Director network ties and the quality of financial oversight

(1)

Director network ties and the quality of

financial oversight

Douwe Boersma- s4118510

Economics Master's Thesis

Specialization Corporate Finance & Control

Supervisor: Dr. M.G. Contreras

Radboud University Nijmegen

2016-2017

(2)

Abstract

The European Union introduced legislation outlining mandatory audit rotation to improve auditor independence and by extent improve the quality of financial oversight. Opponents of this legislation claim that the benefits of potentially increasing auditor independence do not outweigh the costs of the frequent rotation and costly initial audits, and that legislator efforts should be utilized differently. This thesis studies the effects of past and present ties between directors and directors of auditing firms on the quality of financial oversight. The sample consists of firms listed at the FTSE Euronext 100 Index, and studies the years 2008-2012. We hypothesize a negative effect of the amount of ties on financial oversight and expect a stronger relation from ties based on friendship networks than ties based on advice networks. No significant effects are found for ties in our accruals model and contrary to our first hypothesis, a significant positive effect is found for employment ties. When testing for robustness we find a significant negative effect of education of audit fees, while other results remain the same.

(4)

2

1. Introduction

On June 16th 2014, a new European Union provision entered into force. This provision introduced mandatory audit firm rotation to the 28 member states of the European and to Norway, Iceland, and Liechtenstein, which are bound to this legislation because of their membership of the European Economic Area. This mandatory audit firm rotation entails that companies are not allowed to engage with the same statutory auditor or audit firm for an engagement period less than 1 year or longer than 10 years (KPMG, 2016). This new legislation is part of a long series of new rules and regulations which are aimed at increasing auditor independence. Other examples of such new rules and regulations are the prohibition of providing non-audit services by auditors to their clients, the introduction of cooling-off periods for auditors obtaining jobs at their former clients, mandatory audit partner rotation every 5 years, and a recent, more extreme suggestion from the Public Company Accounting Oversight Board(PCOAB) to require the disclosure of the name of the engagement partner, which would publicly link the reputation of the individual to the audits he/she oversees. Most of these rules and regulations are meant to diminish the opportunities for managers to behave opportunistically. They are used as safeguards to prevent managers from influencing auditors, so that the auditor acts in the manager’s interest.

One of the most important concepts underlying the relationship between management and external auditors is that auditors have to be independent. Auditor independence can be defined as the “conditional probability that, given a breach has been discovered, the auditor will report the breach.” (DeAngelo, 1981, p.116), where a breach is a misstatement or an error in the accounting system. Auditors, and the auditing profession in general, have become an essential part in reducing agency costs and providing investors and other stakeholders with information on how a company and its management is performing (Watts & Zimmerman, 1986). It is important to the market that auditors maintain independence because auditor independence influences audit quality and because investors rely on external auditors to monitor and verify information provided by management (Francis, 2004; Francis, 2011; Tepalagul & Lin 2015; Zhang, Zhou & Zhou, 2007; Austin & Herath, 2014). The framework by Tepalagul & Lin (2015) mentions four different ways in which auditor independence may be threatened, being: client importance, non-audit services (NAS), non-auditor tenure and client affiliation with non-audit firms. A lot of research has been done with regards to the first three threats. As these threats will not be the focal point of this

(5)

3

thesis, but are still important factors with regards to auditor independence, they will be briefly discussed by using the literature overview provided by Tepalagul & Lin (2015).

When a client becomes an important part of the total portfolio of an auditing firm, companies may become economically dependent on that client. Studies found that auditing firms are more conservative towards larger clients (Reynolds & Francis, 2000; Hunt & Lulseged, 2007), however the general notion in literature is that client importance does not impact reporting decisions (Tepalagul & Lin, 2015). Tepalagul & Lin (2015) have also discovered that most studies “find no evidence that NAS impair actual audit quality” (p.108). The main issue with auditing firms performing NAS for clients is that users of financial information perceive NAS to be a threat to auditor objectivity. This appears to be a faulty assertion because in some cases empirical evidence is found that NAS actually improves audit quality in some cases (Kinney, Palmrose & Scholtz, 2004; Seethamaran, Sung & Wang, 2011).

The relationship between auditor tenure and financial reporting quality can be viewed from two perspectives. The first being the expectation that auditors will become more lenient towards client management as the engagement becomes lengthier, thereby decreasing audit quality. The other perspective is that longer auditor tenure increases audit quality because the auditor gets a better understanding of the client’s business (Tepalagul & Lin, 2015). Most studies give empirical evidence for the second view, showing that longer audit tenure is associated with higher audit quality (e.g. Myers, Myers & Omer, 2003; Chen, Lin & Lin, 2008).

If we look at the first three threats, we can see that there is evidence that in some cases the audit quality actually improves, in spite of the perception of users of financial statements. This raises the question how effective new regulations are in improving auditing quality. The final threat to auditor independence, client affiliation with audit firms, has not been researched as extensively as the other three and will be the focal point of this thesis. In spite of the limited availability of research, some evidence has been found that affiliated firms (firms become affiliated when an auditor becomes an employee of an auditing client, or if a firm hires it’s executives former CPA firm) have a significantly higher chance of receiving clean audit opinions than non-affiliated firms and that these types of affiliations negatively influence earnings and audit quality (Lennox, 2005; Menon & Williams, 2004; Baber, Krishnang & Zhang, 2014).

(6)

4

This thesis examines the effects of client affiliation with audit firms (as a threat to auditor independence) on audit quality. A similar study was performed by Bruynseels & Cardinaels (2013). They looked at the effect of social ties, by which they mean both ties formed “through

either employment (e.g., shared current directorships or past employment or directorships at other firms), past education (e.g., graduating from the same school), or other non-professional activities” (Bruynseels & Cardinaels, 2013, p.114) between CEOs and members of the audit

committee on the quality of financial reporting. The relationship between CEO and audit committee is similar to the relationship between CEO and external auditor, because both the external auditor and the audit committee are tasked with monitoring (financial) information provided by management. They found that, while it is mandated by the Sarbanes Oxley Act (SOX), approximately 39 percent of audit committee members in their sample are not independent “because of social ties between CEOs and audit committee members” (Bruynseels & Cardinaels, 2013, p.141). They also present evidence that oversight quality is negatively influenced if social ties exist that have been formed through the friendship network of the CEO and they suggest this is because CEOs “build an audit committee that is sympathetic to their

reporting choices” (Bruynseels & Cardinaels, 2013, p.142)

After the introduction of SOX, which states that an independent audit committee is “directly

responsible for the appointment, compensation and oversight of the work of any registered public accounting firm” (Section 301), managers should no longer be able to influence the

decisions mentioned above. This regulation seemingly intends to decrease the threats of auditor independence. Dhaliwal et al. (2015) researched whether management still influenced auditor firm selection. They find evidence that suggests that managers still can influence audit firm selection (in a subset of firms that are audited by one of the Big 4). If management can still influence the selection; they can also attempt to get an external auditor that is sympathetic to their reporting choices which would decrease the quality of financial reporting.

