UNIVERSITY OF GRONINGEN FACULTY OF ECONOMICS AND BUSINESS
Corporate Governance
and Audit Fees
The impact of regulatory oversight
Rita Alexandra HajdúS2251841
r.a.hajdu@student.rug.nl Supervisor: Dr. Dirk Akkermans
2 Abstract
I examine the relationship between internal corporate governance characteristics, such as the size and the diligence of the board of directors and the audit committee, and the amount of external audit fees. The aim of the thesis is to show the impact of regulatory oversight on the above mentioned relationship through a sample comprising financial and non-financial companies in the United States from 2000 till 2004. My results indicate that the association between audit fees and audit committee size and diligence is significant in case of non-financial firms before the Sarbanes-Oxley Act (SOX) in 2002. However, I find no significant evidence for the relationship between the board of directors and external audit fees. In case of a highly regulated business environment – financial firms through the entire sample and non-financial firms in the post-SOX era – the examined relationship is insignificant. My findings suggest that regulatory oversight abolishes the effects of corporate on external audit fees.
4 Table of Contents
1. Introduction ... 5
2. Literature Review and Hypotheses Development ... 7
2.1. Measures of Board of Directors’ Effectiveness ... 9
2.2. Measures of Audit Committee Effectiveness ... 12
2.3. Demand-based Perspective ... 15
2.4. Risk-based Perspective ... 16
2.5. Regulatory Oversight ... 18
3. Research Design ... 19
3.1. Sample and Data ... 19
3.2. Dependent Variable ... 20
3.3. Independent and Control Variables ... 21
3.4. Statistical Model ... 23
3.5. Estimation Method ... 24
3.6. Evaluation of Method Assumptions ... 24
4. Empirical Results ... 27
4.1. Descriptive Statistics ... 27
4.2. Regression Results ... 31
4.3. Robustness Check and Additional Analysis ... 38
5. Conclusions ... 39
References ... 42
5
1. Introduction
The highly publicized financial reporting frauds in the United States, such as Enron and WorldCom, have drawn attention to the audit profession. The auditor of these fraudulent firms was Arthur Andersen, which was among the most prestigious international accounting companies in the world (Big 5). The Enron Corporation went bankrupt in 2001, and Arthur Andersen collapsed in 2002. As a result of this event, the level of concentration in the auditing industry became higher than before, ending in four big firms. The role of the audit committee (AC), as a corporate governance mechanism against fraudulent financial reporting, came to the forefront, due to the above mentioned accounting scandals. The Sarbanes-Oxley Act (SOX) of 2002 was a rapid and uncompromising reaction by the United States to the accounting frauds. The main goal of the SOX was to restore investors’ confidence in financial reporting and to improve the quality of corporate governance (Krishnan & Visvanathan, 2009). This decisive legislation brought fundamental changes in corporate governance, particularly in the role of the audit committee and changed the corporate governance arena in the U.S. (Hoag & Hollingsworth, 2011).
Numerous researchers analyzed the connection between corporate governance characteristics and various accounting and auditing outcomes. In this paper I examine the relationship between certain board of directors and audit committee characteristics and one aspect of the auditor-management relationship, namely, the external audit fees, for a sample comprising industrial companies and financial firms. In general, prior research find a positive association between the board of directors and audit committee size, independence, diligence and expertise and external audit fees (Abbott, Parker, Peters, & Raghunadan, 2003b; Carcello, Hermanson, Neal, & Riley, 2002; Goodwin-Stewart & Kent, 2006; Lee & Mande, 2005; Zaman, Haniffa, & Hudaib, 2011). However, most of the papers excluded financial firms, due to their highly regulated nature. Therefore, in case of banks the direction of the above association cannot be derived straightly from the previous papers. In particular, this study examines the following research question: What is the impact of regulatory oversight on the relationship between various corporate governance factors and external audit fees?
6
by the Sarbanes-Oxley Act. Second, prior research oriented to non-financial firms, while this study includes bank holding companies to shed more light on the nature of the relationship between regulatory oversight, corporate governance and audit fees (Boo & Sharma, 2008b).
My tests are based on a sample of 30 non-financial and 30 financial firms in the United States from 2000 till 2004. I examine the relationship between board of directors’ and audit committee size and diligence, and external audit fees. My regression results for non-financial companies in the pre-SOX era indicate that audit committee size has a positive and significant effect on external audit fees; therefore bigger audit firms require more extensive auditing. This finding is in line with Zaman, Haniffa, & Hudaib (2011) results. However, audit committee diligence is negatively associated with audit fees, suggesting that more active audit committees mitigate the client risk, resulting in less extensive auditing. In terms of the characteristics of the board of directors I find no significant effect. This finding is inconsistent with prior literature that generally finds positive and significant association in case of low-level of regulation. One possible explanation for this outcome is that the thesis examines large firms exclusively, while prior research included small and medium-sized companies, as well.
To investigate the research question I compare the regression results for low-level and high-level of regulation. My findings show that audit committee size and diligence have a significant effect on external audit fees in case of low-level of regulation, while the coefficients become insignificant in the highly regulated business environment. Moreover, my results supports the assumption that in the pre-SOX era there is significant difference between financial and non-financial firms, though in the post-SOX era regulatory oversight eliminates this difference. These results are consistent with Boo and Sharma’s (2008a, 2008b) findings, although they do not examine the relationship after the SOX legislation.
7
2. Literature Review and Hypotheses Development
Before I discuss the two determining perspectives of the extent of external auditing and develop the hypotheses, it is necessary to establish a throughout understanding of the relationship between corporate governance and audit fees. In order to do so, I will first provide an introduction to the audit fee research. Then I review the measures of different corporate governance characteristics, such as the board of directors and the audit committee independence, expertise and diligence. This is followed by a discussion on the demand-based and the risk-based perspective, and the effect of regulatory oversight. Finally, I will conclude this section with the hypothetical specification of the relationship between internal corporate governance and external audit fees.
One of the most essential functions of corporate governance is to ensure the quality of the financial reporting process (Cohen, Krishnamoorthy, & Wright, 2004). The Public Oversight Board (POB, 1993) gave the following definition to corporate governance: "those oversight activities undertaken by the board of directors and audit committee to ensure the integrity of the financial reporting process". This paper focuses on the United States where the Anglo-American setting of corporate governance dominates, which is characterized by dispersed ownership and short-term shareholder commitment (Allen, Chui, & Maddaloni, 2008). The U.S. model is identified as a one-tiered system, where the board is composed of both executive and non-executive directors (Mallin, 2011).
8
responsibility; and (4) the audit committee must approve all non-audit services that are not specifically prohibited by SOX and exceed five percent of the total fees paid to the auditor firm (SOX, 2002).
