Professional and Commercial Incentives in Audit Firms: Evidence on Partner Compensation

Abstract

Audit firms can shape and foster the desired behavior of their partners in accordance with the firm’s strategy and objectives through compensation schemes. Understanding financial incentives is therefore of great importance, especially to ensure that auditors actually deliver high quality work. This study analyzes the elements associated with the compensation of audit partners in Belgium, and specifically contributes to the literature by investigating how compensation schemes vary across firms of different sizes (Big 4, second-tier, and small audit firms). Our results indicate that the mixture of equal sharing and performance based compensation elements varies across firms of different sizes. Partner compensation in larger (Big 4 and second-tier) firms is more heavily based on performance based elements than in smaller audit firms. Furthermore, in firms of all sizes compensation appears to be strongly associated with commercially oriented elements while professionally oriented elements only play a minor role. Professionally oriented elements are, however, slightly more important in larger (Big 4 and second-tier) firms than in smaller firms.

Keywords:

• Partner Compensation
• Financial Incentives
• Equal Sharing
• Performance Based

JEL codes::

• M42
• M52
• J33

Acknowledgements

We thank Juha-Pekka Kallunki (the editor), an anonymous reviewer, Ann Vanstraelen, Joël Branson, Robert Knechel and several practitioners for their helpful comments. This paper has also benefited from the comments of the participants of the 2015 EAA Conference, the PhD workshop of the 8th European Auditing Research Network Symposium and the 2015 Limperg Workshop on Auditing and Assurance Research. We would also like to thank Graydon Belgium and Sanne Janssen for their assistance in collecting data.

Notes

1 The lock-step system is a remuneration system whereby subsets of partners (usually based on seniority) are remunerated equally (Burrows & Black, 1998; Malhotra et al., 2006).

2 We do not claim that brand-name capital is unimportant for second-tier audit firms, as these firms nowadays have also developed into international networks. However, due to their much smaller size, the second-tier firms have considerably less invested capital (partner equity) to protect (Boone et al., 2010). Consistent with Big 4 firms having more brand-name capital than second-tier audit firms, research also suggests that the Big 4 are still higher valued by the market (e.g., Boone et al., 2010; De Franco, Gavious, Jin, & Richardson, 2011).

3 The tension between the traditionally dominant trustee logic and the commercial logic in the audit profession is extensively discussed in both the academic (e.g., Lander et al., 2013; Gendron, Suddaby, & Lam, 2006) and professional literature. Our evidence on audit partner compensation is timely. In September 2014, for example, the Dutch professional body of auditors issued a report entitled ‘In the public interest’ with measures to improve the quality and independence of audits in the Netherlands. This report also mentions measures for partner compensation.

4 Audit firms of course also incentivize professional values in order to counterbalance the conflicting demands of commercialism and professionalism (Johansen & Christoffersen, 2017; Knechel et al., 2013). Qualitative research, however, suggests that the commercial dimension tends to dominate in larger firms (e.g., Coram & Robinson, 2017).

5 We thank an anonymous reviewer for pointing this out. The way in which commercialism is performed might, however, still differ between smaller and larger firms. For example, the provision of non-audit services to audit clients might be more important for larger audit firms.

6 In total there are about 1000 registered auditors in Belgium. A large number of registered auditors, however, are sole practitioners and thus not affiliated with any audit firm (≈15%) or do not sign audit opinions (≈30%). Such auditors are not part of our population of interest. Based upon partnership size, BDO, Mazars, RSM, and PKF are second-tier firms in Belgium.

7 The nominal corporate tax rate in Belgium is 33.99%, whereas an individual’s net taxable income is subject to progressive tax rates between 25% and 50% (plus additional municipal taxes).

8 Audit firms that carry out statutory audits of the annual accounts of public-interest entities have to publish annual transparency reports (European Commission, 2006).

9 The sum of Big 4, second-tier, and small audit partners (245) exceeds the total number of individual auditors in our sample (241) because some individuals switched audit firms during our sample period.

10 To control for selection bias and to examine any differences between the partners in our sample and the excluded partners without a BVBA/SPRL, a comparison of means for all independent variables used in this study was performed. This (untabulated) analysis showed that the average partner in our sample is very similar to the average partner without BVBA/SPRL.

11 A comparison of means (untabulated) suggests that these 25 observations are representative as there were no important differences between these 25 observations and our full sample.

12 On average, our compensation measure contains only 6% of financial revenues. Our results are unchanged if we exclude financial revenues from our compensation metric.

13 As we do not have information on the income source, we acknowledge that our compensation metric may include some income from other activities than audit-related work (e.g., part-time teaching at a university). We overcome this limitation by using a mixed model with random auditor-specific intercept to capture auditor-specific variation that might influence compensation.

14 The Belgian thresholds for audit exemption were until 2015: (i) more than 50 employees, (ii) sales of more than 7.3 million euros, and (iii) total assets of more than 3.65 million euros.

15 Informal conversations with audit partners as well as audit firms’ transparency reports confirm that compensation decisions are taken at the national firm level.

16 As a further sensitivity test we also ran a difference-in-differences model. Results of this analysis (untabulated) were quantitatively and qualitatively similar to those reported later on in the paper. As such model captures unobservable partner-specific income and controls for time-invariant unobserved effects (e.g., client characteristics), this further alleviates concerns about unobservable omitted variable bias.

17 Including random effects accounts for correlations among observations from the same individual and over time. Additionally, random effects indicate the between-individual variability. The more between-individual variability there is, the higher the correlation between observations from the same individual. $b_i$ represents the random intercept for each individual. It is a residual term that measures the distance from each individual’s intercept around the overall intercept ß₀. Rather than calculating an estimate of those distances for every auditor, the model is able to estimate just a single variance ${\sigma }_b^2$. That variance parameter estimate is the between-individual variance. The variance of the residuals is the within-individual variance. Their sum is the total variance in the dependent variable that is not explained by the independent variables.

