Audit Partner Tenure and Independence in a Low Litigation Risk Setting

Abstract We investigate whether long audit partner tenures impair auditor independence, as proxied by the opinion of the audit report, with a sample of Spanish companies for the period: 2002–2010. The Spanish audit market constitutes an ideal setting in which to address this issue, as it is characterized by unusually lengthy engagements with the audit firm. The motivation relies, on the one hand, on the current discussion about the necessity to reinforce the independence of auditors and, on the other hand, on the very limited available research at the partner level. The main result is the lack of significant effects of partner tenure on independence. This finding is robust to various checks. Unlike prior research, we also address the joint effects of firm and partner tenure on independence. Results indicate that partner tenure does not compromise independence, even under long or extremely long audit firm tenures. These findings might have some interesting policy implications, in particular for the intense current debate on auditor rotation regimes which is taking place within the European Union.


Introduction
Auditor independence is not a new issue in accounting research. In the late fifties McLaren (1958) pointed out that excessive familiarity could result in collusion between auditors and clients. Shortly afterwards, Mautz and Sharaf (1961) stated that extended auditor -client relationships could negatively affect independence because auditor's objectivity about a client would be reduced with the passage of time. Similarly, Hoyle (1978) argued that the audit program might become a routine as the auditor would anticipate the client's systems and control procedures. These concerns were summarized by Shockley (1981, p. 789): 'complacency, lack of innovation, less rigorous audit procedures and a developed confidence in the client may arise after a long association with the client'. More than two decades later, Myers, Myers, and Omer (2003) also argued that long relationships might cause auditor complacency about and possibly complicity in the decisions that management makes regarding the presentation of financial statements. As posed by Dopuch, King, and Schwartz (2003), this threat to independence does not limit to the time of the audit report, but it could also affect the judgments made by the auditor during the whole auditing process. Concerns about the negative effects of long tenures have also been expressed by the IFAC Code of Ethics: 'familiarity threat occurs when, by virtue of a close relationship with an assurance client, its directors, officers or employees, a firm or a member of the assurance team becomes too sympathetic to the client's interests' (International Federation of Accountants [IFAC] Code of Ethics ED 2003, p. 18).
However, potential loss of independence associated to long audit tenures needs to be balanced against other arguments suggesting that longer tenures could provide higher levels of audit quality. This was clearly posed by Myers, Myers, Palmrose, and Scholz (2005) who argued that financial reporting problems are more likely to occur early in the auditor-client relationship, when the auditor is less familiar with the client's business, processes and risks. The potential contradictory effects of tenure on audit quality are implicit in DeAngelo's (1981) classical definition of audit quality as the joint probability an auditor will both detect and report material misstatements. Thus, audit quality would be a function of the ability to detect material misstatements (expertise) and the willingness to report detected misstatements (independence). The ability to detect misstatements should be higher when the auditor has deeper client knowledge, and this knowledge increases with tenure through a learning curve effect. However, the auditor's willingness to report detected misstatements could be lower in lengthy engagements. To address these contradictory effects, the Sarbanes-Oxley Act (SOX) required a study by the Comptroller General of the United States (General Accounting Office [GAO], 2003) about the potential effects of imposing the mandatory rotation of auditors. The results of the study did not show a negative effect of extended tenures on the quality of financial reports, and thus, it did not recommend rotation. Nevertheless, the regulator finally established audit rotation at a partner level. Many countries worldwide have adopted similar rotation rules.
