Auditor University Education: Does it Matter?

Abstract

We examine the implications of auditor education for audit quality and audit pricing. We exploit a novel institutional setting in the UK, where auditors major in many different degrees at university and signing auditors are identifiable. Using hand-collected data for a large sample of signing auditors, we establish two main findings. First, auditors whose degrees have a quantitative orientation are associated with higher accruals quality and higher audit fees than those with more qualitative degrees. Second, auditors with degrees directly relevant to accounting are also associated with higher accruals quality and increased audit fees relative to auditors with unrelated university degrees in qualitative subjects. Overall, our study provides evidence that heterogeneity in auditor education is associated with divergent audit outcomes.

Keywords:

• Audit partners
• Education
• Audit quality
• Audit fees

JEL Classifications:

• M41
• M42
• K42

Acknowledgments

We are grateful to Henrik Nilsson and two anonymous referees for their helpful comments and suggestions. We also thank Jan Bouwens, Ulf Brüggemann, David Hay, Rani Hoitash, Chris Humphrey, Robert Knechel, Clive Lennox as well as seminar participants at the 27^th Audit and Assurance Conference, the 40^th European Accounting Association (EAA) Annual Conference and the ESMT Berlin Annual Accounting Conference for their numerous helpful comments. We thank the Cambridge Endowment for Research in Finance (CERF) for research support.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplemental Data and Research Materials

Supplemental data for this article can be accessed on the Taylor & Francis website, doi:10.1080/09638180.2020.1866633

Notes

1 For a recent review of archival studies examining auditing at the auditor level, see Lennox and Wu (2018).

2 See, for example, AICPA (US), HKICPA (Hong Kong), SAICA (South Africa), NZICA (New Zealand), and CICA (Canada).

3 We note that while university-level accounting education in the UK is relatively less technical and specialized compared to some other countries, such as the U.S. (Diamond, 2005), accounting degrees offered in the UK typically do not follow the broader liberal arts model (Gammie & Kirkham, 2008).

4 The Russell Group comprises 24 research-led UK universities. See http://russellgroup.ac.uk/about/our-universities/ for more details.

5 Other outcome variables commonly studied in the audit quality literature, including mandated accounting restatements and litigation, are generally relatively rare in the UK and impossible to identify for our sample period from databases available to us. An alternative audit outcome studied in the literature is the likelihood of a qualified audit opinion (DeFond & Zhang, 2014). In our analysis, we are not able to employ this measure because our data indicates no qualified audit opinions. Accordingly, we perform our audit quality analysis based on accruals proxies for which data is available.

6 For example, accounting education was suspended in China during the Cultural Revolution (Ezzamel & Xiao, 2015) and Western accounting systems were only introduced into college education in China in 1990 (Gul et al., 2013); therefore, the skills developed by auditors through formal education in China are likely different from those acquired in Western countries.

7 We note that 95% of our sample of signing auditors with university-level education have been educated in the UK, with the remaining 5% holding degrees primarily from other western countries.

8 Che et al. (2018) examine the implications of the level of university education (i.e. master vs bachelor’s degrees) for auditor effort. Similarly, Burke et al. (2019) examine the effect of U.S. auditor education on accruals metrics and audit fees, albeit their perspective is the type of the degree awarding institution.

9 A person without a university degree can attain university entrance level in one of two ways. First, her academic or professional qualifications obtained in or outside the UK may be assessed as being of a sufficient standard; or second, she may be assessed using written tests considered adequate for the purpose (Companies Act, 2006, c. 46, p. 683).

10 The FRC is the regulatory body responsible for financial reporting and auditing in the UK.

11 See https://www.frc.org.uk/auditors/professional-oversight/oversight-of-audit/recognition-of-recognised-supervisory-bodies-and-r.

12 For further details see: https://careers.icaew.com/en/find-your-route.

13 For further details see: http://www.accaglobal.com/uk/en/qualifications/glance/acca/who.html.

14 For example, in 2018 the percentage of students with a university (relevant) degree varies from 34% (21%) to 97% (70%), depending on the professional accountancy body (FRC, 2019).

15 Typically, applicants with a business university degree may be eligible for individual module exemptions, whereas applicants without a university degree might be required to complete more extensive training for the qualification because they may have to undertake special programs before entering professional examinations.

16 As discussed, audit firms organize their activities in teams. Clearly, the characteristics, structure and dynamics of these teams are potentially very important. In line with this reasoning, recent research documents that audit team attributes, such as team diversity and experience, may affect audit quality (Aobdia et al., 2019; Cameran et al., 2018). From an empirical perspective, if the education and experience of other audit team members is also important to audit outcomes, our inability to observe and include audit team composition in our model reduces the power our empirical tests; however, it does not bias our tests if audit team attributes are uncorrelated with lead auditor attributes. If this is not the case then we cannot entirely rule out correlated omitted variable bias.

