PRIOR RESEARCH AND HYPOTHESES DEVELOPMENT

1867 be to increase the AF. Conversely, if existence of an audit committee is associated with increased internal control strength, a reduced fee would be expected. Collier and Gregory 1996 examined these propositions and found a significant positive relationship for the first but no significant relationship for the second. The authors conclude that ‗there is no conclusive evidence to suggest that audit committees are effective in engendering a stronger intern al control environment that is reflected in reduced audit fees‘ p.1λ5. Evidence that the proportion of non-executive directors has a positive and significant impact on AF, which is consistent with increased non-executive representation encouraging more extensive auditing, is provided by O‘Sullivan β000 based on an examination of the 1λλβ fees of 40β UK companies. Intriguingly, however, this research did not test whether the presence of an audit committee affects AF, but a study by the same author O‘Sullivan, 1999 using the 1995 audit fees for a sample of 146 UK companies found no evidence that board and audit committee characteristics influence audit pricing. The study by Goodwin-Stewart and Kent 2006 based on a survey of 401 Australian companies found the existence of audit committee, more frequent committee meetings and increased use of internal audit are related to higher AF. However, they considered those variables independently and did not hypothesise them as being a composite measure for ACE and only considered the impact on AF only. We extend prior work by considering UK data and incorporating into our research the interactive effects of the three AC elements viz. independence, financial expertise and diligence number of meetings as a proxy for ACE in both the AF and NAS models. We recognise that various explanatory factors suggested in the literature may intervene in the process and have complex relations with audit and NAS fees. Thus, we included in our models a number of control variables that were missing in the Goodwin-Stewart and Kent 2006 study but which were deemed necessary in AF and NAS fees literature. These variables are further discussed in section 2.2. Given the paucity of research on the relationship between corporate governance quality and auditor remuneration in the UK, we believe that our paper makes an important contribution. Specifically, in this paper we examine the influence of audit committee effectiveness ACE on both audit and NAS fees. Although individually independence of the audit committee, the possession of financial expertise, and frequent meetings are important considerations, we believe it is the interaction of such characteristics that is likely to have most impact on audit quality. We therefore focus on the joint effect of these dimensions as an empirical proxy for ACE. Below we develop our hypotheses and explain the models variables: 1868 i AC Effectiveness ACE – the composite measure We believe that for audit committees to be effective they must at least exhibit three characteristics. First, audit committees must be independent of management. Second, the audit committees must be active, i.e. meet at least certain times a year. Third, the membership of the committee must include a director with relevant financial expertise. We combine some independent variables, explained below, to form a new construct called ACE audit committee effectiveness. ACE is thus a dichotomous variable equal to 1 when i the audit committee membership consists of all independent non-executive directors, ii has a member with financial expertise and iii meets at least 3 times a year. The remainder of this section develop our hypotheses relating to ACE that will be tested in the study. Independence: Directors who are independent of management are expected to be more interested in auditor quality and are more likely to believe that the provision of high levels of NAS by the auditor may impair auditor independence. Prior research suggests that audit committees which are independent of management are likely to be averse to high levels of NAS. Abbott et al. β00γ noted, for example, that independent audit committee‘s concern for auditor independence can have a direct or indirect affect on the purchase of NAS. The perceived threat to independence could result in the audit committee monitoring the provision of NAS. Alternatively, an independent audit committee may not be directly involved in the purchase decision. Nevertheless it is possible that management may voluntarily reduce the level of NAS in anticipation of the potential concern of the audit committee. For this reason, we incorporated the independence variable in our composite measure for ACE. Financial expertise: The financial expertise of the audit committee forms the second component of our test variable, ACE. Prior research has established that for audit committees 1869 to be effective, their membership needs to include a member with relevant financial expertise. Corporate governance codes do vary as to the specification of the level or nature