There is some contradiction between the causes of deterioration of independence in long audit tenures suggested in the literature (e.g., Dopuch et al., 2003;Mautz & Sharaf, 1961;Shockley, 1981) and the methodology used to assess the significance of such causes. Thus, the negative implications of long tenures for independence would usually derive from personal relationships between the company's management team and the incumbent auditor. Since personal relationships cannot be achieved between firms, the main threat to independence should not be long audit firm tenures but long partner tenures. Nevertheless, empirical research has addressed the relationship between tenure and independence almost exclusively at the firm level. Such a situation is mostly explained by the fact that in many countries audit reports do not record the name of the engagement partner. Nevertheless, even in those countries in which audit reports record the name of the audit partner like Spain, some studies have investigated the implications of long audit firm tenures (e.g., Ruiz-Barbadillo, Gómez Aguilar, De Fuentes-Barberá, & García Benau, 2004 but none has addressed the effects of long partner tenures. We investigate whether lengthy partner tenures impair auditor independence. The motivation of our study relies on the lack of research on the relationship between tenure and auditor's opinion at the partner level. There are just a few articles addressing the effects of partner tenure on various proxies of audit quality, and only two of them (Carey & Simnett, 2006 (hereinafter, C&S) and Ye, Carson, & Simnett, 2011) have specifically examined the effects of tenure on the opinion of the audit report. The 2010 Green Paper on Audit Policy by the European Commission (hereinafter 'the Green Paper') explicitly acknowledged the necessity of further research on the tenure -audit quality relationship within the EU. 1 Four years after the release of the Green Paper, the Directive 2014/56/EU amending Directive 2006/43/EC on statutory audits of annual accounts and consolidated accounts and the Regulation (EU) No 537/2014 of the European Parliament and of the Council on specific requirements regarding statutory audit of public-interest entities (hereinafter 'the 2014 EU Regulation') clearly indicate that the independence of auditors constitutes a major issue for European regulators and also for scholars (e.g. Cameran, Francis, Marra, & Pettinicchio, 2015;Cameran, Prencipe, & Trombetta, 2016;Ratzinger-Sakel, Audousset-Coulier, Kettunen, & Lesage, 2013;Ratzinger-Sakel & Schönberger, 2015).
Following previous research, the premise is that a less independent auditor will be less willing to issue a negative report so as not to lose clients (e.g. Chow & Rice, 1982;Craswell, 1988;Krishnan, 1994). We conduct the empirical analysis through a multivariate logistic model in which the dependent variable, the opinion of the audit report, will be explained by the tenure with the lead audit partner and the control variables suggested in prior research. The analysis is based on a sample of Spanish public companies for the nine-year period 2002 -2010. Anticipating our main result, partner tenure does not seem to have significant effects on the likelihood of audit qualifications. This result is robust to a series of checks. Moreover, we do not find different implications of partner tenure on auditor independence depending on the tenure with the audit firm. Since we do not observed loss of independence in longer partner tenures, the accelerated rotation from seven to five years established by the 2015 Spanish Audit Law, would not seem to be justified on the grounds of the reported results.
We aim to contribute to the literature by extending prior research in various ways. Firstly, for the first time, the effects of partner tenure on the opinion of the audit report are studied in a low litigation risk country. Since litigation risk constitutes a major motivation in the auditor reporting decision, auditors' incentives to maintain independence should be stronger in high litigation risk countries than in low litigation ones. Thus, evidence reported in high litigation countries cannot be directly extrapolated to low litigation countries. We aim to fulfill this gap in the literature. Moreover, in the current discussion on auditor independence within the EU, the evidence we report might be useful for other low litigation risk European countries. Secondly, while the scarce available evidence about the effects of partner tenure on audit qualifications has been obtained under voluntary partner rotation, we provide evidence obtained under mandatory partner rotation. Since partner rotation is currently mandatory in many countries, it is necessary to update the available evidence on the partner tenure -independence relationship to this regulatory framework. Thirdly, unlike prior research, we include audit firm tenure into the analysis of the effects of partner tenure. This point enhances the contribution of our paper in two ways. On the one hand, because as firm and partner tenures would show some positive correlation, the omission of audit firm tenure in the analysis could cause misleading results, as partner tenure will also account for the effects of the omitted variable firm tenure. On the other hand, because the inclusion of firm tenure allows to examine the interaction effects of firm and partner tenure. Hence, we can address, for example, whether the effects of partner tenure on independence are more severe under longer tenures with the audit firm.
