The relationship between auditor independence and cost of debt

MSc Accountancy & Control, track Accountancy

Faculty of Economics and Business, University of Amsterdam

Master thesis:

The relationship between auditor

independence and cost of debt

Student name: Germaine Ortmans

Student number: 10661875

Date: 23-06-2014

First supervisor: dr. Alexandros Sikalidis

Abstract

In this study is examined whether a higher level of auditor independence reduces the cost of debt. As a proxy for auditor independence, different measures for provisions of audit services and non-audit services are used. In order to highlight the importance of the client to the auditor, the revenues of the auditor are also considered. In order to measure the dependent variable cost of debt, the interest expenses are divided by the average of total short- and long-term debt during the year (Pittman & Fortin, 2004). An additional regression analysis is conducted to test whether the effect of auditor independence on cost of debt is stronger for younger firms compare to older ones, since younger firms are dealing with a higher level of information asymmetry (Diamond, 1989). Diamond (1989) formulated the theory of reputation and implies that a better credit history results in lower interest rates over time. Myers and Majluf (1984) concluded that information asymmetry between two parties affects the financing decisions of organizations. To test this mediating effect, the initial public offering date of an organization is included as a variable in the second regression model. Both regression models are mainly based on the models of Pittman and Fortin (2004) and Larcker and Richardson (2004) and measure the effect of auditor independence on cost of debt. Two different samples are used in order to conduct the regression analyses (of respectively 1650 and 510 firm-year observations), consisting of organizations listing on the S&P 500. The variable that proxies whether an organization is in its first years after initial public offering had a lot of missing values, therefore two separate samples are used. The results lack sufficient evidence to conclude that auditor independence has a significant effect on cost of debt.The hypothesis expected that a higher level of auditor independence would reduce the cost of debt. Despite the fact that the variables for testing the mediating effect show significant relationships with the dependent variable, we cannot conclude that auditor independence affects cost of debt in firms listed on the S&P 500 in the period from 2008 to 2012.

Table of Contents

1. Introduction ... 4

2. Literature review and hypotheses ... 6

3. Research methodology ... 16 3.1 Sample selection ... 16 3.2 Methodology ... 18 3.2.1 Regression models ... 18 3.2.2 Dependent variable ... 21 3.2.3 Independent variables ... 21 3.2.4 Control variables ... 22 4. Results ... 24 4.1 Descriptive statistics ... 24

4.1.1 Descriptive statistics regression model 1 ... 24

4.1.2 Descriptive statistics regression model 2 ... 25

4.2 Differences in means ... 26

4.2.1 Differences in means regression model 1 ... 26

4.2.2 Differences in means regression model 2 ... 28

4.3 Correlation matrix ... 30

4.3.1 Correlation matrix regression model 1 ... 30

4.3.1 Correlation matrix regression model 2 ... 31

4.4 Regression results ... 36

4.4.1 Regression results regression model 1 ... 36

4.4.2 Regression results regression model 2 ... 38

5. Conclusion ... 41 6. References ... 43          

1. Introduction  

Several Accounting scandals like Enron and WorldCom have shed light on the accounting profession and this has raised debates regarding the quality of the audit and auditor independence. However, Jamal (2012) has concluded that auditor independence is quite an important factor, which contributes to the good functioning of capital markets. In order to improve auditor quality and auditor independence and to recover the trust in audit organizations, the Sarbanes-Oxley Act (2002) and mandatory rotation of audit firms were introduced. For example, the Sarbanes-Oxley act (2002) was introduced to restore the confidence in the external auditing profession and to improve the quality of financial reporting. Besides, there are some concerns with regard to threats to auditor independence. The SEC (Securities and Exchange Commission) is concerned that the provisions for non-audit services can act as a threat and therefore reduces non-auditor quality and independence. According to Levitt (2000), the auditing profession received increasingly negative critique due to the growing amount of non-audit provisions received by the auditors. These growing amounts might negatively influence the independence and opinion of the auditor.

This study focuses on the effect of auditor independence on cost of debt. Prior literature already shows some results with regard to the effect of audit quality on cost of debt. There’s for example a negative and significant relation between auditor quality and financing costs (Pittman & Fortin, 2004; Mansi, Maxwell & Miller, 2004). Furthermore, Chang, Dasupta and Hilary (2009) also found that financing decisions of companies are affected by auditor quality. Empirical evidence shows that firms audited by a Big6 auditor will have a lower cost of debt in comparison with firms audited by non-Big6 firms (Chang, Dasupta & Hilary, 2009). They found that companies audited by bigger audit firms are able to make larger equity issues. However, the results show that this difference between the Big6 and non-Big6 auditors will decrease when the market conditions improve (Chang, Dasupta & Hilary, 2009). This can be explained by the fact that companies audited by Big6 auditors issue less capital in comparison with companies audited by non-Big6 auditors (in case of more favourable market conditions).

In this study, an important aspect of audit quality will be examined, namely auditor independence. According to DeAngelo (1981a), audit quality consists of two very important attributes, namely competence and independence. These two attributes reflect the probability that an auditor will detect an offence and the probability that the auditor will do this correctly

(DeAngelo, 1981a). According to prior research, there might be some concerns that the quality of the auditor is harmed: “Regulators, financial statement users, and researchers are concerned that auditors compromise their independence by allowing high fee clients more financial statement discretion relative to low fee clients” (Ashbaugh et al., 2003, p. 611). This research will investigate the effect of auditor independence, one of the aspects of audit quality, on financing costs. The aim of this study is to examine whether auditor independence has an impact on cost of debt and additionally examines whether newly public organizations benefit from higher quality auditors. This is leading to the following research question: What is the effect of auditor independence on cost of debt?

In order to answer this research question, a hypothesis will be defined in the next chapter of this study. Another hypothesis will be formulated in order to test whether the effect of auditor independence on cost of debt is stronger for younger organizations compared to older organizations. These hypotheses will be tested by regression models composed of variables from prior research including a dependent variable, testing variables and several control variables (Pittman & Fortin, 2004; Larcker & Richardson, 2004 and Dhaliwal et al., 2008). Results show that there’s not sufficient evidence to conclude that auditor independence has a significant effect on cost of debt. It was expected that a higher level of auditor independence would reduce the cost of debt. It was also expected that this effect would be stronger of younger firms compared to older firms because of a higher level of information asymmetry (Diamond, 1989). Diamond (1989) formulated the theory of reputation and implies that a better credit history results in lower interest rates over time. Myers and Majluf (1984) concluded that information asymmetry between two parties affects the financing decisions of organizations. The results for both hypotheses do not show enough significant relations between the variables, since only one of the measures for auditor independence reported a significant p-value (in the first hypothesis). In the second hypothesis, none of the measures for auditor independence showed a significant relationship. On the other hand, one of the variables for testing the mediating effect of newly public firms showed a significant relation with the dependent variable for measuring interest expenses. The dependent variable is interest rate and is calculated as: “the interest expense divided by the average of total short- and long-term debt during the year” ( Pittman & Fortin, 2004, p. 125). However, it cannot be concluded that there’s an association between auditor independence and cost of debt.

