Investors' perceptions of nonaudit services and their type in Germany: The financial crisis as a turning point

2.1 Litigation risk and oversight setting

German investors' perceptions of auditor-provided NAS may be influenced by certain specifics of the German regulatory environment, particularly auditors' litigation risk, public oversight of auditors, and the regulation of the provision of NAS, including the disclosure of auditor fees. These German specifics are presented in the following and serve, together with prior research, to motivate the hypotheses.

The literature indicates that the investor protection environment and auditors' litigation exposures affect financial reporting and audit quality (e.g., Ball, Kothari, & Robin, 2000; Ball & Shivakumar, 2008; Choi & Wong, 2007; Djankov, McLiesh, & Shleifer, 2007; Francis, Khurana, & Pereira, 2003; Francis & Wang, 2008; Gul, Zhou, & Zhu, 2013; Leuz, Nanda, & Wysocki, 2003). Investor protection is considered being weak in Germany (Djankov et al., 2007; Gul et al., 2013; La Porta, Lopez-de-Silanes, Sheifer, & Vishney, 1998; La Porta, Lopez-de-Silanes, Sheifer, & Vishney, 2000). For example, Gul et al. (2013) derive scores of investor protection based on a factor analysis on four investor protection indexes (anti-director rights index, index of disclosure requirement, index of liability standard, and index of public enforcement). They report a score of −2.071 for Germany, while the USA and the UK score 0.253 and 0.939 respectively. Only two of the 30 countries in the sample have a lower investor protection score than Germany.

Litigation exposure is viewed as perhaps the most effective mechanisms to discipline auditors. The German Commercial Code caps auditors' liabilities for negligent misconduct toward audit clients; set at €4 million for listed clients. The scope for third parties to pursue actions against auditors is very limited; the German Civil Code requires that an intentional violation is established. The intent requirement severely restricts investors from suing auditors for tort actions. Given the nature of an audit, it is extremely difficult to prove that the statutory auditor acted intentionally. In addition to the Civil Code, case law may hold auditors liable to third parties for negligent misconduct.⁵ Judicial decisions to compensate for auditors' negligence based on previous legal cases are, however, very rare (relating to a very restricted set of circumstances and only to other cases with similar issues or facts) and normally apply a similar liability cap as previously mentioned (Gietzmann & Quick, 1998; Schmidt & Feldmüller, 2014). Thus, the German Civil Code and case law hold auditors liable only to a very limited degree for negligent misconduct. We believe, therefore, that the low risk of auditor litigation in Germany may limit investors' comfort that auditors resist client-induced biases in financial reporting. It is possible, however, that even in a low-litigation risk environment investors' concerns may be compensated for by investors' reliance on auditors' incentives to avoid reputation losses (DeFond, Raghunandan, & Subramanyam, 2002; Hope & Langli, 2010). Other compensating mechanisms, such as effective public oversight of auditors, may also discipline auditors' behaviors and bolster investors' trust in auditor independence and the fairness of audited financial statements.

A professional body, the German Chamber of Auditors, monitors auditor compliance with their professional duties, including auditor independence and objectivity.⁶ The Chamber organizes a system of external quality controls, which are peer reviews, of audit practices that perform statutory audits and is responsible for disciplinary observance. The Chamber sanctions auditor misconduct, but disciplinary actions are normally not made public and the identity of the disciplined party is never disclosed. Moreover, severe sanctions, such as suspension or exclusion from the profession, are rarely applied (Wirtschaftsprüferkammer, 2014). Until recently, a public body, the AOC, supervised the activities of the Chamber, including the quality control process and disciplinary observance. Members of the AOC held only part-time positions. From 2007 the AOC became responsible for inspections of PIE audits and audit firms with PIE clients; annually for audit firms with more than 25 PIE clients and at least every third year for other firms. The Chamber's inspectors performed these inspections on the AOC's behalf.⁷ The modest role of the public oversight of auditors in Germany in our sample period, in addition to the lack of transparency in disciplinary cases against auditors, calls into question the effectiveness of the German system of oversight of auditors in addressing investors' concerns about auditors' incentives to control client-induced biases in financial reporting.

Although the investor protection environment, auditors' litigation exposure, and public oversight of auditors are commonly believed to be fundamental mechanisms for disciplining auditors, the potential perceived adverse effects of auditor-provided NAS may also be controlled by restricting the provision of NAS. Further, adequate disclosure of auditor fees may be crucial in the formation of investors' perceptions (DeAngelo, 1981a).