The goal of this study is to show how regulatory oversight affects the impact of corporate governance on audit fees. In order to do so, it includes both financial and non-financial firms. These two groups vary since bank holding companies differ extensively from non-financial firms in several ways. Bank's liquidity production function is based on a maturity mismatch between the two sides of the balance sheet; they are highly leveraged institutions with rather opaque balance sheets, which is one of the reasons why most of the scholars exclude financial firms from their samples. Banks have a specific role in the economic system and are subjects to creditor runs. This threat means that, in case of a creditor run, they might go bankrupt, which makes their operation more risky than the operation of non-financial firms (Mülbert, 2009). Both the public and private firms in the industry must have an external auditor, which adds costs and complexity to the audit requirements in case of banks (Fields, Fraser, & Wilkins, 2004). In addition, banks are heavily regulated and supervised entities in the United States, both before and after the Sarbanes-Oxley Act in 2002. The regulatory and supervisory institutions for banks in the U.S. are the follows: the Securities and Exchange Commission (SEC), the Office of the Controller of the Currency (OCC), the Federal Deposit Insurance Corporation (FDIC), and the Federal Reserve System (FRS) for commercial banks, and the Office of Thrift Supervision (OTS) for thrifts.
The demand for external auditing derives from the desire to reduce the information asymmetries that arise from the separation of ownership and control (Jensen & Meckling, 1976). Due to the significant role of dispersed shareholders in the Anglo-American system, this asymmetry is especially noteworthy, since the most relevant agency conflict in the U.S is between owners and managers. Effective external auditing has been examined in several ways, such as the auditor type, auditor fees, going concern reporting and audit committee-auditor cooperation (Abbott et al., 2003b; Beasley & Petroni, 2001; Carcello et al., 2002; Carcello & Neal, 2003; DeZoort and Salterio 2001). In this study I focus on one aspect of effective external auditing, namely, the audit fees.
9
1996). In case of an outside auditor it is hard to observe the accurate level of work. In order to solve this problem scholars generally assume that, among others, the costs of a more extensive and overall auditing are billed to the client. After 2002, the structure of the auditor industry became even more concentrated with the fail of Arthur Andersen, the auditor of Enron and WorldCom. Learning from the case of Arthur Andersen, the remaining Big 4 firms (PricewaterhouseCoopers, Deloitte Touché Tohmatsu, Ernst & Young, and KPMG) apply more extensive auditing to identify financial and accounting frauds in order to protect their reputation. These increased efforts lead to higher audit fees. Given the above assumption researchers use the amount of the audit fees as an indicator of quality (Abbott et al., 2003b; Carcello et al., 2002; Goodwin-Stewart & Kent, 2006; Lee & Mande, 2005; Zaman et al., 2011).
2.1. Measures of Board of Directors’ Effectiveness
The board of directors is considered as the highest level control mechanism of a company (Fama & Jensen, 1983). Adams, Hermalin and Weisbach (2010) argue that, since 2000, there have been substantial changes in the structure of the boards, especially after the SOX, which requirements increased the workload of the outside directors. In their study they conclude that the "boards have become larger, more independent, have more committees, meet more often, and generally have more responsibility and risk" (p. 81). The effectiveness of the board is influenced by its characteristics, such as size, expertise, independence, diligence and CEO-Chairman duality. Table 1 summarizes the proxies used in prior literature to measure these characteristics.
10 Table 1
Measures of Board of Directors' Effectiveness Characteristic Proxies
Size Number of board members
Expertise Percentage of directors with relevant experience Percentage of directors who sit in multiple boards
Dummy variable indicating whether the board has a majority of experienced directors or not
Independence Percentage of outside (non-executive) directors
Dummy variable indicating whether the board has a majority of outside directors or not
Diligence Number of meetings held during the financial year
Dummy variable indicating whether the board is active or not
Duality Dummy variable indicating whether the CEO is the chairman of the board or not
The effects of expertise have been examined in several ways in prior literature. Since there is no exact definition for board member expertise, scholars use different proxies to capture this variable: percentage of directors with relevant experience, percentage of directors who sit in multiple boards, a dummy variable indicating whether the board has a majority of experienced directors or not. The importance of board member expertise is based on the monitoring and controlling role of the board (Cohen, Krishnamoorthy, & Wright, 2002). Directors, who sit in multiple boards, have more to lose in case of financial fraud; therefore they require more extensive monitoring (Cohen et al. 2002).
11
board of directors or the audit committee is independent if he or she is not an employee of the firm, a retired employee, a former employee, a relative of an employee, or a service provider engaged by the firm. Researchers use two kinds of proxies to measure board independence: one is the percentage of outside (or non-executive) directors, the other is an indicator variable showing whether the board has a majority of outside directors or not.
The diligence of the board can indicate its effectiveness. Carcello et al. (2002) argues that the more time the board members devote to fulfill their duties the more effective the board will be in terms of its crucial role, namely the monitoring. Moreover, more diligent boards seek to an increased level of financial reporting oversight; therefore they might purchase higher-quality audit services. On the other hand, if the board performs a satisfying monitoring role, than internal risk decreases and there is no need for extensive auditing. Prior research uses two proxies to measure the diligence of the board: the number of meetings held during the financial year and a dummy variable indicating whether the board is active or not.
An additional aspect of the analysis of the board of directors is leadership duality. In that case the power is concentrated in the hand of the CEO (who is also the chairman of the board), which may reduce the monitoring effectiveness and the independence of the board, and as a result the financial reporting quality, as well. In the presence of leadership duality managerial opportunism might raise, which results in higher demand for external auditing in order to protect shareholders and higher audit fees (Fama & Jensen, 1983). Tsui, Jaggi, & Gul (2001) present empirical support for the above assumption, specifically that there is a significant and positive relationship between leadership duality and audit fees. To measure this association, scholars use a dummy variable indicating whether the CEO is the chairman of the board or not.
12
To summarize, prior studies on non-financial firms find that board independence, expertise, diligence and leadership duality have a positive effect on audit fees (Abbott et al., 2003b; Carcello & Neal, 2000; Carcello et al., 2002; Krishnan & Visvanathan, 2009). Independent directors put greater emphasis on audit quality, since they are threatened by reputation damages. Moreover, those boards that meet more frequently are seeking to higher control over the financial reporting process (Carcello & Neal, 2000; Carcello et al., 2002). 2.2. Measures of Audit Committee Effectiveness
The audit committee is an integral part of a company's corporate governance, especially in terms of audit quality and oversight of financial reporting (He, Labelle, Piot, & Thornton, 2009). The goal of the audit committee is to maintain the quality of financial reporting in order to strengthen investor certainty in financial markets (BRC, 1999). The Sarbanes-Oxley Act (2002) defines audit committees as "a committee (or equivalent body) established by and amongst the board of directors of an issuer for the purpose of overseeing the accounting and financial reporting processes of the issuer and audits of the financial statements of the issuer" (p. 3). Levitt (2000) argues that ''audit committees play an indispensable role in changing those practices that have the potential to undermine the quality of financial reporting". Under the Sarbanes-Oxley Act, the AC is responsible for the whole external audit function, such as the selection, compensation, work and independence of the external auditor (Bédard & Gendron, 2010).