This thesis studies the extent to which social ties, as used by Bruynseels & Cardinaels (2013), between CEOs and external auditors influence financial reporting quality. To achieve this, the following research question will be answered:

“How does the number of social ties between directors and directors of external auditors

(7)

5

This research question is scientifically relevant because it provides insights into the influence which ties between directors and external auditors have on the quality of financial reporting. This thesis adds to scientific literature about auditor independence, and to literature that studies the influence of social ties on the quality of financial reporting (Bruynseels & Cardinaels, 2013). Specifically, it links social ties between CEOs and external auditors to the quality of financial reporting. As was mentioned before, relatively little attention has been given to client affiliation with audit firms and its effect on audit quality, this is surprising since auditor independence is one of the main reasons why new rules and regulations are being implemented. This thesis provides empirical evidence for the theoretical assertion that auditor independence decreases reporting quality. Most of the existing literature (e.g. Lennox, 2005; Menon & Williams, 2004; Baber, Krishnang & Zhang, 2014) focuses on the U.S., this thesis contributes to existing literature by looking at data from the FTSE Euro 100. By looking at a different sample of firms, evidence is found to support or disprove generalizability of earlier results.

Besides its scientific relevance, this thesis also has practical relevance. Firstly it can provide information to managers on how social ties between CEOs and external auditors influence the quality of financial reporting. They can use this information to improve the quality of financial information they provide. Secondly this thesis has practical relevance for investors, when investors know how social ties influence financial reporting and they observe existing ties, they can include this information in their investment decision. Lastly, this thesis has practical relevance for policy makers. Current policy is aimed at increasing auditor independence to increase objectivity and the quality of financial reporting. The results of this thesis offer new insights to policymakers which can help them to adjust or maintain the current policy.

In order to provide an answer to the research question, the rest of this research is structured as follows. Chapter 2 will provide a review of existing literature which will be used to formulate hypotheses at the end of the chapter. Chapter 3 will describe the research methodology, the variables that are studied in this research, and the research model that is used to test the hypotheses. Chapter 4 will present the results of the research. Finally, chapter 5 will provide a discussion, a conclusion, a number of limitations of this research and some suggestions for possible future research.

(8)

6

2. Theoretical Background

2.1 Auditor Independence

Capital markets play a fundamental role in the functioning of our economy. Proper allocation of funds is a “fundamental job of the economy” (Wurgler, 2000, p 188). This entails that capital needs to be taken from sectors with low profitability and allocated to sectors that are expected to have higher profitability in the future. In order for the reallocation of funds to be possible, investors need to be convinced that their initial investment is returned to them with a profit. Another way of putting this is that the efficiency of capital markets depends on the reliability and completeness of information provided by entities that want to attract capital (Fama, 1970; Moore et al., 2006). In a utopian setting, all actors in a market have access to complete information, this means that investors do not have to doubt information provided by managers and can rationally outweigh all options they have. Unfortunately it doesn’t work like that in the real world. In the real world the relationship between investors and managers is influenced by agency theory. Possibilities for opportunistic behavior decrease the reliability of the information provided by firms and thereby compromise the investment decision (for a complete overview of agency theory see: Eisenhardt, 1989). Healy & Palepu (2001) argue that agency problems and information asymmetry lead to a demand for financial reporting and disclosure and that the credibility of disclosure can be “enhanced by regulators, standard setters, auditors and other

capital market intermediaries” (Healy & Palepu, 2001, p.406). Some authors argue (e.g. Jensen

& Meckling, 1976) that audits add value to a firm since they decrease agency costs by improving the credibility of disclosures. The argument that auditors can enhance credibility of disclosures is precisely why many governments require firms to hire an independent auditor to audit their financial reports. Auditor independence is seen as a key element of the external auditing process and the American Institute of Certified Public Accountants (AICPA) Council have even stated that: “Independence, both historically and philosophically, is the foundation of the public

accounting profession’s strength and its stature” (Carrey, 1970, p.182).

Many regulators, auditing professionals, and academics agree that auditor independence increases the credibility of financial reporting, nevertheless there is some disagreement about how auditor independence should be defined and operationalized (Dopuch, King & Schwartz, 2003). Regulators such as the Securities and Exchange Commission (SEC) and the Independence Standards Board (ISB) and also professional organizations such as the AICPA have generally

(9)

7

“partitioned independence into two dimensions: independence in fact, and independence in

appearance” (Dopuch, King & Schwartz, 2003, p. 84). The problem with independence is that it

is not directly observable for users of financial statements. This means that a breach of the independence in appearance of an auditor may lead investors to believe that they can’t rely on the financial reporting audited by that particular auditor, even though he may still be independent in fact(Lindberg & Beck, 2004).

The primary reason for maintaining auditor independence is the credibility of financial reports since this allows investors to make proper investment decisions. However, there are also other motives for auditors to maintain their independence (both in appearance and in fact). Accountants that compromise their independence become vulnerable to litigation risk and may also suffer a loss in reputation (Bonner, Palmrose & Young, 1998; Watts & Zimmerman, 1983; Simunic & Stein, 1996; Farmer, Rittenberg & Trompeter, 1987). As was stated in the introduction, Tepalagul & Lin (2015) elaborate on four threats to auditor independence. This thesis will focus on the client affiliation threat, which will be discussed next.

2.2 Auditor-Client affiliation

Imhoff (1978) discusses three possible reasons why auditor independence may be compromised with regards to the client-auditor relationship. The first reason that auditor independence may be impaired, is that auditors may view employers as potential future employers, which may cause the auditor to be more lenient. Secondly, auditors might get the idea that they work for and with management, while their actual employers are the shareholders. This may be caused by the “physical and emotional distance between the auditor and his true employer—the shareholder” (Imhoff, 1978, p.870). The third and final aspect of client affiliation issues that Imhoff discusses is that it may be difficult for auditors to maintain independence when their former colleagues sit across the table as managers of a client firm.

Other studies have looked into the effect of affiliations that executive officers have with CPA firms on audit quality. Lennox (2005) for example, discerns three different ways in which executive officers may be affiliated with CPA firms. The first one he mentions is the ‘employment affiliation’ which entails that an auditor gains employment with a client. The second type is the ‘alma mater’ affiliation. This means that an executive officer persuades the board of his or her new firm to hire the audit firm where the executive was formerly employed.

(10)

8

The third affiliation that Lennox mentions is the ‘chance affiliation’ meaning that there is “no

causal factor underlying the affiliation’ (Lennox, 2005, p. 202). Lennox finds that firms that

employ executives that have affiliations with CPA firms receive significantly more clean audit opinions than unaffiliated firms. Lennox suggests that this may be either due to decreased auditor independence or due to the familiarity of the executive with the testing methodology of the executive officer, allowing him to find ways to circumvent the controls (see also: Koh & Mahathevan, 1993). With regards to the employment affiliation issue, parts of SOX were aimed at minimizing any effects that this type of affiliation may have by requiring a one-year cooling off period before an auditor is allowed to be employed by a former client.

With regards to the employment affiliation, Menon and Williams (2004) have found that firms that employ a former partner of their auditing firm report larger accruals than firms that do not have this affiliation. Furthermore the affiliated firms are more likely to just meet analysts’ expectations when compared to unaffiliated firms. This suggests that affiliated firms benefit from employing former partners of their auditor.

Lennox and Park (2007) further research the ‘alma mater’ affiliation. They find that in a sample of firms who appointed one of the Big 5 auditing firms between 1995 and 2000 were more likely to appoint a particular firm if either the chief executive officer or the accounting/financial officer were alumni of that particular auditing firm. Of course, this might again mean that existing relationships and/or existing knowledge about auditing methodologies decreases the quality of oversight. For this thesis, the alma mater affiliation will not be studied but we will primarily focus on past and current employment relations which directors may have or have had with auditing firms.

