitte) could have a positive influence on the audit quality, which means that firms with a auditor of one of these audit firms will possibly have a higher audit quality (DeFond et al., 2002; Choi et al., 2010). The variable BIG 4 equals 1 for a big four auditor and 0 otherwise. Several other common control variables will include to see if those variables will affect the audit quality, like TOTAL ASSETS, which represents the size of a firm. Larger firms are expected to have more sophisticated or more accurate financial-reporting systems. Better financial-financial-reporting systems will result in higher audit quality (Johnson et al., 2002; Ball & Shivakumar, 2005). Also LEVERAGE (ratio of total liabilities to total assets) will be include as a control variable, because a higher degree of leverage could be a incentive to manage earnings in a profitable way (Vander Bauwhede et al., 2000; Ball & Shivakumar, 2005; Al-Thuneibat et al., 2011). In addition to LEVERAGE, I control for ROA (return on assets) to capture the possible effect of financial risk on audit quality (Hope & Langli, 2009).

(23)

4. Results

4.1 Descriptive statistics and correlations

Table 2 summarizes the descriptive statistics of the private firms subsample. 72 private firms have an auditor tenure of 8 years or more. These 72 private firms represent 58% of the private firms in the subsample. Thus 58% of the total private firms has a longer relationship with the auditor than the Dutch government will allow for PIEs from 1 January 2016. So if the new provision regarding auditor tenure would also take effect for private firms, it would result in much more auditor changes in private firms than nowadays. This shift could lead to a change in the quality of the audits, because auditor rotation could have a positive and/or a negative effect on the audit quality, as described in section 2.4.

N Range Minimum Maximum Sum Mean Std.

Deviation AUDIT QUALITY 125 168093862 2048 168095909 1001563999 8012512 20655294 AUDITOR TENURE 125 1 0 1 72 .58 .496 BIG 4 125 1 0 1 86 .69 .465 TOTAL ASSETS 125 1232022727 4777273 1236800000 9591259063 76730073 151791154 LEVERAGE 125 1.1417208224 .1048754218 1.2465962441 77.2532761917 .6180262095 .2145498522 ROA 125 119.360 -50.900 68.460 698.750 5.59000 11.815631

Table 2: Descriptive statistics private firms

Table 3 shows that the percentage for public firms with an auditor tenure of 8 years of longer is even higher than for private firms; 71%. So the new regulation will have a major effect on these firms, because only 29% of the public firms, measured in 2012, complies with the new auditor tenure provision. This will lead to a great shift in auditors in 2016 when the new audit firm rotation provision will take effect for PIEs.

The provision of non-audit services, which is only observable for public firms, differs a lot for the firms within the public subsample. The highest observed non-audit fee is 7,6 million euro’s and the lowest just 4000 euro’s, apart from firms without any non-audit

services. This large difference is also coming back in the measures of non-audit services. The ratio of non-audit services is as highest 73% against 0,004% as lowest and the natural

logarithm of non-audit fees is as highest 15,8437 against 8,29405 as lowest, apart from firms without any non-audit services (table 3).

(24)

N Range Minimum Maximum Sum Mean Std. Deviation AUDIT QUALITY 86 29732658 229967 2975565488 21299546664 247669147 475359620 NASRATIO 86 .7301 .0000 .7301 18.14152 .210948 .1873832 LNNAS 86 15.8437 .0000 15.8437 855.4256 9.946809 4.7278288 AUDITOR TENURE 86 1 0 1 61 .71 .457 BIG 4 86 1 0 1 74 .86 .349 TOTAL ASSETS 86 46156918000 9082000 46166000000 265995051318 3092965713 7220358574 LEVERAGE 86 1.3566303327 .0976560601 1.4542863928 49.17641 .5718187385 .1932687268 ROA 86 84.620 -43.790 40.830 293.2 3.40930 11.030730

Table 3: Descriptive statistics public firms

The control variables demonstrate several relative similarities between the private and public subsamples. Both firm ownership types have a relative high percentage of ‘BIG 4’ auditors, which indicates KPMG, E&Y, PWC or Deloitte, respectively 69% for private firms and 87% public firms. This also shows that public firms are more likely to hire a ‘BIG 4’ auditor than private firms. As expected, the public firms are in general larger than private firms measured in total assets. This difference is quite large, 77 million euro’s for private firms and 3 billion euro’s for public firms. This is remarkable, because all private firms are large or very large firms following the Bureau van Dijk database8, except of three firms of medium size. The percentages of leverage are comparable with the both types of firm ownership. Remarkably, the return on assets (ROA) differs a lot between both subsamples. The ROA for private firms is almost 2,2 percentage points higher than for public firms. Tables 4 and 5 provide the Pearson correlations for the private and public subsamples. ‘BIG 4’ auditors correlate

positively with the absolute value of discretionary accruals, the measurement of audit quality, for both public and private firms, but for both types of firm ownership is this positive

correlation insignificant. Compared to prior literature is this positive correlation remarkable, because most research indicates an opposite effect (DeFond et al., 2002; Choi et al., 2010). An opposite effect compared to prior studies is also observable for the size of the firm, measured in total assets. The size of the firm correlates positively with the absolute value of discretionary accruals. This could be the result of the link between the size of the firm and the number and amount of the transactions. Larger and more transactions will also lead to more accruals in total. The return on assets (ROA) correlates positive too.