My study contributes to the existing literature by using another measure for auditor independence than is commonly used. A frequently used measure for auditor independence in prior literature is the ratio of the audit fees or non-audit fees to the total audit fees (for example Ashbaugh et al., 2003; Dhaliwal et al., 2008; Frankel et al., 2002). Though, this resulted in approximately the same paid amounts for audit and non-audit services by the firms. A focus on this ratio measure, ignores the size of the fee payments to the auditor firm (‘the importance’). Therefore, also auditor independence will be measured by focussing on the importance of a client to the audit firm. This ratio measures the non-audit fees and total audit fees paid by the client firm relative to the total revenue of the auditor for that year (Larcker & Richardson, 2004).

The remainder of this research in structured in the following way: in the second section the literature review is provided followed by the development of the hypotheses. In chapter three the research design is described, followed by the results, which are explained in the fourth section. Finally, chapter five concludes, supplemented by limitations of this study and recommendations for future research.

2. Literature review and hypotheses  

This chapter provides a theoretical framework with regard to prior literature and existing theories. The framework is going to work towards the development of the hypotheses of the study.  

2.1 Audit quality and agency theory

According to the Financial Reporting Council (2006), audit quality has to do with the responsibility of the auditor to provide financial information and that the users of financial reports can rely on this information (since the financial reports are reported from a true and fair view opinion). DeAngelo (1981a) extensively examined auditor quality and concluded that it consists of two quite important aspects: competence and independence. These aspects reflect the probability that an auditor will detect an offence and the probability that the auditor will do this correctly.  

A commonly used theory which explains the role of the auditor is the agency theory. According to Jensen and Meckling (1976), there’s a relationship where the principal engages the agent in order to perform services on their behalf. Both the agent and the principal might

maximize their utility and this can lead to the result the agent doesn’t always act in the best interest of the principal. The principal can try to mitigate these differences in interest by introducing monitoring costs and providing incentives for the agent. Besides,  the principal can

pay bonding costs to the agent to prevent that the agent will take harmful decisions. Like mentioned before, in the agency theory exists a contractual relationship between on the one hand the principal and on the other hand the agent. In the study about auditor independence of Antle (1984), we see a relationship between the owner-principal and the manager-agent. According to Antle (1894), the auditor has the function to produce information that can be used in contracts with the manager. The owner of a company hires the auditor. This means that the auditor performs the function of the agent. Based on the agency theory, “this auditor-agent is assumed to behave as if she/he maximizes expected utility while taking investigative acts and making reports under conditions of moral hazard” (Antle, 1984, p. 1).

An important and well-known theoretical concept with regard to auditor quality is adverse selection, which can be explained as a type of information asymmetry between for example managers and outsiders. Prior literature shows that information asymmetry can affect the financing decisions of companies: Myers and Majluf (1984) explained for example how adverse selection led to forgoing profitable projects. Chang et al. (2009) also based their research framework on adverse selection. Their research framework involved that in case of poor auditor quality, companies rely more on debt rather than equity, in comparison with companies audited by audit firms of higher quality (Chang et al., 2009).

As earlier explained, information asymmetry between two parties can affect the financing decisions of organizations (Myers & Majluf, 1984). This might result in very high financing costs for organizations. According to Melumad and Thoman (1990), organizations with a higher level of information asymmetry face a higher risk of becoming insolvent and go more frequently bankrupt. However, auditing might be a solution for this explained problem (Becker, DeFond, Jiambalvo & Subramanyam, 1998). Auditing gives outsiders the allowance for verifying the validity and credibility of annual reports and financial statements and this could be a solution for the problem of information asymmetry (Becker et al., 1998). However, the effectiveness of auditing might vary with the quality of the auditor, since auditors of high quality are more likely to detect materiality and misstatements in financial statements compared to low quality auditors.

2.2 Auditor independence

DeAngelo (1981a) defined auditor independence as the conditional probability that the auditor has the function to report a breach when he has discovered a breach. She made the assumption that the conditional probability that a breach will be discovered, is positive and fixed (DeAngelo 1981a). Auditor independence is often measured by focusing on audit fees. In particular, prior literature measured the ratio of the audit fees or non-audit fees relative to the total audit fees (Ashbaugh et al., 2003; Dhaliwal et al., 2008). Dhaliwal et al. (2008) however used three other different measures for audit fees scaled by the total assets: “the total auditor fees divided by the square root of total assets, the audit fees divided by the square root of total assets and the non-audit fees divided by the square root of total assets”. Larcker and Richardson (2004) used some scaled variables that measured the importance of the client to the auditor.

According to Beattie et al. (1999), we can distinguish two types of auditor independence: independence in fact and independence in appearance. Independence in fact is recognized as the mental attitude of the auditor that should be nonbiased and independence in appearance is the assumption by a reasonable observer that there exists no relationship between the auditor and the audit client that can result in a conflict of interest (AICPA, 1993). This leads to the result that we cannot reliably measure independence in fact (since it has to do with the attitude of the auditor). However, we can measure auditor independence in appearance (Beattie et al., 1999). Also Majoor (2000) explains us the two notions with regard to the definition of independence. He links independence in mind (this is ‘independence in fact’ in the study of Beattie et al., 1999) often to the concepts objectivity and integrity. Independence in appearance however, is linked to the way independence in perceived by a stakeholder that is well-informed to a certain level.    

2.3 Revision of the Commission’s Auditor Independence Requirements (U.S. SEC)  

The independence standard of the U.S. SEC states that: “An auditor is not independent if a reasonable investor, with knowledge of all relevant facts and circumstances, would conclude that the auditor is not capable of exercising objective and impartial judgment” (SEC, 2000). During the 1990s the SEC had some concerns with regard to the growth of non-audit fees with respect to audit fees. “The SEC's concern that the growth in the provision of non-audit services compromises audit firm independence is based on the premise that the provision of

non-audit services increases the fees paid to the audit firm, thereby increasing the economic dependence of the audit firm on the client” (Ashbaugh et al., 2003).

Dhaliwal et al. (2008) concluded that the auditor fee disclosure rules of the SEC have stated that both audit fees and non-audit fees affects the perceptions of investors with regard to auditor independence. The intention of the SEC is providing information for the investors in order to evaluate the independence of the auditor: “First, the more the auditor has at stake in its dealings with the audit client, the greater the cost to the auditor should he or she displease the client, particularly when the non-audit services relationship has the potential to generate significant revenues on top of the audit relationship. Second, certain types of non-audit services, when provided by the auditor, create inherent conflicts that are incompatible with objectivity” (SEC The final rule: Revision of the Commission’s Auditor Independence Requirements, 200).  

