Auditor independence impairment: bonding between clients and individual engagement partners

ABSTRACT

In this study, I investigate whether the economic bond between an individual audit engagement partner and a client threatens auditor independence (and thus audit quality). As the propensity of the auditor to issue going-concern modified audit opinions (GCOs) is lower when there is auditor independence impairment, I examine to which extent fee dependency leads to a lower propensity to issue GCOs. Using archival data from Belgian private companies, I empirically examine a sample of 18,638 companies for the period 2008–2010 to test for the effect of the auditor-client bond on the audit opinion. None of my results support common concerns that client importance undermines auditors’ GCO reporting decisions, neither at the firm level nor at the partner level. Specifically, I find no evidence for a negative association between the likelihood that an auditor issues a GCO and the economic importance of that client to the auditor (firm or partner). These results are supported by analyses on an additional sample of 5,934 companies for the period 2011–2013. Despite broad concerns among academics and regulators about auditor independence and economic bonding, this is one of the first studies to investigate this issue at the partner level and for private clients.

Resumen

En este trabajo se analiza si el vínculo económico entre un socio individual contratado para realizar la auditoría y su cliente compromete la independencia del auditor (y consecuentemente la calidad de la auditoría). Como la propensión del auditor a emitir una opinión con salvedades por gestión continuada disminuye cuando hay deterioro de la independencia del auditor, se investiga hasta qué punto la dependencia económica de los honorarios de ese cliente reduce la propensión a emitir opiniones con salvedades. Utilizando datos de sociedades privadas belgas, se examina empíricamente una muestra de 18.638 sociedades en el período 2008 – 2010 para comprobar el efecto del vínculo entre auditor y cliente sobre la opinión del auditor. Los resultados no confirman la preocupación habitual de que la importancia del cliente socava la decisión del auditor, ni a nivel de la empresa ni del socio. En concreto, no se encuentra evidencia de una asociación negativa entre la probabilidad de que el auditor emita una opinión con salvedades por gestión continuada y la importancia económica del cliente para el auditor (empresa o socio). Análisis con un muestra adicional de 5.934 sociedades en el período 2011–2013 apoyan estos resultados. A pesar de la amplia preocupación entre académicos y reguladores sobre la independencia del auditor y el vínculo económico, este es uno de los primeros trabajos en investigar este asunto a nivel del socio y para clientes privados.

KEYWORDS:

• Client importance
• audit partners
• audit quality
• auditor independence
• going-concern

PALABRAS CLAVES:

• Importancia del cliente
• socios de auditoría
• calidad de la auditoría
• independencia del auditor
• informes con salvedades

Acknowledgments

This paper has benefited from the comments of Nieves Carrera (the editor), two anonymous reviewers, Marleen Willekens, Robert Knechel, Reiner Quick, Diane Breesch, and Joël Branson. I would also like to thank the workshop participants at the 6^th European Auditing Research Network (EARNet) Symposium (2011), AFI seminar at Katholieke Universiteit Leuven (2012), and the Auditing Section Midyear Conference (2012). I gratefully acknowledge the help of Eric Van den Broele (Graydon Belgium), Marloes Beijer, Valerie Henrion, Jan De Muylder, Marie-Laure Vandenhaute, and Sanne Janssen for their assistance in collecting data.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1. See Table 1 in Li (2009) and Table 7.1. in Sharma (2014) for overviews of prior studies on fee dependence and auditor independence.

2. Consistent with the general believe that auditor independence impairment is more likely for private firms, a recent study by Bell et al. (2015) shows that fees for non-audit services have differential effects on quality in audits of private (negative association) than in audits of listed firms (positive association).

3. The dependency of an auditor to any single client in China might be exaggerated because the study of Chen et al. (2010) did not include private companies.

4. In settings where data on audit fees are unavailable, it is common to use total assets as a surrogate to measure client importance. This might seem appropriate given that client size generally accounts for a large portion of the variation in audit fees, but it inevitably introduces noise in the measures of client importance. In the context of auditor industry specialisation, Audousset-Coulier, Jeny, and Jiang (2016) showed in fact that client-size based measures are poor surrogates for audit-fee based measures and therefore recommend the use of audit-fee based measures. In my sample, the correlation between client importance measured by total assets and client importance measured by total fees is only 0.67 (at the partner level).

5. Based upon actual compensation data, Knechel, Niemi, and Zerni (2013) showed that in Sweden audit partners’ compensation is positively associated with their revenue potential (i.e. the size of their personal client portfolio and the acquisition of new clients). A recent paper by Vandenhaute, Hardies, and Breesch (2019) reports similar findings for Belgium.

6. Findings from experimental settings might not extrapolate to naturally occurring settings, for example because it might be that quality control mechanisms successfully prevent inherent biases from ultimately impacting audit outcomes (Church, Jenkins, McCracken, Roush, & Stanley, 2015). Archival studies on the relation between economic dependence and auditor independence (at the partner level) therefore remain necessary.

7. Hardies et al. (2018) report that since 2000 there have been only five lawsuits in Belgium in which an auditor was involved. In three of these cases, charges against the auditor were dismissed. The two other cases ended in a settlement. Litigation risk is also limited due to the existence of a liability cap; furthermore, this cap is much lower for audits of private clients than for audits of listed clients (€3,000,000 vs. €12,000,000).

8. During the sample period, firms were considered large if they meet at least two of the following criteria: (1) turnover (revenue) (excl. VAT) >€7,300,000; (2) total assets >€3,650,000; and (3) number of employees (yearly average) >50. These criteria are considered on a consolidated basis if the firm belongs to a group that publishes consolidated statements or if the firm is a holding or a listed firm. Listed firms and firms with more than 100 employees were always considered large.

9. This restriction criterion minimises potential measurement error in CI^F and CI^A. When firms have just a few partners it is difficult to separate firm-level effects from partner-level effects (within the sample of firms with less than three audit partners, the correlation between CI^F and CI^A is .879). My results are robust to applying even more restrictive inclusion criteria (i.e. requiring a minimum of four or more partners per audit firm).

10. To minimise the potential measurement error in CI^F, I used all companies in Belgium for which fee data were reported in the period 2008–2010 to compute the base $\sum _{i=1}^{n}LnTFE{E}_i$.

11. Results are unaffected by the use of different cut-off values for EXPERT^F and EXPERT^A.

12. Results are unaffected if portfolio size is measured by total audited total assets rather than by number of clients.

13. Before computing the VIF in OLS regression, I first ruled out nonlinearity using Box-Tidwell Transformation.

14. The area under the curve (AUC) summarises the discrimination ability of a model (i.e. the ability of the model to correctly classify those firms with and without a GCO), with higher values indicating higher discrimination ability.

15. Table 6 presents results for subsample analyses where client importance is measured based on total fees (CI^F and CI^A), but results are similar for the other measures of client importance.

16. All Belgian companies that employ on average more than 100 employees are required by law to have a works council (i.e. a body with parity representation of employers and employees, with the purpose of implementing social legislation).

17. Results are similar if I use a 1 or 3 year bankruptcy cut-off.

18. The confusion matrix shows that the performance of this model is high, with an overall classification accuracy of 84.6% (see Appendix Table A3). The true positive rate is 69.1% and the true negative rate is 92.4%.

Additional information

Funding

This work was supported by the Research Fund of the University of Antwerp (Grant #29447).