The article is organized as follows. Section 2 outlines a review of the literature on the tenureindependence relationship and develops the hypotheses. Section 3 presents the design of the research and our sample selection. The discussion of results and sensitivity analyses are addressed in Section4. Finally, the conclusions and implications of our findings are drawn in Section5.

Background and Hypothesis Development
While a number of articles have studied the effects of audit firm tenure on the opinion of the audit report (e.g. Carcello & Neal, 2000;Gul, Basioudis, & Ng, 2011;Louwers, 1998), only C&S and more recently Ye et al. (2011) have examined the issue at partner level. Since the probability of switching the audit firm increases after a qualified opinion, the issuance of qualified reports can be viewed as an exercise of independence. C&S found that the auditor's propensity to issue going-concern modified opinions (GCMOs) diminishes over the audit partner's tenure, after controlling for other factors which influenced this propensity. The authors, therefore, Accounting in Europe 407 concluded that long tenures are negatively associated to independence. They acknowledged, however, that the deterioration of independence was confined to non-Big 4 audit firms. Consistent with C&S findings, Ye et al. (2011) found that longer engagement partner tenure is associated with a reduction in the propensity to issue GCMOs. As both papers investigated the same audit market, the reporting of similar results was to be expected. Beyond the tenure -audit qualifications relationships, scholars have also investigated the effects of partner tenure on other proxies of audit quality, such as the quality of earnings. With a private sample of US companies, Manry, Mock, and Turner (2008) concluded that audit quality increases with tenure, as discretionary accruals were significantly and negatively associated to the lead audit partner's tenure. However, this significant effect was observed only for relatively small clients having fairly lengthy partner tenures. In the same line, evidence available for Taiwan (Chen, Lin, & Lin, 2008;Chi & Huang, 2005;Chi, Myers, Omer, & Xie, 2011) and China (Gul, Wu, & Yang, 2013) supports a positive effect of partner tenure on earnings quality. Conversely, C&S 2 and Fargher, Lee, and Mande (2008) found lower manager's accounting discretion in the initial years of tenure of a new audit partner, thus suggesting a negative effect of partner tenure on audit quality in Australian. A final set of papers has not directly examined the implications of partner tenure on audit quality but the related issue of the implications of the auditor rotation regime. Hence, Chi, Huang, Liao, and Xie (2009) concluded that mandatory partner rotation does not enhance audit quality in Taiwan. It should be noted, however, that they examined the quality of earnings only during the first year of the engagement with the new partner. More recently, Firth, Rui, and Wu (2012) found that firms under mandatory partner rotation show significantly higher likelihood of modified opinions than no-rotation firms, in China. Therefore, contrary to Chi et al. (2009), they concluded that mandatory partner rotation does enhance audit quality.
Although, as posed by Bamber and Bamber (2009), the effects of audit firm rotation cannot be extrapolated to partner rotation, 3 evidence reported at the audit firm level might provide some insights into the effects of partner tenure on independence. Thus, next we summarize available evidence at the audit firm level. Empirical research has, for the most part, rejected that long firm tenures threaten independence, although research evidence would be inconclusive (Ewelt-Knauer, Gold, & Pott, 2013). With samples of US financially distressed firms, Louwers (1998) and Carcello and Neal (2000) do not support a significant effect of tenure on the auditors' reporting decision. A similar conclusion was reached by Vanstraelen (2002) and Knechel and Vanstraelen (2007) for the Belgian market. The former included both financially distressed and non-distressed companies, while the latter was limited to companies entering bankruptcy. Ruiz-Barbadillo et al. (2004 investigated the issuance of GCMOs to financially distressed companies in Spain. The authors did not report loss of independence with tenure. However, more recent research has provided some contradictory results. Hence, Lim and Tan (2010) reported a positive effect of tenure on the propensity to issue GCMOs to financially distressed firms. Conversely, Gul et al. (2011) concluded that auditors were willing to forgo their independence by issuing fewer GCMOs when auditor tenure was long. All the aforementioned studies share the focus on GCMOs. The studies by Vanstraelen (2000) for the Belgian market, and Ruiz- Barbadillo, Gómez Aguilar, and Biedma (2005) for Spain represent exceptions to the mainstream approach, as they include all types of modified opinions into the analysis and do not limit the sample to financially distressed firms. Hence, Vanstraelen (2000) concluded that long-term auditorclient relationships significantly increase (decrease) the likelihood of unqualified reports (auditor's independence) in the Belgian market. Conversely, Ruiz-Barbadillo et al. (2005) did not observe loss of independence with tenure, as the likelihood of modified opinions was in fact higher in longer tenures.