2.4 The economic bond between audit firms and clients

Based on prior research, one can conclude that auditor independence is reduced by the strength of the economic bond between the audit firm and the client (DeAngelo, 1981a; Beck et al., 1988; Magee & Tseng, 1990). DeAngelo (1981a) stated that an audit firm would be more dependent on the client, in case of an increase of the bond between the audit firm and the client. Besides, this bond between audit firms and clients will increase further by means of the provisions non-audit fees, when the portion of audit-firm wealth increases (Simunic, 1984; Beck et al., 1988). While audit standards normally include a prohibition on contingent fees, Magee and Tseng (1990) argue that clients can create contingent fees. When clients use non-audit fees as contingent fees, the independence of the non-auditor might be threatened. Ashbaugh et al. (2003) argued that: “the sum of audit and non-audit fees, i.e., total fees, best captures the explicit economic bond between the audit firm and client” (p. 614).

2.5 Audit fees, non-audit fees, and total auditor fees

Dhaliwal et al. (2008) examined the association between audit fees, non-audit fees and total auditor fees and the cost of debt for firms. The audit fees are the fees (which could be either audit of audit fees) that are paid to the auditors. Since the SEC asserts that audit and non-audit services fees weaken non-auditor independence, they also examined the relationship between firms’ cost of debt and financial statement information: “We test whether the relation between

financial statement information and the cost of debt is affected by the fees paid to the auditor” (Dhaliwal et al., 2008, p.21). Their results provided evidence for the probable relationship between the non-audit provisions and cost of debt.

Dhaliwal et al. (2008) discussed the various implications of audit fees, non-audit fees and total auditor fees:  

1. Audit fees: Like mentioned before auditor independence is reduced by the strength of the economic bond between the audit firm and the client, caused by the economic rents from audit fees (DeAngelo, 1981a; Beck et al., 1988; Magee & Tseng, 1990). Dhaliwal et al. (2008) expected that: “investors will perceive the auditor’s independence to be impaired and the firm’s financial statements to be less reliable when the audit fee charged by the auditor is large relative to the size of the firm” (p. 4).

2. Non-audit fees: Dhaliwal et al. (2008) defined non-audit fees as: “those fees paid to a firm’s auditor that are related neither to the audit services performed for the purposes of financial statement, nor to the review services that are customary under generally accepted auditing standards” (p. 5). They include for example fees for pensions and benefit plan accountings and tax assistances.

3. Total auditor fees: Since auditor independence is weakened by the economic bond (implied by audit and non-audit fees), the total auditor fees are the best measure for auditor independence (Dhaliwal et al., 2008): “We include both the main effect of total auditor fees on the cost of debt and the effect of total auditor fees on the relation between financial statement information and the cost of debt” (p. 6).

Like mentioned before, Dhaliwal et al. (2008) expected that the financial statements become less reliable in case of a high charged audit fee relative to the size of the firm. However, Carcello et al. (2002) suggest that high audit fees enclose reliable financial statement information since larger audits are purchased by independent boards. These competing assumptions of the studies can lead to different hypotheses with regard to the association between audit fees and cost of debt.

2.6 Threats to auditor independence

We can distinguish several categories of threats to auditor independence (Eilifsen et al., 2006; Hayes et al., 2004 and IFAC Handbook on international auditing, assurance and ethics, 2008). These threats can reduce the independence of the auditor, which might result in a decrease of auditor quality. I think that the fourth threat ‘familiarity threats’ can be linked with the earlier explained concept ‘independence in appearance’, since both ‘familiarity threats’ and ‘independence in appearance’ highlight the risk of the auditor being in a relationship with the audited client.

1. Self-interest threats: This kind of threat to auditor independence can be the result from the auditor preferring its self-interest to the interest to perform a professional and neutral audit (Eilifsen et al., 2006). The auditor can for example have direct financial interest in the client (Hayes et al., 2004).

2. Self-review threats: A self-review threat can arise when it’s hard to evaluate the work of the auditor or the audit firm without any bias (Eilifsen et al., 2006). This can be the result of having the auditor (or anyone else from the audit team) an influence on the subject of the audit (Chia-Ah & Karlsson, 2010, master thesis).

3. Advocacy threats: This threat occurs when the auditor is advocating for or against the client and thus the auditor is acting in bias. In this case, the judgment of the auditor is subordinate to the judgment of the client (Chia-Ah & Karlsson, 2010, master thesis). 4. Familiarity threats: When the auditors are in a close relationship with a client (this

can be for example a family member), they might jeopardize their objectivity that can result in a familiarity threat (Chia-Ah & Karlsson, 2010, master thesis).

5. Intimidation threats: “This arises when the auditor is deterred from acting objectively because they are being or perceived to being prevented either overtly or covertly by the client or interested parties” (Chia-Ah & Karlsson, 2010 p. 27, master thesis).

2.7 Cost of debt

The effective rate that an organization has to pay on its current outstanding debt is called the cost of debt or financing costs. The cost of debt of an organization is part of its capital structure, next to the cost of capital. The level of the effective interest rate can depend on several factors. Information asymmetry is an important factor that can affect the level of the interest rate, since a higher level of information asymmetry between the lender and borrower can result in more risk for the lender. Interest rates can for example be based on the level of risk for the lender and a higher level of risk in an organization will obviously lead to a higher interest rate. Moerman (2005) has tested this indicator for the level of interest rates and found a positive and significant relationship between information asymmetry and cost of debt. From this study, one can conclude that a higher level of information asymmetry between the borrower and lender will lead to a higher interest rate (Moerman, 2005). Besides, the level of information asymmetry can depend on the quality of the disclosed information and the quality of the financial reports. According to prior research, organizations have a lower interest rate when they have high disclosure quality (Sengupta, 1998). A higher level of disclosure quality and high quality financial reports might reduce information asymmetry (Ball, 2001). This is also in accordance with the earlier explained conclusion that there exists a positive relationship between the level of information asymmetry and cost of debt.

The level of the effective interest rate of an organization depends also on its reputation. Prior studies found that a higher reputation will reduce the firm’s interest rate (Diamond, 1989 and Datta, Iskandardatta & Patel, 1999). When the lender has more guarantees that he will receive his money from the borrower and has confidence in the borrower, the risk of non-receiving is reduced and this might lead to a lower interest rate.

As explained, Moerman (2005) found evidence for this explanation and found a significant and positive relationship between cost of debt and information asymmetry. However, engaging an auditor of high quality and thus an auditor with a higher level of auditor independence might reduce this information asymmetry. Since auditing has a monitoring function in contracts, a high quality auditor can for example reduce the information asymmetry in a contract between a lender and borrower. According to prior research, engaging a high quality auditor might reduce the problems between lenders and borrowers with regard to contracting, since the auditor can provide valuable information to the lender and therefore can reduce the risk (Jensen & Meckling, 1976 and Watts & Zimmerman, 1986).

Since higher audit quality contributes to a higher level of credibility of the reported financial statements, the cost of contracting and therefore the cost of debt might be reduced (Ball, 2001). This conclusion is also in accordance with the evidence found in the studies of Pittman and Fortin (2004) and Mansi, Maxwell and Miller (2004). From their studies one can conclude that there exists a significant and negative relationship between auditor quality and cost of debt, since engaging a higher quality auditor results in a lower interest rate for an organization and so the cost of debt.