Table 2
Measures of Audit Committee Effectiveness Characteristic Proxies
Size Number of audit committee members
Expertise Percentage of AC members with financial expertise
Dummy variable indicating whether the AC has at least one member with financial expertise or not
Independence Percentage of outside (non-executive) directors
Dummy variable indicating whether the board has a majority of outside directors or not
Diligence Number of meetings held per year by the AC
Dummy variable indicating whether the AC is active or not
13
effectiveness. The Blue Ribbon Committee (1999) recommends that audit committees should have at least three members. Prior research suggests that a larger audit committee is more likely to enforce its will – high audit quality – within an organization, than a smaller AC (Zaman et al., 2011). Moreover, larger audit committees have more resources to identify fraudulent reporting, thus to improve audit quality.
The Public Oversight Board (1993) argues that the relevant, field specific expertise of the audit committee members is the most salient characteristic in terms of AC effectiveness. In the United States, the Sarbanes-Oxley Act (2002) requires that all audit committee members have financial literacy (ability to read and understand financial statements) and one member must be considered as a financial expert, suggesting that the regulators understood the importance of the association between expertise and audit committee effectiveness. The Securities and Exchange Commission also requires companies to determine whether at least one financial expert is a member of the audit committee and to disclose the name of that individual (Securities and Exchange Comission, 2003). According to Section 407 of SOX the following attributes are needed for the AC financial expert: “(1) an understanding of generally accepted accounting principles and financial statements and (2) experience in (a) the preparation or auditing of financial statements of generally comparable issuers; and (b) the application of such principles in connection with the accounting for estimates, accruals, and reserves, (3) experience with internal accounting controls, and (4) an understanding of the audit committee functions”. Abbott et al. (2003b) and DeZoort and Salterio (2001) argue that AC members with financial expertise provide greater support for the external auditor during the negotiation of auditing issues with the management, since they understand the risk associated with a lower quality audit. AC expertise is captured in two kinds of proxies: percentage of AC members with financial expertise and a dummy variable indicating whether the AC has at least one member with financial expertise or not.
14
Johnstone, 2004; Klein, 2002). Abbott et al. (200b), Carcello et al. (2002) and Lee and Mande (2005) argue that independent audit committees seek differentially higher level of audit quality, in order to avoid the undetected and unreported financial misstatements, which can harm their reputation. This greater expectation and assurance from the external auditor results in an increased amount of audit work, thus higher audit fees (Boo & Sharma, 2008). However, the audit risk perspective suggests, that more effective audit committees have greater oversight to the internal controls and financial reporting process (Klein, 2002). In that case effective ACs is expected to have a negative effect on external audit fees.
The fourth analyzed AC characteristic is meeting frequency, which is a signal of the audit committee diligence. Regular meetings serve as an opportunity for the audit committee to monitor audit quality. Prior research uses two proxies to measure AC diligence: first, the number of meetings held per year by the AC; second, an indicator variable showing whether the AC is active – meets at least four times a year – or not. The Blue Ribbon Committee suggests that the audit committee should meet at least four times a year in order to be effective. Lee and Mande (2005) and Krishnan and Visvanathan (2009) find a positive relationship between audit fees and audit committee diligence in their analysis of US firms. Goodwin-Stewart and Kent (2006) analyze Australian companies and find further evidence that higher audit fees are associated with more frequent audit committee meetings.
The majority of the research dealing with the association between AC characteristics and external audit fees found that higher quality audit committees are associated with higher audit fees, greater auditor independence and selection of a Big 4 auditor (Bédard & Gendron, 2010). Audit committee members suppose, that the selection of a high quality large auditor has a positive effect on the financial reporting quality, since Big 4 firms are more likely disclose the discovered errors than smaller or domestic auditor firms (Cohen et al., 2004). Abbott et al. (2003b) analyze the effects of the audit committee characteristics on audit fees for non-regulated firms. Their results support the prediction that more independent and expertise audit committees are requiring high quality audit, which results in higher audit fees.
15
audit work that is performed by external auditors. Strong corporate governance either increases the demand for external auditing resulting in higher audit fees or reduces the risk of the auditors, which leads to lower audit fees (Carcello et al., 2002). I review prior findings according to the demand-based and the risk-based perspectives before I propose my hypothesis.
2.3. Demand-based Perspective
The demand-based perspective argues that firms with strong corporate governance – effective board and audit committee - demand additional assurance and higher audit quality from the external auditors in order to preserve their reputation and avoid potential litigation, thus resulting in higher audit fees. This perspective suggests a positive association between effective boards and audit committees and external audit fees (Goodwin-Stewart & Kent, 2006). The majority of prior research on corporate governance and audit fees find evidence for the demand based perspective in case of non-financial firms in the pre-SOX era.
16 Figure 1. Demand-based perspective
Figure 1 illustrates the hypothesized relationship between the board of directors, the audit committee and the external audit fees. Given the above argument I propose four sub-hypotheses referring to the non-financial firms in the pre-SOX era according to the demand-based perspective:
H1a: According to the demand-based perspective there is a positive association between the board of directors’ size and external audit fees.
H1b: According to the demand-based perspective there is a positive association between the board of directors’ diligence and external audit fees.
H1c: According to the demand-based perspective there is a positive association between audit committee size and external audit fees.
H1d: According to the demand-based perspective there is a positive association between audit committee diligence and external audit fees.
2.4. Risk-based Perspective
17
Prior research shows that there is a significant and positive association between the strength of the board of directors and the audit committee and the quality of internal controls and financial reports (Beasley, 1996; Dechow, Sloan, & Sweeney, 1996; Klein, 2002; Abbott et al., 2003b; Krishnan, 2005). Given this argument, if the audit firm faces an effective board and audit committee, than they conclude that the internal control is working properly, which is leading to reduced audit effort, therefore lower audit fees (Boo & Sharma, 2008b). Cohen and Hanno (2000) find that in the presence of stronger government, auditors were more willing to reduce the amount of work and testing, thus resulting in lower audit fees.
According to the risk-based perspective, the board’s involvement in strengthening internal control may lead the external auditor to reduce the assessed level of control risk (Collier & Gregory, 1999; Tsui et al., 2001). Since auditors respond to client risks, they would consequently reduce their audit testing due to higher internal or regulatory control, leading to lower audit fees. The same conclusion can be drawn for the audit committees. More effective audit committees have greater oversight to the internal control and financial reporting process, leading to less extensive external auditing, resulting in lower fees (Klein, 2002).