of financial expertise. Nonetheless, research shows that audit committee members without financial experience may not be strong enough to protect auditor quality Knapp, 1987; DeZoort et al. 2002, Turley and Zaman, 2004. We expect that audit committees which have at least one member with financial expertise is likely to be concerned about audit quality and thus have a positive association with AF and a negative association with NAS fees. Thus, it is necessary to incorporate this variable in our composite measure for ACE. Meetings: For an audit committee to be effective, it must be active. Audit committee meetings are thus the third component of our test variable ACE. Prior research has established the importance of active committees for the oversight of the financial reporting and auditing process. Regular meetings provide opportunity for the audit committee to monitor audit quality. Meeting frequency can be a signal of audit committee diligence Menon and Williams, 1994 and has been associated with reduced likelihood of fraud Beasley et al., 2000 and financial restatement Abbott et al., 2003. We believe that active committees are likely to exert a positive influence on audit scope, which in turn will be reflected in higher AF and a negative influence on NAS. Thus, it is necessary to incorporate this factor as part of our ACE measure. Overall, based on the above discussions, we expect our composite measure, ACE, to have a negative effect on NAS fees and a positive effect on AF. Hence, our two hypotheses are as follows: H 1 : Ceteris paribus, there is a negative relationship between ACE and the level of non-audit services fees. H 2 : Ceteris paribus, there is a positive relationship between ACE and the level of audit fees. 1870 In testing the above hypotheses we wish to control for additional corporate governance, in particular board of director related, factors which may potentially have an influence on the level of audit andor NAS fees. In this respect our paper has a further secondary aim which is to examine whether ACE has influence on auditor remuneration after taking into account board of director effects. Here we note that the US study of Carcello et al. 2002 on board characteristics and AF based on a sample of Fortune 1000 companies with fiscal years ending between April 1λλβ and εarch 1λλγ found that ―audit committee variables provide no incremental explanatory power when the board characteristics are included in the model‖. Their results showed that ―none of the audit committee variables is significantly related to audit fees‖. We find the Carcello et al β00β result puzzling. We believe that given that audit committees have specific oversight responsibility for financial reporting and external audit, after controlling for the board director and additional audit committee related characteristics, our test variable ACE will have a significant positive association with AF and a negative significant association with NAS fees. ii Control variables In testing for the influence of ACE we control for the following corporate governance and agency related variables. Our first two control variables are actually related to the audit committee chair which may potentially affect its independence and expertise. First, we include a variable ACCS audit committee chair holds shares in the company to test whether the shareholding has any affect on AF and NAS fees. Second, we also test whether the number of additional directorships TAD held by the audit committee chair exerts an influence on the level of audit and NAS fees. The holding of additional directorships has been used in prior research as a proxy for audit committee chair‘s expertise Carcello and Neal, 2003; however its effect on auditor remuneration has not been investigated. Carcello et al 1871 2002 have used a similar concept, directorships – average number of outside directorships in other firms held by outside directors, in examining the influence of board characteristics on audit fees and found that it had a positive significant effect. The extension of our model to test for the effects of both TAD and ACCS is particularly relevant given that over time, governance codes have sought to tighten the definition of audit committee independence and restrict or at least discourage non-executive directors from holding shares in the company and from taking on too many directorship appointments in other companies see Collier and Zaman, 2005. The latest Combined Code on Corporate Governance FRC, 2006 in the UK for instance recommends that audit committees should consist entirely of independent directors with no interest in the company they serve. In addition to the above two audit committee chair variables incorporated into in our model, we also control for the effect of number of board meetings. Similar to our discussion earlier about audit committee meetings, the number of board meetings BDM can indicate the level of diligence exercised by the board of directors. Carcello et al. 2002 confirmed high frequency of board meetings could indicate higher level of