2.3 Audit Tenure

It is important to note here that auditor tenure may play (or at least may be perceived to play) a role in ensuring auditor independence. From a theoretical standpoint it may be argued that longer tenure can result in more affiliation between management and auditors which would lead to lower financial reporting quality. Many studies however have shown that audit tenure actually has positive effects on audit quality and perceived audit quality in terms of accruals, cost of debt and analyst perceptions (Myers et al., 2003; Johnson, Khurana & Reynolds, 2003; Ghosh & Moon, 2005; Mansi, Maxwell & Miller, 2004). Furthermore on the website of the Public

(11)

9

Company Accounting Oversight Board, many people who wrote comment letters express their concern that mandatory firm rotation may have costs that outweigh the benefits (e.g. Pattillo, 2011). These claims are supported by Geiger and Raghunandan (2002) who find evidence that “audit reporting failures were more likely to occur in the initial years of an audit engagement” (Geiger & Raghunandan, 2002, p.75). Furthermore, it may be argued that the mandatory rotation of audit firms is a redundant measure if the affiliation between auditors and clients still exists through the presence of different types of social ties, which will be discussed in the following section.

2.4 Different types of social ties

Bruynseels & Cardinaels (2013) studied what effect different types of social ties between CEOs and members of the audit committee have on the quality of oversight. In order to study this they distinguish between three different types of ties, the first being ties formed through current or past employment, the second being ties formed through past education, and the final type of ties originates in “other non-professional activities”(Bruynseels & Cardinaels, 2013, p. 114). They suggest in their research that managers tend to surround themselves with people who are favorable to their policies. Other research has suggested that people use different types of networks in different situations (Saint-Charles & Mongeau, 2009; Caroll & Teo, 1996; Gibbons, 2004), they distinguish between advice networks and friendship networks. Generally speaking, ties formed through current or past employment and past education are considered to be part of advice networks, while friendship networks consist of ties formed by people being members of “leisure clubs, societies, charitable organizations, and so on (Bruynseels & Cardinaels, 2013, p.117). It is easier to discuss controversial topics when discussing them within friendship networks because people have closer and stronger ties (Plickert, Cote & Wellman, 2007). This suggests that managers can surround themselves with people who they are connected to through friendship networks to increase the likelihood that those connections will be favorable to their plans and policies, even if they are dubious in nature.

Ties in general are expected to decrease the quality of financial oversight. A critical mindset is essential to the process of financial oversight. When directors are connected to auditing firm directors in each of the three ways, it can be argued that they will have similar opinions which would diminish their ability to maintain a critical mindset. Studies have shown that when people are part of the same social network, their opinions start to converge through a process called

(12)

10

value congruence (e.g. DeGroot, 1974; Salancik & Pfeffer, 1978; Carley, 1991; Meglino, Ravlin & Adkins, 1989). These findings suggest that directors who are in the same network may eventually obtain the same opinion about certain subjects as the other members of their network and Gibbons (2004) finds that both friendship and advice networks play "discrete and significant

roles in shaping individuals' professional values." (Gibbons, 2004, p. 21). This is not inherently

a bad thing which negatively influences oversight, but in light of this thesis it may mean that the professional values of auditing firm directors are influenced by directors from other firms in their networks. The directors from audited firms may have financial incentives, e.g. stock options, to conceal bad news (Imhoff, 2003) or incentives to give some firms more effort than others based on the prestige of their directorship (Masulis & Mobbs, 2014). If the professional values of auditing firm directors are influenced by these incentives, this may lead to a lower quality of financial oversight. Barnea & Guedj (2007) "analyze several aspects of network that together

capture the connections board members have with other firms, and we show that these connections affect firm governance" (Barnea & Guedj, 2007, p.45). They find that firms with

directors who are more connected to directors in other firms show weaker governance in terms of CEO compensation, CEO pay-performance and CEO turnover. They also wonder "whether a

director who feels highly committed to a group of people but at the same time is supposed to govern individuals from this group, may, on the margin, be less critical of these individuals"

(Barnea & Guedj, 2007, p. 2). Bruynseels & Cardinaels (2013) provide some evidence that ties between CEOs and members of the audit committee formed through friendship networks adversely affect oversight quality. This thesis will focus on social ties that directors of firms in our sample have with auditing firm directors. We will then be able to see whether or not social ties between audit firm directors and directors from the firms in our sample have an effect on audit quality. In order to discover whether ties between directors and auditing firm directors have similar effects on governance (more specifically financial oversight) as results found by previous studies, the following hypothesis will be researched:

H1: A higher number of ties between directors and auditing firm directors leads to lower quality of financial disclosure oversight

Different situations may call for different networks to be used. Studies have shown that people draw on different networks in different situations (Gibbons, 2004; Saint-Charles & Mongeau,

(13)

11

2009). These different types of networks are used depending on what the situation requires. These different ties are drawn from various different types of networks. As we mentioned before individuals can draw on advice networks and friendship networks. Research has shown that ties based on friendship networks may create stronger bonds than ties based on advice networks (McPherson, Smith-Lovin & Cook, 2001). These stronger types of ties may in turn lead to directors feeling connected and feeling a stronger obligation to someone from a friendship network than to someone they are connected to through an advice network. In other words, individuals may be more lenient or more dedicated towards a friend than to a former colleague or fellow student. Gibbons (2004) found that values of individuals in friendship networks moved closer than with those in advice networks, and that "Influence through the advice network was

more restricted" (Gibbons, 2004, p. 22). This leads to the belief that ties formed through

friendship networks may have a stronger influence on the value congruence between directors. Following the expectations posed by Bruynseels & Cardinaels (2013), this thesis assumes that different types of network ties differ in the strength of the effects they have on oversight quality. This leads to the second hypothesis:

H2: The quality of oversight will decrease more for social ties based on friendship networks than for ties based on advice networks.

The next section will explain the methodology used for this thesis.

3. Research Method

3.1 Sample

This thesis will look at firms that are listed in the FTSE Euronext 100 Index. The shifting nature of the index makes it difficult to determine which firms were in the index for a given moment or period of time (since the index consists of the “100 largest companies within the FTSEurofirst

300 Eurozone Index (subject to Rules 7.3.3 and 7.3.4) which qualify as eligible for inclusion in the index.”(FTSE Russell, 2017) it has a shifting nature). To avoid such issues and shifting

samples, this thesis will look at the firms that were included in the index in March of 2017 and study those firms for the period 2008-2012. This index is used because most previous literature focuses on U.S. markets. By studying an index that looks at European firms, it is possible to compare outcomes with regards to social ties with evidence found in U.S. firms. The period has been selected because in April 2014 the EU first adopted legislation with regards to mandatory

(14)

12

audit firm rotation (PWC, 2015). The above mentioned legislation was adopted in order to increase auditor independence and reduce the perceived (negative) effects of long auditor tenure. (By looking at the prior period, evidence may be discovered concerning the appropriateness of adopted legislation. Data has been collected from BoardEx and ThomsonOne.

Firstly we exclude 80 firm-years from financial firms (SIC codes 6000-6999) because they are not comparable to other firms in the sample in terms of assets and financial reports. A lot of missing data was encountered for the index that was chosen for this thesis. To prevent the elimination of too many firm-years from the sample, a great deal of information was hand-collected from annual reports, this was only done for financial data as opposed to information such as board size, auditor tenure, CEO tenure, etc. Nevertheless another 25 firm-years had to be removed from the sample due to missing data in either Thomson-One or BoardEx. Which eventually resulted in a final sample of 395 firm-years for the ADA , another 21 firm-years were removed for the audit fee model since audit fee information was missing. Table 1 provides a general overview of the distribution of firms in our sample by Two-Digit SIC and year. Many of the firms in our sample are concentrated in the manufacturing industry. We find a quite equal distribution of the firms in our sample by years. That is because most firms that had available data, had the data for all years we were studying. Some audit fee data was missing for some firms, hence the discrepancy there.

