The pricing of initial audit engagements by big 4 and leading mid-tier auditors

Abstract

The recent investigation of the UK audit market by the Competition Commission testifies to perennial regulatory concerns regarding increasing supplier concentration, big 4 dominance of large company audits and the capacity of mid-tier auditors to compete. Against this backdrop, this paper presents new evidence on whether there is competitive pricing for initial audit engagements by big 4 auditors relative to their next four largest mid-tier (mid 4) counterparts for the UK quoted and private corporate sectors. Based on data from FAME for 2007 and 2010, the evidence indicates that larger quoted companies switching between the big 4 benefit from substantial discounts, with smaller discounts attracted by clients switching to the mid 4. Coupled with evidence that fees for both audit and non-audit services recover in subsequent periods, and consistent with the theoretical framework, the paper concludes that big 4 discounting is a competitive outcome aimed at securing future economic rents. New evidence demonstrates that smaller clients switching to big 4 or mid 4 auditors do not benefit from low-balling.

Keywords:

• audit competition
• audit fees
• initial audit engagements
• fee discounting
• price recovery

Acknowledgements

I am grateful to two anonymous reviewers for detailed comments, insights, suggestions and advice on earlier drafts. I am also grateful to Mark Clatworthy for helpful discussions and comments. Any errors are those of the author.

Notes

1. For example, Clatworthy and Peel (2007) reported that only 7.8% (7.7%) of the audits of a large sample of private companies were conducted by big 4 (mid 4) auditors.

2. In fact, DeAngelo (p. 126) concludes that these property rights (rents) lower the optimal level of independence and ‘are a necessary condition for lessened independence’. I am grateful to reviewer for drawing attention to this and number of pricing issues in Section 2.

3. For example, based on quoted corporate survey data, Beattie and Fearnley (1995, p. 235) report that some auditor switches may be motivated by opinion shopping since ‘Disagreements concerning accounting principles were cited by 9% of companies as a reason for considering auditor change, with 4% citing disagreement with the audit opinion’.

4. Importantly, Ghosh and Lustgarten (2006) test this hypothesis with reference to non-big to non-big 4 switches. The current paper examines non-big to mid 4 auditor switches; but empirical findings in this paper (including for private companies) are still contrary to Ghosh and Lustgarten's (2006) general conjecture.

5. Start-up costs include the new audit firm familiarising itself with the company's management and personnel and its risk and control systems, together with the requirement to verify opening balances (DeAngelo 1981). These costs should decline in subsequent periods as efficiency improves in line with standard learning curve effects.

6. In this study, it was not possible to decompose mean discounts to reflect differentials determined by such factors as market share, degree of competition (bidding) and industry expertise. For instance, for the latter, Hay and Jeter (2011, p. 189) find ‘strong evidence of fee premiums to city-level specialists in New Zealand and no evidence of premiums to national specialists’. A more complex research design, such as a time series panel (including more switch observations), would facilitate a more refined analysis.

7. For instance, two potential auditors with high levels of industry expertise may still compete vigorously for a new audit.

8. In fact, the year dummy variables are statistically insignificant for all quoted company models.

9. An attempt was made to discover some of the causes for non-matching cases by cross referencing them to the FAME database. In total, 1112 companies (88% of them private) appeared to have switched to non-mid 4 or non-big 4 auditors or to have no auditor. Furthermore, the size limits for small and medium-sized private companies have been relaxed further. Given the recession, more companies may have taken advantage of the audit exemption. Other causes of sample attrition include company mergers and corporate dissolutions, amalgamations and failures, which may not appear on the updated (or for the years selected) FAME discs.

10. Including these variables (FAIL and DISS) also controls for potential survivorship bias, particularly in the matched samples.

11. Both AIM and Ofex were launched in 1995 for smaller, riskier companies. Relative to the Main Market, both markets are less regulated, the stocks quoted on them less liquid and they have a lower profile. Ofex is the least regulated and liquid market, with the lowest profile. In October 2006, Ofex changed its name to the ‘PLUS-quoted’ market.

