Effect of regulatory oversight on the association between internal governance characteristics and audit fees

2.1. Corporate governance and audit fees

The statutory audit is an important governance mechanism through which shareholders monitor management, given the separation of ownership from management. Results on the association between corporate governance and audit fees are generally mixed in prior studies. Studies that focus on corporate governance risk argue and find that corporate governance is associated with lower audit fees (e.g. Bedard and Johnstone, 2004). This stream of literature proposes that external auditors consider the quality of monitoring provided by internal governance in their risk assessment and audit planning. Bedard and Johnstone (2004) argue that external auditors perceive greater director independence is associated with stronger internal controls and lower risk of financial misreporting. Consequently, external auditors are likely to reduce the extent of audit tests.

In contrast, studies that adopt a demand-based perspective argue that directors with greater reputation capital at risk demand a more extensive audit to protect their reputation capital and reduce the risk of litigation (e.g. O'Sullivan, 1999, 2000; Carcello et al., 2002; Abbott et al., 2003; Goodwin-Stewart and Kent, 2006). This stream of literature uses director independence as a measure of directors’ desire to protect their reputation capital. Therefore, a potential contribution to the literature relates to alternative explanations of the effects of board/audit committee independence on audit fees. We extend the literature by dissecting apart the effects of board/audit committee independence, reputation and regulatory oversight on audit fees. As elaborated later in the present paper, we theoretically and empirically distinguish the constructs of board/audit committee reputation from independence and their effects on audit fees. In doing so, we address concerns raised by DeFond and Francis (2005) to develop theoretical explanations regarding how and why the board and audit committee influence the financial reporting and audit processes.

Following recent studies (Goodwin-Stewart and Kent, 2006; Yatim et al., 2006) and as the audit committee is taking a more active and independent role in selecting the auditor and monitoring the audit process, we include characteristics of the audit committee in addition to those of the board. Cohen et al. (2002) suggest that as the audit committee needs the strong support of the board to discharge its duties effectively, examining audit committee effects without the board introduces confounding effects. We investigate three constructs of the board of directors and the audit committee – independence, reputation and size – in the context of regulatory oversight.

2.2. The effect of regulatory oversight

Cohen et al. (2004) present a corporate governance framework that encompasses both internal monitoring mechanisms (e.g. corporate board, audit committee, and internal and external auditing) and external monitoring agencies (e.g. regulators, analysts and stockholders). Their framework suggests the interrelationships among the internal governance agencies are affected by outside forces, such as regulators. The existence of monopoly power, externalities and informational asymmetries create a potentially constructive role for government regulation to reduce market failures and enhance social welfare. Because financial institutions and utility companies are highly complex businesses and enjoy market and political power, they are difficult to monitor. Absence of effective monitoring can lead to expropriation and market failure (e.g. bank runs, when depositors withdraw money from banks because they think they will fail, could cause a domino effect on other industries). Such concerns favour government intervention to regulate economic activities (Pigou, 1938; Stigler, 1971) as exemplified by industry-specific regulation of financial institutions (Sierra et al., 2006) and utility companies (Russo, 1992).

Government regulatory oversight could impact the demand for additional assurance from the external auditor. Prior research (e.g. Dunn and Mayhew, 2004; Bryan and Klein, 2005) suggests that regulatory oversight provides close monitoring that could potentially reduce the role of external auditing as a control mechanism. Financial and utility companies are subject to closer regulatory oversight as they are essential services and they play a critical role in economic development. This and other differences between the nature of regulated and non-regulated companies have led some to view regulatory oversight of the industry as a substitute for corporate governance, or at least to view internal governance as less critical to the conduct and operation of regulated companies (Stoll, 1998; Adams and Mehran, 2003). Although we do not subscribe to the view that regulatory oversight reduces the importance of corporate governance, we believe that regulatory oversight could affect board/audit committee's demand for external audit monitoring. In other words, as stricter requirements are imposed on regulated companies, economic theory suggests that regulated companies’ investment in compliance and risk-reduction strategies could accommodate a reduced level of costly external audit monitoring.

