Audit Team Attributes Matter: How Diversity Affects Audit Quality

Abstract

How audit teams are structured and function plays a crucial role in determining the level of audit service quality. Despite this claim, little empirical research has been conducted on this effect. Using private data from two of the Big 4 audit firms, we fill this gap and document how diversity of audit teams influences audit quality. By combining the existing work in psychology and sociology with that in auditing, we develop our model by arguing that teams are composed not simply of single auditors but of sub-teams of individuals whose various combinations affect team performance. Starting from this premise, we study how the diversity of audit teams in terms of the different mix of work assigned to staff, seniors, managers, and partners influences audit quality and how this effect varies depending upon years of tenure. We also show that the proportion of leading auditors characterized by a common educational background and the percentage of female leading auditors affect audit quality. As an additional analysis, we examine how team diversity affects audit efficiency. The same elements found relevant for audit quality also affect audit efficiency.

Acknowledgements

We are grateful to Hervé Stolowy (editor) and Liesbeth Bruynseels (associate editor) and the two anonymous reviewers for their valuable comments and suggestions. We also thank Michael Favere-Marchesi, Sandra C. Vera-Muñoz and Stephen Penman for the insightful feedback on earlier drafts of the paper and the participants to the European Accounting Association 2015 Annual Congress and to the American Accounting Association 2017 Management Accounting Section Midyear Meeting.

Notes

1 There are different ways of defining diversity to understand its effects on team dynamics and outcomes. The conceptualization of diversity adopted here is linked to the idea of proportions and focuses on the role that proportions of different sub-teams have in affecting the relationships between team members (Mannix & Neale, 2005).

2 In audit firms, teams are normally composed of partners, managers (leading levels), seniors, and staff (junior levels).

3 Sub-teams are groups of individuals in a team that share some common attributes (De Vaan et al., 2015). Thus, for example, a team made up of 9 members – 2 partners, 3 managers, 2 seniors and 2 staff – could be described as a team composed of 4 sub-teams: 1 sub-team of 2 partners, 1 sub-team of 3 managers, 1 sub-team of 2 seniors, and 1 sub-team of 2 staff. Moreover, if in the same team, 1 of the 2 partners is female and 1 partner is male, the 3 managers are males, the 2 seniors are females, and the 2 staff are males, we could say that the team also comprises two other types of sub-teams: a sub-team of 3 females and a sub-team of 6 males.

4 One additional factor that might affect the diversity of the team and in turn its performance is represented by differences in national culture. For a comprehensive review of the literature on the effect of cross-cultural differences on auditors’ judgment and decision-making, see Nolder and Riley (2014). This variable has not been incorporated into our model because, in the empirical setting we analyze, cross-cultural differences are almost non-existent.

5 In fact, in the setting in which data are collected, leading audit team members hold the same type of university degree. According to national (i.e. Italian) regulation, signing partners should have a three-year university degree in Business Administration or the equivalent. Such partners should be enrolled in a public register (Registro dei Revisori Contabili) after at least three years of practical experience, followed by a State exam. This process is explained in more detail in Section 4.

6 Concerning signing partner gender, using GCOs as an indicator of audit quality, Hardies et al. (2016) document for a sample of private Belgian companies that females are more likely to issue GCOs than are males. The authors conclude that their findings indicate higher audit quality by female auditors. Based on a sample of Finnish and Swedish listed companies, Ittonen et al. (2013) document that female partners are associated with higher accrual quality. Similar results are obtained by Niskanen et al. (2011). However, using a large sample of Chinese data, Gul et al. (2013) do not find that partner gender matters in explaining audit quality. Moreover, experiments are used for exploring auditor gender’s effect (Chung & Monroe, 2001; O’Donnell & Johnson, 2001). Overall, the results maintain that auditor gender affects audit judgment and that female auditors are more accurate and effective in information processing and less prone to be influenced by unverified explanations from clients. In addition, efficiency in audit judgment is positively related to female auditors.

7 In particular, Italy is assigned a litigation risk score equal to 6.22, Nordic European countries (Denmark, Finland, the Netherlands, and Norway) show an average score of 5.22, and Germanic countries (Germany and Austria) have an average score of 4.91. A higher score is assigned to Canada (8.07), the UK (10), and the US (15) (Wingate, 1997). An example of anecdotal evidence concerning the Italian litigation environment is that of the well-known Parmalat scandal (similar to the Enron and Worldcom cases), which ended with the removal of Parmalat’s auditor (Italaudit SpA, previously Grant Thornton SpA) from the list of audit firms allowed to audit listed companies (Melis, 2005). Furthermore, after the scandal, the Italian stock market reacted strongly; Parmalat’s stock price collapsed, and related trading activity was suspended for several days. A loss of 70% was recorded by food industry stocks from mid to late December 2003, i.e. following the Parmalat scandal (Velucchi, 2009).

8 To consider these features, we add to our investigation a number of sensitivity analyses to control for both partner and audit firm mandatory rotations.