TABLE 1

Panel A: Distribution of Sample Firms per Industry

ADA Model Audit Fee Model n Percentage n Percentage

10-14 Mining 5 1.27 5 1.34

15-17 Construction 15 3.80 15 4.01

20-29 Manufacturing Part 1 95 24.05 90 24.06

30-39 Manufacturing Part 2 100 25.32 100 26.74 40-49 Transportation, communication, utilities 75 18.99 64 17.11

50-51 Wholesale trade 5 1.27 5 1.34

52-59 Retail trade 30 7.59 28 7.49

(15)

13

Panel B: Distribution of Sample Firms per Year

ADA Model Audit Fee Model n Percentage n Percentage 2008 79 20 75 20.05 2009 79 20 76 20.32 2010 79 20 75 20.05 2011 79 20 74 19.79 2012 79 20 74 19.79 Total 395 100 374 100

3.2 Director-Auditor Social Ties

In order to measure the amount of ties that directors have with major auditing firm directors this thesis will use a similar approach as Bruynseels and Cardinaels (2013) and Fracassi and Tate (2012). Where Bruynseels and Cardinaels (2013) used the proportion of directors with ties, we will take a somewhat different approach. The way in which this thesis will differ from their methods is that we will look at all ties, past and present that the directors from the firms in our sample have, or have had with either Big 4 directors or directors from medium sized firms such as Mazars and BDO. Employment ties will arise when a link is found between a director of one of the firms in our sample and a director of a major auditing firm. In order to discover if any ties originate from education, we will look at the education that directors in our sample have had and we will combine this information with the educational institutions which the auditing firm directors have enjoyed. Any overlaps that are found will be viewed as educational ties for the purposes of this research, this is loosely based on the approach that Cohen, Frazzini and Malloy (2008) use. A comparable approach will be used to measure ties that arise through other non-professional activities, we will look at all the directors in our sample and any other activities (such as charities, leisure clubs etc.) they partake in, we will look for any overlaps with auditing firm directors in the data from BoardEx and we will use this information to identify other activities ties. As we have discussed in previous sections, ties formed through friendship networks are expected to decrease the quality of financial oversight because it may be easier to discuss sensitive issues.

3.3 Other Independent Variables

Previous studies have suggested that there are also other governance variables that influence audit effectiveness. The empirical models in this thesis will include governance variables that pertain to CEO strength, such as tenure (Hill & Phan, 1991) and whether or not the CEO is chair

(16)

14

of the board (Core, Holthausen, and Larcker, 1999), governance variables that pertain to board strength such as board size, board independence, audit committee size and the number of financial experts on the audit committee (Abbot et al., 2003; Klein, 2002; Bruynseels & Cardinaels, 2013; Krishnan, 2005). A previous study by Davis, Soo and Trompeter (2002), has also found empirical evidence for an association between earnings management and auditor tenure. Therefore an additional variable measuring auditor tenure is also included in the model. 3.4 Empirical Models

3.4.1 Absolute Discretionary Accruals

The first empirical model in this thesis is based on Klein (2002) and Bruynseels and Cardinaels (2013). Discretionary accruals will be estimated to proxy for the amount of earnings management. Congruent with Bruynseels and Cardinaels (2013), the method by Ball and Shivakumar (2006) will be used. In order to measure the amount of discretionary accruals, the following model is estimated:

(1) where;

TACCi,t total accruals for firm i in fiscal year t, calculated as income before extraordinary

items minus operating cash flow

AVTAi,t average total assets for firm i in year t and year t-1

∆REVi,t change in revenues for firm i in fiscal year t

PPEi,t gross property, plant and equipment for firm i in fiscal year t

DCFOi,t dummy variable equal to 1 if cash flow from operations is negative, and 0 otherwise

CFOi,t Cash flow from operations for firm i in fiscal year t

(Bruynseels & Cardinaels, 2013)

After estimating the model above, discretionary accruals are obtained by calculating the difference between the total accruals deflated by AVTA and the fitted values obtained above (Bruynseels & Cardinaels, 2013). It would be good practice to collect at least 15 years per

(17)

two-15

digit SIC code industry and year, unfortunately this was not possible for our index and the accruals calculation was performed for the entire sample.

Following Klein (2002) and Bruynseels and Cardinaels (2013) a number of control variables are added to the model. These are included because previous studies have found that it is important to control for financial performance when researching earnings management (e.g. Dechow, Sloan & Sweeney, 1995; Cohen, Dey & Lys, 2008) Combining these with the abovementioned independent variables, the following model is estimated:

(2) where;

VARIABLE DESCRIPTION

ADA Absolute discretionary accruals as calculated by the Ball & Shivakumar method (2006)

TIES Number of ties between company directors and directors from auditing firms; either ALLTIES, which is a summation of the three separate types of ties or OTHERACT (ties from other activities), EMPLOY (ties arising from employment), EDUCATION (ties arising from

directors having enjoyed education at the same institution).

CEOCHAIR Indicator variable showing 1 if the CEO is also chairperson of the board and 0 otherwise

CEOTENURE Number of years CEO has been in office

BOARDSIZE Number of directors on the board

INDEP Number of independent directors on the board

ACSIZE Size [Number of people?] of the audit committee

AUDITORTENURE Duration of current engagement with auditor in years

BIG4 Indicator variable showing 1 if external auditor is a Big4 firm, 0 otherwise

(18)

16

LNTA Natural log of total assets

CHANGENI Absolute value of change in net income between t-1 and t in EUR

LOSS Indicator variable equal to 1 if the firm has two or more consecutive years of negative income and 0 otherwise

LTDTA Long term debt divided by last year’s total assets

INDUSTRY Industry dummies

YEAR Year dummies

3.4.2 Audit Fee Analysis

The variables used in the estimation of the audit fee model are based upon previous work that has studied the determinants of audit fees (Bruynseels & Cardinaels, 2013; Craswell, Francis & Taylor, 1995; Defond, Raghunandan & Subramanyam, 2002; Higgs & Skantz, 2006). Again auditor tenure is included in the model. Hay (2013) combined findings from the US General Accounting Office (GAO, 2003) and from Hay et al. (2006) and concluded that audit firms with short tenures charge lower fees than audit firms with longer tenures. Again, besides some financial indicators the model below includes variables concerning governance characteristics. The control variables included in the model control for factors that influence audit fees due to the fact that they require more audit effort, thereby causing higher audit fees (e.g. initial audits, high proportion of receivables/inventory). (3) where; VARIABLE DESCRIPTION

AUDITFEE Natural log of the audit fee paid to the incumbent auditor

TIES Number of ties between company directors and directors from auditing firms; either ALLTIES, which is a summation of the three separate types of ties or OTHERACT (ties from other activities), EMPLOY

(19)

17

(ties arising from employment), EDUCATION (ties arising from directors having enjoyed education at the same institution).