Given the substitution effect of regulatory oversight for external auditing (Dunn and Mayhew, 2004; Bryan and Klein, 2005), we posit that the extensiveness of the external audit is lower in regulated companies with stronger corporate governance than non-regulated companies. However, the demand for the extent of the external audit is dependent on the risks and incentives facing the directors serving on the board and audit committee. Similarly, the level of audit effort applied by the auditor is dependent on the auditor's assessment of client-related risks. We develop theoretical propositions from demand-based and risk-based perspectives and hypothesize how regulatory oversight moderates the association between audit fees and board/audit committee independence, multiple directorships and size.

2.3. Board/audit committee independence

Management has incentives to misrepresent financial results for opportunistic reasons given shareholder–manager conflict of interests (Jensen and Meckling, 1976). The potential for such opportunistic behaviour could be reduced if there is monitoring by independent directors (Beasley et al., 2000; Carcello and Neal, 2000, 2003; Carcello et al., 2002; Abbott et al., 2004). The literature suggests that independent directors are effective monitors as they do not have financial interest in the company and psychological ties to management. Therefore, they are in a better position to objectively challenge management (e.g. Beasley, 1996; Klein, 2002; Abbott et al., 2004) and support the auditor. Carcello et al. (2002) and Abbott et al. (2003) support this view by providing empirical evidence of an association between the proportion of independent directors and higher audit fees. Bedard and Johnstone (2004) also argue that higher independent director representation on the board and audit committee provides more vigilant oversight of the financial reporting process. This view is corroborated by prior studies that investigate the relationship between director independence and financial misreporting (e.g. Beasley, 1996; Klein, 2002; Abbott et al., 2004; Anderson et al., 2004).

We extend the literature by examining how regulatory oversight affects the association between board/audit committee independence and audit fees from both demand and supply perspectives. From a demand perspective, we posit that independent boards/audit committees in regulated companies have a lesser need for external audit monitoring given the partial substitution of regulatory oversight for external auditing (Dunn and Mayhew, 2004; Bryan and Klein, 2005). From a supply perspective, external auditors perceive lower informational asymmetries and risk of financial misreporting in regulated companies given stringent regulatory oversight and increased controls. By relying on effective external and internal monitoring processes, auditors could reduce the extent of costly testing procedures in regulated companies. Therefore, we expect regulatory oversight to weaken the association between independent board/audit committee and higher audit fees:

2.4. Board/audit committee multiple directorships

Fama and Jensen (1983) contend that multiple board appointments signal director quality. Kaplan and Reishus (1999), Vafeas (1999) and Abbott and Parker (2000) likewise suggest that the number of directorships is a measure of a director's reputation capital. These authors reason that directors effective in their monitoring of management influence firm performance and are consequently offered more board seats. Ferris et al. (2003) provide more recent evidence of this association.

To protect their reputation capital, directors with multiple directorships demand additional assurance from the auditor. Directors serving on multiple boards are more visible in the directorate market, have more responsibilities across companies and, hence, face higher risk of litigation (Carcello et al., 2002). Multiple directorships could also reduce directors’ time and focus they could devote to each board; hence, constraining their monitoring effectiveness (Shivdasani and Yermack, 1999). To compensate for this constraint, such directors demand more assurance from the auditor. From a supply perspective, auditors are likely to perceive audit risk to be higher and consequently perform more audit work when directors dilute their attention and focus by serving on multiple boards.