2.2 German nonaudit services regulations

The German Commercial Code regulates the provision of NAS by the statutory auditor to audit clients and requires auditor fee disclosure in the notes to the financial statements. From June 2016, additional restrictions on NAS are imposed on German PIEs as a result of the new EU regulatory decision, and these are discussed in Section 2.3. The Commercial Code reflects a mixture of principle-based and rule-based approaches, and many legal terms in the Code are open for interpretation. See Appendix 1 for prohibited specific NAS in the Code and in our sample period.⁸

In the paper we refer to the services beyond the statutory audit as assurance NAS, tax NAS, and other NAS. The Code does not require finer specification within NAS categories, and no official guidance exists for which service to include in a category. Sometimes, however, companies voluntarily disclose in the notes the types of services within a given category.

Assurance NAS typically include voluntary audits of consolidated entities and statutory audit-related services such as review of interim financial reporting and assurance of pro-forma financial information. Inspection of fee disclosure notes to the financial statements of our sampled companies reveals that assurance NAS may occasionally include services such as due diligence, assurance related to bond issues, assurance of internal control systems, assurance of implementation of information technology (IT) systems, issue of comfort letters, and confirmation of debt covenants.

The financial statement notes also show that commonly demanded tax services include tax declaration support, review of tax assessment notes, tax planning, tax advice regarding transactions, transfer pricing issues, tax advice to German employees with temporary appointments abroad, tax due diligence, tax advice regarding a restructuring, or support with regard to tax audits.

Our category, “other NAS” includes all NAS other than assurance NAS and tax NAS. This residual category may include a variety of NAS (e.g., training and consulting with regard to IT systems, expert opinions on managerial problems, financial due diligence, merger and acquisition advice, and bond issue advice) (Sattler, 2011, p. 118; IDW, 2012, p. 776). The other NAS fee information is normally reported as the aggregate without further disclosure of its types.

Based on the foregoing discussion, we argue that the flexibility inherent in the NAS regulations, including the disclosure requirements of NAS fees, may not be very effective in comforting investors of potential negative effects of NAS on audit quality.

2.3 EU nonaudit services regulations for public interest entities and the German application

The EU 8th Directive requires that all statutory auditors and audit firms are subject to principles of professional ethics, addressing their public-interest function, their integrity and objectivity, their professional competence, and the concept of due care (EU, 2006). The directive refers to the EC's Recommendation on Statutory Auditors' Independence in the EU: A Set of Fundamental Principles (EC, 2002). The recommendation, like the international Code of Ethics for Professional Accountants, adheres to the conceptual framework approach to auditor independence and serves as a benchmark of good practice.

In contrast, the new EU regulations on NAS to PIEs are rule based and ban specific NAS, known as the “black list.” Appendix 2 summarizes the “EU black” list for PIEs (EU, 2014). Germany decided to use the member state option to deviate from the “black list” and to allow the provision of certain tax services (preparation of tax forms, identification of public subsidies and tax incentives, and provision of tax advice) and the provision of valuation services whenever they are immaterial or have no direct effect on the financial statements. The EU also introduced a mandatory cap for NAS; that is, total NAS fees shall not exceed 70% of the average of the audit fees paid in the last three consecutive financial years. Germany did not use the option to deviate from the upper limit cap of 70%.

The new EU and German regulations for PIEs are more restrictive than those of the Commercial Code. The “black list” has a wider scope and prohibits additional NAS, such as payroll services, designing and implementing internal control or risk management procedures related to the preparation and/or control of financial information, and certain human resources services. Further, the “black list” also specifies the prohibited NAS in more detail, leaving less room for judgment. Finally, a fee cap for PIEs on NAS did not exist previously in Germany.

4.1 Sample selection

The sample consists of companies at the Deutsche Börse CDAX index in the 11-year period 2005–2015. The CDAX index includes all stocks on the Frankfurt Stock Exchange that are listed in the General Standard or Prime Standard market segments. Stock market and accounting data were collected from Worldscope and Datastream. The audit data were hand-collected data from annual reports. As reported in Table 1, panel A, the initial number of company-year observations was 5,632. We deleted observations from financial-sector companies (927), from incomplete data sets (1,173), and for other reasons (809).¹⁵ Our final sample consists of 2,723 company-year observations for 379 individual companies for the 11-year period.

Panel B of Table 1 defines the variables involved in testing the hypotheses.