After the review of prior research, we pose our research question as follows: Does auditor independence, measured through the opinion of the audit report, decrease with auditor tenure?
As discussed before, long tenures might weaken auditor's objectivity (Mautz & Sharaf, 1961), strengthen confidence in the client (Shockley, 1981), and, as a result, make qualified reports less likely. In this line, evidence reported by C&S and Ye et al. (2011) has shown lower likelihood of audit qualifications in longer partner tenures in Australia. Thus, even in countries characterized by high litigation risk, independence seems to be impaired in long engagements with the audit partner. Since auditor's litigation risk is lower in Spain than it is in Australia, auditors' incentives to maintain independence in longer tenures would also be expected to be lower. Accordingly, we predict a weakening of independence in lengthy engagements with the audit partner, and thus the first hypothesis states: Hypothesis #1 (H1): The likelihood of a qualified report will decrease with audit partner tenure.
Although our main interest is the impact of partner tenure, the effects of audit firm tenure are also examined. The above discussion on H1 can be used to develop the hypothesis on the effects of audit firm tenure. As posed by Bamber and Bamber (2009), when the audit partner is changed by another partner of the same audit firm, the rest of the audit team remains with the client. Thus, the same factors which could explain a negative impact of lengthy engagements with the audit partner on independence would also explain the effects of lengthy engagements with the audit team. In this line, the Green Paper warns that, even when lead partners are regularly rotated, the threat of familiarity persists. While prior evidence for the Spanish audit market has provided mixed results, we predict a negative effect of audit firm tenure on the likelihood of a qualified report. This expectation relies on two main points. On the one hand, on the concerns expressed by the Green Paper and the 2014 EU regulation regarding the negative effects of long audit firm tenures on audit quality within the EU. If European regulators consider that long audit firm tenures impair independence at the EU level, this problem should be more serious in low litigation countries such as Spain. On the other hand, on the relatively long audit firm tenures in Spain by international standards. Longer tenures provide more room for a bonding effect in the auditor-client relationship. Thus, our second hypothesis states: Hypothesis #2 (H2): The likelihood of a qualified report will decrease with audit firm tenure.
Unlike prior research, we also address the interaction effects of firm and partner tenures on auditor independence. We wonder whether, let's say, five years of partner tenure involve the same potential implications for auditor independence, under 5 or 20 years of tenure with the audit firm. As discussed above, when the audit partner is changed but not the audit firm, the rest of the audit team remains with the client, and neither the audit methodology nor the client's procedures tend to change (Bamber & Bamber, 2009). Thus, the potentially negative implications of partner tenure on independence could also depend on the tenure with the audit firm, as the marginal effect of one more year of partner tenure on the partner's involvement with the client might be stronger under long tenures with the audit firm. In long audit firm tenures, long-time standing audit teams would be expected to facilitate the bonds between a new lead audit partner and the client's management team. Hence, the 'familiarity threat' Accounting in Europe 409 (IFAC, 2003) of long tenures with the audit partner would be expected to be more severe under long audit firm tenures. Accordingly, the third hypothesis states: Hypothesis #3 (H3): The negative effects of partner tenure on the likelihood of a qualified report will be stronger under long audit firm tenures.