control in the company and thus could be associated with audit fees. The composition of the board of directors is also a potentially important factor affecting audit quality. We also control for the proportion of non-executive directors on the board PNED. Non-executive directors have an interest in protecting their reputation and avoiding potential financial loss that may result from litigation by increasing audit quality see Young 2007 for a discussion of non-executive directors. It is possible that in companies with a high proportion of non-executive directors, there will be a high level of concern about audit quality which in turn would be reflected in a positive association with audit and a negative association with NAS fees. We control for a third factor relating to board of directors – duality whether the board chair is also the company’s chief executive officer that may potentially influence audit quality. As noted by Collier and Gregory 1996 board duality can have a potentially adverse influence on audit quality and audit committee activity. Finally, following the literature on AF we control for a number of company related variables which have been hypothesised in the literature to be associated with auditor remuneration. These include: auditor type BigFour, company size lnTA, company complexity lnSubs, level of risk proxied by leverage long term debt to total assets Lev and whether the company made any acquisitions Acq or incurred a loss Loss in the previous two years. Our final control variable relates to the concentration of ownership. Consistent with Firth 1997 we use a variable number of shareholders with 5 or more shareholdings NSH5 to test if it has an influence on audit and NAS fees. 1872 iii Regression Models Consistent with prior literature on AF we use a single equation approach to test our hypotheses relating to the influence of ACE on audit and NAS fees. The OLS regression models are as follows:            3 1 6 5 4 3 2 ln i it it it it it itx it Duality BM PNEDB ACCS TAD ACE NAS        it it it it it it Loss Big Sub Lev Size NSH 12 11 10 9 8 7 4 ln 5             it Acq 13   Model 1            3 1 6 5 4 3 2 ln i it it it it it itx it Duality BM PNEDB ACCS TAD ACE AF        it it it it it it Loss Big Sub Lev Size NSH 12 11 10 9 8 7 4 ln 5             1 13 m it Acq     Model 2 Where: Independent variables lnNAS lnAF – natural logarithm of non-audit services fees – natural logarithm of audit fees Dependent variables: ACE: – Audit Committee Effectiveness. An audit committee is effective when [ACI = 1 + ACX = 1 + ACM ≥ γ] : ACI = Audit committee independence. Dichotomous with 1 if all AC members are non- executive directors and 0 otherwise. ACX = Audit committee‘s financial expertise. Dichotomous with 1 if the audit committee contains a member with financial expertise and 0 otherwise. ACM = Number of audit committee meetings held during the financial year. TAD – Audit committee chair‘s total additional directorships. The number of additional directorships the audit committee chair holds, including executive and non-executive positions. TAD = [ACCXD = AC chair holds additional executive position in another company + ACCnXD = AC chair holds additional non-executive position in another company]. ACCS – Audit committee chair‘s shareholdings. Dichotomous with 1 if the audit committee chair holds the company‘s shares and 0 otherwise. PNEDB – The proportion of non-executive directors to total number of directors on the board of the company. BM – Number of board meetings held in the financial year. Duality – Chief executive is also chair. A dummy variable equal to 1 if the chief executive concurrently holds the position of chairman, 0 otherwise. NSH5 – Number of shareholders with= 5 Shareholdings. This is the number of shareholders holding 5 or more of the compan y‘s shares. 1873 Size – The natural log of total assets. Lev – Leverage of the company measured by the ratio of long-term debt to total assets. lnSub – Natural logarithm of number of subsidiaries. Big4 – Big 4 as auditor. A dummy variable equal to 1 if the company employs a big 4-auditing firm as their auditor, 0 otherwise. Loss – Whether the company made a loss in the 2 previous financial years. A dummy variable equal to 1 if the company made a loss in the 2 previous financial years, 0 otherwise. Acq – Whether the company made an acquisition in the 2 previous financial years. A dummy variable equal to 1 if the company made an acquisition, 0 otherwise. 