4.2 Descriptive statistics

Table 2, panel A, presents descriptive statistics for the entire investigation period (2005–2015). The average stock return (RETURN) is 14.3% and the average return on equity (EARN) is about 1.9%. The average nonaudit fee ratio (NAF) is 27.9%, ranging from 11.1% to 42.3% from the first to the third quartile. The means of the three categories of nonaudit fees ratios are quite close, with the highest average of 10.5% for other nonaudit services (NAF3). The average fee ratio for other assurance services (NAF1) and tax services (NAF2) is the same and 8.7%.¹⁶ Debt capital is on average the most important source of capital (LEV, 62.3%), and the average market-to-book ratio (MBV) is 196.4%. The percentage of companies reporting a loss (LOSS) is 24%, and 65.8% of the companies are audited by one of the four big audit firms (BIG4).

Table 2, panel B, presents correlations between selected variables. The stock market return (RETURN) is significantly positively correlated with the level of earnings (EARN) and significantly negatively correlated with other assurance services (NAF1) and the existence of a loss (LOSS). Following from the variable definitions, EARN is significantly negatively correlated with LOSS. In addition, there is a significant positive correlation between EARN and NAF3. Significant positive correlations are observed between the NAS fee ratio (NAF) and its three component ratios (NAF1, NAF2, and NAF3). LOSS correlates significantly positively with other assurance services (NAF1) and significantly negatively with EARN and tax services (NAF2). This may reflect decreased focus on tax issues and the greater demand for other assurance services for loss-making companies. Finally, there are significant positive relations between BIG4 audits and NAF, NAF1, and NAF3.

4.3 Test of hypotheses and research question

To test the hypotheses, we use the ERC from earnings-response regression models as a proxy for investor perceptions of earnings quality (Eilifsen & Knivsflå, 2013; Francis & Ke, 2006; Frankel et al., 2002; Ghosh et al., 2009; Gul et al., 2006; Krishnan et al., 2005, 2013; Lim & Tan, 2008).

In the basic model, ERC is the estimated effect of reported earnings on stock returns (Easton & Harris, 1991).

(1)

(2)

Equation 1 can be expanded to allow for variables, other than EARN, that may affect RETURN, and interactions between EARN and these other variables; NAF is one such potential variable. Variables other than EARN and NAF variables that may affect RETURN are defined as X_i, where i = 1 − m:

(3)

Our primary interest when testing hypothesis  is the interaction between EARN and NAF, β₃ (i.e., the effect of NAF on ERC, cf. Equation 2).¹⁷

Similarly, when testing Hypotheses 2(a), 2(b), and 2(c), the primary interest is the interactions between EARN and the fee ratio components, NAF1, NAF2, and NAF3 (β₅, β₆, and β₇ respectively).

(4)

Following previous research, we include in the main tests the following control variables (X_i): LEV and STDRET to control for firm risk (Collins & Kothari, 1989), MBV as a proxy for growth prospects (Lipe, Bryant, & Widener, 1998), LOSS as a control variable for earnings persistence (Hayn, 1995), and SIZE to control for size effects (see Francis & Ke, 2006). Since large audit firms may moderate investors' concerns of independence from providing NAS (Eilifsen & Knivsflå, 2013; Francis, 2004; Gul et al., 2006), we include BIG4. Year dummies (YEAR) are used to control for time series variations. Finally, to capture potential industry-specific factors that are not completely covered by the other firm-specific controls, industry dummies (INDUSTRY) are included in the regressions.¹⁸

Based on Equation 3, we use the following regression to test hypothesis :

(5)

Based on Equation 4, we use the following regression to test Hypotheses 2(a), 2(b), and 2(c):

(6)

Table 3 reports the regression results for Equations 5 and 6 in two separate main columns. As expected, reported income (EARN) is positively and significantly associated with stock market return (RETURN). The coefficient of our test variable for hypothesis  (EARN × NAF) is negative but not significant. Hypothesis  is not supported. Contrary to our expectation, we cannot conclude that investor perceptions are negatively associated with the magnitude of auditor-provided NAS for our entire sample period.

The coefficients of our test variables for the components of NAS, EARN × NAF1, EARN × NAF2, and EARN × NAF3, are negative, positive and negative respectively, but none are significant. The components of NAS do not associate with the magnitude of auditor-provided NAS and hypotheses 2(a), 2(b), and 2(c) are not supported. Surprisingly, all our test variables turned out as insignificant and we do not find support for any of the four hypotheses for the entire sample period 2005–2015.¹⁹

Next, we turn to our research question: Did the financial crisis affect investor perceptions of auditor-provided NAS and its components? We partition our sample in the pre-financial crisis period 2005–2007, the financial crisis period 2008–2009, and the post-financial crisis period 2010–2015.

Table 4, panel A, presents the descriptive statistics for the three subsample periods, while panel B presents the correlation matrix between selected variables for each subsample period.