The ramifications of non-compliance and misreporting are greater for regulated industries as failure of regulated companies, such as banks and utilities, results in significant losses to investors, and disrupts social welfare and the economy (Tadesse, 2006). Regulatory oversight increases both the likelihood of detection and the gravity of consequences arising from directors’ failure in discharging their governance responsibilities. Boards/audit committees in regulated companies have greater incentives to protect their reputation and they can do this by demanding additional assurance from the auditor. As the consequences of an audit failure are severe in regulated companies, the external auditor is likely to design and execute a more extensive audit. If auditors perceive multiple directorships to increase client-related risks as the directors devote insufficient time to their role, then the risk-based perspective suggests that the auditors would perform more extensive audit tests. The demand and risk-based perspectives together suggest that regulatory oversight strengthens the association between board/audit committee multiple directorships and higher audit fees:

2.5. Board/audit committee size

Jensen (1993, p. 865) states that: ‘as groups increase in size they become less effective because the coordination and process problems overwhelm the advantages from having more people to draw on’. Larger boards are less effective monitors because of potential free riding, communication breakdowns and inefficiencies (Dechow et al., 1996; Bushman et al., 2004). Yermack (1996) provides empirical evidence consistent with these views. He shows that companies with smaller boards have higher market values. Beasley (1996) similarly finds that the likelihood of fraud increases with board size. Drawing on this literature and consistent with the risk-based perspective, we posit that larger boards/audit committees heighten the risk of material misstatements; hence, requiring a more extensive audit. Therefore, we expect larger board/audit committee size to increase audit fees.

We acknowledge that the size–audit fees association can be non-linear as extremely small boards/audit committees lack a sufficient pool of expertise to govern effectively (Vafeas, 2000). Unreported results confirm that only 3.0 per cent of our sample companies have fewer than 7 board members (minimum and median of 5 and 11 members, respectively) and 3.6 per cent have fewer than 3 audit committee members (minimum and median of 2 and 4 members, respectively). Given that boards/audit committees are adequately staffed in our sample comprising large, listed companies, it is not empirically meaningful to test the notion of non-linearity in our study.

The presence of strong internal monitoring and external regulatory oversight mitigate informational asymmetries and agency problems; hence, diminishing the need for an extensive audit (Smith and Watts, 1992; Bryan and Klein, 2005). The substitution effect of regulatory oversight for external auditing suggests regulated companies with larger boards/audit committees are associated with lower audit effort. Therefore, we propose that regulatory oversight weakens the association between board/audit committee size and higher audit fees:

3.1. Model specification

Drawing from prior literature (e.g. Craswell and Francis, 1999; Tsui et al., 2001; Carcello et al., 2002), we estimate the following audit fee ordinary least-squares regression (OLS) model to test our research hypotheses:

where:

The purpose of the OLS model is to test the moderating effect of regulatory oversight on the association between audit fees and board/audit committee independence (IND), multiple directorship (ODIR) and size (SIZE). We investigate whether the coefficients on the interaction terms are statistically significant in the expected directions. The moderator variable, REG, is set to 1 for companies in regulated industries (i.e. financial and utility companies), and 0 otherwise. Consistent with prior research (e.g. Beasley, 1996; Klein, 2002), a member of the board or 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 (e.g. lawyer) engaged by the firm.

We extract variables relating to both the board and audit committees from proxy statements. We then factor-analyse these variables (deriving varimax-rotated factor scores) for three reasons. First, Cohen et al. (2004) argue that board and audit committee characteristics should be considered together because of considerable interactions and overlap between the board and audit committee. Second, the board and audit committee variables are highly correlated.² Third, we reduce the six governance variables to derive a parsimonious model. The three factors extracted with eigenvalues greater than 1 account for 80 per cent of the total variance. The board/audit committee independence factor explains 24 per cent, the multiple directorship factor explains 31 per cent, and the board/audit committee size factor explains 25 per cent of the total variance. All the factor loadings are greater than 0.82.