Research Design
We address the effects of partner tenure on auditor independence, measuring independence through the ability of the auditor to issue a qualified report. As prior research, we assume that a more independent auditor will be more willing to issue a qualified report to a client. Yet, it should be noted that most prior research measuring independence by the opinion of the audit report has been conducted in high litigation risk countries (e.g. DeFond, Raghunandan, & Subramanyam, 2002;C&S;Ye et al., 2011). In these countries, potential litigation costs should provide strong incentives to auditors to preserve independence. Hence, we wonder whether this framework is still valid in a low litigation risk country such as Spain. We consider that even in low litigation risk countries, and more specifically in Spain, there are incentives for auditors to maintain independence. This view is supported by prior research conducted in low litigation risk countries measuring audit quality by the opinion of the audit report (e.g. Laitinen & Laitinen, 1998 for Finland;Vanstraelen, 2000Vanstraelen, , 2002Knechel & Vanstraelen, 2007 Lam and Mensah (2006), the maintenance of a high litigation risk environment does not appear to be a necessary pre-requisite for high quality audits. Moreover, in the specific case of the Spanish audit market it should be noted that, on the one hand, the supervisory agency ICAC (Institute of Accounting and Auditing) has imposed some penalties to audit firms for low quality audits (e.g. Deloitte received a 12 million E fine after the Bankia scandal) and, on the other hand, the market shows an extreme concentration by Big 4 firms (firms with strong incentives to maintain reputation, and therefore independence). Hence, we conclude that, although incentives to maintain independence are expected to be stronger in high litigation counties, in low litigation countries and, particularly in Spain, auditors still face incentives to maintain independence and, consequently, to provide high quality audits.
While most studies on auditor independence have restricted the analysis to financially distressed companies and have examined only GCMOs (e.g. C&S and Ye et al., 2011), some papers have followed a more comprehensive approach and have included all types of firms and audit qualifications into the analysis (e.g. Chi & Chin, 2011;Firth et al., 2012;Laitinen & Laitinen, 1998;Ruiz-Barbadillo et al., 2005;Vanstraelen, 2000). Such a view would be supported by empirical evidence reporting negative effects of a qualified report for the auditor client (e.g. Ball, Walker, & Whittred, 1979;Choi & Jeter, 1992;Chow & Rice, 1982;Levinthal & Fichman, 1988) and therefore a higher probability of losing the client after a qualified report (e.g. Chow & Rice, 1982b;Craswell, 1988;Krishnan, 1994). Accordingly, the issuance of qualified opinions (not only GCMOs) by the auditor represents an exercise of independence. Besides, this approach presents some advantages in terms of generalization of the reported results and also allows to address the classical role of the auditor which is not bankruptcy prediction. Therefore, as Laitinen and Laitinen (1998), Vanstraelen (2000), Ruiz-Barbadillo et al. (2005), Chin (2011) andFirth et al. (2012), we consider audit reports with either qualified, unfavorable, disclaimer of opinion, or with explanatory paragraphs expressing doubts about the future of the company, collectively as qualified reports.
To test the first two hypotheses posed in the former section (H1 and H2) we will estimate the model given by (1).
where: Dependent Variable: OPINION: 1 if company receives a modified opinion in the audit report and 0 otherwise. Experimental Variables: FIRMTEN: the number of consecutive years audited by the same audit firm. PARTEN: the number of consecutive years the same partner has been signing the audit report of the company.
Control Variables: PBANK: probability of bankruptcy as measured by adjusted Zmijewski score, with the weights proposed by Carcello, Hermanson, and Huss (1995); SIZE: natural logarithm of total assets of the company; AGE: natural logarithm of the number of years since listing on the supervisor of the Spanish stock market; LEV: total liabilities divided by total equity; CLEV: change in LEV during the year; LLOSS: 1 if client reported negative net income for the previous year and 0 otherwise; INVEST: current assets less debtors and inventories divided by current liabilities; AUDFIRM: 1 if the company is audited by a Big 4 audit firm and 0 otherwise; Year controls: Eight year dummies indicating years 2002, 2003, 2004, 2005, 2006, 2007, 2008 and 2009. Industry controls: We follow the sector classification provided by the Madrid Stock Exchange which includes six main industries. Thus, five dummies variables indicating the firm's industry are included in (1).