2.8 Auditor quality and cost of debt in prior literature

Prior literature has examined the relation between auditor fees and cost of debt, since the SEC had some growing concerns with regard to auditor independence (Dhaliwal et al., 2008). These growing concerns have arisen from the assumption that non-audit fees reduce auditor independence. Their results supported the presumption that there exists a positive relationship between non-audit fees and a firms’ cost of debt. However, this result only holds for investment-grade firms (Dhaliwal et al., 2008).

Pittman and Fortin (2004) have examined the association between debt pricing and auditor choice for newly public firms: “We report economically significant and statistically robust evidence that hiring a Big6 auditor lowers firms’ cost of debt capital less with age” (p. 114). Their study suggested that younger firms could lower their interest rates if they choose a Big6 auditor. Since the choice for a Big6 auditor enhances the credibility of financial statements, debt-monitoring costs can be reduced (Pittman & Fortin, 2004).

Chang, Dasupta and Hilary (2009) examined the relationship between the financing decisions of companies and auditor quality. Their study provided some statistical evidence that auditor quality affects financing decisions. Chang et al. (2009) found that companies audited by a Big6 auditor are more inclined to issue equity (as opposed to debt). Chang et al. (2009) also found that companies audited by bigger auditors are able to make larger equity issues (compared to companies that are audited by smaller audit firms). However, the results of their study have shown that this difference between the Big6 and non-Big6 auditors will decrease when market conditions improve. This can be explained by the fact that companies audited by Big6 auditors issue less capital in comparison with companies audited by a non-Big6 auditors (in case of more favourable market conditions). Chang et al. (2009) found that when companies are audited by Big 6 auditors, their debt ratios are less affected by market conditions.

Finally, Lennox and Pitman (2011) have examined the effect of the choice for on the one hand voluntary audits and on the other hand mandatory audits on a firms’ credit rating, which can also be used as a proxy for cost of debt. They inspected the signalling impact of this change in the audit regime and examined whether this has impact on firms’ credit ratings, since the research of Lennox and Pittman (2011) has found that auditors can be important information providers in debt contracting. They found strong evidence that companies had significantly higher credit ratings if they remain audited after the switch from mandatory audit to voluntary audit. This indicates that when an organization is still audited in a voluntary regime, quality of financial statement information is higher resulting in higher credit ratings. These higher credit ratings might reduce the cost of debt. Besides, their analysis provided evidence that opt-out companies were less likely to choose one of the Big4 audit firms and paid lower audit fees during the mandatory audit period, in comparison with companies who remained audited after the introduction of optional auditing (Lennox & Pitman, 2011).

2.9 Hypothesis 1 – development

Based on prior literature and the earlier explained theories in this literature review, the following hypothesis (H1) is formulated in order to examine whether auditor independence, as an aspect of auditor quality, affects the cost of debt:

H1: The higher the level of auditor independence, the lower the cost of debt.

This relationship between these two variables is expected since prior literature found for example a negative and significant relation between auditor quality and financing costs (Pittman & Fortin, 2004; Mansi, Maxwell & Miller, 2004). Besides, Dhaliwal et al. (2008) found evidence for their presumption that there exists a positive relationship between non-audit fees and a firms’ cost of debt. Chang et al. (2009) found that companies non-audited by a Big6 auditor are more inclined to issue equity (as opposed to debt).

2.10 New public organizations

To extend this research in examining the effect of auditor independence on cost of debt, the mediated effect of new public organizations on this relationship will be investigated. According to prior research, the level of information asymmetry in younger firms (so firms that are in the first period after public initial offering) is higher compared to the level of information asymmetry in older firms (Diamond, 1989). Diamond (1989) formulated the

theory of reputation formation in debt markets and this theory implies that gathering a good credit history might result in lower interest rates over time. In the initial years after public offering, organizations do not have an own reputation on which they can build. They also don’t have a very long credit record where lenders can build on. There’s a higher level of information asymmetry among others stakeholders and lenders since the level of disclosure of information is low in the initial years. According to Pittman and Fortin (2004), hiring a high quality auditor (for example a Big4 auditor) is important for newly public firms in order to reduce the level of financing costs. However, from their research one can conclude that this effect will decline over time when organizations have their own reputation on which they can build. Pittman and Fortin (2004) argue that the level of information asymmetry will decline over time. Besides, prior research found that new organizations are more dependent on external financing sources compared to older organizations (Rajan & Zingales, 1998). Furthermore, in case of a higher level of uncertainty of surroundings, it’s harder for younger organizations to rely on their own developed reputation (Hughes, 1986 and Datar et al., 1991).  

2.11 Hypothesis 2 – development

Based on prior literature and the earlier explained theories in this literature review and especially the previous paragraph, the following hypothesis (H2) is formulated in order to check whether auditor independence, as an aspect of auditor quality, has a stronger effect on the cost of debt for newly public organizations compared to older companies:

H2: The effect of auditor independence on cost of debt is stronger for newly public organizations in comparison with older organizations.

As mentioned before, a negative relationship between auditor independence and cost of debt is expected since prior literature found for example a negative and significant relation between auditor quality and financing costs (Pittman & Fortin, 2004; Mansi, Maxwell & Miller, 2004). Besides, according to the reputation theory of Diamond (1989), the level of information asymmetry in younger organizations is higher compared to older organizations. This might lead to the expectation that the effect of auditor independence on cost of debt is stronger for newly public organizations in comparison with older organizations.

3. Research methodology

In this chapter the process of data collection and sample composition will be elaborated. Furthermore the methodology concerning auditor independence and the strengthened effect of newly public firms on cost of debt will be explained.

3.1 Sample selection

The aim of this study is to test whether auditor independence has an effect on the cost of debt. In order to conduct this study an empirical research will be executed since I think the variables can be measured best by means of an empirical and mainly database research. This study will be focused on a period of 5 years and as recent as possible data will be used. Therefore data is collected from any point between 2008 and 2012. The sample is based on the Standard & Poor’s 500 (or shortly the S&P 500) which is the stock market index of the United States based on the market capitalizations of the 500 largest organizations. These large companies are listed on the New York Stock Exchange (NYSE), American Stock Exchange (AMEX) or NASDAQ. In order to compose the sample, data will be mainly collected from two databases namely Audit Analytics and Compustat. Data with regard to the level of audit fees and non-audit fees is collected from the Audit Analytics database. The revenues of the auditors are manually collected from global annual reports obtained through the Internet or Company.info (access is obtained through an internship at EY). The remaining data (information form the balance sheets, income statements, cash flow statements etc.) is collected from Compustat North-America Fundamental Annual Updates.

According to prior research, observations from financial industries, insurance industries and real-estate industries are excluded from the sample (Dhaliwal et al., 2008; Pittman & Fortin, 2004). “Explicit (or implicit) investor insurance schemes such as deposit insurance may strongly influence the credit decisions of firms such as banks and insurance companies” (Pittman & Fortin, 2004, p. 117). Firms in the S&P economic sector with code 800 (financials) are eliminated from the sample. This means that every company with a S&P industry sector code between 810 and 850 is excluded from the sample. These industry and economic sector codes are obtained through Compustat.