CEOCHAIR Indicator variable showing 1 if the CEO is also chairperson of the board and 0 otherwise

CEOTENURE Number of years CEO has been in office

BOARDSIZE Number of directors on the board

INDEP Number of independent directors on the board

ACSIZE Size of the audit committee

AUDITORTENURE Duration of current engagement with auditor

FINEXP Proportion of financial experts on the audit committee

LNTA Natural log of total assets

INITIAL Indicator variable equal to 1 if the auditor is in the first year of engagement, and 0 otherwise

BIG4 Indicator variable equal to 1 for a Big 4 auditor, and 0 otherwise

DA Ratio of liability to assets

LIQ Ratio of current assets to current liabilities

ROA Ratio of operating income to assets

LOSS Indicator variable equal to 1 if net income in fiscal year t or t-1 is less than 0, and 0 otherwise

INVREC Ratio of inventory and receivables to total assets

BM Book-to-market ratio at the end of the audit-fee year, calculated as common equity to market capitalization

GROWTH One-year growth in sales

INDUSTRY Industry dummies

YEAR Year dummies

3.4.3 Multi-Collinearity

Before we proceed we need to examine whether collinearity is an issue in these models. The correlation coefficients can be found in table 2. This needs to be examined because collinearity may adversely influence research outcomes. Collinearity may “yield less precise estimates, can

induce parameters to switch signs, and affect R2”(Mela & Kopalle, 2002, p. 675). To test

(20)

18

obtained. We find that our variable LNTA shows significant collinearity with three other variables, namely TIES, BOARDSIZE and INDEP, this occurs in both the ADA model (0.403, 0.511, and 0.506 respectively) and the Audit Fee Model (0.403, 0.511, and 0.506 respectively). When performing the regression without the natural log of total assets, there is no significant change in the effect of the variables with which it is correlated. Additional analysis is performed by looking at the variance inflation factor (VIF) is analyzed. There is no universal standard for when a VIF is deemed too high, nevertheless “generic cutoff values, such as VIF ≥ 5 or VIF ≥

10, are commonly used to determine if the collinearity is strong enough to require remedial measures” ( Craney & Surles, 2002, p. 392). In the case of our models, no VIF exceed 3.49,

therefore no remedial measures are deemed necessary. Taking these findings into consideration we will in this case decide not to omit the variable since it controls for the size of our firms which is deemed too important to not include in the models.

It is important to note about both Pearson correlation analyses is that there is collinearity between the OTHERACT, EMPLOY, and EDUCATION variables. A theoretical explanation for this occurrence may be that directors with more ties through one of the three possible sources will have a larger network through other sources as well.

(21)

19

Table 2: Pearson Correlations Panel A: ADA Model

ADA ALLTIES OTHERACT EMPLO

Y

EDUCATIO N

CEOCHAIR CEOTENURE BOARDSIZE INDEP ACSIZE

ADA 1 ALLTIES -0.00175 1 OTHERACT -0.0906 0.625*** 1 EMPLOY 0.00555 0.989*** 0.562*** 1 EDUCATION -0.0160 0.664*** 0.586*** 0.551*** 1 CEOCHAIR 0.291*** -0.00167 -0.201*** 0.0230 -0.0990 1 CEOTENURE 0.156** -0.153** -0.169** -0.136** -0.175*** 0.274*** 1 BOARDSIZE 0.182*** 0.178*** 0.0905 0.145** 0.296*** 0.205*** -0.0112 1 INDEP 0.155** _0.276*** _0.373*** _0.252*** _0.249*** _0.0266 _-0.0903 _0.489*** ₁ ACSIZE 0.0519 0.222*** 0.0889 0.200*** 0.269*** 0.0499 -0.176*** 0.335*** 0.184*** 1 AUDITOR- TENURE 0.0474 0.0489 0.129* 0.0724 -0.125* 0.0637 0.0944 -0.128* 0.0407 -0.0804 BIG4 0.0665 0.0194 0.00100 0.0229 -0.00487 0.0486 0.0120 0.00260 0.00503 -0.0237 FINEXP -0.242*** 0.0902 0.228*** 0.0427 0.278*** -0.302*** -0.135** -0.121* 0.121* 0.241*** LNTA 0.0848 0.407*** 0.493*** 0.365*** 0.415*** 0.152** -0.239*** 0.501*** 0.494*** 0.276*** CHANGENI 0.0198 -0.0656 -0.0717 -0.0572 -0.0778 -0.0464 0.0696 -0.0125 -0.0517 0.049 LOSS 0.0308 -0.0488 -0.0477 -0.0541 0.00135 0.112* -0.0804 0.00305 -0.0916 0.120* LTDTA -0.0860 -0.0190 -0.129* -0.0563 0.238*** 0.0356 -0.0799 -0.0111 -0.157** 0.0165

AUDITORTENURE BIG4 FINEXP LNTA CHANGE

NI LOSS LTDTA AUDITORTENURE 1 BIG4 0.0763 1 FINEXP -0.130* 0.00476 1 LNTA -0.0552 0.0165 0.000408 1 CHANGENI -0.0342 -0.000671 0.00507 -0.101 1 LOSS 0.0244 0.0201 0.0888 -0.0187 -0.0569 1 LTDTA -0.173*** 0.00136 0.138** -0.0410 -0.0163 0.0190 1

(22)

20

Panel B: Audit Fee Model AUDIT

FEES

ALLTIES OTHERACT EDUC EMPLOY CEOCHAIR CEOTENURE BOARDSIZE INDEPDIR ACSIZE AUDITOR

TENURE AUDITFEES 1 ALLTIES 0.323*** 1 OTHERACT 0.400*** 0.623*** 1 EDUC 0.220*** _0.666*** _0.592*** ₁ EMPLOY 0.307*** 0.989*** 0.560*** 0.553*** 1 CEOCHAIR 0.143** -0.000910 -0.206*** -0.0916 0.0228 1 CEOTENURE -0.188*** -0.145** -0.158** -0.176*** -0.127* 0.281*** 1 BOARDSIZE 0.428*** 0.181*** 0.0968 0.286*** 0.150** 0.230*** -0.00754 1 INDEPDIR 0.439*** 0.282*** 0.389*** 0.237*** 0.260*** 0.0537 -0.0969 0.471*** 1 ACSIZE 0.225*** 0.228*** 0.0961 0.262*** 0.209*** 0.0888 -0.187*** 0.318*** 0.149** 1 AUDITORTE NURE 0.136** _0.0421 _0.126* _-0.129* _0.0655 _0.0644 _0.105* _-0.123* _0.0454 _-0.0822 ₁ FINEXP 0.0112 0.0972 0.245*** _0.273*** _0.0506 _-0.285*** _-0.156** _-0.149** _0.0924 _0.211*** _-0.140** LNTA 0.731*** 0.403*** 0.490*** 0.415*** 0.360*** 0.153** -0.228*** 0.511*** 0.506*** 0.287*** -0.0638 INITIAL -0.103 -0.0365 -0.0417 0.0812 -0.0536 -0.0273 -0.0356 0.0808 -0.0657 0.0290 -0.208*** BIG4 0.0859 0.0198 0.00175 -0.00514 0.0234 0.0491 0.0111 0.00200 0.00449 -0.0245 0.0778 DA 0.0632 0.126* -0.0349 0.209*** 0.107* 0.183*** -0.0527 0.180*** -0.122* 0.137** -0.156** LIQ -0.191*** -0.213*** -0.122* -0.210*** -0.196*** -0.157** 0.0129 -0.209*** 0.0333 -0.143** 0.0370

FINEXP LNTA INITIAL BIG4 DA LIQ ROA LOSS INVREC BM GROWTH

FINEXP 1 LNTA 0.00951 1 INITIAL 0.0629 -0.00859 1 BIG4 0.00402 0.0171 0.00789 1 DA 0.0676 0.247*** 0.0521 0.0168 1 LIQ -0.0337 -0.337*** -0.0954 0.0507 -0.566*** 1 ROA -0.0252 -0.0369 -0.0100 0.00789 -0.0629 -0.00438 1 LOSS 0.0976 -0.0225 -0.00439 0.0208 0.114* 0.0809 -0.0611 1 INVREC -0.182*** -0.118* 0.0362 0.0510 -0.134* 0.225*** -0.0483 0.164** 1 BM -0.173*** 0.293*** 0.0812 -0.0212 -0.00993 0.00844 0.0542 0.251*** 0.242*** 1 GROWTH 0.00802 0.132* -0.00732 0.00579 -0.0111 -0.0229 0.00255 -0.0591 -0.00819 0.000963 1 t statistics in parentheses *_{p < 0.05,}**_{p < 0.01,}***_{p < 0.001}

(23)

21

4. Results

This section will present the empirical results found by estimating the two models described above.

