Experience of Audit Committee Members and Audit Quality

Abstract

We investigate whether the experience of audit committee members is associated with audit quality. In order to comprehensively analyse the experience of audit committee members, we include audit committee member tenure, age and multiple-directorships in our analysis. Using observations from 2001 to 2012, we undertake analysis on 13,155 firm-year observations and find that all our proxies of audit committee member experience are positively associated with audit fees. A range of additional tests, including using discretionary accruals as an alternative measure of audit quality and differences-in-differences analysis, support our main findings and our results consequently make a number of contributions to both the literature and policy making. One possible policy contribution is that regulators may wish to consider audit committee characteristics representing experience when framing recommendations to improve audit quality and thereby, financial reporting by firms.

Keywords:

• Audit committee member tenure
• audit committee member age
• audit committee member multiple-directorships
• audit quality
• audit fee

JEL codes:

• M41
• M42
• M48

Acknowledgements

The authors thank Juha-Pekka Kallunki (the Editor) and two anonymous reviewers for both their insights and constructive suggestions in improving the paper. We also thank Ross Taplin, Roger Simnett and Elizabeth Carson for their valuable comments and gratefully acknowledge the beneficial remarks from both the discussants and participants at the 22nd International Symposium on Audit Research 2016 and the Accounting and Finance Association of Australia and New Zealand 2015 conference.

ORCID

Nigar Sultana http://orcid.org/0000-0003-4273-2034

Harjinder Singh http://orcid.org/0000-0002-8250-5808

Asheq Rahman http://orcid.org/0000-0002-3245-4512

Notes

1 Audit fees have several advantages over other proxies measuring audit quality. First, audit fees are continuous and able to capture subtle variations in audit quality. Second, the literature has developed, particularly post-Simunic (1980), sophisticated audit fee models with R-squares often exceeding 60% which alleviate concerns about correlated omitted variables.

2 Additionally, Dao et al. (2013) utilize proxies for cost of equity that are difficult to calculate and also do not address the issue of audit quality which, as we discuss above, is a key responsibility of an audit committee.

3 Our choice of the three variables underpinning audit committee experience reflects a number of audit committee member skills and competencies that increase the effectiveness of audit committees: for example, (1) audit committee members’ board tenure reflects members’ accumulated firm-specific knowledge, (2) audit committee member age suggests both risk-aversion and a broader range of life experiences and (3) audit committee members’ multiple-directorships suggest experience (including governance experience) and a broader range expertise.

4 We undertake this analysis to determine whether the demand or supply side of an audit engagement is likely to be driving higher audit fees. Please refer to footnote 25 for more details.

5 To date, corporate governance rules and regulations have mainly focused on audit committee member independence, meeting frequency and financial expertise.

6 As an example, Beck, Mauldin, and Elaine (2014) utilize a sample of 9,214 firm-year observations over a three year period 2006–2009 when examining CEO and audit committee influence on audit fees. All other notable studies examining audit committee and audit fees have research designs less robust than Beck et al. (2014).

7 For example, SOX 2002 now requires audit committees to also approve audit and non-audit services, review auditors’ reports on critical accounting policies and engage in discussions with management on alternative GAAP (US Congress. Sarbanes-Oxley Act, 2002).

8 For a detailed discussion of studies that examine the association between audit committee effectiveness and financial reporting quality, please refer to Pomeroy and Thornton (2008) and DeZoort et al. (2002).

9 However, Knechel and Willekens (2006) acknowledge that they only examine such relationships from the demand side of an audit engagement whereas we offer an analysis that encompasses both demand and supply side aspects of an audit.

10 These include oversight over the internal audit function and related internal controls, auditor selection, determination of both audit and non-audit fees, audit scope and maintaining auditor independence (Sharma & Iselin, 2012).

11 We do not include the expertise of audit committee members as an experience measure in our study due to the fact that expertise and experience can be two different constructs in that expertise in a particular area (such as accounting and auditing) does not necessarily increase the experience capability of an audit committee member. Given that experience encompasses factors such as firm-specific knowledge, expertise will not always represent experience.

12 All of our variables are winsorised at the 1 and 99 percentiles.

13 All of our multivariate analyses report coefficient estimates and White-Huber t-statistics clustered at firm-level. Standard errors of these regressions are clustered by firms to mitigate the serial correlation arising from our multiple year observations.

14 All of our regression results reported include variance inflation factor scores to address multicollinearity concerns.

15 Given that our data are pooled over 2001–2012, the same firm is likely to appear more than once in the sample, possibly resulting in autocorrelation leading to an underestimation of standard errors and an overestimation of significance levels. We therefore compute the t-statistics using robust standard errors corrected for firm-levels clustering and heteroscedasticity consistent with Gow, Ormazabal, and Taylor (2010) and Petersen (2009).

16 We also seek to determine if older audit committee members have the right type of experience rather than just have greater levels of experience. We therefore undertake additional analysis in which we go through our dataset, randomly select 250 firms from our dataset and examine the biography of each audit committee member provided in the annual reports of these firms. This equated to examining the biography of more than 1,000 audit member members. After examining each audit committee member’s biography, we are confident that the older audit committee members have wider experience pertaining to the industry that the firm operates within, greater interactions with the external audit function and corporate governance in general than younger audit committee members. This additional analysis that we undertake is also consistent with past published literature that indicates older directors are more experienced (Kilduff, Angelmar, & Mehra, 2000; Mahadeo, Soobaroyen, & Hanuman, 2012; Serfling, 2014).

17 The suest test is a generalised Hausman specification test available within Stata used, among other things, to check the significance of regression coefficients.