As audit fees are predominantly influenced by size, complexity and risk of the audit client, we use proxies for these characteristics as control variables in our regression model. Client size is measured as the natural logarithm of total assets (e.g. Craswell and Francis, 1999; Tsui et al., 2001; Carcello et al., 2002). Consistent with prior studies (e.g. Carcello et al., 2002), we include the number of subsidiaries, the number of business segments and the proportion of foreign subsidiaries as proxies for audit client complexity. Return on assets and loss-making in any of the last 3 years are proxies for audit risk (e.g. Craswell and Francis, 1999; Tsui et al., 2001). We include non-audit fees because non-audit fees are significantly associated with audit fees (Whisenant et al., 2003; Hay et al., 2006). Non-audit fees paid to the auditor are measured as the natural logarithm of total non-audit fees (e.g. Davis et al., 1993; Whisenant et al., 2003). BIG5 controls for differences in auditor quality (e.g. Craswell and Francis, 1999; Tsui et al., 2001; Carcello et al., 2002; Clarkson et al., 2003).³

3.2. Data

Our sample comprises 469 US listed companies with total assets exceeding $US1bn in the fiscal year 2001. We use a more homogeneous sample restricted to larger companies because larger regulated companies are subject to intense scrutiny and the prior literature indicates that company size is related to corporate governance (Abbott et al., 2003). We limit our sample to 2001 to minimize the regulatory effects of the Sarbanes-Oxley Act of 2002. We acknowledge our results might not be generalizable beyond our sample period and to smaller companies. We hand-collect the corporate governance data from proxy statements filed with the Securities and Exchange Commission. Based on our selection criteria, we obtain 252 regulated companies comprising 32 utilities and 220 bank holding companies⁴ from the Standard & Poor's (S&P) 500 and Federal Deposit Insurance Corporation (FDIC)⁵ listings, respectively, and 217 non-regulated companies⁶ from the S&P 500 list.

3.3. Descriptive statistics

Table 1 provides descriptive statistics for the continuous and indicator variables in Panels A and B, respectively. Fifty-four per cent of the sample are regulated companies. Ninety-five per cent of the sample are audited by the Big Five auditors and 16 per cent experience a loss during the prior 3 years. The mean (median) audit fee is $US1.73m ($US0.78m) and it ranges from $US0.05m to $US21.6m. The mean (median) book value of total asset is $US20.5bn ($US5.8bn). On average, the sample companies have audit committees and boards that are predominantly independent (97 and 78 per cent, respectively). On average (median), audit committee members sit on 1.50 (1.33) other boards, and board directors sit on 1.46 (1.50) other boards. The mean (median) audit committee size is 4.20 (4) and mean (median) board size is 11.54 (11).

Table 2 shows t-test results for differences between the regulated and non-regulated companies. Results in Table 2 generally show significant differences between the two groups except for size, measured by the logarithm of total assets, and profitability, measured by return on assets. Consistent with prior studies (e.g. Simunic, 1980; Maher et al., 1985; Turpen, 1990; Gist, 1994; Hay et al., 2006), regulated companies incur lower audit fees than non-regulated companies. This could also be explained by regulated companies having fewer subsidiaries and business segments and lower proportion of foreign subsidiaries. Although regulated companies have larger boards and audit committees, they have fewer independent directors. Directors in regulated companies, on average, sit on fewer outside boards.

Consistent with our expectations, the correlation matrix in Table 3 shows that the correlation coefficient between regulatory oversight (REGUL) and audit fees is negative and significant. Except for return on assets, all control variables are significantly correlated with higher audit fees. None of the explanatory variables are highly correlated (r > 0.80), which suggests that multicollinearity is not a problem (Neter et al., 1987). In addition, the maximum variance inflation factor (VIF) of 4.96 in the OLS is less than 10, which substantiates the absence of multicollinearity (Neter et al., 1987).

4.1. Multivariate statistics

Table 4 reports results of the regression analyses. The adjusted R² of all three models are at least 86 per cent. With the exception of RA and LOSS, the sign and significance (p < 0.05 or lower) of the coefficients on the control variables across all three models are consistent with prior research. LOSS is significant at p < 0.05 in Model 3 and p < 0.10 in Models 1 and 2. Overall, these data suggest the OLS models are structurally stable.