According to H1 (H2) we expect a negative and statistically significant coefficient associated to PARTEN (FIRMTEN). To test H3 we will estimate the model given by (2). To define long audit firm tenures (LONGFT: 1 if tenure is 10 years or more and 0 otherwise) we use the median value of FIRMTEN as the cutoff point. This model includes the same variables as model (1) but also the interaction variable PARTEN * LONGFT. According to H3 we expect a negative and statistically significant coefficient associated to this variable. OPINION = a 0 + a 1 * PBANK + a 2 * SIZE + a 3 * AGE + a 4 * LEV + a 5 * CLEV + a 6 * LLOSS + a 7 * INVEST + a 8 * AUDFIRM + a 9 * FIRMTEN + a 10 * PARTEN + a 11 * PARTEN * LONGFT + (z * YEAR) Control variables in (1) and (2) are similar as those in C&S, Laitinen and Laitinen (1998) and Chi and Chin (2011) to estimate the auditor's probability of issuing a qualified report, consisting Accounting in Europe 411 basically in financial ratios. Although C&S limit the analysis to GCMOs and financially distressed firms, similarly to Laitinen and Laitinen (1998) and Chi and Chin (2011), we do not restrict the analysis to financial distressed firms and consider audit reports with either qualified, unfavorable, disclaimer of opinion, or explanatory paragraphs, collectively as qualified reports. PBANK measures the probability of bankruptcy based on Zmijewski (1984), where higher values indicate a higher probability of bankruptcy. Thus we predict a positive effect on PBANK on the likelihood of audit qualifications. SIZE is included to control for the impact it can have on the propensity of the auditor to be independent. Hence, small clients would face higher risk of bankruptcy and this should increase the likelihood of qualified reports (Francis & Krishnan, 1999;Reynolds & Francis, 2001). Besides, small clients have lower negotiating power with the auditor to avoid audit qualifications. However, on the other hand, since large clients involve potentially higher litigation costs, auditors might be more willing to issue qualified reports to these clients (Craswell, Stokes, & Laughton, 2002). Consequently, the sign on the coefficient for SIZE could be either positive or negative. AGE accounts for a higher risk of financial distress for younger companies (Dopuch, Holthausen, & Leftwich, 1987). Accordingly, we predict a negative effect of AGE on the likelihood of a qualified report. LEV measures the risk associated to higher levels of debt, while CLEV captures changes in leverage that may make companies to unsustainable levels of debt. Hence, financial leverage will raise litigation risk as it makes bankruptcy more likely. In addition, highly levered firms tend to present lower accounting quality (Aharony, Lin, & Loeb, 1993). Thus, we predict a positive effect of both LEV and CLEV on the likelihood of a qualified report. LLOSS and INVEST are included because companies with losses and low levels of liquidity are more likely to enter bankruptcy (Reynolds & Francis, 2001) and also, similarly to what occurs with leverage, to manipulate earnings. Hence, we expect positive (negative) effects of LLOSS (INVEST) on the likelihood of audit qualifications. Finally, AUDFIRM accounts for a potential higher propensity to issue qualified reports by Big 4 auditors.