18 The professional and commercial logic can co-exist within an audit firm and firms’ compensation schemes may thus contain both elements grounded in a professional and in a commercial logic. Furthermore, incentives aimed directly at increasing audit quality (i.e., incentives stemming from a professional logic) might positively contribute to profitability, or vice versa. The essential issue, however, is whether increasing audit quality is the primary goal (professional logic) or a by-product (commercial logic) of the compensation element.

19 We acknowledge that all audit quality measures are subject to interpretation and measurement error. We, therefore, include two different proxies that capture different dimensions of audit quality.

20 Issuing a going-concern opinion without client failure within the next 12 months is labeled as a Type I error and issuing a clean opinion with client failure within the next 12 months is labeled as a Type II error.

21 As one might argue about classifying industry specialization as a professionally oriented element, we performed additional sensitivity tests in which industry specialization was categorized as commercially oriented (see infra).

22 We classify compensation for auditing listed clients as a commercially oriented element due to its direct link with revenue generation. However, as one might argue that compensating for auditing listed clients might be a way for audit firms to incentivize ‘public-client specialization’ (in order to improve audit quality for such clients), we performed additional sensitivity tests in which the number of listed clients was categorized as professionally oriented (see infra).

23 AUDITFEES_F, NASFEES_F, VAR_TOT_FEES_F, and PROBF_CLIENTS_F are thus calculated based upon data from all audited Belgian companies. Likewise, OFFICE_SIZE represents the number of auditors within an office (irrespective of whether these auditors are included in our sample to estimate partner compensation).

24 There are ≈150 listed companies in Belgium (NBB, 2015).

25 Similar to Knechel et al. (2015), we use mean levels of signed abnormal accruals in auditor-conducted prior year audits as a proxy for audit quality at the partner-level. Accruals are calculated based on the modified Jones model, with an additional control for firm performance (Kothari, Leone, & Wasley, 2005).

26 The difference in compensation between second-tier and small firm partners is not statistically significant. Similar to studies on law firms (e.g., Leibowitz & Tollison, 1980), our results suggest that Big 4 partners earn more than partners in non-Big 4 firms. However, due to the fact that our compensation measure does not account for operating costs (see supra), the relative difference between Big 4 and second-tier firms is exaggerated.

27 The Bonferroni post-hoc test shows no statistical difference between the financial health of Big 4 and second-tier partners’ clients.

28 This is consistent with the descriptive results of prior Belgian studies (e.g., Hardies, Vandenhaute, & Breesch, 2018) and can be explained by the fact that Big 4 firms also have more financially unhealthy clients.

29 This seemingly surprising result can be explained by the fact that errors correlate strongly with number of clients and with fees.

30 The significant values of the LR test in all models (p < .01) confirm the importance of using models with random effects.

31 Unstandardized coefficients for elements of audit partner compensation for both pooled and separate regressions for Big 4, second-tier, and small audit firms are presented in Appendix A.

32 We use absolute values because we are interested in the relative strength of the regression coefficients.

33 Private discussions with partners indicate that profit in second-tier firms is often distributed with delay (sometimes even only after several years). This could explain why we do not observe a positive association between profit and compensation within the subsample of second-tier firms. Additional analyses (untabulated) in which we replace current year profits by delayed profits (i.e., profits in year t-1 or average profits since year t-3) support this assertion.

34 Results are similar when we measure OFFICE_SIZE as the natural logarithm of the number of registered auditors in office f in year t.

35 The relationship between compensation and experience is in fact non-linear, as is evident from the negative coefficient for the variable measuring the square of EXPERIENCE (untabulated). This variable is excluded from the tabulated model to avoid unstable coefficients due to multicollinearity.

36 We know from prior literature as well as from informal conversations with partners that larger audit firms adopt mandatory retirement policies for partners at age 60 or below (V. L. Brown, 2017).

37 As one might debate on whether industry specialization (number of listed client) is an indication of professional (commercial) performance, we ran sensitivity tests in which industry specialization (number of listed clients) was categorized as a commercial (professional) performance indicator. The results (untabulated) were qualitatively and quantitatively similar to those reported, confirming the robustness of our results.

38 For individually generated (non-)audit fees we considered years t and t-1 as there might be a time lag between the generation of fees and its effect on compensation. Furthermore, compensation might depend on the evolution in performance or the achievement of new goals based on previous performances. In Big 4 and small firms compensation is associated with fees of last year (LOLD_FEES_P), in second-tier firms with current year fees (LFEES_P). For other commercial performance based variables, the results are qualitatively similar when using year t or t-1.

39 FEMALE is insignificant in the Belgian sample, but it is likely that the number of women in the sample is too small for this coefficient to reach statistical significance. The coefficient is, however, negative and statistically indifferent from the coefficient in the Swedish sample. Furthermore, previous studies show that the gender pay gap is larger in Sweden than in Belgium (European Union, 2013).

40 To facilitate comparison, we define the measure of IND_SPEC in the same way as KNZ.

41 MEDIAN_RISK is insignificant in the Swedish sample, but the coefficient sign is in the same direction. The sign of the coefficient of STD_PARTNERS_ASSETS for the Belgian sample in Table 7 is positive (while being negative in Sweden). This is likely to be due to correlated omitted variables. Results in Table 7 suggest that compensation in Belgian Big 4 firms increases with the financial health of firm-wide client portfolio. A possible explanation for the different coefficient in our main analysis is that we do not control for fees in the former model and therefore observe a positive association.

42 During our sample period only four audit partners switched audit firms.