Joint Audits: Does the Allocation of Audit Work Affect Audit Quality and Audit Fees?

Abstract

Audit quality and cost consequences of joint audits have been continually discussed, especially since the publication of the European Commission’s Green Paper in 2010. We provide new empirical evidence for the French audit market. We show that a more balanced audit work allocation between the engaged audit firms reduces the audit quality and enhances the audit fees as compared to an unbalanced work allocation. We measure the quality effects following the concept of abnormal accruals and the concept of cosmetic earnings management. As unbalanced joint audits have parallels to single audits, our results have interest to those debating the benefits and costs of joint audits as compared to single audits.

Keywords:

• allocation of audit work
• audit fees
• audit quality
• earnings management
• joint audit

JEL codes:

• M42
• M48

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 The Big Four audit firms are Deloitte Touche Tohmatsu, PricewaterhouseCoopers (PwC), Ernst & Young and KPMG.

2 This is why we observe just a small amount of non-audit fees received by the engaged audit firms (Autorité des marchés financiers (AMF), 2011). Nevertheless, some fees are still related, e.g. fees for fiscal services to foreign subsidaridies (André et al., 2016, p. 249).

3 It has to be mentioned that we do not know exactly how well the allocation of the audit fees between the joint auditors reflect the allocation of audit work between them. We use the fee allocation as a proxy for the work allocation as we think that having a larger stake in the audit leads to higher audit costs resulting in higher audit fees. However, we do not have any internal information about e.g. work hours.

4 Measuring audit quality by the concept of abnormal accruals is often critizised especially because the model demonstrates a poor ability to separate discretionary accruals from the normal (not subject to manipulation) accrual component of earnings. Furthermore, tests that use abnormal accruals are susceptible to misspecification because they do not consider correlated variables. See for advantages and disadvantages of the concept of abnormal accruals Dechow et al. (2012).

5 Thinggaard and Kiertzner (2008) note that their findings could be the result of the special competitive situation in Denmark, where the audit firms might compete for position as the preferred audit firm after the abolition of the joint audit requirement in Denmark.

6 See Ratzinger-Sakel, Audousset-Coulier, Kettunen, and Lesage (2013) for a joint audit literature overview.

7 The fact that after the joint audit abolishment in Denmark the majority of companies opted for single audits (Holm & Thinggaard, 2014, p. 140) shows that they do not see advantages in joint audits.

8 The risk that the managment will play off one auditor against each other was also one of the arguments for abolishing the joint audit requirement in Denmark in 2005 (Thinggaard & Kiertzner, 2008, p. 145). It has to be noted that auditors are appointed for a six-year period in France. Therefore the possibilities for clients to put pressure on their auditors should be rather small (André et al., 2016, p. 249).

9 See for the comparison of the audit evidence precision of single audits to joint audits Deng et al. (2014), p. 1043.

10 This may be due to the size of the companies in our dataset as former studies show that there is a significant relationship between client size and choosing a Big Four auditor. We analyze CAC60 and CACMid60 data which include in general large companies.

11 This result is in line with Kermiche and Piot (2016), p. 5.

12 We have kept companies with more than two audit firms in the database in order to avoid distortions due to excluding especially companies with the highest average fees in their respective group. A robustness test shows that the results are robust against reducing the database to 452 observations with only two auditors.

13 See for a review of audit fee research studies and the used control variables with a high explanatory power Holm and Thinggaard (2006) or Hay (2012).

14 Simunic (1980) uses this variable to model the sole auditor liability (Simunic, 1980, p. 173).

15 The audit fee model faces in the Hausman-test a prob = 0.0004 and a χ² = 38.95. Thus, the fixed effect model is the right specification for our empirical analyses. Random effects and pooled OLS estimators would be inconsistent (Cameron & Trivedi, 2010, p. 266).

16 It has to be noted that Chaney, Jeter, and Shivakumar (2004) show that for their dataset of unlisted firms Big Five audit firms are self-selected as, e.g. companies choose Big Five audit firms because they are best suited to the companies characteristics. In detail, they find that client firms choosing Big Five audit firms generally would have faced higher fees had they chosen non-Big Five audit firms, given their firm-specific characteristics. Consequently, it is not sure whether the companies in our dataset pay a Big Four premium or whether they generally would have faced higher fees had they chosen non-Big Four auditors.

17 We do not use the model of Jones (1991) and its extensions as it requires more data and would cause a large reduction in the number of observations.

18 Newcomb (1881) was the first researcher who identified the different frequencies of numbers. However, Benford (1938) rediscovered this law, and it was named after Frank Benford.

19 For details regarding the applicability of the Benford’s Law to reported earnings figures, see Van Caneghem (2004).

20 Also Van Caneghem (2004) uses the chi-square-test to verify the results from the z-value-test.

21 The chi-square-test must exceed 16.92 to be significant at the 5% level with 9 degrees of freedom. This threshold is not reached for one of the two subsamples. However, this does not disrupt our initial conclusion. A significant chi-square test only means that the observed distribution as a whole deviates from the expected distribution (Carslaw, 1988, p. 324). Such a deviation is not necessary to detect CEM. The chi-square test is never significant for a certain subsample when the threshold is 70%, 75% and 80%. The discussed results are robust against replacing audit fees by total fees.