Sample and Dataset
We perform the empirical analysis on the basis of non-financial companies quoted in the Spanish Stock Exchange (Sistema de Interconexión Bursátil Español) during the research period 2002 -2010. In addition, to be able to compute the length of audit partner tenure, for the companies in the sample, information about the auditor opinion and the name of the engagement partner should be available for the period 1995 -2010. Information regarding audit reports is obtained from the Comisión Nacional del Mercado de Valores (CNMV), while data about independent variables in the models are provided by Thomson Reuters Knowledge. Our dataset is formed by 83 firms and, given the nine-year research period, by 747 firm-year observations. However, in 12 cases, information about at least one variable in the models was not available. Therefore, the sample is finally formed by 735 observations. We examine 735 audit reports, 600 of them unqualified and 135 have a qualified opinion. GCMOs have been relatively scarce in our research period, as only 13 out of the 135 qualified reports have GCMOs. Besides, none of the audit reports has an adverse opinion or a disclaimer of opinion.
In the Report on the Review of the Annual Financial Reports filled with the CNMV (Comisión Nacional del Mercado de Valores [CNMV], 2009), the regulator of the Spanish stock market classifies audit qualifications into two major groups: Quantified and unquantified. In addition, quantified qualifications are also classified into two subgroups depending on whether they affect profit and losses or equity. Similarly, unquantified qualifications are also classified into 'Uncertainty and others' and 'Limitations'. Among qualifications due to uncertainties, the most serious ones are those concerning the continuation of business, in which the auditor expresses its doubts about the future of the company. However, uncertainties can also have less dramatic effects; for example, they can be associated to the firm's ability to recover some tax credits. On the other hand, qualifications for limitations on scope show that the auditor has not had enough information to apply the procedures required by the technical auditing standards.
Regarding the regulatory context of this research, the 1988 Spanish Audit Law imposed the mandatory rotation of the audit firm after nine years. Nevertheless, as a result of a subsequent legal reform in 1995, the mandatory rotation was abolished. Thus, the mandatory rotation of the audit firm was, in fact, never applied in Spain. After the reform, auditors could be engaged for an initial period between three and nine years, but after the expiration of the initial engagement the company could renew the contract with the same auditor on a yearly basis. Researchers have widely agreed that Spanish legislation has not been particularly strict regarding safeguards to strength auditor independence (e.g. Gonzalo Angulo, 1995;Paz-Ares, 1996;Ruiz-Barbadillo et al., 2004). With the same aim as the SOX in the US, the Spanish Financial Law was passed in 2002. An amendment was included during the Law's approval process including the mandatory rotation of the audit firm after 12 years. According to this amendment, the change of the audit firm would be mandatory after the expiration of this term, and a minimum three-year period was required to re-hire the audit firm. Similar to the 1988 firm rotation rule, this amendment led to strong criticism from the auditing profession (Ruiz-Barbadillo, Gómez Aguilar, & Carrera, 2006), causing its final withdrawal. Mandatory rotation after seven years was finally imposed, but limited to the audit team not to the audit firm. In addition, a minimum two-year period was required before the partner could be reappointed. The maximum partner tenure of seven years has also been maintained by the 2010 reform of the Spanish Audit Law, however, without imposing the rotation of the rest of the audit team. Finally, in 2015 a new Audit Law was passed to adapt the Spanish regulation to the 2014 EU Regulation. The new regulation imposed the mandatory rotation of the audit firm after 10 years and accelerated the mandatory rotation of the lead audit partner from seven to five years. Table 1 shows some descriptive statistics about the variables used in the models. Qualified audit reports represent 15% of the total sample. Average partner tenure is three years, rather below the maximum tenure established by Spanish legislation. It should be noted that the maximum tenure in our sample is nine years, two years over the maximum of seven years established by the Spanish law. Regarding the tenure with the audit firm, the average of 10 years with a maximum of 24 years would show the relatively long audit firm tenures in Spain by international standards. 4 On average, companies in the sample have been reporting to CNMV for 16 years. Regarding the type of audit firm, companies audited by Big 4 auditors represent 92% of the sample, thus showing an extreme concentration of the Spanish audit market by Big 4 auditors by international standards. 5 Table 2 shows Pearson's correlation coefficients, with significance levels, between pairs of variables. In general, correlation levels are rather low, with a maximum value of 0.56 between LEV and CLEV. The correlation pattern of OPINION with the independent variables strongly supports the discussion carried out at the end of the previous section. Hence, qualified opinions would be negatively associated to the firm's financial health (LEV, CLEV, LLOSS and INVEST) and to SIZE, while no significant relationship is observed with PBANK, AGE or AUDFIRM. Moreover, although no significant association between OPINION and PARTEN is reported,