12. In general, unlimited companies need not file their accounts at CH, though some voluntarily do so.

13. This compares to a difference in total audit fee income between the largest and fourth largest mid-tier (mid 4) firms of only £9.6 m. Subsequent analysis revealed that these auditors maintained their dominance in the rankings over subsequent years.

14. Current and previous editions of the same FRC source, together with Accountancy magazine (both of which tabulate mergers of accountancy firms), were examined and Internet searches were conducted to ensure this was the case.

15. None of the models produced more than one value >1, with remaining values being relatively low.

16. For example, when these observations were omitted the SWITCH coefficient in Model 1 changed from −0.0644 to −0.0641.

17. Note that there are too few observations to estimate separate sub-market models for unlimited and public unquoted companies. However, these companies are included in Model 1 together with their associated variables: UNLIM and UNQPLC.

18. Discounts are calculated via the transformation (e^x) − 1.

19. Following Francis (1984), Models 1–3 were re-estimated excluding switching companies. The coefficients from each model were then used to predict the audit fees of companies in the associated switching samples. Residuals in the switching sample were then calculated. Evidence of significant (mean) negative residuals is consistent with discounting since the model excluding switching companies would tend to predict larger audit fees in the switching sample than were actually charged. The results are highly congruent with the original findings. Mean residuals differ from the SWITCH coefficients reported in Table 3 to only the third or fourth decimal place. They are: 0.06434 (Model 1); −0.0199 (Model 2); and −0.2045 (Model 3). Based on t-tests, the estimated mean residuals for Models 1 and 3 differed significantly from zero at p ≤ 0.01. This level of conformity employing two different methodologies is not due to a statistical tautology. It suggests a high degree of model robustness of the estimated discounts.

20. To obtain this result, private and quoted companies were pooled to test whether the discount for quoted switches was higher than the (insignificant) one for private companies. The difference in discounts (18.1%) is statistically significant at p = 0.01.

21. Note also that whereas the coefficient of SBIG4B4 is significantly different from zero (Model 3) that of SNBIG4B4 is not.

22. The mean difference in total assets (LNASSET) is significant at p = 0.01. The medians (Mann–Whitney test) for total assets also differ significantly between the two samples (p = 0.01).

23. A regression model with the log of NAS fees as the dependent variable in Model 3 confirmed this, with the SWITCH coefficient again exhibiting a negative and highly significant coefficient (p = 0.01).

24. Whilst it is reasonable to treat companies that had switched auditors in earlier periods as proxies to test for evidence of price recovery, it is possible that initial audit discounts were not received (as assumed) by these companies.

25. When added together (SWITCH1 + SWITCH4) and included in Model 6 (as reported) the coefficient (p-value) was 0.161(0.63).

26. The coefficient of SWITCH14 was positive (0.120) and highly insignificant (p = 0.79) when the log of NAS fees was specified as the dependent variable. In an alternative specification, the coefficients (p-values) of SWITCH2 and SWITCH3 were −0.688 (0.57) and 0.185 (0.71), respectively. For the model with log of total fees as the dependent variable, they were −0.224 (0.32) and 0.069 (0.48).

27. Note that UK APB Ethical Standard 4 states: ‘The audit engagement partner shall ensure that audit fees are not influenced or determined by the provision of non-audit services to the audited entity’. The EU is currently examining proposals to ban joint audit and NAS appointments.

28. For instance, although recognising the need to balance the costs and gains from tendering and switching (p. 201), the CC (2013) is concerned about low switching rates and has proposed remedies regarding mandatory rotation and tendering. A further proposed CC remedy is the prohibition of big 4 only clauses in loan agreements with the aim of reducing barriers to entry and increasing companies' willingness to switch via a larger pool of available auditors.