Figure 2. Risk-based perspective
Figure 2 illustrates the hypothesized relationship between the board of directors, the audit committee and the external audit fees. Given the above argument I propose four sub-hypotheses referring to the non-financial firms in the pre-SOX era according to the risk-based perspective:
H2a: According to the risk-based perspective there is a negative association between the board of directors’ size and external audit fees.
18
H2c: According to the risk-based perspective there is a negative association between audit committee size and external audit fees.
H2d: According to the risk-based perspective there is a negative association between audit committee diligence and external audit fees.
2.5. Regulatory Oversight
Relatively high-level of regulatory oversight – direct monitoring and oversight by regulators – might have an effect on corporate governance, thereby changing the nature of the relationship between corporate governance and audit fees (Levine, 2004). The Sarbanes-Oxley Act of 2002 brought significant changes to the business environment of the U.S. and made it relatively highly regulated for both financial and non-financial firms. In the pre-SOX era non-financial firms are considered as a low regulated industry in the United States, since these companies had greater variance in their inner corporate governance measures. Financial firms assumed to be a highly regulated industry even before the SOX legislation.
Bryan and Klein (2005) argue that the regulatory oversight partially substitutes corporate governance in case of financial firms, thus reducing its importance, which corresponds to the results of Boo and Sharma (2008a). The same assumption holds for non-financial companies in the post-SOX era. If regulatory oversight is a perfect substitution of corporate governance, there will be no significant relationship between corporate governance characteristics and audit fees in the post-SOX era. Boo & Sharma (2008b) investigate audit committee and board characteristics, as well as leadership duality in the pre-SOX environment. They find a significant and negative relationship between audit committee independence and audit fees in case of bank holding companies, which underlies the audit risk perspective. However, none of the other variables are significant in the analysis of Boo and Sharma (2008b). These results support that firms' corporate governance and regulatory oversight are rather substitutes than complements.
19
2003; Bryan & Klein, 2005). Therefore, auditors decrease the amount of testing and rely more on the internal control mechanisms of the audited firm (Boo & Sharma, 2008a). Since the focus of this study is board and audit committee size and diligence, I expect that regulatory oversight eliminates the relationship between these variables and external audit fees:
H3: The association between board of directors/audit committee size and diligence and external audit fees is insignificant in case of regulatory oversight.
3. Research Design
In this section I first introduce the sample, which I use for the empirical test of the above specified hypotheses. Then I describe the dependent, independent and control variables. Following this I specify the statistical model and estimation method. In the last part of this section I present the empirical results.
3.1. Sample and Data
20 3.2. Dependent Variable
Prior research use external audit fees as an indicator of audit quality, since the costs of a more extensive, thus more accurate auditing are billed to the client (Abbott et al., 2003b; Carcello et al., 2002; Goodwin-Stewart & Kent, 2006; Lee & Mande, 2005; Zaman et al., 2011). Following the above assumption the dependent variable of this study is external audit fees, specifically the natural logarithm of audit fees. The purpose of using the natural logarithm of the external audit fees is to access a normally distributed dependent variable. "Having least squares estimators with normal or approximately normal distributions is important for the construction of interval estimates and the testing of hypotheses about the parameters of the regression model" (Hill et al., 2011, p.178). Table 3 shows the skewness and kurtosis test for normality for the audit fee variable and the natural logarithm of it. The results indicate that the basic dependent variable is not normally distributed, but the second variable (natural logarithm of audit fees) satisfies the normality criteria. Figure 3 shows that audit fees are concentrated in one rank of the histogram, which is the result of the sample selection – only large firms. In order to access a bell shaped normal distribution, I will use the natural logarithm of the audit fees. In Figure 4 the audit data is plotted against a theoretical normal distribution. The plotted line of the audit fees differing from the straight line. However, the plotted line of the natural logarithm of the audit fees approximately follows the straight one. Figure 3 and 4 clearly show the importance of the usage of the natural logarithm of the audit fees instead of the audit fees for the dependent variable.
Table 3
Skewness / Kurtosis tests for normality
Variable Observations Pr(Skewness) Pr(Kurtosis) Prob>Chi2
AUDFEE 300 0 0 0
21 Figure 3
Histogram of audit fees and natural logarithm of audit fees
Figure 4
Normal probability plot of audit fees and natural logarithm of audit fees
3.3. Independent and Control Variables
Explanatory Variables
I distinguish between two different elements of corporate governance: board of directors and audit committee. The explanatory variables of interest are size and diligence. The size of the board is measured by the number of board members in the financial year. A reasonably large board represents the ability for strong internal monitoring, which results in a reduced need for extensive external auditing, especially in case of highly regulated firms (Boo & Sharma, 2008b). The diligence of the board is measured by the number of meetings held by the board of directors in a financial year. Carcello et al (2002) argue that an increase in the number of meetings may lead to an increase in board effectiveness, since a more diligent board inquires a higher level of financial reporting.
0 1 .0 e -0 8 2 .0 e -0 8 3 .0 e -0 8 4 .0 e -0 8 D e n sit y
0 1.000e+08 2.000e+08 3.000e+08 4.000e+08
external audit fees
0 .1 .2 .3 .4 D e n sit y 12 14 16 18 20
the natural logarithm of external audit fees
0 .0 0 0 .2 5 0 .5 0 0 .7 5 1 .0 0 N o rm a l F [(A U D F EE -m )/ s] 0.00 0.25 0.50 0.75 1.00
Empirical P[i] = i/(N+1)
0 .0 0 0 .2 5 0 .5 0 0 .7 5 1 .0 0 N o rm a l F [(L AU D F EE -m )/ s] 0.00 0.25 0.50 0.75 1.00
22
The size of the audit committee can determine its effectiveness. The Blue Ribbon Committee (1999) recommends that the AC has at least three members. Prior research suggests that a larger audit committee is more likely to enforce its will – high audit quality – within an organization (Zaman et al., 2011). Moreover, larger audit committees have more resources to identify fraudulent reporting, thus to improve audit quality. The number of meetings held by the audit committee is a proxy for diligence. High frequency of the board and AC meetings can indicate a higher level of control within a company (Carcello et al., 2002).