8

‘Company category’ variable in Bureau van Dijk.

24

(25)

Correlations Private Firms

**. Correlation is significant at the 0.01 level (2-tailed). *. Correlation is significant at the 0.05 level (2-tailed).

Table 4: Pearson correlations private firms

Correlations Public Firms

AUDIT QUALITY

NAS-RATIO LNNAS AUDITOR TENURE BIG 4 TOTAL ASSETS LEVERAGE ROA AUDIT QUALITY 1 NASRATIO .097 1 LNNAS .298** .647** 1 AUDITOR TENURE .192 -.180 .114 1 BIG 4 .153 .160 .280** .192 1 TOTAL ASSETS .526** .081 .343** .199 .182 1 LEVERAGE .205 .005 .034 .000 .105 .143 1 ROA .010 -.067 .043 .159 -.073 -.004 -.261* 1

**. Correlation is significant at the 0.01 level (2-tailed). *. Correlation is significant at the 0.05 level (2-tailed).

Table 5: Pearson correlations public firms

4.2 Main tests results

4.2.1 Simple t-tests results

Before I perform the mediating test for auditor tenure, I will take a closer look on the auditor tenure and audit quality relation. As showed in tables 2 and 3 the difference between public firms and private firms in the absolute value of discretionary accruals is large. Table 6 also shows that the means in absolute value of discretionary accruals differs between the two values of auditor tenure. Especially for the public firms is this difference large. Public firms

AUDIT QUALITY AUDITOR TENURE BIG 4 TOTAL ASSETS LEVERAGE ROA AUDIT QUALITY 1 AUDITOR TENURE -.037 1 BIG 4 .047 .051 1 TOTAL ASSETS .582** -.187* .135 1 LEVERAGE -.202* -.003 -.038 -.145 1 ROA .259** .056 .045 .006 -.212* 1 25

(26)

with auditor-client relationships of 8 years or more have 3 times more discretionary accruals on average than public firms with a shorter auditor-client relationship. Furthermore, table 8 shows that the mean of absolute value of discretionary accruals of auditor tenure of more than 8 years is significantly (.006) larger than the mean of absolute value of discretionary accruals of auditor tenure less than 8 years for public firms. For private firms this significant (.681) difference is not observable (table 7). Taking both types of firm ownership together a

significant (.003) difference between the two values of auditor tenure is observable (table 9). In conclusion, when the auditor-client relationship is 8 years or more than is the audit quality, on average, lower than when the auditor-client relationship is less than 8 years. This confirms hypothesis 2; auditor tenure is negatively associated with audit quality.

PRIVATE FIRMS AUDITOR TENURE Mean N 0 8901630 53 1 7358022 72 Total 8012512 125 PUBLIC FIRMS AUDITOR TENURE Mean N 0 91206839 25 1 311793044 61 Total 247669147 89

Table 6: Means absolute value discretionary accruals

ABSOLUTE VALUE DA AUDITOR TENURE 1 ABSOLUTE VALUE DA AUDITOR TENURE 0 Mean 7358022 8901630

Variance 4,317E+14 4,26539E+14

N 72 53

df 113

T -0,411903682

Sig. 0,681190317

Critical region of t-test: 2-tailed 1,981180296

Table 7: Independent t-test for two samples (AUDITOR TENURE: 1 & 0) for private firms

(27)

Variance 2,92247E+17 3,37297E+16

N 61 25

df 82

T 2,81507373097981

Sig. 0,00610483637007824

Critical region of t-test: 2-tailed 1, 989318521

Table 8: Independent t-test for two samples (AUDITOR TENURE: 1 & 0) for public firms

Variance 156258E+17 122957E+16

N 133 78

df 165

T 3,06010454488682

Sig. 0,00258317134101137

Critical region of t-test: 2-tailed 1,97444559270437

Table 9: Independent t-test for two samples (AUDITOR TENURE: 1 & 0) for both firm types together

4.2.2 Auditor tenure test results

As described in section 3.2.1 I followed the model of Baron and Kenny (1986) with a mediating variable to test the effect of auditor tenure on audit quality of private and public firms.

Mediating variable step 1:

Model Coefficients t Sig.