Since this study is going to test two hypotheses by means of two different composed regression models, two different samples of different sizes will be used. For the testing

variable NEW_COMPANY, which indicates whether an organization is in its first period after public offering, there were a lot of missing values that had to be excluded. Therefore different samples are used for both regression models. Only organizations that are listed on the S&P 500 index during the whole research period (2008-2012) are included in the samples, for example organizations that went bankrupt are excluded.

In order to test the two hypotheses separately, we collected two different datasets from Compustat North-America Fundamental Annual Updates. For the first dataset without the variable NEW_COMPANY (which measures whether an organization is in its first nine years after initial public offering), 2467 firm-year observations were collected from the Compustat database. All missing values are excluded from the sample and this is leading to a dataset of 1912 firm-year observations. After excluding the organizations that did not provide financial information for the whole period of 5 years (2008-2012), 1660 firm-year observations were left. Finally, financial industries, insurance industries and real-estate industries were excluded from the sample leading to a final first sample of 1650 firm-year observations. This sample is merged with the Audit Analytics dataset in order to include audit fees-related data. Table 1 summarizes the composition of the first sample:

Table 1: Composition of the first sample

Number of firm-year observations

2467 Collected from the Compustat database 1912 After excluding missing values

1660 After excluding companies that did not report all five years 1650 After excluding financial companies

Also for the second sample including the variable NEW_COMPANY, 2467 firm-year observations were collected from Compustat. All missing values are excluded from the sample and this is leading to a dataset of 590 firm-year observations. After excluding the organizations that did not provide financial information for the whole period of 5 years (2008-2012), 515 firm-year observations were left. Finally, financial industries, insurance industries and real-estate industries were excluded from the sample leading to a final second sample of 510 firm-year observations. Also the second sample is merged with the Audit Analytics

dataset in order to include audit fees-related data. Table 2 summarizes the composition of the second sample:

Table 2: Composition of the second sample

Number of firm-year observations

2467 Collected from the Compustat database 590 After excluding missing values

515 After excluding companies that did not report all five years 510 After excluding financial companies

Finally, the variables in both samples are winsorized at the 5th and 95th percentiles of the initial distribution. In order to overcome the effect of noisy proxies, it is often used in prior research to drop observations and this might decrease the effect of outliers (Dechow, 1994). This winsorizing means that every observation with a value smaller than the 5th percentile is replaced by the value at the 5th percentile and observations with a value larger than the 95th percentile are replaced by the 95th percentile.

3.2 Methodology

This study examines how auditor independence has an effect on the cost of debt and additionally how this effect is strengthened for new public organizations. The two formulated regression models include several variables of different kind of interest to test the two hypotheses. The models include a dependent variable, several testing variables and a couple of control variables.

3.2.1 Regression models

The first regression model tests the effect of auditor independence on cost of debt and the model consists of the following variables:

Model 1:

INTEREST = β0 + β1*RATIO + β2*NONAUDFEE + β3*TOTFEE + β4*BIG_FOUR + β5*SIZE + β6*LEVERAGE + β7*CASH_FLOW + β8*ASSET + β9*ROA + ε

The second regression model tests whether the effect of auditor independence on cost of debt is strengthened for organizations that are in one of the first nine years after initial public offering:

Model 2:

INTEREST = β0 + β1*RATIO + β2*NONAUDFEE + β3*TOTFEE + β4*NEW_COMPANY + β5*INTERACTION + β6*BIG_FOUR + β7*SIZE + β8*LEVERAGE + β9*CASH_FLOW + β10*ASSET + β11*ROA + ε

For both regression models, the dependent variable INTEREST and all control variables except for the variable ROA, are based on the study of Pittman and Fortin (2004). In their study is the association between debt pricing and auditor choice for newly public firms examined. They found statistically significant evidence that hiring a Big6 auditor affects firms’ cost of debt and younger firms might lower their interest rate by choosing a Big6 auditor (Pittman & Fortin, 2004). Hiring an auditor of higher quality (this implies a Big4 auditor) might reduce the level of financing costs for newly public firms but this effect declines over time when a firm has built their own reputation (Pittman & Fortin, 2004). This is leading to a lower level of information asymmetry. Pittman and Fortin (2004) measure the dependent variable based on the level of the interest rate. In order to measure auditor independence, prior literature used several ratio measures (for example Larcker & Richardson, 2004 and Dhaliwal et al., 2008). They calculated the fee ratio by dividing the amount of non-audit fees by the amount of total audit fees (Larcker & Richardson, 2004 and Dhaliwal et al., 2008). In order to focus on the importance of the client to the corresponding audit firm, Larcker and Richardson (2004) incorporated the size of the payment. Based on these scaled variables, I’ve chosen my independent variables for my regression models. Explanation of the variables:

• INTEREST = “Interest rate is interest expense divided by the average of total short- and long-term debt during the year” (Pittman & Fortin, 2004, p. 125).

• RATIO = “The ratio of non-audit fees to total fees paid to the auditor” (Larcker & Richardson, 2004, p. 630).

revenue of the auditor for that year” (Larcker & Richardson, 2004, p. 632).

• TOTFEE = “The ratio of total fees (both audit and non-audit) paid by the client of the firm to the total revenue of the auditor for that year” (Larcker & Richardson, 2004, p. 632).

• NEW_COMPANY = This variable indicates the age of an organization based on the initial public offering date (IPO) (Pittman & Fortin, 2004). This indicator variable has the value of one when an organization is in one of its first nine years after initial public offering and zero otherwise.

• INTERACTION = This interaction variable is the product of the variable RATIO (the ratio of non-audit fees to total fees paid to the auditor) and NEW_COMPANY (indicator whether an organization is in its initial years after public offering).

• BIG_FOUR = “This indicator variable has a value of one when the firm retains a Big4 auditor; zero otherwise” (Pittman & Fortin, 2004, p. 125) (In the research period of 2008-2012, we still have four big audit firms).

• SIZE = “Firm size is the natural logarithm of one plus total assets” (Pittman & Fortin, 2004, p. 125).

• LEVERAGE = “Leverage is the book value of total short- and long-term debt deflated by firm market value (sum of the market value of equity and the book value of total debt)” (Pittman & Fortin, 2004, p. 125).

• CASH_FLOW = “Cash flow is cash flow from operations scaled by total assets” (Pittman & Fortin, 2004, p. 125).

• ASSETS = “Asset structure is total property, plant and equipment scaled by total assets” (Pittman &Fortin, 2004, p. 125).

• RETURN ON ASSETS (ROA) = “Income before extraordinary items divided by total assets at the end of year t – 1” (Dhaliwal et al., 2008, p. 13).