4.1 Absolute Discretionary Accruals Model

4.1.1 Descriptive Statistics—Absolute Discretionary Accruals Model

Table 2 presents the descriptive statistics for the ADA model. The mean of our absolute discretionary accruals is 0.0665. We can see that a lot of ties are found for the sample and approach we use. The mean for our accruals is 0.0665. The mean for ties formed through employment (301.36) are much higher than ties formed through education (68.61) and other activities (14.85). Most firms (99.75%) have a Big 4 auditor, this may be influenced by the fact that all firms in our sample are listed firms. In 46.58% of our observations, the CEO is also the chair of the board of directors. CEO tenure has a mean of 5.0785 with a lot of variance ranging from CEOs who have just begun and a maximum tenure of 25 years. An average audit committee in our sample consists of 4.365 directors. An interesting thing to note is that auditor tenure is on average much longer (11.55 years) than the proposed period of four years which auditors are allowed to audit firms in the new legislation. In our sample, on average firms have 1.06 financial experts in their board of directors.

4.1.2 Empirical results—ADA Model

In this section the main results of the regression analysis of the ADA model will be presented and will be analyzed in light of the hypotheses. Table 3 presents the results of the ADA model. As was specified earlier, two models will be estimated for each dependent variable. One model where ALLTIES is included in the model, and one model where ties are grouped by type, either OTHERACT, EMPLOY, or EDUCATION. All our independent variables show very low correlation with the dependent accruals variable. This can be partially explained since this thesis use the number of social ties as a metric instead of the proportion of tied directors as was done by Bruynseels & Cardinaels (2013). Nevertheless, our regression finds no significant correlation for any of our independent variables, and in fact even finds a negative correlation for our ALLTIES variable. These results suggest that social ties between directors of firms in our sample and directors of major auditing firms have no significant effect on absolute discretionary accruals. Additionally for Model 2, we find that EMPLOY shows a negative correlation with

(24)

22

ADA, suggesting that more ties stemming from employment decrease the amount of accruals, which is interpreted as a proxy for higher quality of financial oversight.

Table 3: Descriptive Statistics ADA Model (n = 395)

mean sd min max

ADA .0665 .0777 -.1600 .3020 ALLTIES 384.0506 349.3217 31 2187 OTHERACT 14.8481 15.9449 0 90 EMPLOY 301.3671 307.6281 12 2043 EDUCATION 68.6076 52.8109 9 320 CEOCHAIR .4658 .4995 0 1 CEOTENURE 5.0785 4.7052 0 25 BOARDSIZE 13.2205 3.5095 4 29 INDEP 6.2026 2.5366 0 12 ACSIZE 4.3650 1.3737 2 15 AUDITORTENURE 11.5570 5.3410 1 25 BIG4 .9975 .0503 0 1 FINEXP 1.0651 1.0212 0 4 LNTA 9.5103 1.3300 6.4672 12.4795 CHANGENI -134.26 1230.896 -8907 6873.8 LTDTA .2130 .1598 0 1.0048 LOSS .1367 .3440 0 1 INDUSTRY1 .0125 .1112 0 1 INDUSTRY2 .0375 .1902 0 1 INDUSTRY3 .2375 .426 0 1 INDUSTRY4 .25 .4336 0 1 INDUSTRY5 .1875 .3908 0 1 INDUSTRY6 .0125 .1112 0 1 INDUSTRY7 .075 .2637 0 1 INDUSTRY8 .175 .3804 0 1

It is interesting to note that both CEOCHAIR and CEOTENURE show a positive and significant effect on accruals in our model. This suggests that firms that employ CEOs who are also chair of the board of directors and who have longer tenure, will have larger accruals. Contrary to expectations, INDEP and ACSIZE also have a positive effect on ADA. These results imply that firms who have more independent directors and a larger audit committee also have more accruals. A final interesting result is the positive correlation between our indicator value LOSS and ADA, indicating that firms who report a loss will have more accruals than firms who don't report a loss. R-squared and adjusted R-squared for Model 1 (Model 2) are 0.2890 (0.2894) and 0.2402 (0.2362) respectively. We find no significant support for hypothesis 1 that more ties lead to lower quality of financial disclosure oversight. With regards to hypothesis 2 we find that EDUCATION shows a stronger (while still insignificant) effect on ADA than OTHERACT

(25)

23

does. This means that no evidence is found to support hypothesis 2 that ties from friendship networks decrease the quality of financial oversight more than ties stemming from advice networks do.

Table 4: Regression Results for ADA Model

ADA Model 1 ADA Model 2

ALLTIESDIR -0.00000128 (-0.10) OTHERACT 0.00000151 (0.00) EMPLOY -0.00000551 (-0.33) EDUCATION 0.0000421 (0.38) CEOCHAIR 0.0253** (3.05) 0.0258** (3.03) CEOTENURE 0.00261** (3.00) 0.00259** (2.95) BOARDSIZE -0.00105 (-0.74) -0.00118 (-0.79) INDEP 0.00849*** (4.50) 0.00854*** (4.48) ACSIZE 0.00885** (2.79) 0.00878** (2.71) AUDITORTENURE -0.00116 (-1.42) -0.00113 (-1.27) BIG4 0.0972 (1.41) 0.0971 (1.40) FINEXP -0.0205*** (-4.95) -0.0209*** (-4.84) LNTA -0.00585 (-1.43) -0.00607 (-1.40) CHANGENI -0.000000124 (-0.04) -3.43e-08 (-0.01) LOSS 0.0230* (2.00) 0.0227* (1.97) LTDTA -0.0276 (-0.92) -0.0314 (-0.97) INDUSTRY1 0.0103 (0.29) 0.00961 (0.26) INDUSTRY2 0.0114 (0.51) 0.0119 (0.52) INDUSTRY3 0.0308* (2.41) 0.0295* (2.21) INDUSTRY4 0.0312* (2.50) 0.0306* (2.42) INDUSTRY5 0.0596*** (4.05) 0.0589*** (3.95) INDUSTRY6 0.0625 (1.88) 0.0622 (1.87) INDUSTRY7 0.111*** (6.46) 0.110*** (6.17) YEAR1 0.0128 (1.10) 0.0130 (1.10) YEAR2 0.00151 (0.13) 0.00169 (0.15) YEAR3 0.0170 (1.49) 0.0171 (1.49) YEAR4 0.0154 (1.38) 0.0155 (1.38) Constant -0.0851 (-1.11) -0.0814 (-1.05) N 395 395 R Squared 0.2890 0.2894 Adjusted R squared 0.2402 0.2362 t statistics in parentheses *_{p < 0.05,}**_{p < 0.01,}***_{p < 0.001}

4.2 Audit Fee Model

4.2.1 Descriptive Statistics—Audit Fee Model

Table 4 presents the descriptive statistics for the Audit Fee Model. The means for the ties are similar to the means found for the ADA model (employment ties: 309.72, education: 68.70,

(26)

24

otheract: 15.35). Similar auditor tenure is found (11.81 years) when compared to the ADA model, which is again longer than the proposed period of four years. In 13.9% of observations, firms reported a loss. The descriptive statistics for most variables are quite comparable to the statistics in our ADA model since only a small portion of firms in our sample had missing audit fee data.