Control Variables
Audit fee models predict that the prominent determinants of external audit fees are factors relating to size, complexity and riskiness of the audited firm (Carcello et al., 2002; Simunic & Stein, 1996). I use proxies for these characteristics as control variables in my regression model. Client size is measured as the natural logarithm of total assets (Carcello et al., 2002; Tsui et al., 2001). Consistent with prior studies (Carcello et al., 2002), I include the number of subsidiaries and the number of business segments as proxies for audit client complexity. Return on assets is a proxy for audit risk (Boo & Sharma, 2008a, 2008b; Tsui et al., 2001). I include non-audit fees because non-audit fees are associated with audit fees. From 2000, SEC registrants are required to disclose both audit and non-audit fees paid to their auditors. The Sarbanes-Oxley Act of 2002 imposed restrictions on non-audit services that audit firms can provide. Furthermore, it requires audit committees to pre-approve all non-audit service fees greater than 5 percent of the total fees paid to the incumbent auditor in the prior year. Non-audit fees paid to the Non-auditor are measured as the natural logarithm of total non-Non-audit fees (Abbott et al, 2003a; Lee & Mande, 2005).
Moreover, consistent with prior research I add other corporate governance variables, as controls, such a board composition, CEO-Chairman duality and audit committee independence.1 The board composition is measured by a percentage of outside, non-executive directors in the board. CEO-Chairman duality is a dummy variable indicating whether the CEO is also the chair of the board (1) or not (0). Audit committee independence is measured by the percentage of independent committee members. This variable has the less variation in
1 Audit committee expertise is excluded from the model due to data availability reasons. The SOX requires
23
the sample, since even before the SOX it was common to hire independent directors. Almost all of the sample firms have hundred percent independent audit committees in the pre-SOX era, as well. Moreover, the SOX made it compulsory to have entirely independent audit committees, so the AC independence variable is constant in the post-SOX environment. The dummy variable, financial, is set to 1 for companies in the financial industry and 0 otherwise. 3.4. Statistical Model
Drawing from prior research, I estimate the following two audit fee ordinary least square regression (OLS) models to test my hypotheses:
Where:
LAUDFEE = the natural logarithm of external audit fee.
SIZE = the size of the company based on the natural logarithm of total assets.
ROAA = return on average assets defined as operating income divided by total assets. SRSUB = square root of the number of subsidiaries.
SEGMT = number of business segments.
LNONAUD = natural logarithm of total non-audit fees.
FINANCIAL = an indicator variable equal to 1 when the firm is a financial company, 0 otherwise.
BDSIZE = the total number of directors on the board.
BDMEET = number of board meetings held in the financial year. BDCOMP = the percentage of independent directors on the board.
DUALITY = a dummy variable equals to 1 if the CEO is also chair of the board, 0 otherwise. ACSIZE = the total number of audit committee members.
24 3.5. Estimation Method
The relevant literature in the research field uses multiple regression technique in order to examine the relationship between audit fees and corporate governance characteristics. The purpose of the OLS model is to test the moderating effect of regulatory oversight on the relationship between external audit fees, and board and AC characteristics. Since prior research resulted in significant models and significant coefficients between the dependent and the independent variables, I use the same method for analyzing the association between financial and non-financial firms' external audit fees and board and AC size and diligence.
In order to eliminate the drawback of the fixed-effect OLS model that it does not allow the estimation oftime-invariant variables, I use the fixed effect decomposition method developed by Thomas Plumper and Vera E. Troeger (2007). They suggest a three-stage procedure for the estimation of time-invariant and rarely changing variables in panel data models with unit effects. "The first stage of the proposed estimator runs a fixed-effects model to obtain the unit effects, the second stage breaks down the unit effects into a part explained by the time-invariant and/or rarely changing variables and an error term, and the third stage re-estimates the first stage by pooled OLS (with or without autocorrelation correction and with or without panel-corrected SEs) including the time-invariant variables plus the error term of stage 2, which then accounts for the unexplained part of the unit effects" (Plumper & Troeger, 2007, p.1). However, it is a three stage method I only include the results of the last stage in the thesis, since the statistical program (STATA) allows to do the fixed effect decomposition method with one command and only shows the outcome of the final step. William Greene (2010) expressed strong criticism against the method of Plumber and Troeger. Greene argues that the efficiency gain from the fixed effect decomposition method is due to the fact that the OLS is consistent in the fixed effect model. However, prior research use the fixed effect decomposition method, thus I use it in order to measure the difference between financial and non-financial firms (De Jong, Phan, & Van Ees, 2011). To count with Greene’s criticism, I will conduct a Chow test as a robustness check in Section 4.3.
3.6. Evaluation of Method Assumptions
25
The assumption of homoscedasticity implies that the variance of the error term is constant across the dataset. In case the error variance for all observation is not the same, the dataset is considered heteroskedastic. However, heteroskedasticity does not lead to biased estimates of the coefficients; the standard errors are distorted,
which might lead to incorrect hypotheses tests, p-values and confidence intervals (Hill et al., 2011). In order to avoid misleading results I plot the residuals, which estimates the errors, against the firm size and I conduct a Breusch-Pagan test to
detect heteroskedasticity. Figure 5 shows that the variance is approximately similar for the different firm sizes, thus I conclude that heteroskedasticity is not an issue in the sample. In order to surely declare that the dataset is homoscedastic I run a Breusch-Pagan test (Table 4), which supports my expectations; therefore I determine that heteroskedasticity is not a problem in the dataset.
Figure 5
Graphical Analysis of Homoscedasticity
The second assumption of OLS estimation is that there is no serial correlation in the dataset. “Autocorrelation in the error can arise from an autocorrelated omitted variable, or it can arise if a dependent variable y is autocorrelated and this autocorrelation is
-1 0 1 2 3 R e si d u a ls 22 24 26 28
the size of the company based on the natural logarithm of total assets (in USD m
Table 4
Breusch-Pagan / Cook-Weisberg Test for Heteroscedasticity H0: Constant variance Variable: Size chi2(1) = 1.13 Prob>chi2 = 0.2874 Table 5
Wooldridge test for autocorrelation in panel data H0: no first-order autocorrelation
26
not adequately explained by the x’s and their lags that are included in the equation” (Hill et al., 2011, p.347). To detect autocorrelation on panel data, I conduct a Wooldridge test (Table 5). According to the results I have to reject the null hypothesis that there is no first-order autocorrelation in the dataset. In order to avoid misleading conclusions, I will correct for autocorrelation in the estimation method.
OLS assumes that the error term is uncorrelated with the independent variables, thus the explanatory variables are exogenous. Given this criteria OLS method can avoid the overestimation of the effect of the independent variables on the dependent variable. If endogeneity occurs and not corrected before the hypotheses tests than it leads to invalid conclusions (Hill et al., 2011). Two-stage least-squares
(2SLS) regressions with instrumental variables or generalized least square (GLS) estimations might give a solution to the problem of endogeneity. To detect the
possibility of endogeneity, I perform a Durbin–Wu–Hausman test (Table 6). The results show that the p-value is significantly high to conclude that the OLS estimation is consistent, thus the independent variables are exogenous.