3.2.2 Dependent variable

In both composed regression models, the cost of debt is the dependent variable. According to prior research, several proxies are used in order to measure the cost of debt: examples are the interest rate on a firm’s outstanding debt (Pittman & Fortin, 2004), yield spread (Datta et al., 1998) or the credit rating (Lennox and Pittman, 2011). In order to measure the independent variable cost of debt in this study, the model of Pittman and Fortin (2004) will partially be used and the variable INTEREST is created. Pittman and Fortin (2004) used the interest rate on the firm’s outstanding debt to measure cost of debt. They calculated the cost of debt as: “its interest expense for the year divided by its average short-term and long-term debt during the year” (p. 125). Organizations that don’t have an outstanding debt in a particular year, will report an interest rate of zero percent. Since the proxy interest rate is indicated as quite noisy, the extreme observations should be excluded from the sample (Pittman & Fortin, 2004). This resulted in winsorizing the 5th and 95th percentiles of the observations, which was already explained in the sample selection.

3.2.3 Independent variables

Furthermore, the model contains several testing variables to test the two hypotheses. As a proxy for auditor independence, audit fees, non-audit fees and total auditor fees will be used. Most of the time, prior literature measured auditor independence by calculating a fee ratio (RATIO), which is the ratio of the non-audit fees to the total audit fees (the total audit fees are the sum of the audit fees and non-audit fees paid to the auditor). However, prior literature has shown that mean (median) firms pay approximately the same amounts for audit and non-audit services (Frankel et al., 2002). According to Larcker and Richardson (2004), this ratio measure (RATIO) measures the linkage between the auditor and client, but this measure doesn’t pay attention to the size of the fee payments to the auditor firm. “That is, a client with $1 of audit and non-audit payments produces the same score as a client with $10 million of audit and non-audit payments” (Larcker & Richardson, 2004, p. 632). Larcker and Richardson (2004) use three measures in order to measure auditor independence. Therefore this study focuses on the importance of a client to the audit firm and measures this importance by calculating the variable NONAUFEE: “this is the ratio of non-audit fees paid by the client firm to the total revenue of the auditor for that year” (Larcker & Richardson, 2004, p. 632).

TOTFEE is calculated by using the total audit fees (audit and non-audit) to measure client importance, instead of only using non-audit fees.

Moreover, the relation between auditor independence and cost of debt for newly public firms will be tested. Like earlier explained in this study, prior research found that there’s a higher level of information asymmetry in younger firms compared to older firms (Diamond, 1989). In a younger firm, there’s less information available about an organization, which is resulting in no opportunity for having a longer credit record and organizations cannot build on their own reputation. Since lenders cannot rely on the reputation of the younger organization, they have to build on for example the quality of the auditor (e.g. whether the organization is audited by a Big4 auditor or not). In the initial years after public offering, the quality of the external auditor is quite important in relation to the level of financing costs. However, according to prior research this effect will decline over time when information asymmetry can be reduced due to the development of an organization’s own reputation (Pittman & Fortin, 2004). Cost of debt are expected to be higher in case of a new public company (Pittman & Fortin, 2004). The indicator variable (NEW_COMPANY) is added to the regression model in order to distinguish younger firms from older firms. This variable has the value of 1 when the organization is in one of its first nine years after public offering and 0 otherwise.

An interaction variable is created in order to test if the effect of auditor independence on cost of debt is different for organizations that are in one of the first nine years of public offering (so whether the organization is a company). This interaction variable (INTERACTION) is the product of the variables RATIO (“The ratio of non-audit fees to total audit fees paid to the auditor” (Larcker & Richardson, 2004, p.630)) and NEW_COMPANY and indicates whether the effect of auditor independence on financing costs is different for new organizations. The effect of auditor independence on cost of debt is expected to be higher for new organizations.

3.2.4 Control variables

The regression model of Pittman and Fortin (2004) also consisted of a number of explanatory control variables, like for example Big6 auditor. This variable indicates whether the firm retains a Big6 auditor. Since prior literature found that audit quality depends on auditor size, it might be useful to make a distinction between bigger and smaller audit firms and therefore this control variable will be included (DeAngelo, 1981b and Watts & Zimmerman, 1981). The variable BIG_FOUR can either take the value 1 when the organization is audited by a Big4 auditor and the value 0 otherwise. All of the observations in the sample are lying in the period

2008-2012. In this period the Big4 audit firms consist of EY (Ernst & Young), Deloitte, KPMG and PwC (PriceWaterhouseCoopers). Therefore the variable is named BIG_FOUR, instead of BIG SIX (Pittman & Fortin, 2004).

Furthermore, a couple of other control variables are included in the regression models. Firm-specific characteristics are included in order to the test for variation in cost of debt next to the variables for auditor independence and newly public organizations. Firm size (SIZE) is measured as the natural logarithm of one plus total assets (Pittman & Fortin, 2004). According Carey et al. (1993) an inverse relationship between firm size and interest rates is predicted, since larger firms are perceived to be less risky by creditors and debt production costs include economies of scale.

Leverage (LEVERAGE) is measured as the total short-term and long-term debt scaled by firm market value (Pittman & Fortin, 2004). Firms’ market value is calculated as the sum of market value of equity and the book value of debt (Pittman & Fortin, 2004). According to the research of Petersen and Rajan (1994), interest rates are supposed to be increasing in leverage since there will be an increase in interest rate if an organization uses more debt in order to finance its assets. They’re expecting a positive relationship between leverage (LEVERAGE) and a firms cost of debt (Petersen & Rajan, 1994).

The control variable CASH_FLOW is calculated as the cash flow from operations scaled by the total assets and can be used as a kind of performance measure to control for the profitability of an organization (Pittman & Fortin, 2004). According to prior research, it’s predicted that cash flows are negatively related to interest rates since organizations are generally better able to repay their debts when they’re able to generate more cash (Petersen & Rajan, 1994). Therefore the interest rate is expected to be lower in case of higher cash flows. The structure of the variable ASSET is calculated as total property, plant and equipment scaled by total assets (Pittman & Fortin, 2004). According to prior research, interest rates are suggested to increase in collateral (for example John et al., 2003). Collateral reduces the risk of the lender, since the lender can acquire a claim on the assets of the borrower without a reduction in debt payments. This suggestion is in accordance with the evidence found in the study of Morsman (1986). He formulated the perception in the banking industry that more risky borrowers have to provide security for their loans (Morsman, 1986). Therefore a

positive coefficient on control for asset structure is expected which implies that the higher the amount of property, plant and equipment, the lower the interest rate (Pittman & Fortin, 2004). Return on assets (ROA) is calculated as “Income before extraordinary items divided by total assets at the end of year t – 1” (Dhaliwal et al,. 2008, p. 13). This control variable in included as a control for profitability. Dhaliwal et al. (2008) included this control variable in their regression model and predicted a negative coefficient for return on assets (ROA), since more profitable organizations have a lower default risk and take advantage of lower borrowing costs. A negative coefficient for this variable is expected, since organizations are more likely to pay back their debt in case of favorable performances (Chang et al., 2009).

4. Results  

In this chapter the results of the study will be explained. In order to test the two hypotheses, two regression models are tested, which already are explained in the third section of this study. Like mentioned in the methodology part of this study, the final sample for the first model consists of 1650 firm-year observations and the final sample for the second model consists of 510 firm-year observations. These are the results after winsorizing the data at the 5th and 95th percentiles of the initial distribution (just like explained in the third section).