Table 5: Descriptive Statistics Audit Fee Model (n=374)

mean sd min max

AUDITFEES 2.0013 1.1769 -.7361 7.4254 ALLTIES 392.9545 355.6911 31 2187 OTHERACT 15.3476 16.2232 0 90 EMPLOY 309.7193 313.1189 12 2043 EDUCATION 68.7032 53.7671 9 320 CEOCHAIR .4759 .5001 0 1 CEOTENURE 4.9679 4.7640 0 25 BOARDSIZE 13.1096 3.4348 4 29 INDEP 6.2273 2.4711 0 12 ACSIZE 4.3566 1.3635 2 15 AUDITORTENURE 11.8075 5.2268 1 25 BIG4 .9973 .0517 0 1 FINEXP 1.0771 1.0108 0 4 LNTA 9.5493 1.3433 6.4672 12.4795 INITIAL .0214 .1449 0 1 DA .6183 .1653 0 1.1776 LIQ 1.2843 .5741 .3679 3.7188 ROA .1044 .5788 -.1047 9.28 LOSS .1390 .3464 0 1 INVREC .2642 .1386 .0159 1.2636 BM .7006 .6395 -.1797 5.2684 GROWTH 969.8465 10241.09 -112473.4 79383.2 INDUSTRY1 .0132 .1142 0 1 INDUSTRY2 .0396 .1952 0 1 INDUSTRY3 .2375 .4261 0 1 INDUSTRY4 .2639 .4413 0 1 INDUSTRY5 .1689 .3751 0 1 INDUSTRY6 .0132 .1142 0 1 INDUSTRY7 .0739 .2619 0 1 INDUSTRY8 .1768 .3820 0 1 YEAR1 .1979 .3989 0 1 YEAR2 .2005 .4009 0 1 YEAR3 .1979 .3989 0 1 YEAR4 .1953 .3969 0 1 YEAR5 .1953 .3969 0 1

4.2.2 Empirical Results--Audit Fee Model

Table 5 presents the results for our Audit Fee model, again with the distinction between two models. One including ALLTIES and the other including OTHERACT, EMPLOY, and

(27)

25

EDUCATION. ALLTIES shows no significant correlation with our dependent variable. In Model 2 on the other hand, we find that EMPLOY shows a significant positive correlation with our dependent variable, suggesting that more ties stemming from employment result in higher audit fees, which we interpret as a higher effort and therefore higher quality of financial oversight. We find that the coefficient for OTHERACT (-0.00216) is more negative than the coefficients for EDUCATION (-0.00200) and EMPLOY (0.000454), which may be interpreted as support for our second hypothesis, however as was stated before, this result is not significant. Similar to the ADA Model, our other variables show no significant correlation with our dependent variable, so again we find no evidence to support either of our hypotheses. We find that CEOCHAIR has a significant positive correlation with Audit Fees in both models. It is interesting to note that CEOTENURE also has a significant correlation but in this case a negative effect is estimated. BOARDSIZE shows significant positive correlation for both models, indicating that firms with more members on their board of directors pay higher audit fees and consequently in our reasoning have a higher quality of financial oversight. Our results also imply that more financial experts on the board of directors leads to higher audit fees. It is interesting to note that AUDITORTENURE increases audit fees. This may be caused by several reasons, firstly it may imply that longer tenure leads to increased effort because auditors become more invested in their client, another explanation may be that auditors discount their offers for initial audits to obtain new clients and increase their rates after the initial audit (Huang, Raghunandan & Rama, 2009). As was the case in the ADA Model, we find little to no evidence in the regression results of our Audit Fee Model that supports either the first or second hypothesis.

4.3 Robustness Test

In the regression results presented above, we find little evidence to support either of the two hypotheses. In studying the data that was used for this thesis, there were two firms in our final sample that stood out because of the large number of ties that were found compared to other firms in the sample. In order to test whether our results are robust to excluding these outliers, we perform another regression, this time excluding those firms with extremely high numbers of ties. Only firms with extremely high numbers of ties are removed as opposed to both firms with high

(28)

26

Table 6: Regression Results for Audit Fee Model Audit Fee Model 1 Audit Fee Model 2 ALLTIES 0.000192 (1.49) OTHERACT -0.00216 (-0.50) EMPLOY 0.000454** (2.64) EDUCATION -0.00200 (-1.79) CEOCHAIR 0.220* (2.42) 0.189* (2.05) CEOTENURE -0.0226* (-2.37) -0.0219* (-2.30) BOARDSIZE 0.0572*** (3.64) 0.0636*** (3.91) INDEP -0.0241 (-1.14) -0.0263 (-1.23) ACSIZE -0.0435 (-1.26) -0.0414 (-1.20) AUDITORTENURE 0.0364*** (4.03) 0.0381*** (3.98) FINEXP 0.122** (2.73) 0.148** (3.21) LNTA 0.613*** (12.87) 0.634*** (12.68) INITIAL -0.567* (-2.07) -0.494 (-1.81) BIG4 1.229 (1.68) 1.219 (1.67) DA -0.862** (-2.69) -0.814* (-2.52) LIQ -0.0925 (-1.04) -0.0806 (-0.90) ROA 0.0748 (1.13) 0.0735 (1.11) LOSS -0.0995 (-0.79) -0.0689 (-0.55) INVREC 0.860* (2.18) 0.733 (1.84) BM -0.194** (-2.64) -0.223** (-3.00) GROWTH -0.00000273 (-0.71) -0.00000292 (-0.77) INDUSTRY1 0.545 (1.42) 0.672 (1.65) INDUSTRY2 0.211 (0.88) 0.231 (0.96) INDUSTRY3 -0.133 (-0.95) -0.0448 (-0.31) INDUSTRY4 0.158 (1.09) 0.207 (1.42) INDUSTRY5 -0.400* (-2.54) -0.329* (-2.06) INDUSTRY6 0.200 (0.57) 0.218 (0.62) INDUSTRY7 -0.347 (-1.85) -0.328 (-1.72) Year1 0.0966 (0.76) 0.115 (0.90) Year2 -0.0322 (-0.26) -0.0312 (-0.26) Year3 -0.0457 (-0.37) -0.0515 (-0.42) Year4 -0.107 (-0.89) -0.106 (-0.88) Constant -5.455*** (-6.41) -5.703*** (-6.66) N 374 374 R squared 0.6613 0.6673 Adjusted R squared 0.6318 0.6361 t statistics in parentheses *_{p < 0.05,}**_{p < 0.01,}***_{p < 0.001}

numbers of ties and low numbers of ties since those firms are approximately within one standard deviation from the mean. Results of the robustness test are presented in Table 6. Panel A presents the results for our ADA Model. In our new estimation instead of a negative insignificant effect, we find a significant positive effect of ALLTIES on our dependent variable ADA in model 1, and a significant positive correlation between EMPLOY and ADA in model 2. These results suggest that more ties through employment lead to higher accruals. Also the signs for

(29)

27

OTHERACT and EDUCATION reverse. Other results appear to be robust and show minimal differences when compared to our initial estimation.

