Do (Fe)Male Auditors Impair Audit Quality? Evidence from Going-Concern Opinions

Abstract

Recent research indicates that there may be a relationship between the characteristics of the audit engagement partner and audit quality. In this paper, we examine the relationship between audit quality and the presence of a female or male audit engagement partner. We use the likelihood that an auditor issues a going-concern opinion (GCO), conditional on the client's financial situation, as an indicator of audit quality. Using a sample of 7105 financially distressed, private Belgian companies, we find that female auditors are, ceteris paribus, more likely to issue GCOs than male auditors. Our results also show that this effect is stronger when clients are either important (i.e. represent a material portion of the auditor's revenues) or high-risk (i.e. associated with increased uncertainties and risks). Collectively, these results indicate higher audit quality by female auditors.

Acknowledgements

We would like to thank Ann Vanstraelen (the associate editor), two anonymous reviewers, Marleen Willekens, Alison Woodward, Bruno Heyndels, Will Ciconte, Jean C. Bedard, and W. Robert Knechel for comments on an earlier version of this paper. This paper has also benefited from the comments of the participants of the 34th Annual Congress of the European Accounting Association (2011) and the workshop participants at the Université Catholique de Louvain (2010). We also thank Eric Van den Broele (Graydon Belgium), Marloes Beijer, and Valerie Henrion for their assistance in collecting data.Data availability: The data are publicly available from the sources identified in the paper.

Notes

¹Arguably, the propensity to issue GCOs is a more direct test of auditor independence than measures of earnings management (DeFond et al., 2002; Hope & Langli, 2010) because (1) audit opinions are directly and unambiguously measurable and (2) the auditor is solely responsible for the audit opinion. Therefore, in this paper, we use the propensity to issue GCOs, rather than the extent of earnings management as an indicator of auditor independence impairment (and thus audit quality).

²Audit firms have contracting and monitoring mechanisms in place to mitigate the moral hazard inherent to their decentralised organisational structures (e.g. firm-wide profit-sharing rules that lower incentives for individual partners to take on risky clients; Bedard, Reis, Curtis, & Jenkins, 2008; Burrows & Black, 1998).

³The tone at the top is ‘the ethical environment within the firm created through management practices and espoused values’ (Douglas, Davidson, & Schwartz, 2001, p. 107). Together with an audit firm's culture, the tone at the top establishes whether the firm is quality-orientated (i.e. the audit is a high quality service) or revenue-orientated (i.e. the audit is a mere commodity) (Jenkins, Deis, Bedard, & Curtis, 2008).

⁴Note that we do not argue that such a relationship might exist because female and male auditors are inherently different (as a consequence of inherent differences between men and women; see Hardies and Khalifa (2014) for a detailed critique of such arguments).

⁵Belgian Company Law (Article 96) requires management to justify in the annual report the application of valuation rules in the assumption of continuity if the balance sheet shows an accumulated loss or if the profit and loss account shows a bottom-line loss in two successive years.

⁶Companies are considered to be large if they meet at least two of the following criteria: (1) turnover (excluding VAT) >7,300,000 euros; (2) total assets >3,650,000 euros; and (3) number of employees (yearly average) >50. These criteria must be considered on a consolidated basis if the company belongs to a group that publishes consolidated statements or if the company is a holding or a public company. Public companies and companies with more than 100 employees are always considered to be large.

⁷Rather than using our sample of 7105 financially distressed companies, we used the full sample of 17,066 companies to calculate the various specifications of CLIENT_IMP. The variables SPECFIRM, SPECAP, PORTFOLIO, and BUSY were also calculated based upon the full sample of 17,066 companies.

⁸We measure client importance at the individual partner level, rather than at the firm level, because client importance measured at the individual partner level seems to be more powerful in capturing audit reporting decisions (Chen et al., 2010). In supplementary analyses, we also discuss the results of client importance variables measured at the audit firm level.