4.1 Descriptive statistics

4.1.1 Descriptive statistics regression model 1

Table 3 shows us an overview all variables used in the first regression model and provides the descriptive statistics. From this table we can conclude that on average the ratio between the interest expenses and the average total short- and long-term debt is 0.0608. This is resulting in an average interest rate of 6.08 per cent in the first sample with a standard deviation of 3.09 per cent. The variable RATIO explains the proportion of the total fees that is related to the non-audit fees. With a mean of 0.1805, one can conclude that 18.05 percent of the total audit fees comprises non-audit fees. The variable RATIO reports a standard deviation of 0.1160. The dummy variable BIG_FOUR shows whether the organization in audited by a Big4 audit firm. The mean of 0.9964 implies that 99.64% of the organizations are audited by a Big4 audit firm, which is very high. This result is quite logic to explain since this study focuses on S&P500 companies. The average LEVERAGE ratio is 0.2950 and this implies that the organizations in the sample consist on average more of assets than of debt. However, the maximum ratio of 0.8976 for LEVERAGE is quite high.

Table 3: Descriptive statistics model 1 (1650 firm-year observations)

Variable Mean Minimum Median Maximum Standard deviation

INTEREST 0.0608 0 0.0600 0.1156 0.0309 RATIO 0.1805 0.0197 0.1615 0.4193 0.1160 NONAUDFEE 0.0001 4.19e-06 0.0001 0.0006 0.0001 TOTFEE 0.0007 0.0001 0.0005 0.0025 0.0006 BIG_FOUR 0.9964 0 1 1 0.0602 SIZE 9.4046 7.6681 9.3528 11.2896 1.0304 LEVERAGE 0.2950 0.0424 0.2160 0.8976 0.2332 CASH_FLOW 0.1213 0.0375 0.1107 0.2473 0.0576 ASSET 0.3182 0.0493 0.2402 0.7768 0.2370 ROA 0.0679 -0.0219 0.0616 0.1683 0.0503

4.1.2 Descriptive statistics regression model 2

In table 4 we can see the descriptive statistics of all the variables that are used in the second regression model. In the second table with regard to the descriptive statistics, we can see that the ratio between the interest expenses and the average total short- and long-term debt is on average 0.0554. This results in an average interest percentage of 5.54 percent, which is a little bit lower than in the first sample and has a standard deviation of 0.0341. In both the first and the second sample the minimum value of the variable INTEREST is 0 and this means that the sample comprises organizations without outstanding debt in the particular year (Pittman and Fortin, 2004). The variable RATIO is one of the used measures in order to measure the level of auditor independence and shows which part of the total audit fees is related to the amount of non-audit fees. RATIO has a mean of 0.1623 and this implies that 16.23 percent of the total audit fees is charged for non-audit services. The new testing dummy variable NEW_COMPANY that is measured by means of the initial public offering date (IPO), has an average of 0.0686 and this means that 6.86 percent of the firm-year observations are in one of their first nine years after public offering. The dummy variable BIG_FOUR has again a very high average and this means that 98.82 percent of the organizations is audited by a Big4 audit firm. The explanation for this result is the same as the explanation for the first sample. The measured LEVERAGE ratio has an average of 0.2135 in this sample and this also implies that the organizations consist on average more of assets than of debt. The maximum value of the

LEVERAGE ratio is 0.6947 for the second regression model and is significantly lower compared to the maximum ratio of regression model 1 (0.8976).

Table 4: Descriptive statistics model 2 (510 firm-year observations)

Variable Mean Minimum Median Maximum Standard deviation

INTEREST 0.0554 0 0.0544 0.1156 0.0341 RATIO 0.1623 .0139 0.1434 0.3933 0.1117 NONAUDFEE 0.0001 1.89e-06 0.0001 0.0003 0.0001 TOTFEE 0.0005 0.0001 0.0003 0.0016 0.0004 NEW_COMPANY 0.0686 0 0 1 0.2531 INTERACTION 0.0046 0 0 0.0769 0.0177 BIG_FOUR 0.9882 0 1 1 0.1079 SIZE 8.8870 7.3588 8.8152 10.5755 0.9580 LEVERAGE 0.2135 0.0242 0.1539 0.6947 0.1844 CASH_FLOW 0.1344 0.0439 0.1225 .2577 0.0610 ASSET 0.2448 0.0350 0.1672 .6724 0.1934 ROA 0.0785 -0.0511 0.0757 .1951 0.0607 4.2 Differences in means

4.2.1 Differences in means regression model 1

For both samples, independent T-tests are performed in order to consider the differences in means. The outcomes of the independent sample T-tests might be an indication for the acceptance or not of the hypotheses. The first sample is used for regression model 1, which examines the effect of auditor independence on the cost of debt. The second regression model examines whether a mediating effect on the relationship between auditor independence and cost of debt can be measured and the second sample is used for testing this regression model. Looking at the initial public offering date of the organizations tests this mediating effect. Both sample one and two are split in two groups based on the mean of the dependent variable INTEREST. In order to assume whether the two groups have equal or rather unequal variances, Levene’s F-test is performed. For the first sample the mean of the interest expense divided by the average short-term and long-term debt during the year is 0.0608 (as can be read in table 3). This results in group 1 consisting of 843 firm-year observations with an average

interest rate smaller than 0.0608 and group 2 consisting of 807 firm-year observations with an average rate larger than or equal to 0.0608.

A lower level of auditor independence might increase the interest rate and this implies a negative relationship between auditor independence and cost of debt. However, auditor independence is proxied by the level of audit and non-audit services. This is leading to the conclusion that the higher the amounts of audit fees paid and the higher the values of the testing variables, the higher the interest expenses. Therefore a positive relationship between the proxies for independence and cost of debt is expected.

As we can read in table 5, almost all variables in model 1 have significant differences in means if we compare group 1 and group 2 with each other (tested at a significance level of 0.05). This means that one can conclude with a certainty of 95% that these means significantly differ from each other. All of the ratio measures that are used in order to measure the level of auditor independence differ significantly from each other if we compare the two groups.

However, when we look at the differences in means for the three ratio measures RATIO, NONAUDFEE and TOTFEE, we can see that the means are lower in the second group, compared to the first group (for example the variable RATIO has a mean of 0.1924 in group 1 and a mean of 0.1680 in group 2). Instead of the expected positive relationship that was expected, we can see a negative relationship here. This is inconsistent with the formulated hypothesis and this might be an indication for not adopting the first hypothesis. For the control variables LEVERAGE, CASH_FLOW, ASSET and ROA, the predicted sign is in accordance with the outcomes of the independent T-test (for example LEVERAGE has a predicted negative sign and has a higher mean in the second group, compared to the first group).

The means for the variable BIG_FOUR do not significantly differ from each other. For the variable BIG_FOUR is this non-significant outcome quite logic to explain. As we can read in the descriptive statistics in table 3 is 99.64 percent of the organizations audited by a Big4 company. This implies that the portion of observations that is audited by a Big4 company won’t differ that much from each other between the two groups (since the percentage of 99.64 is extremely high). This can be explained again by the fact that the sample consists of organizations listed on the S&P500.