9 Italian regulation states, ‘the shareholders’ meeting, on a reasoned proposal of the supervisory body, is made responsible for the statutory audit and determines the amount payable to the statutory auditor or the statutory auditing company for the entire duration of the appointment and any criteria for the adjustment of this amount during the appointment’ (art. 13 of the Legislative Decree no. 39 of 27 January 2010).

10 In the traditional Italian firms’ corporate governance system largely adopted by Italian listed companies, audit committee functions are the responsibility of two different committees: the Internal Control Committee and the BSA (Rizzotti & Greco, 2013). The latter is also in charge of monitoring audit firm activity and results during the audit engagement.

11 We interviewed some partners of different Italian Big 4 audit firms to understand how audit firm proposals are typically structured. The Big 4 partners stated that there is not a homogeneous behavior or a ‘standard practice’, even at the audit firm level. Some proposals describe audit team members (at the partner and senior manager levels, citing names and enclosing their CVs) in detail; other proposals report only the hour allocations among the different professional levels inside the teams.

12 Based on the literature, we consider the quality of reported earnings the results of a negotiation process in which the auditors influence clients (Antle & Nalebuff, 1991). Therefore, observed outcomes such as earnings quality are used as proxies for audit quality (e.g. Becker, DeFond, Jiambalvo, & Subramanyam, 1998; Francis & Michas, 2013; Francis et al., 2014).

13 We also run the regressions using the ratio of hours allocated to partner only vs. senior and staff and, as an additional alternative, number of hours allocated to partners over the total number of hours allocated to that specific engagement. All results, which are described later in Section 7, are confirmed.

14 Consistent with the literature, we take the natural logarithm of audit fees.

15 The literature has thus far provided inconsistent evidence on the relationship between audit fees and audit quality. Some papers have, in fact, suggested a positive relationship between audit fees and the quality of the service provided (Larcker & Richardson, 2004; Srinidhi & Gul, 2007), whereas others find a negative (Antle, Gordon, Narayanamoorthy, & Zhou, 2006) or even no relationship (Ashbaugh, LaFond, & Mayhew, 2003; Chung & Kallapur, 2003).

16 For privacy reasons, we do not disclose directly the names of the audit firms.

17 As a robustness test, we also control for managers’ expertise; in particular, we collected managers’ ages and added them as additional control variables in logarithm form. All results obtained (and described later in Section 7) remained confirmed, and the natural logarithm of managers’ ages was not significant.

18 Given that the dependent variable is in absolute terms, we use the absolute value of cash flow from operations as an independent variable.

19 We used an industry classification based on the Italian classification named ATECO. This classification is the Italian version of the European nomenclature, Nace Rev. 2, published in the Official Journal of 20 December 2006 (Regulation (EC) no 1893/2006 of the European Parliament and of the Council of 20 December 2006).

20 The number of observations we consider in this paper is similar to the number used in published research that employ proprietary data (e.g. Schelleman & Knechel, 2010, examine 119 audit engagement from a single audit firm, and Knechel et al., 2009, run their models on 226 engagements).

21 Using data published from Osservatorio di Revisione (www.sdabocconi.it/osservatoriorevisione), we calculated that in the time span covered by our paper, there are approximately 300 listed companies in Italy, and an average of 10% are foreign companies. Excluding foreign and financial companies, the average number of the listed firms for the years examined in the manuscript is approximately 200. Considering the average market share of the two audit firms for which we have data in the years indicated, we have on average 55 engagements per year for the (summed) two audit firms, implying an average maximum of approximately 220 engagements (55 * 4 = 220).

22 All variables were winsorized at a 1% level of the distribution to control for outliers. We also tried a different winsorization threshold of 0.5% and the locked adaptive computationally efficient outlier nominator (BACON) algorithm proposed by Billor, Hadi, and Velleman (2000) and suggested by Weber (2010) as an alternative way to treat outliers. All results described below hold. We therefore believe that our results are not sensitive to outlier treatment.

23 Although proxies of risk have been included as control variables in our analyses, a different interpretation of our results cannot be wholly ruled out. Those engagements that are characterized by specific risks might in fact need more hours of leading auditors to interact more effectively with the client, and such engagements might be more likely to have a lower level of financial reporting quality, which we use as a proxy of auditing quality. We thank one of the anonymous reviewers for noting this possibility.

24 The first-stage regression is presented in appendix. Results hold also when adding number of subsidiaries as an additional instrumental variable.

25 The only exception is variable Same_uni, which is positive but not significant (p-value = .318). Given the low frequency of partner id variables, we were not able to run a logit regression (i.e. Mb as a dependent variable) to add partner id dummies.

26 Variable Aud_spec was computed on a larger sample that includes all listed non-financial companies and therefore not on the restricted dataset for which we have proprietary data.

27 That is ‘each Big 4 audit firm has its own unique set of internal working rules that guide and standardize the auditor’s application of auditing and accounting standards’ (Francis et al., 2014, p. 606).