⁹It is worth noting that the interpretation of coefficients of interaction variables is not as straightforward in logistic models as it is in linear models because logistic regressions calculate changes in the odds ratio of the dependent variable, not changes in the dependent variable itself. An interpretation that is useful in this study is that an interaction between SEX and another variable Y tells us by how much the effect of Y differs between female and male auditors in multiplicative terms (Buis, 2010). Because negative coefficients (β) lead to odds ratios (e^β ) less than one, negative coefficients for SEX * CLIENT_IMP would indicate that for female auditors the effect of CLIENT_IMP on the odds that a client receives a GCO is less than one time the effect of CLIENT_IMP on the odds that a client receives a GCO for male auditors.

¹⁰Positive coefficients lead to odds ratios greater than one, so positive coefficients for SEX * CLIENT_RISK would indicate that for female auditors the effect of CLIENT_RISK on the odds that a client receives a GCO is more than one time the effect of CLIENT_RISK on the odds that a client receives a GCO for male auditors.

¹¹In this model (the ‘FITO-metric’), eight variables are first logit-transformed and then equally weighted: (1) Gross added value/personnel employed, (2) Net return on total assets before taxes, (3) Net return on equity after taxes, (4) Self-financing level, (5) General level of financial independence, (6) Short-term financial debt level, (7) Free cash flow, and (8) (Cash + short-term investments – short-term financial debt)/current assets (Ooghe & Spaenjers, 2005). A higher score indicates a healthier company.

¹²Our results are unchanged if we use alternative definitions of EXPERIENCE such as the natural logarithm of the number of years the company's auditor has been legally authorised to sign audit opinions or dummy variables based on, for example, 5 or 10 year cut-off points.

¹³Belgium is divided into two large regions, the Dutch-speaking region of Flanders in the north and the French-speaking southern region of Wallonia. Registered auditors have to indicate their linguistic affiliation.

¹⁴Before computing the VIF in an OLS regression, we first ruled out nonlinearity using Box–Tidwell transformation.

¹⁵As explained in footnote 9, the interaction variable SEX * CLIENT_IMP can be interpreted as the extent to which the effect of CLIENT_IMP differs between female and male auditors in multiplicative terms. For example, the effect of CLIENT_FEES on female auditors issuing a GCO is only 0.38 times (= e^−.982) the effect of CLIENT_FEES for male auditors. For male auditors each unit increase in CLIENT_FEES decreases the odds that a GCO is issued by 0.40 (= e^−.909). For female auditors each unit increase in CLIENT_FEES only decreases the odds that a GCO is issued by 0.15 (= 0.40 * 0.38).

¹⁶These results are obtained in the same way as those for CLIENT_IMP (see footnote 15). For example, the effect of LEVERAGE is 1.6 times (= e^.491) larger for female than for male auditors. For male auditors each unit increase in LEVERAGE increases the odds that a GCO is issued by 2.7 (= e^1.006). For female auditors each unit increase in LEVERAGE increases the odds that a GCO is issued by 4.5 (= 1.6 * 2.7).

¹⁷We wish to thank an anonymous reviewer for bringing this to our attention.

¹⁸The results of the (unreported) control variables are almost identical to those reported in Table 4.

¹⁹In addition, we also used client importance measures for which the denominator was not based on the auditor's total revenue from all clients. Specifically, we performed additional sensitivity tests in which client importance was measured as (1) the ratio of a client's audit fee to the auditor's revenue from audit services from all clients and (2) the ratio of a client's non-audit fees to the auditor's revenue from non-audit services from all clients. We performed these sensitivity tests both at the individual partner level and the firm level. The results (not tabulated) were qualitatively and quantitatively similar to those reported.

²⁰We wish to thank an anonymous reviewer for bringing this to our attention.

²¹By excluding certain companies from our sample (e.g. public companies, financial institutions), our sample contains somewhat more voluntarily audited companies (32.7%) than the population as a whole; Sarens, Reheul, Van Caneghem, De Vlamincken, and Dierick (2012) report that 4561 Belgian companies (25%) voluntarily appointed an auditor in 2009.