18 We are grateful to two anonymous reviewers for helping us with a number of these tests.

19 Principal components analysis was chosen for factor extraction given the need for a structure to calculate factor scores.

20 When measuring audit committee multiple-directorship, we also calculate the proportion (rather than the number) of audit committee members serving on multiple boards and re-run our main analysis. Our main results remain quantitatively similar.

1 Consistent with Dao et al. (2013), we also test the sensitivity of using alternative cut-off points of board tenure by employing two other split points for audit tenure, namely five and ten years. Furthermore, in addition to calculating audit committee member tenure in terms of service to a particular firm, we also calculate tenure in terms of the number of years an individual has been an audit committee member at any firm. We find that our main results in column 1 table 3 strengthen further with this variable having a positive coefficient with a higher t-statistic of 4.702 (compared to 4.273 in column 1 table 3).

22 By converting our audit committee member experience predictors into dummy variables, we also mitigate the effect of possible extreme values in these variables.

23 We use another continuous variable to test the sensitivity of our multiple-directorships measure by using the average other directorships of audit committee members instead of the number of audit committee members serving on multiple boards for firm i in time period t. Our main results remain quantitatively similar.

24 We also re-run our analysis excluding audit committee chairperson tenure, age and multiple-directorships and find no change to our results.

25 It is possible that any increase in audit fees may not be due to the beneficial aspect of audit committee member experience (such as audit committee members with multiple-directorships seeking greater monitoring) but from the detrimental aspects of audit committee member experience (audit committee members with multiple-directorships being too busy to discharge their responsibilities) resulting in the auditor assessing higher control risk and consequently doing more audit work and driving up audit fees. As such, it could be the auditor driving increases in audit fees rather than experienced audit committee members. Consequently, we undertake additional triangulation analysis. We thank an anonymous reviewer for this suggestion.

26 Consistent with DeFond and Zhang (2014), we select discretionary accruals as it reflects an input-based measure to audit quality compared to audit fees which represent an output-based measure of audit quality. DeFond and Zhang (2014) point out that no single input or output-based measure can paint a complete picture of audit quality and recommend selecting proxies from each of these two categories of audit quality thus better capturing the multi-dimensional aspect of audit quality.

27 Kothari et al. (2005) model is derived from the cross-sectional version of the Dechow, Sloan, and Sweeney’s (1995) modified Jones model using previous year’s return on assets as an additional regressor. The use of lag of return on assets as a control for firm performance in the performance adjusted model mitigates heteroscedasticity and avoids severe misspecification issues associated with the Jones and modified Jones models in estimating discretionary accruals (Kothari et al., 2005). Following the approach adopted by Kothari et al. (2005), we also include a constant (α₀) to address the econometric issue of a missing intercept term in the modified Jones model.

28 Given that we use firm-fixed effects to address the issue of unobservable omitted variable bias in our earlier analyses, we use a differences-in-differences design as supplementary analysis to our firm-fixed effects examination.

29 This value of 1 therefore for all new audit committee appointments will cause any DID analysis between the change in audit committee member tenure from an outgoing member to a new one with a value of 1 year unusable.

30 We do not split the greater number of multiple-directorships into categories of greater by one or two or three or more multiple-directorships in order to minimize the attrition to our sample.

31 The model consequently uses each firm as its own control.

32 Additionally, we also re-run our analyses seeking to determine if audit fees reduce when (1) an audit committee member with greater multiple-directorships is replaced by one with lesser multiple-directorships; and (2) an older audit committee member is replaced by a younger one. Supporting our main tests, in both cases, there was a decrease in ΔLnAFees_it (β −.031, t-statistic 9.257 for the former and β −.088, t-statistic 8.164 for the latter). However for the purposes of brevity, we do not tabulate these results.

33 It is also possible that the use of average measures of experience for the entire audit committee is disguising information in our results. As such, we undertake additional analysis in which we examine the diversity of the audit committee given that any decision made by the audit committee reflects the dynamics of the committee. In this respect, we undertake the analysis along two fronts. Initially, we assess whether the diversity characteristics (i.e., gender and financial expertise) of the audit committee chairperson and also the experience levels of that chairperson is associated with audit fees. Second, we then analyse whether non-chairperson audit committee member diversity characteristics (also gender and financial expertise) are associated with our three audit committee experience variables.

Untabulated results indicate that audit fees are higher for firms where the audit committee chairperson is a financial expert (t-statistic 4.470). We also find that, although female chairpersons do not work more effectively in increasing audit quality than their male counterparts, female chairpersons who are financial experts may demand more audit coverage (t-statistic 2.593). Furthermore, in all instances, audit committee chairpersons’ age, tenure and multiple-directorships are significantly associated with audit fees (t-statistics 4.624, 5.749 and 7.034 respectively). When examining whether experienced non-Chairperson audit committee members with diversity characteristics (i.e., gender and financial expertise) are better monitors, we find that female audit committee members with longer tenure and multiple-directorships may demand better audit coverage than their male counterparts (t-statistics 1.799, and 4.445 respectively). Finally, we find that firms with either older audit committee members, longer tenure or more multiple-directorships, and who are also financial experts have higher audit fees (t-statistics 2.491, 2.882 and 5.885 respectively). As such, our additional analyses indicate that the diversity dynamics of the audit committee are beneficial to firms as diversity and experience appear to complement each other making the audit committee more effective.

34 We acknowledge that director fees also reflect other skillsets of directors apart from experience but use director fees only as additional analysis to substantiate our main results.

35 In the course of collecting data on directors’ fees, we had a small number of missing observations equating to 1,451 resulting in our final sample for this analysis being 11,704 firm-year observations.