The fourth assumption of OLS is that the values of the independent variables are not exact linear functions of other explanatory variables (Hill et al., 2011). If the criterion of multicollinearity is violated, meaning that the independent variables are collinear; it is difficult to isolate the effects of the explanatory variables on the dependent variable. Multicollinearity can be detected in two ways: the examination of the correlation matrix and calculation of the variance inflation factor (VIF). Table 10 in Section 4.1 shows the Spearman correlation among the dependent and the independent variables of the model. Since there is no exact collinearity (none of the correlation equals to one), moreover, none of the correlations are higher than 0.8, the OLS method can be applied. Furthermore, I conduct a VIF test for multicollinearity. The variance inflation factor shows how many times bigger the variance of the regressor will be for multicollinear data. Since all the computed VIF-values (Table 7) are
Table 6
Durbin-Wu-Hausman test for endogeneity laudfee_res = 0
F( 1, 298) = 0.37 Prob > F = 0.5414
Table 7
Test for Multicollinearity
Variable VIF 1/VIF
27
below the cut-off value of 10 suggested by Neter, Wasserman, Nachtsheim & Kutner (1985), I conclude that multicollinearity is not a problem.
4. Empirical Results
4.1. Descriptive Statistics
Table 8 and Table 9 provide descriptive statistics for the dependent and independent variables for the total sample and the financial and non-financial firms separately. The sample is equally distributed between the two types of firms and hundred percent of the sample are audited by the Big 5 auditors. The mean audit fee for the total sample is $US9.3, while it is $US7.12 million for financial firms, which is lower than the mean for the non-financial companies ($US11.5 million). The mean value of the total assets is $US187 billion for financial firms and $US90.6 billion for non-financial ones. The mean number of subsidiaries is significantly higher for financial firms (183.6) than for non-financial companies (107.1). In terms of business segments, there is no significant difference between the two groups. Consistent with audit fees, non-audit service fees are notably higher for non-financial firms than for financial companies.
28 Table 8
Descriptive statistics (n=300) Panel A: Continuous variables
Variable Mean Standard Deviation Minimum Maximum
Audit fees 9308721 11200000 149328 78200000 Total assets 139000000000 236000000000 2730000000 1480000000000 ROAA 4.3628 6.2474 -59.8200 23.0300 Subsidiaries 145.3500 265.9385 1 1826 Segments 3.9933 1.9629 1 14 Non-audit fees 13400000 19200000 43545 138000000 Board size 13.8733 3.7698 6 26 Board composition 79.3819 11.9714 11 100 Board diligence 8.3067 3.0699 3 24 AC size 4.6800 1.3919 3 11 AC independence 99.7778 2.7168 66.6700 100 AC diligence 7.6833 3.6009 0 16
Panel B: Dummy variables
Mean Standard Deviation Median Number of firms
coded '1'
Duality 0.6867 0.4646237 1 206
Financial 0.5000 0.5008354 1 150
Table 9
Descriptive data by financial and non-financial sub-sample (total n=300)
Financial (n=150) Non-financial (n=150)
Variable Mean Standard Deviation Mean Standard Deviation
29 Figure 6
Audit and Non-audit Fees (yearly average in USD)
Consistent with prior research the dataset indicates that financial firms incur lower audit fees than non-financial companies. Figure 6 reports the yearly average of audit and non-audit fees separating financial and financial firms. During the examined time period, non-financial firms pay substantially higher fees for both audit and non-audit services. The trend of the amount of fees paid for audit and non-audit services goes together, hence if the audit fees paid by financial firms increase, the amount billed to the non-financials also increases. A clear pattern can be seen in the figure: audit fees were relatively low before the SOX regulation, while non-audit services were relatively high. In the post-SOX era audit fees significantly and constantly increase, whereas non-audit fees decrease dramatically. The reason behind this phenomenon is that SOX requires more extensive auditing, which results in higher audit fees. Furthermore, after 2002 regulators limited the possible range of services, provided by the auditor firm to the client. This resulted in huge decrease in non-audit services and three of the Big 4 firms sold off their consulting practices between 2000 and 2002 (Charles et al., 2010). The data indicates that clients are fairly loyal to their chosen auditor firm. Among the sixty companies in the sample fifty-five did not change its auditor firm during the time period. From the remaining nine firms, five had to modify the auditor because of the failure of Arthur Andersen. Therefore, only four firms, less than 7% of the sample, changed its auditor due to managerial decisions.
0 5.000.000 10.000.000 15.000.000 20.000.000 25.000.000 30.000.000 2000 2001 2002 2003 2004
30 Table 10
Spearman Correlation Matrix
Natural logarithm of audit fees
Size ROAA Square root of
31
The Spearman correlation matrix for the whole sample is presented in Table 10. Appendix 1 and 2 presents the correlation matrix for the non-financial and financial firms separately. There is no correlation index above 0.8; therefore none of the independent variables are seriously correlated with each other. Consistent with my previous expectations the correlation between financial firms and audit fees is negative (-0.3513). Moreover, the following control variables are positively correlated with audit fees: the natural logarithm of total assets, the return on average assets, the square root of subsidiaries, the number of business segments and the natural logarithm of non-audit fees. These correlations are in line with my expectations drawn from prior research.
There is a positive correlation between board size and board diligence, which is in line with the theory and the hypotheses development. However, audit committee size is negatively correlated to AC diligence. This is contrary with my expectations, and might have an effect on the interpretation of the AC related hypotheses. This relationship remains negative for both the non-financial and the financial firms, meaning that larger audit committees meet less often than smaller ones. One possible explanation can be the sample selection of the thesis, namely the examination of large firms. If the audit committee has several members than the arrangements of the meetings become challenging, and the members rather contact through informal meetings and perform more non-meeting activities (Carcello & Neal, 2000). There are some differences between the correlation signs of the explanatory variables in case of non-financial and non-financial firms. These differences emphasize the variance between the two groups.
4.2. Regression Results
The theory presented in Section 2 suggests that non-financial firms considered as a relatively low-regulated industry in the US before the SOX. Therefore, consistent with the hypotheses development I only examine the non-financial firms in the pre-SOX era to decide whether the demand or the risk-based perspective dominates in the non-financial sample.
32
Company complexity - number of subsidiaries, number of business segments – is positively related to audit fees, although only the number of subsidiaries is significant (p<0.05). Non-audit fees have a positive and significant (p<0.01) effect on Non-audit fees in case of non-financial firms in the pre-SOX era. This result is in line with prior literature on non-financial firms (Abbott et al., 2003b; Carcello & Neal, 2000; Carcello et al., 2002; Krishnan & Visvanathan, 2009). The coefficient of return on assets has a positive sign, which corresponds to the demand-based perspective. The model is highly significant with a p-value of 0.000, and the goodness of fit is 78.98%.