1

(Constant) .295 .768

SIZE .363 5.635 .000

Table 10: Mediating step 1 (SIZE)

The regression coefficient (0.363) of SIZE on AUDIT QUALITY is significant (.000). So it meets the requirement for a possible mediating variable of auditor tenure on the relation SIZE – AUDIT QUALITY. This mediating step also gives an answer on the first hypothesis. Table 9 implies that public firms have a significant higher absolute value of discretionary accruals

(28)

than private firms, which is interpretable as a lower audit quality for public firms. This rejects hypothesis 1; audit quality for public firms is higher than for private firms.

1

(Constant) 13.402 .000

SIZE .136 1.980 .049

The regression coefficient (0.136) of SIZE on AUDITOR TENURE is significant (.049). So it meets the requirement for a possible mediating variable of auditor tenure on the relation SIZE – AUDIT QUALITY.

1

AUDITOR TENURE .166 2.438 .016

The regression coefficient (0.166) of AUDITOR TENURE on AUDIT QUALITY is significant (.016). So it meets the requirement for a possible mediating variable of auditor tenure on the relation SIZE – AUDIT QUALITY.

Model Unstandardized Coefficients Coefficients t Sig.

B Std. Error (Constant) SIZE AUDITOR TENURE -38110332.602 36813243.390 -1.035 .302 228982533.785 42682805.148 .347 5.365 .000 80074382.982 43449428.072 .119 1.843 .067

(29)

The effect of AUDITOR TENURE is not longer significant (0.067) adding SIZE in the regression. This means that the variable AUDITOR TENURE, equals 1 when the length of the auditor–client relationship is 8 years or more and 0 otherwise, is not a mediating variable for the relationship between SIZE, equals 1 for public firms and 0 otherwise, and AUDIT QUALITY, absolute value of discretionary accruals.

Adding the control variables to this last mediating step does not change the conclusion that we cannot observed a mediating effect of auditor tenure on the relationship between SIZE and AUDIT QUALITY.

(Constant) -.988 .324 SIZE .163 2.958 .003 AUDITOR TENURE .062 1.194 .234 BIG 4 .045 .866 .387 TOTAL ASSETS .592 10.964 .000 LEVERAGE .029 .555 .579 ROA -.038 -.712 .477

Table 14: Mediating step 4 with control variables (SIZE)

The control variable TOTAL ASSETS is very significant in table 12. In the Pearson correlations tables (table 4 and 5) a same pattern is observable with correlations of .582 (private firms) and .526 (public firms) both significant at the 0.01 level (2-tailed). Taking a closer look on the total assets, we observed a huge difference in the means, respectively 77 million euro’s for private firms and 3 billion euro’s for public firms (table 2 and 3). This could lead to the strong explainable power of TOTAL ASSETS in the possible mediating relationship. To test this assumption I ran a similar mediating variable model with TOTAL ASSETS as independent variable. The low values of TOTAL ASSETS represent the private firms and the high values represent the public firms in this way.

1

(Constant) 2.732 .007

TOTAL ASSETS .659 12.679 .000

Table 15: Mediating step 1 (TOTAL ASSETS)

(30)

1

(Constant) 17.872 .000

TOTAL ASSETS .136 1.978 .049

1

AUDITOR TENURE .166 2.438 .016

The use of TOTAL ASSETS as mediating variable results also in three significant relations in the first three steps of the Baron and Kenny (1986) model, which implies that a mediating variable could be possible in this case. In addition to that, hypothesis 1 is also rejected by this method in the first mediating variable step. To see if the mediating variable is present in this method I run the last mediating step.

Model Unstandardized Coefficients Coefficients t Sig.

B Std. Error (Constant) TOTAL ASSETS AUDITOR TENURE 15495134.969 27700045.841 .559 .576 .044 .004 .649 12.396 .000 52622322.038 35156212.199 .078 1.497 .136

Also in this mediating model with TOTAL ASSETS as independent variable the effect of AUDITOR TENURE is not longer significant (0.136) after controlling for TOTAL ASSETS. This means that the variable AUDITOR TENURE, equals 1 when the length of the auditor– client relationship is 8 years or more and 0 otherwise, is not a mediating variable for the relationship between TOTAL ASSETS, the total assets of the firms, and AUDIT QUALITY, absolute value of discretionary accruals. So the results of both methods indicate that auditor tenure has not got any significance influence on the difference in audit quality between

(31)

private and public firms.