17 Continuing professional development (CPD) courses attended by certified auditors may also affect their performance (Che et al., 2018). We are unable to observe and measure this type of educational activity. However, CPD typically applies to all qualified auditors in the UK; moreover, 88% of our sample auditors are certified by ICAEW and therefore are subject to the same CPD requirements (see https://www.icaew.com/-/media/corporate/files/members/cpd/quick-guide-to-cpd.ashx). Accordingly, performance or pricing differences between auditors in our sample are unlikely driven by differences in CPD training. It is possible, however, that auditors may receive different in-house training, depending on the requirements of their audit firm. Again, we cannot observe such training differences. In order to mitigate concerns related to this issue, we re-estimate all our baseline models after including audit firm-fixed effects, in addition to industry- and year-fixed effects. The inclusion of the audit firm-fixed effects allows us to control for time-invariant audit firm-specific factors (e.g. training courses provided to their staff) that might affect our outcomes. Untabulated analysis reveals that all our primary inferences persist (See Online Appendix, Table OA1).

18 Our discussion on the potential importance of quantitative or accounting education focuses on technical auditor competencies. Soft skills can also play an important role in mitigating accounting errors and misstatements. Unfortunately, we are not able to develop informed priors regarding the differential effects of different degree programs on soft skills development. Development of soft skills is more likely in specific types of course (some of which may not be mandatory) and through specific assessment methods (e.g. group/presentation coursework). Since we have no detailed information on specific degree programs and syllabi of courses within programs, we cannot form well-founded priors about soft skills development. Accordingly, we adopt an agnostic stance on the differential indirect effects of education on soft skills development, and hence on audit quality and pricing.

19 Implicitly, we argue that educational skills learned a long time ago are actually transferable and therefore potentially important in determining audit quality and pricing. There are at least two reasons justifying this argument. First, university education may develop core skills (e.g. numerical skills) that not only do not fade away with the passage of time, but instead may strengthen further through auditors’ professional qualification, CPD courses and in-house training. Second, university education may develop a certain ‘mind-set’ that also does not fade away with the passage of time (e.g. the ability to apply systematically certain audit methodologies or a body of well-embedded, foundational accounting knowledge). Nevertheless, as discussed in our research design section below, we document the implications of auditor education after controlling for the effect of auditor experience.

20 We assume that the hourly charge rate does not differ among auditors. This is confirmed by feedback we received from UK accounting practitioners, who stated that there can be differences in the audit hours spent (but not in the charge rate/hour) conditioning on several auditor characteristics, such as her skills, knowledge and experience. Similarly, we assume that all other factors affecting liability costs, and in particular the litigation regime, remain constant. We believe that this is a reasonable assumption given that all auditors in our sample audit UK listed companies and therefore, they are subject to the same regulatory and enforcement `requirements.

21 When we use absolute total accruals as an audit quality proxy, our results are consistent with those of Gipper et al. (2020) in their recent examination of absolute total accruals as a measure of audit quality - we fail to find a significant relation between audit quality and our test variables. However, total accruals reflect the combined effects of operating, investment and financing activities and their relation to past, current and future economic performance. It is possible that the absolute total accruals proxy has low power as an audit quality construct because it does not adequately control for economic performance. We are unaware of models in the literature providing theoretical guidance for modelling absolute total accruals.

22 Following the literature, we set the minimum observation threshold to run the industry-year estimation model to ten and define industry sectors by SIC 2-digit industry codes.

23 Although the number of courses taught in each university degree may be informative, we are not able to observe relevant data on this potential factor, and hence cannot include it in our empirical analysis.

24 The mandatory rotation rules in the UK require a change at the individual auditor level every five years; and then the auditor has to have a five-year gap (APB Ethical Standard 3, Auditing Practices Board [APB], 2009). As a result, there is insufficient time-series heterogeneity in auditor assignment to clients during our four-year sample period; and therefore, we cannot include firm-fixed effects. For example, in our audit fees sample only 13.5% of firm-years involve lead auditor switches. More importantly, our research design requires not only an auditor change but also a switch between different educational subjects; for example, we observe that only 1.96% (1.55%) of our audit fee sample observations relate to an auditor change and a switch from QUANT to HSS (from HSS to QUANT). Similarly, we cannot employ a differences-in-differences research design in order to exploit auditor switches. Further, as discussed below, we cannot extend the sample period back before 2011 because auditor identity data is not available.

25 The requirement for engagement auditors to sign audit reports became effective for financial statements ending in April 2009 or later (PricewaterhouseCoopers Legal, 2010). However, FAME began reporting the auditor identity in 2011; prior to 2011 the coverage of signing auditors’ identity is very limited. A further limitation of the FAME database is that it reports a company’s public status at its most recent value only. Following Lennox and Li (2012) we correct this problem by using historic data from the London Share Price Database to identify publicly traded companies in each sample year.