Table 7: Robustness Test

Panel A: Absolute Discretionary Accruals Models

ADA Model 1 t statistics ADA Model 2 t statistics

ALLTIESDIR 0.0000401* (2.10) OTHERACT -0.000767 (-1.65) EMPLOY 0.0000949** (2.83) EDUCATION -0.0000122 (-0.11) CEOCHAIR 0.0276** (3.28) 0.0232** (2.68) CEOTENURE 0.00289** (3.28) 0.00329*** (3.66) BOARDSIZE -0.00113 (-0.79) -0.00124 (-0.83) INDEP 0.00879*** (4.59) 0.00923*** (4.79) ACSIZE 0.00879** (2.72) 0.00761* (2.31) AUDITORTENURE -0.00149 (-1.77) -0.00114 (-1.27) BIG4 0.0937 (1.36) 0.0897 (1.30) FINEXP -0.0206*** (-4.96) -0.0176*** (-4.02) LNTA -0.00940* (-2.18) -0.00819 (-1.86) CHANGENI -0.000000535 (-0.18) -0.000000715 (-0.24) LOSS 0.0205 (1.78) 0.0202 (1.76) LTDTA -0.0216 (-0.72) -0.0174 (-0.54) INDUSTRY1 0.000313 (0.01) 0.0225 (0.60) INDUSTRY2 0.0165 (0.73) 0.0232 (1.02) INDUSTRY3 0.0290* (2.26) 0.0375** (2.78) INDUSTRY4 0.0345** (2.75) 0.0412** (3.18) INDUSTRY5 0.0570*** (3.79) 0.0604*** (4.00) INDUSTRY6 0.0673* (2.02) 0.0683* (2.06) INDUSTRY7 0.113*** (6.50) 0.109*** (6.10) YEAR1 0.0106 (0.90) 0.0123 (1.03) YEAR2 0.000562 (0.05) 0.00174 (0.15) YEAR3 0.0168 (1.45) 0.0178 (1.54) YEAR4 0.0157 (1.38) 0.0160 (1.42) Constant -0.0610 (-0.79) -0.0777 (-1.00) N 385 385 R squared 0.3031 0.3119 Adjusted R Squared 0.2538 0.2588 t statistics in parentheses *_{p < 0.05,}**_{p < 0.01,}***_{p < 0.001}

(30)

28

Panel B: Audit Fee Model

Audit Fee Model 1 Audit Fee Model 2 ALLTIESDIR 0.0000843 (0.40) OTHERACT -0.00385 (-0.77) EMPLOY 0.000816* (2.25) EDUCATION -0.00267* (-2.36) CEOCHAIR 0.200* (2.17) 0.154 (1.64) CEOTENURE -0.0228* (-2.35) -0.0182 (-1.84) BOARDSIZE 0.0591*** (3.75) 0.0681*** (4.18) INDEP -0.0155 (-0.72) -0.0164 (-0.76) ACSIZE -0.0565 (-1.61) -0.0593 (-1.68) AuditorTenure 0.0399*** (4.31) 0.0403*** (4.17) FINEXP 0.115* (2.56) 0.156** (3.27) LNTA 0.604*** (11.94) 0.604*** (11.84) INITIAL -0.518 (-1.89) -0.418 (-1.53) BIG4 1.200 (1.64) 1.177 (1.62) DA -0.713* (-2.18) -0.707* (-2.14) LIQ -0.0739 (-0.82) -0.0494 (-0.55) ROA 0.0778 (1.17) 0.0765 (1.16) LOSS -0.0868 (-0.69) -0.0636 (-0.51) INVREC 0.836* (2.11) 0.584 (1.45) BM -0.196* (-2.56) -0.200** (-2.65) GROWTH -0.00000221 (-0.57) -0.00000260 (-0.68) INDUSTRY1 0.592 (1.53) 0.715 (1.75) INDUSTRY2 0.175 (0.72) 0.265 (1.08) INDUSTRY3 -0.154 (-1.10) -0.0487 (-0.33) INDUSTRY4 0.129 (0.89) 0.216 (1.45) INDUSTRY5 -0.483** (-2.99) -0.418* (-2.58) INDUSTRY6 0.142 (0.40) 0.199 (0.57) INDUSTRY7 -0.391* (-2.07) -0.359 (-1.89) YEAR1 0.0994 (0.77) 0.112 (0.87) YEAR2 -0.0286 (-0.23) -0.0284 (-0.23) YEAR3 -0.0489 (-0.40) -0.0568 (-0.46) YEAR4 -0.104 (-0.85) -0.107 (-0.88) Constant -5.436*** (-6.31) -5.552*** (-6.48) N 364 364 R squared 0.6603 0.6690 Adjusted r squared 0.6298 0.6371 t statistics in parentheses * p < 0.05, ** p < 0.01, *** p < 0.001

Panel B of Table 6 shows the results for the robustness test of our Audit Fee Model. We find that the effect of EMPLOY doubles but becomes less significant when compared to the original model. The effect of EDUCATION becomes significant and increases slightly. OTHERACT also increases slightly but remains insignificant. For both models, the hypotheses would still be rejected so the test are considered to support the robustness of results.

(31)

29

5. Conclusion and Discussion

Auditor independence is sometimes viewed as the panacea for improving audit quality. Fraud cases such as Enron, Ahold, and more recently Imtech, have been the proponents behind new rules and regulations such as mandatory audit firm rotation, and the separation of audit services and non-audit services and these regulations have been implemented to ensure higher degrees of auditor independence. It remains unsure whether this mandatory rotation achieves it goals and whether its benefits outweigh its costs. DeFond and Zhang (2014) state in their literature review that "there is little evidence that several commonly perceived threats to audit quality actually

pose serious threats" (Defond & Zhang, 2014, p. 313). This study focuses on ties that may exist

between directors of audited firms and directors of auditing firms to see whether any adverse effects could be found that decrease audit quality. Some evidence was found for a positive correlation between ties formed through employment and audit fees, and by proxy, financial oversight. We expected social ties to have a negative effect on financial oversight, through finding a positive correlation between accruals and the number of ties and a negative relation between audit fees and the number of ties. In our results we found no evidence to support a significant adverse relationship between ties and financial oversight quality. As was stated earlier in this thesis, studies have shown that other perceived threats to audit quality, such as auditor tenure have positive effects on audit quality (e.g. Myers et al., 2003; Johnson, Khurana & Reynolds, 2003; Ghosh & Moon, 2005; Mansi, Maxwell & Miller, 2004). These results may indicate that the approach to improving audit quality should be changed. Instead of looking at and trying to remedy perceived threats it may be beneficial to thoroughly study what the determinants of audit quality actually are and translating the findings of those studies into policy, instead of acting upon the perceived threats. The latter seems to be common practice for policymakers and regulators. In this thesis some evidence was also found for other determinants of audit quality. CEO tenure and whether the CEO was also the chair of the board of directors were both found to increase accruals and designating financial experts was found to improve audit quality for both models. As was stated before, there are two main components of auditor independence, namely independence in fact and independence in appearance, and it seems that current policy focuses more on independence in appearance than in fact. This thesis will not argue that there are no issues with auditor independence, merely that both components deserve and require equal attention in policy and legislation.

Director network ties and the quality of financial oversight