The average stock return (RETURN) ranges from 10.9 (crisis period) to 16.8%, with the highest mean value for the pre-crisis period and a slightly lower mean for the post-crisis period (14.2%). The average return on equity (EARN) is almost equal for pre-- and post-crisis period (3.8% and 3.5% respectively), but turns negative for the crisis period (−5.4%). The average nonaudit fee ratio (NAF) decreases for each period, from 31.5% for the pre-crisis period, to 27.8% for the crisis period, and to 26.1% for the post-crisis period. A similar pattern is observed for the three categories of nonaudit fees, (NAF1, 9.2, 9.0, 8.2%; NAF2, 10.1, 8.9, 8.0%; and NAF3, 12.1, 9.9, 9.9%). Total liabilities to total asset ratio (LEV) ranges from 53.2 (pre-crisis period) to 69.5% (post-crisis period). The pre-crisis period shows the highest mean value for the market-to-book ratio (MBV) (290.2%) and with the lowest for post-crisis period (159.1%). For the crisis period, the percentage of companies reporting a loss (LOSS) is 32.1%, while this is around 22% for the pre- and post-crisis periods. The Big 4 firms become more dominant for each period (60.9, 65.3, and 68.6%).

Table 4 presents the Pearson correlation matrix for the dependent and selected independent variables regarding the subsamples. The stock market return (RETURN) is significantly positively correlated with EARN for all three periods, and for the pre- and post-crisis periods negatively correlated with NAF1 (marginally significant for the post-crisis period) and LOSS. Following from the variable definitions, EARN is negatively correlated with LOSS throughout all three periods. Moreover, for the years before and after the crisis, a positive correlation is observed between EARN and NAF3. For all periods, NAF is significantly positively associated with NAF1, NAF2, and NAF3. For the pre-crisis period, LOSS correlates significantly positively with NAF1 and NAF3, and significantly negatively with NAF2. For the crisis period, LOSS correlates weakly negatively with NAF2. For the post-crisis period, LOSS correlates negatively with NAF (weakly significant), NAF2 and NAF3 (weakly significant). Finally, BIG4 is positively associated with NAF (weakly significant) for the crisis period and post-crisis period, NAF1 for all three periods, and NAF3 for the post-crisis period, and is negatively correlated with NAF2 (weakly significant) for the pre-crisis period.

Table 5, panel A, reports the regression results for Equation 5, where our test variable is EARN × NAF for each sub-period; the pre-crisis period 2005–2007, the crisis period 2008–2009, and the post-crisis period 2010–2015.

As expected, reported income (EARN) is positively and significantly associated with stock market return (RETURN) for all three periods. Before and during the crisis, the coefficient of our test variable EARN × NAF is negative (−1.179 and −0.450 respectively) and significant (p = 0.003 and 0.045 respectively); that is, a large NAS negatively associates with ERC. This is line with the prediction of hypothesis . The result indicates that investors perceive large NAS adversely. Investors' concerns for auditor independence seem to dominate any perceived knowledge spillover benefits to the audit from NAS. For the period following the crisis, however, the coefficient of EARN × NAF is significantly positive (0.259 with p = 0.015).²⁰ This suggests positive perceptions of NAS by investors. The finding may reflect that the crisis and the reform process starting immediately in the aftermath of the crisis changed supervisory boards', audit committees', and auditors' awareness of potential negative effects on audit quality of auditor-provided NAS. In turn, this may have eased investors' concerns from auditor-provided NAS to the extent that their perceived knowledge spillover benefits now dominate their independence concerns.

Table 5, panel B, reports the regression results for Equation 6, where our test variables are EARN × NAF1, EARN × NAF2, and EARN × NAF3, for each sub-period. Before the crisis, the coefficients of our test variables EARN × NAF1, EARN × NAF2, and EARN × NAF3 are all negative (−1.218, −2.501, and −1.230 respectively) and clearly significant (p = 0.029, 0.004, and 0.009 respectively).²¹ The results show that investors perceive each individual audit fee category negatively in the pre-crisis period. For the crisis period 2008–2009, the coefficients of our test variables EARN × NAF1 and EARN × NAF3 are again significantly negative (−1.050 with p = 0.028 and −0.670 with p = 0.012 respectively). We do not observe a significant association for the variable EARN × NAF2 in the crisis period.²² For the post-crisis period, only the coefficient of the test variable EARN × NAF2 is significant (p = 0.007). The coefficient has a positive sign (0.486); that is, the provision of tax NAS increases investors' confidence in reported earnings.²³