Block 2 introduces the board variables and Block 3 presents the audit committee variables separately. The audit fee regression tests the hypotheses regarding the board of directors, namely that there is a positive (H1a & H1b) or a negative (H2a & H2b) association between the board of directors’ size and diligence and external audit fees. In the pre-SOX era, there is no significant relationship between the board characteristics and the external audit fees in the sample of the thesis.2 Therefore, H1a, H1b and H2a, H2b are not proved. One possible reason behind the insignificance of the board variables might be that the sample contains solely large firms. This factor influences the characteristics of internal corporate governance and the allocation of the responsibilities within a firm. In case of large firms, the board might assign more tasks to the audit committee, especially in terms of auditing. The positive (0.2501) correlation between board size and audit committee size in case of non-financial firms supports the above argument. This means that large boards are correlated with relatively large audit committees, resulting in higher AC responsibilities.
Block 3 presents the coefficients of the audit committee variables regarding the non-financial firms in the pre-SOX era. Conversely with the board variables, the coefficients for the audit committee variables are significant. The significance of AC variables also supports the previous discussion on the allocation of responsibilities within a large company. The results suggest that in case of large non-financial firms, the audit committee has a significant effect on the amount of the external audit fees, whereas the board of directors does not influence this relationship. However, the sings of the AC variable coefficients are ambiguous. Audit committee size and independence have a positive effect on external audit fees,
2 In order to explore whether this result holds through the total time period of the non-financial sub-sample, I
33
indicating that larger audit committees with independent directors seek to more extensive auditing in order to identify accounting frauds and protect themselves from a possible reputation loss (Abbott et al., 2003b; Bédard & Johnstone, 2004; Klein, 2002; Zaman et al., 2011). This result supports the demand-based perspective and proves H1a. Meanwhile, audit committee diligence has a negative coefficient on external audit fees. That means that a more diligent audit committee represents a stronger and more reliable factor in internal corporate governance; thus it lowers the client risk, resulting in less extensive auditing (Boo & Sharma, 2008b). The risk-based perspective supports these results, suggesting that in case of AC diligence, Hypothesis 2d is proved.
Table 11
Audit fee regression models for non-financial firms in the pre-SOX era (dependent variable is natural logarithm of audit fees)
Block Variables Controls Board of Directors Audit Committee
1 Constant 1.410 2.806 1.690 (1.971) (2.283) (2.373) Size 0.369*** 0.299*** 0.333*** (0.0940) (0.111) (0.0961) ROAA 0.00351 0.00457 0.00002 (0.0118) (0.0129) (0.0114)
Square root of the subsidiaries 0.0368** 0.0399** 0.0412**
(0.0167) (0.0168) (0.0166)
Segments 0.0102 0.0150 -0.0157
(0.0376) (0.0399) (0.0382)
Natural logarithm of non-audit fees 0.278*** 0.268*** 0.196*** (0.0724) (0.0785) (0.0758) 2 Board size 0.0249 (0.0309) Board composition -0.00108 (0.00799) Board diligence 0.0342 (0.0327) Duality -0.161 (0.198) 3 AC size 0.114** (0.0615) AC independence 0.0192 (0.0160) AC diligence -0.0835** (0.0457) N 60 60 60 F-statistics 92.19 102.74 100.21 Model p-value 0.000 0.000 0.000 R2 0.7898 0.7985 0.7774
34
In order to address the effect of regulatory oversight I use two methods: the first uses the fixed effect decomposition method to capture the effect of firm type (non-financial or financial); the second method separates the sample according to the level of regulation. Non-financial firms in the pre-SOX era are considered as low regulated companies, while non-financial firms in the post-SOX era and non-financial firms through the entire time period are considered as highly regulated companies.3
The results of the fixed effect decomposition method are presented at Table 12. The most important variable in this model for the purpose of the study is the financial indicator variable. Through the three blocks, the coefficient of the financial variable remains qualitatively similar and significant at a 1% level. This finding indicates that financial firms pay comparatively lower audit fees than non-financial ones. This result is in line with the theoretical argument that regulatory oversight is partially substitutes for extensive external audit monitoring (Bryan & Klein, 2005; Boo & Sharma, 2008b). Moreover, the external auditors might reduce their efforts due to the lower risk, which results in less extensive testing, leading to lower audit fees (Adams & Mehran, 2003). The coefficients of the board of directors' variables remain insignificant in the total sample, as well. On the other hand all the AC variables have a significant effect on audit fees in the regression. The fixed effect decomposition models are notably significant (p-value = 0.000), and the adjusted R2 is outstandingly high (93.5% for the board model and 95.2% for the AC model).
3 Boo and Sharma (2008a) use factor-analysis to capture the effect of regulatory oversight. In their study, they
35 Table 12
4
The number of observations reduces (from 300 to 240) in the regressions due the usage of the ar1 option which counts for autocorrelation in the dataset.
Audit fee regression models (n=300)
(dependent variable is natural logarithm of audit fees)
Block Variables Controls Board of directors Audit Committee
1 Constant -0.3547 -0.8027 -6.4477*** (4.056) (3.865) (1.4824) Size 0.778*** 0.794*** 0.875*** (0.228) (0.219) (0.0382) ROAA -0.00316 -0.00405 -0.000856 (0.00843) (0.00806) (0.00665)
Square root of the subsidiaries 0.0589*** 0.0593*** 0.0356**
(0.0175) (0.0171) (0.0155)
Segments 0.0926 0.110 0.0778**
(0.0690) (0.0761) (0.0385)
Natural logarithm of non-audit fees -0.244*** -0.237*** -0.183***
(0.0882) (0.0732) (0.0341) Financial -1.457*** -1.423*** -1.471*** (0.304) (0.260) (0.137) 2 Board size -0.00974 (0.0295) Board composition -0.00241 (0.00364) Board diligence 0.0147 (0.0141) Duality 0.105 (0.114) 3 AC size -0.0429* (0.0258) AC independence 0.0279*** (0.0105) AC diligence 0.0557*** (0.0165) N4 240 240 240 F-statistic 7.821146 8.653298 936.8373 Model p-value 0.000 0.000 0.000 Adjusted R2 0.935 0.935 0.952
36
The second method divides the sample according to low and high level of regulation and runs the board and audit committee regressions on the sub-samples. Table 13 presents the regression results. Two control variables change significantly between the two time periods (2000-2001 and 2002-2004). Return on average assets has a positive effect in case of low-level of regulation (non-financial firms in the pre-SOX era), supporting the demand-based perspective. However, in the presence of regulatory oversight the relationship between ROAA and external audit fees is negative, suggesting that profitable clients take a larger part from the total risk shared between the auditor and client, thus they are charged lower fees (Boo and Sharma, 2008b; Tsui et al., 2001). The other transformation in control variables appears in terms of non-audit fees. The SOX (2002) legislation made severe restrictions on the scope of non-audit services provided by the auditor firm. The regression results provide clear support for the effect of this limitation. Non-audit fees are significantly (p<0.01) and positively associated with audit fees in case of low-level of regulation, whereas this relationship becomes negative and insignificant in the highly regulated business environment. The coefficient on the financial indicator variable is insignificant in case of high-level of regulation, indicating that the SOX eliminated the difference between non-financial and financial firms.