1 m  – Standardised residuals m1 : The computed residuals for model 1. The dependent variable of interest lnNAS is the logarithm of NAS fees. Data regarding audit and NAS fees were obtained manually from the notes to the accounts in each of the companies‘ annual reports. For the AF related hypotheses, we replace the lnNAS dependent variable in the above equation with the log of audit fees lnAF. Since, most factors influencing lnAF also influence lnNAS fees and that the relationships between AF, NAS fees and the other explanatory variables suggested in the literature are complex, we therefore incorporated the standardised residuals 1 m  of lnNAS from Model 1 into Model 2 to control for the incremental effect of NAS fees on level of AF. Our test variable ACE is coded 1 if the following conditions are met: the audit committee is composed of non-executive directors, at least three audit committee meetings are held per year; and the audit committee has at least one member with financial expertise. The Smith Report 2003 defined expertise as follows: At least one member of the audit committee should have significant, recent and relevant financial experience, for example as an auditor or a finance director of a listed company. It is highly desirable for this member to have a professional qualification from one of the professional accountancy bodies para 3.16, p.9. For the purpose of our models we regarded the audit committee has having financial expertise if one of its member had experience as an auditor, finance director or had a professional accounting qualification. To summarise, the dependent variables used in the test is the level of NAS fees lnNAS and level of AF lnAF. The main independent variable of interest is ACE audit committee effectiveness which consists of AC independence ACI, financial expertise ACX, and number of meetings ACM. The control variables include: AC chairman’s total additional directorships TAD and whether they hold shares in the 1874 company ACCS. Board characteristics related control variables included the proportion of NEDs on the board PNEDB, number of board meetings BM, and duality Duality. We also control for ownership, the number of shareholders holding 5 or more of the company’s shares NSH5. Finally, company related control variables include the size of the company lnTA, its leverage Lev, the number of subsidiaries lnSub, whether the company uses a big-four auditor Big4 and whether in the last 2 years the company made a loss Loss, or an acquisition Acq.

3. EMPRICAL RESULTS AND ANALYSIS

i Descriptive Statistics The sample used in the paper is drawn from non-financial companies in the FTSE-350 which represents a good mix of the largest UK companies and relatively smaller companies and covers the period 2001 to 2004. We use a random sample of 400 company-year observations to test our hypotheses relating the influence of audit committee effectiveness ACE on audit and NAS fees. Table 1 provides the descriptive statistics for the dependent and continuous independent variables used in our models. TAKE IN TABLE 1 The analysis of residuals, plots of the studentised residuals against predicted values as well as the Q-Q plot indicate no problems of homoscedasticity and linearity. Residuals of standard tests on skewness and kurtosis indicated a problem with the normality assumption and therefore the dependent variables lnSub is transformed into normal scores. Table 2 presents the correlation matrix for the dependent and the continuous independent variables. It does not indicate any multicollinearity problem, as the correlations are relatively low. TAKE IN TABLE 2 ii Regression Results Model 1 tests which independent variables viz. the composite measure of ACE, AC chair‘s total additional directorships TAD and shareholding in the company ACCS, the proportion of NEDs on the board PNED, number of board meetings BM, role duality Duality , number of shareholders holding 5 or more of the company‘s shares NSH5, and control variables viz. size of the company lnTA, its leverage Lev, number of subsidiaries 1875 lnSub, type of audit firm Big4, loss made within the last 2 years Loss and acquisition Acq, are associated with NAS fees. We run the model three times, referred to as Model 1a, 1b, and 1c. The results are shown in Table 3. The F-value for each model is significant at the 1 level and the adjusted R 2 for each of the three models is between 21 and 23. TAKE IN TABLE 3 Results in Model 1a show that board meetings, role duality, company size, type of audit firm and the test variable ACE are all found significantly associated with level of NAS fees but no significant relationship for the other variables tested. The fact that board meetings, role duality, company size and type of audit firm are significant suggest that large companies chaired and managed by the same director, audited by big-four and with frequent board meetings tend to buy more NAS. As predicted the negative coefficient and significance level for the test variable indicate that the probability of higher purchase of NAS decreases with ACE. In Model 1b, we substituted TAD and ACE with ACCXD and ACCnXD for the former and ACM, ACX and ACI for the latter. As can be seen in Table 3, the overall findings are similar to the earlier model except that ACX was found to be significant and negatively associated with lnNAS. The result suggests that financial expertise in the audit committee team is the main driver for reducing the demand on NAS. As for model 1c, all variables are similar to model 1a except for TAD and ACE. In this model, we substituted TAD with ACCXD and ACCnXD as in model 1b and replaced ACE with a new refined ACE i.e.