Table 5: Differences in means model 1 (1650 firm-year observations)

Variable Mean group 1 INTEREST < 0.0608

Mean group 2 INTEREST >= 0.0608

Difference T-stat. Sig.

RATIO 0.1924 0.1680 0.0244 -4.299 0.000* NONAUDFEE 0.0002 0.0001 0.0001 -6.957 0.000* TOTFEE 0.0008 0.0006 0.0002 -7.168 0.000* BIG_FOUR 0.9900 1.0000 -0.0100 1.606 0.109 SIZE 9.3123 9.5011 -0.1888 3.753 0.000* LEVERAGE 0.2102 0.3836 -0.1734 16.168 0.000* CASH_FLOW 0.1357 0.1063 0.0294 -10.735 0.000* ASSET 0.2286 0.4118 -0.1832 16.914 0.000* ROA 0.0880 0.0469 0.0411 -18.228 0.000* * Significant at level 0.05

4.2.2 Differences in means regression model 2

Also for the second regression model, an independent sample T-test is conducted in order to test whether there are significant differences in means in the sample. Again, this test is performed to provide an indication for the acceptance or not of the second hypothesis. The results are presented in the sixth table. In this second regression model is tested whether the effect of auditor independence on cost of debt is stronger for younger organizations. In the second sample the mean of the interest expense divided by the average short-term and long-term debt during the year is 0.0554. This results in group 1 consisting of 262 firm-year observations with an average interest rate smaller than 0.0554 and group 2 consisting of 248 firm-year observations with an average rate larger than or equal to 0.0554. Again Levene’s F-test is performed in order to assume whether the two groups have equal or rather unequal variances. Both groups are compared with each other in order to test whether the variables significantly differ from each other and try to make a prediction in adopting the hypothesis. In the second regression model, there are less variables in the model that differ significantly from each other in comparison with the independent sample T-test of regression model 1. When we look at the differences in means for the three ratio measures RATIO, NONAUDFEE and TOTFEE, we can see that the mean is higher in the second group, compared to the first group or the means remained the same (for example the variable RATIO

has a mean of 0.1620 in group 1 and a mean of 0.1627 in group 2). However, the means do not significantly differ from each other and therefore we cannot provide an indication for adopting the second hypothesis. All control variables excluding the variable BIG_FOUR significantly differ from each other. For the variable BIG_FOUR the same explanation can be used that is earlier mentioned in this research.

In the second regression model, the mediating effect of newly public firms on the relationship between auditor independence and cost of debt is tested. According to Pittman and Fortin (2004), a negative relationship is expected. Younger organizations have no (good) reputation yet, resulting in higher interest rates. However, both variables NEW_COMPANY and INTERACTION don’t report a significant p-value. Besides, the outcome of the independent T-test reports a positive relation instead of the expected negative relation. Since there’s no significant difference in mean for those two variables, we cannot provide any indications whether the second hypothesis could be adopted or not. For the control variables LEVERAGE, CASH_FLOW, ASSET and ROA, the predicted sign is in accordance with the outcomes of the independent T-test (for example LEVERAGE has a predicted negative sign and has a higher mean in the second group, compared to the first group).

Table 6: Differences in means model 2 (510 firm-year observations)

Variable Mean group 1 INTEREST < 0.0554

Mean group 2 INTEREST >= 0.0554

Difference T-stat. Sig.

RATIO 0.1620 0.1627 -0.0007 0.076 0.940 NONAUDFEE 0.0001 0.0001 0 1.780 0.076 TOTFEE 0.0005 0.0005 0 1.887 0.060 NEW_COMPANY 0.0600 0.0700 -0.0100 0.343 0.732 INTERACTION 0.0044 0.0049 -0.0005 0.318 0.751 BIG_FOUR 0.9800 1.0000 -0.0200 1.605 0.109 SIZE 8.6667 9.1198 -0.4531 5.488 0.000* LEVERAGE 0.1511 0.2795 -0.1284 8.335 0.000* CASH_FLOW 0.1515 0.1163 0.0352 -6.820 0.000* ASSET 0.1859 0.3070 -0.1211 7.368 0.000* ROA 0.0981 0.0578 0.0403 -7.943 0.000* * Significant at level 0.05

4.3 Correlation matrix

4.3.1 Correlation matrix regression model 1

In table 7 the correlations between the dependent, independent and control variables are presented in the Pearson correlation matrix. The correlations are presented for the first regression model, which examines the relation between auditor independence and cost of debt. In the first column, the correlations are presented between the independent variable (column variable) and the dependent and control variables (row variables). From this first column we can conclude that every relationship is significant at the levels 5% and/or 1% respectively and that on average the correlation coefficients are not very high.

For each of the independent variables RATIO, NONAUDFEE and TOTFEE, the correlation matrix presents significant negative correlation coefficients (correlations of 0.105, 0.131, -0.132 respectively, each significant at a level of 1%). However, positive relationships between these variables and the dependent variable are expected and therefore this might be an indication that the first hypothesis cannot be adopted according to the results.

As mentioned before, on average the correlations are not very high in the first column: the highest correlations are between the control variables LEVERAGE, CASH_FLOW, ASSET and ROA and the dependent variable (correlations of 0.435, -0.322 0.444 and -0.485 respectively, each significant at a level of 0.01). For each of these control variables, the sign of the correlation coefficients (positive or negative correlation) is in accordance with the predicted sign in prior literature (Pittman & Fortin, 2004 and Dhaliwal et al., 2008). The control variable SIZE had a significant positive relationship of 0.147 and is significant at a level of 0.01. However SIZE was predicted to have a negative correlation (Pittman & Fortin, 2004).

Table 7: Pearson correlation matrix model 1 (1650 firm-year observations)

INTEREST RATIO NONAUD FEE

TOTFEE BIG_FOUR SIZE

INTEREST 1.000 RATIO -0,105*** 1.000 NONAUDFEE -0.131*** 0.628*** 1.000 TOTFEE -0.132*** 0.223*** 0.823*** 1.000 BIG_FOUR 0.052** 0.047* 0.018 -0.093*** 1.000 SIZE 0.147*** 0.027 0.490*** 0.638*** 0.069*** 1.000 LEVERAGE 0.435*** -0.099*** 0.008 0.100*** 0.009 0.287*** CASH_FLOW -0.322*** -0.025 -0.081*** -0.117*** -0.026 -0.259*** ASSET 0.444*** -0.234*** -0.235*** -0.201*** 0.039 0.273*** ROA -0.485*** 0.082*** 0.003 -0.048* 0.02 -0.255***

LEVERAGE CASH_FLOW ASSET ROA LEVERAGE 1.000

CASH_FLOW -0.502*** 1.000

ASSET 0.288*** -0.053** 1.000

ROA -0.614*** 0.661*** -0.288*** 1.000 *, **, *** Significant at 0.1, 0.05 and 0.01 levels respectively