The results show that in case of relatively low-level of regulation, there is no significant association between the board of directors’ characteristics and the external audit fees. In the highly regulated environment, the signs of the coefficients on the board variables change, except the one on board diligence, but the all the coefficients remain insignificant. The audit committee regression for the pre-SOX era results in significant AC size and diligence coefficients. Audit committee variables are significant for the pre-SOX era, but the SOX regulation diminished their effect on external audit fees. In case of high level of regulation none of these variables are significant5, supporting Hypothesis 3 that in the presence of regulatory oversight the association between board of directors and audit committee size and diligence, and external audit fees is insignificant. None of the corporate governance variables are significant in the post-SOX environment, indicating that regulatory oversight is a perfect substitute for corporate governance.
5 In the post-SOX era audit committee independence is omitted due to the fact that the SOX requires that audit
37 Table 13
Audit fee regression models for low-level and high-level of regulation (dependent variable is natural logarithm of audit fees)
Low-level High-level
Block Variables Board of
Directors Audit Committee Board of Directors Audit Committee 1 Constant 2.806 1.690 -2.7432* -2.306* (2.283) (2.373) (2.951) (2.159) Size 0.299*** 0.333*** 1.844*** 1.643*** (0.111) (0.0961) (0.412) (0.191) ROAA 0.00457 0.00002 -0.00440 -0.000958 (0.0129) (0.0114) (0.963) (0.479)
Square root of the subsidiaries 0.0399** 0.0412** 0.0580* 0.0450*
(0.0168) (0.0166) (0.162) (0.103)
Segments 0.0150 -0.0157 0.0700 0.0312
(0.0399) (0.0382) (0.744) (0.385)
Natural logarithm of non-audit fees 0.268*** 0.196*** -0.132 -0.131
(0.0785) (0.0758) (0.301) (0.238) Financial -1.6171 -1.666 (5.5077) (2.711) 2 Board size 0.0249 -0.0356 (0.0309) (0.159) Board composition -0.00108 -0.00773 (0.00799) (0.0206) Board diligence 0.0342 0.0236 (0.0327) (0.0763) Duality -0.161 0.217 (0.198) (1.534) 3 AC size 0.114** -0.0397 (0.0615) (0.189) AC independence 0.0192 (Omitted) (0.0160) AC diligence -0.0835** 0.0345 (0.0457) (0.0484) N 60 60 140 142 F-statistics 92.19 102.74 190.75 340.3286 Model p-value 0.000 0.000 0.000 0.000 R2 0.7898 0.7985 0.963 0.964
Standard errors in parentheses
38
4.3. Robustness Check and Additional Analysis
I conduct some sensitivity analyses to assess the robustness and the validity of my findings. First, in order to address Greene (2010) criticism on the fixed effect decomposition method I run separate regressions on the financial and the non-financial sub-samples. Appendix 4 presents the results of these regression models. The coefficients of the financial and non-financial firms are different, which indicates that the fixed effect decomposition method resulted in robust coefficients. The coefficient of the financial indicator variable is negative and significant in all models, which is in line with the results of these separate regressions. To further investigate this question I conduct a Chow test. With this method, I can test for the equivalence of two regressions, namely the financial and the non-financial models. The results of the Chow tests for the board (Table 14) and audit committee (Table 15) indicate that the coefficients of the two equations are different; thus the financial dummy’s coefficient in the previous regressions is supported.
Second, I check whether the clients of any one of the audit firms influenced the results. To do so, I form five separate sub-samples according to the Big 5 auditors. Appendix 5 and Appendix 6 present the outcome of this separation to the board and to the audit variables, respectively. In case of Arthur Andersen there is insufficient number of observations, thus I only conduct the regressions for the remaining four auditor firms. In case of Deloitte, the coefficient on board size become significant at the 5% level and audit committee size turn out insignificant. Beside this change, the regression results of these sub-samples are qualitatively similar to the reported ones. Therefore, I conclude that none of the audit firms influenced the results significantly.
Third, I consider an additional auditor related variable, indicating whether the contract between the client and the auditor firm is in the first year or not. Prior research suggest that auditors generally charge lower fees in the first year of cooperation, regardless of the fact that initial engagements involve high start-up costs (DeAngelo, 1981; Simon & Francis, 1988). When I include this dummy variable in my audit fee models, it is significant in case of the control and the audit committee models, but the original coefficients remain substantially the Table 15
Chow test for Model 2 ( 1) acsize1 - acsize2 = 0 ( 2) acmeet1 - acmeet2 = 0 ( 3) group1 - o.group2 = 0 chi2( 3) = 13.35 Prob > chi2 = 0.0039 Table 14
Chow test for Model 1 ( 1) bdsize1 - bdsize2 = 0 ( 2) bdmeet1 - bdmeet2 = 0 ( 3) group1 - o.group2 = 0 chi2( 3) = 12.42
39
same. Appendix 7 shows the results of the regression models including the ‘first year of the contract’ indicator variable. The coefficient of this additional variable is negative and significant at a 5% level in case of the audit committee model. Moreover, audit committee size becomes insignificant, but AC independence and diligence remains significant at a 5% and at a 1% level, respectively. The signs of the coefficients do not change. These results support the assumption of DeAngelo (1981) and Simon and Francis (1988) that audit firms apply lower fees in the first year of the contract, than later on.
Finally, I introduce an audit committee effectiveness variable suggested by Abbott et al. (2003b). This joint variable is a dummy variable, equal to 1 if the audit committee is composed entirely of independent variables and meets at least four times during the financial year, and 0 otherwise. I include this variable into the audit fee regressions instead of AC independence and AC diligence (Appendix 8). Audit committee effectiveness has a positive and significant (p<0.1) effect on external audit fees, such as independence and diligence separately. However, this indicator variable is less significant than the audit committee variables independently. Beside the fact, that AC size becomes insignificant, the results of this model are qualitatively similar to the original one. The robustness tests suggest that AC diligence is highly stable – it is the only AC variable that has the same sign in all the models presented in the thesis – both in terms of the value and the sign of the coefficient, but the significance of AC size is sensitive to the additional independent variables.