26 For further details on the subjects included in humanities see https://en.wikipedia.org/wiki/Humanities; and in social sciences see https://en.wikipedia.org/wiki/Social_science#Methods. We note that we exclude the field of economics from social sciences given its apparent quantitative focus.

27 The classification of subjects into broad academic fields requires us to exercise a degree of subjectivity; for example, the subject of ‘physics & mathematics’ could be classified as mathematics-related instead of physics-related (the choice we make). However, in our empirical analysis we employ indicator variables based on the five consolidated academic subject-groups; therefore, to the extent that potential ‘misclassifications’ relate to the same aggregate subject-group (e.g. sciences) our empirical findings are not affected.

28 We note that the descriptive statistics reported in Table 2 are similar to those related to the samples used in the regression analyses.

29 We group all sample firms into 12 industry sectors based on the Campbell (1996) classification. The total number of observations is slightly smaller than that of the full sample (i.e. 4254 compared to 4293) due to a small number of missing SIC industry codes.

30 Similar to Table 2, untabulated analysis reveals a very similar sample distribution by audit firm at the auditor-level. Accordingly, we report results based on the firm-year-level only.

31 In sensitivity tests, we also employ alternative definitions of industry expertise based on total assets of clients instead of audit fees as well as based on binary than continuous variables of industry expertise (e.g. Goodwin & Wu, 2014). Our inferences remain qualitatively unchanged (see Online Appendix, Table OA3).

32 We re-estimate all models in Table 6, Panel A after decomposing the ACCOUNTING test variable into its 4 components, i.e. Accounting only; Accounting & Finance; Accounting-related; and Finance/Banking (see Table OA2). Untabulated analysis reveals that our primary inferences persist; that is, in most cases auditors with some type of an accounting degree are associated with higher accruals quality, whereas in all cases of an accounting degree auditors are associated with higher audit fees (see Online Appendix, Table OA4). We note, however, that the definition of a relevant degree adopted by the UK professional bodies (the RQBs) is not restricted to a particular type of Accounting degree (e.g. the pure Accounting subject); instead, the definition of relevant degrees embraces Accounting usually combined with other subjects, such as Finance, Economics or Business Administration (FRC, 2019). The fact that Accounting is almost always studied in combination with other subjects and that this is reflected in degree titles leads us to conclude that our definition of Accounting degrees mapped out in Table OA2 is appropriate.

33 Following Chen et al. (2018), we examine the relation between |AB_ACC| and our three test variables after including the first stage variables of Equation (1). Results remain qualitatively unchanged. Also, we repeat our empirical tests after controlling for operating variability (i.e. standard deviation of sales or operating cash flows) in the accruals models (Hribar & Nichols, 2007); again, all findings persist and, moreover ACCOUNTING becomes significantly negative even under |AWCA|. Finally, in line with prior literature (e.g. Burke et al., 2019; Carcello & Li, 2013) we re-estimate our models after employing signed accruals by splitting the accruals samples into companies with positive accruals and companies with negative accruals. Un-tabulated analysis reveals similar results; that is, auditors with a quantitative or accounting educational background are associated with smaller positive and less negative accruals, suggesting higher audit quality for the aforementioned auditors compared to their HSS peers (see Online Appendix, Tables OA5, OA6 and OA7).

34 We also re-estimate Equations (2) and (3) after including both QUANT_NO_ACCT and ACCOUNTING. In doing so, we are able to: (a) compare the audit quality and audit fees implications of QUANT_NO_ACCT and ACCOUNTING with those of HSS (as in Table 6); and (b) compare the corresponding effects between the above test variables. Our baseline results remain qualitatively unchanged. Notably, the coefficients of ACCOUNTING and QUANT_NO_ACCT are very similar and significant under both accruals measures. But we do not find a significant difference between the estimated coefficients of QUANT_NO_ACCT and ACCOUNTING; in other words, we find that auditors with a quantitative education and those with an accounting-related degree are equally likely to be associated with enhanced accruals quality and increased audit fees (see Online Appendix, Table OA8).

35 The HSS control group is approximately 12 percent of sample observations, raising possible concerns that there is insufficient information in the control group to identify the counterfactuals reliably. To mitigate this concern, we also perform entropy balancing by reversing the process, using HSS (i.e. the qualitative degree holders) as the treatment group (Hainmueller & Xu, 2013). In this case the estimate on HSS is significantly positive (significantly negative) for the |AB_ACC| (AUD_FEES) model. Finally, we repeat the above sensitivity tests using the propensity score matching (PSM) approach. Our findings are qualitatively very similar to those reported in Table 7 (see Online Appendix, Tables OA9 and OA10, respectively).

Additional information

Funding

This work is supported by the Cambridge Endowment for Research in Finance (CERF) under a small research grant.