The two methods, respectively SIZE and TOTAL ASSETS, did not find a mediating variable of auditor tenure in relation to audit quality, but both results of the mediating step 4 (tables 13 and 18) indicates that the audit quality for public firms is lower than for private firms. The unstandardized coefficients9 show in both outputs a positive coefficient for SIZE and TOTAL ASSETS. In case of SIZE is this value 228982534, which indicates that a public firm will have, on average, 228982534 more discretionary accruals than a private firm, if all other variables remain equal. In the case of TOTAL ASSETS the value is .044, which indicates that if the total assets will grow with 1 the absolute value of discretionary accruals will grow with 0.044, if the other variables remain equal. I linked low values of total assets to private firms and high values of total assets to public firms. So both methods result in a larger negative effect on audit quality for public firms compared to private firms in this case. This confirms hypothesis 3; auditor tenure has a larger negative effect on audit quality for public firms than for private firms.

4.2.3 Non-audit services tests results

Table 19 shows a positive association between AUDIT QUALITY and NASRATIO. Audit quality is measured as the absolute value of discretionary accruals. This implies that when the NASRATIO becomes higher the absolute value of discretionary accruals becomes larger. This indicates that the audit quality is negatively related to an increased proportion of non-audit fees to total audit fees, because high discretionary accruals are related with high levels of earnings management. The significance of this relation is .199, which indicates an

insignificance. Thus we cannot conclude with certainty that NASRATIO is the booster of the discretionary accruals.

9

By standardizing a dummy, the interpretation gets lost. Therefore I did not use the standardized dummy regressor coefficients.

31

(32)

Model Coefficients t Sig. (Constant) -.948 .346 NASRATIO .116 1.296 .199 AUDITOR TENURE .140 1.532 .130 BIG 4 .051 .568 .571 TOTAL ASSETS .570 6.473 .000 LEVERAGE .060 .674 .502 ROA -.078 -.868 .388

Table 19: Linear regression on AUDIT QUALITY (NASRATIO)

As in the auditor tenure test the TOTAL ASSETS variable is very significant (.000) again. This could have a big influence on the relation. This is the result of the multicollinearity between the total assets and the absolute value of discretionary accruals. To exclude this effect I decided to run the same regression without TOTAL ASSETS (table 20).

(Constant) NASRATIO AUDITOR TENURE BIG 4 LEVERAGE ROA -1.351 .180 .166 1.516 .134 .236 2.125 .037 .117 1.060 .292 .107 .980 .330 -.081 -.729 .468

Table 20: Linear regression on AUDIT QUALITY (NASRATIO) excluding TOTAL ASSETS as control variable

The significance of the NASRATIO coefficient changes in this way. NASRATIO becomes less insignificant, but is still not significant. The auditor tenure variable in this regression without total assets is significant, which implies that an auditor(audit firm)-client relationship of 8 years or more for public firms will result in higher absolute value of discretionary

accruals and thus a lower audit quality.

The second non-audit services test makes use of LNNAS as measure of the provision of non-audit services. As observable in the Pearson correlations LNNAS is significant at the 0.01 level (2-tailed). The regression with the control variables shows a different result in table 20. The significance disappears and results in an insignificance of .082 with a positive

association with the discretionary accruals (table 21). If we excluded the TOTAL ASSETS variable again to take the multicollinearity in account, we observe a significant (.004) positive association between the natural logarithm of non-audit fees and the absolute value of

(33)

discretionary accruals (table 22). Thus LNNAS is associated with lower audit quality for public firms. The two measures of the provision of non-audit services show a same pattern. Both indicate a positive association for public firms. LNNAS is significant, but NASRATIO is not. This is in line with the effects of LNNAS and NASRATIO, because LNNAS captures the ‘economic binding’, which is the negative effect, and NASRATIO captures the

‘knowledge spillovers’, which is the positive effect. Overall, the provision of non-audit services has a negative effect on the audit quality for public firms. The control variables BIG 4, LEVERAGE and ROA have no significance association with absolute value of the

discretionary accruals in all tests to non-audit services. This is in contrast with prior literature. The levels of insignificance of these control variables are large. The lowest insignificance is .292 (table 20), the other values are all higher. The sample size could have influence on this result. Compared with other researches, this study uses a relative small sample, which makes it harder to detect small differences in the variables. Apart from that, hypothesis 4 is

confirmed by these results; the provision of non-audit services is negatively associated with audit quality.

(Constant) -1.210 .230 LNNAS .161 1.760 .082 AUDITOR TENURE .114 1.286 .202 BIG 4 .036 .394 .695 TOTAL ASSETS .536 5.925 .000 LEVERAGE .059 .662 .510 ROA -.091 -1.010 .315

Table 21: Linear regression on AUDIT QUALITY (LNNAS)

(Constant) LNNAS AUDITOR TENURE BIG 4 LEVERAGE ROA -1.931 .057 .311 2.956 .004 .188 1.795 .076 .066 .614 .541 .102 .969 .336 -.102 -.951 .344

Table 22: Linear regression on AUDIT QUALITY (LNNAS) excluding TOTAL ASSETS as control variable

The effect of firm ownership type on audit quality in relation to auditor tenure and the provision of non-audit services : evidence from the Netherlands