Model 1 regresses audit fees (LAUDFEE) on eight control variables and the regulatory indicator variable (REG). After controlling for client-related risks, client size and complexity, we find a negative and significant (p < 0.01) coefficient on REG. This suggests that regulated companies pay comparatively lower audit fees than non-regulated companies. This is consistent with the argument that regulatory oversight partially substitutes for external audit monitoring; hence, reducing the demand for an extensive audit. In addition, the external auditors might reduce the extent of audit tests as they perceive stronger internal controls and internal monitoring in regulated companies.

Model 2 introduces the board/audit committee independence (IND), multiple directorships (ODIR) and size (SIZE) variables. The coefficients on IND and ODIR are positive and significant (p < 0.05 or lower), whereas the coefficient on SIZE is not significant. The coefficient on REG remains negative and significant. These findings are consistent with the demand-based argument that independent boards/audit committees and directors with higher reputation capital demand additional assurance from the auditor to ensure effective oversight over financial reporting and to protect their reputation capital (Carcello et al., 2002; Abbott et al., 2003). Because multiple directorships could also reduce directors’ time and focus they could devote to each board and, hence, constrain their monitoring effectiveness (Shivdasani and Yermack, 1999), such directors are likely to demand more assurance from the auditor. In addition, auditors are likely to perceive audit risk to be higher and consequently perform more audit work when directors dilute their attention and focus by serving on multiple boards.

Model 3 tests our hypotheses and introduces interaction terms. The coefficient on the interaction term IND_REG (board/audit committee independence by regulatory oversight) is negative and significant (p < 0.05). This result is consistent with our expectation in Hypothesis 1, which posits industry-specific regulatory oversight diminishes the association between board/audit committee independence and audit fees. The coefficient on the interaction term ODIR_REG (multiple directorship by regulatory oversight) is positive and significant (p < 0.01). This result is consistent with our expectation in Hypothesis 2, which posits regulatory oversight strengthens the association between board/audit committee reputation and audit fees. Finally, the coefficient on the interaction term SIZE–REG (board/audit committee size by regulatory oversight) is negative and significant (p < 0.01). This result is also consistent with our expectation in Hypothesis 3, which posits regulatory oversight diminishes the association between the size of the board/audit committee and audit fees.

4.2. Sensitivity analyses

We conduct several sensitivity analyses to assess the robustness of our findings. First, our results remain qualitatively similar after excluding outliers and influential points.⁷ Second, we test whether our results are sensitive to possible endogeneity between audit fees and non-audit fees. Prior evidence suggests that audit and non-audit fees could be jointly determined (Whisenant et al., 2003; Hay et al., 2006). Results of our two-stage least-squares show that the coefficients on SIZE–REG and ODIR–REG remain significant in the expected directions (p < 0.01, one-tailed), but the coefficient on IND–REG is not significant. Except for board/audit committee independence, our results are robust to the joint determination of audit and non-audit fees.

Third, we estimate our OLS models by including variables representing director independence, multiple directorships and size, and their interactions with regulatory oversight for the board of directors and excluding audit committee variables, and vice versa. We do not observe any qualitative changes to the results. However, as the VIFs are greater than 10 (VIF > 50 for the board variables and > 200 for audit committee variables), we question the validity of the results. Likewise, it is not meaningful to interpret analyses containing both sets of board and audit committee variables and their interaction variables with regulatory oversight because of high multicollinearity (Neter et al., 1987).

Fourth, we derive factor scores from board and audit committee independence, multiple directorships, size, financial expertise and the number of meetings (proxying for diligence) and interact them with regulatory oversight. OLS results show that the coefficients on SIZE–REG and ODIR–REG remain significant in the expected directions (p < 0.01, one-tailed) and the coefficients on the other interaction variables are not significant.

Finally, we perform our analysis based only on 217 non-regulated companies and 220 bank holding companies after omitting the 32 utility companies. Our results are qualitatively similar. IND–REG is marginally significant (p = 0.06, one-tailed) and both SIZE–REG and ODIR–REG remain significant (p < 0.01, one-tailed) in the expected directions.