To summarize, the results for the sub-periods indicate that investors before and during the crisis perceive that large NAS weaken earnings quality. Investors' concerns of biased financial reporting dominate possible knowledge spillover benefits from providing large NAS. Also, the results for the years 2005–2009 indicate that this is the case for the three NAS components individually. In contrast, the findings for the post-crisis period suggest that investors' perceived improved earnings quality with larger NAS fees. This is consistent with perceived knowledge spillover benefits from NAS dominating investors' independence concerns. Tax NAS are the driver of this finding. The results show that after the financial crisis investors' perceptions of auditor-provided NAS substantially changed. The positive perceptions of NAS after the crisis may be associated with the lower NAS fees in this period.²⁴ Investors may interpret this reduction in such a way that it indicates that boards are less willing to take unwarranted independence risks.

4.4 Robustness and additional tests

We perform robustness tests for the sub-periods by changing the specification of our empirical model (additional control variables) and using alternative nonaudit fee definitions. In addition, we simulate the effect of the new German (EU) cap on NAS fees for our first two subsamples to evaluate its effectiveness if applied in the subsample periods.

The first robustness test includes two additional control variables proposed in the literature: AUDITOR CHANGE (indicator variable that equals 1 if company j changed audit firms for the fiscal year t, and 0 otherwise) and INDUSTRY LEADER (indicator variable that equals 1 if the auditor of company j had the highest audit fees in the related industry for the fiscal year t, and 0 otherwise). Table 6 reports the results.

For the years preceding the crisis, the regression coefficient for the interaction EARN × NAF is negative (−1.139) and strongly significant (p = 0.005). The coefficients for the components other assurance services fees (NAF1), tax services fees (NAF2) and other services fees (NAF3) are also negative (−1.119, −2.494, and −1.217 respectively) and clearly significant (p = 0.047, 0.004, and 0.010 respectively). For the crisis period, the coefficient for the test variable EARN × NAF is again negative (−0.385), but just missed significance (p = 0.107). For other assurance services fees (NAF1) and other services fees (NAF3), the impact is again negative (−1.102 and −0.913 respectively) and clearly significant (p = 0.038 and 0.004 respectively). As in the main test, the coefficient for tax services fees (NAF2) is positive but insignificant (0.286 with p = 0.396). After the crisis, the regression coefficient for the interaction EARN × NAF is positive (0.263) and significant (p = 0.013). Tax services fees (NAF3) exert a significantly positive effect (0.483 with p = 0.008), while the results do not show a significant relationship for the two other fee components, NAF1 and NAF3 (−0.069 with p = 0.811 and 0.241 with p = 0.210 respectively). Hence, the results for the alternatively specified regressions are basically the same as those in the main tests.

The second set of robustness tests applies alternative nonaudit fees variable definitions to the ratio between nonaudit fee and total auditor fee. Table 7 reports the results for NAF scaled by the market value of equity (panel A), for unscaled NAF (panel B), and for the natural logarithm of NAF (panel C).

Again, EARN × NAF is significantly negative in the pre-crisis period for the three alternative nonaudit fees variable definitions. The negative sign of the interaction coefficients extent into the crisis period but the coefficients are insignificant or weakly significant (for scaled NAF, panel A). In the post-crisis period EARN × NAF is positive but insignificant (for unscaled NAF and natural logarithm of NAF) or significant negative (for scaled NAF). The latter contrasts the findings in the main test. Additional results for alternative definitions of nonaudit fees for NAS categories (not tabulated) show associations quite similar to the main tests but in some cases with less significance, mainly related to tax services fees. The results from the robustness tests of alternative audit fees variable definitions confirm the previous results from the basic models except that in the post-crisis period the results become less clear.

To simulate the effect of the new German (EU) 70% cap of NAS fees relative to the audit fee for our two subsamples before and during the crisis, we exclude all company-year observations when NAF is above 0.4112.²⁵ For the pre-crisis period, this reduces the sample from 769 to 518 company-year observations. The coefficient for the test variable EARN × NAF is still negative (−2.940) and strongly significant (p = 0.002). Additional calculations indicate that the test variable becomes insignificant when NAF does not exceed 0.36, which implies a cap of about 56% (NAF =  − 0.560 with p = 0.684). Thus, for this subsample, the “optimal” cap may be lower than the 70%. For the crisis period, the sample is reduced from 504 to 377 company-year observations. The coefficient for the test variable EARN × NAF is again negative (−0.126) but not significant (p = 0.770). Thus, for the crisis period, the results do not indicate that the cap is too moderate.