Accounting in Europe 413
OPINION would be negatively related to FIRMTEN. This would suggest loss of independence in longer tenures with the audit firm though not in longer tenures with the audit partner. Focusing on independent variables, correlation matrix shows some rather obvious results, as the correlation pattern of PBANK with LEV, CLEV, LLOSS and INVEST. However, some other results are more interesting. For example, companies audited by Big 4 firms tend to be relatively larger and more profitable. These companies also show longer firm tenures but shorter partner tenures, the later indicating that partner rotation is higher among clients of Big 4 firms. This probably reflects both, a greater willingness of Big 4 firms to rotate partners, as well as a better chance of doing so compared to non-Big 4 firms. Finally, results for variables measuring tenure would show the expected positive correlation between PARTEN and FIRMTEN. Table 3 shows the results of the univariate analysis. As the Shapiro-Wilk test strongly rejects the hypothesis of normality for each independent variable, the Mann -Whitney test of differences of medians is performed in order to assess the statistical significance of these differences. Median values of independent variables across subsamples of firms, according to the opinion of the audit report, and significance levels (from the Mann -Whitney test for continuous variables and the chi-square test for dichotomous variables) are provided. As shown by the table, PARTEN is not significantly associated to the opinion of the audit report. Conversely, firms with unqualified reports tend to show significantly longer engagements with the audit firm. In both cases, results would support the correlation pattern shown in Table 2. Results regarding control variables are, in general, far from surprising. Hence, firms with qualified reports show relatively lower levels of solvency (PBANK, LLOSS and INVEST) and tend to be smaller than firms with unqualified reports. However, we do not find significant results for, AGE, LEV, CLEV or AUDFIRM.  Note: * , * * , * * * Significant at the 10%, 5% and 1% levels, respectively. OPINION: 1 if company receives an 'unclean' audit report and 0 otherwise; PBANK: probability of bankruptcy; SIZE: natural logarithm of total assets of the company; AGE: natural logarithm of the number of years since listing on the supervisor of the Spanish stock market; LEV: total liabilities divided by total equity; CLEV: change in LEV during the year; LLOSS: 1 if client reported negative net income the previous year and 0 otherwise; INVEST: current assets less debtors and inventories, divided by current liabilities; AUDFIRM: 1 if the company is audited by a Big 4 audit firm and 0 otherwise; FIRMTEN: number of consecutive years audited by the same audit firm; PARTEN: number of consecutive years the partner has been signing the audit report of the company.

Multivariate Analysis
In this subsection we address the joint effect of partner tenure and the proposed control variables on the likelihood of audit qualifications through logistic regression models. Results of the estimation of (1) are shown in Table 4 (column A). Unlike C&S and Ye et al. (2011) who used cross-sectional datasets, we follow a panel data approach. Datasets with panel structure allow to control for individual unobserved heterogeneity better, and therefore to reduce the likelihood of reporting purely spurious effects due to an omitted variable bias. In the particular case of audit qualifications, given the relatively low explanatory power of the models proposed, the omitted variable bias could be serious. In keeping with the panel structure of the dataset we perform panel data logistic regressions. As expected, the likelihood ratio test supports the use of a panel data approach to estimate the model (p , .000) over a classical pooled regression. Besides, the model is statistically significant (p , .000), indicating that observations are well fitted by the model. Other indicators of the model's goodness of fit are the pseudo R 2 and the percentage of cases correctly predicted. Hence, the model explains 28% of the total variance and correctly classifies 90% of cases. Although correlation coefficients between pairs of independent variables shown in Table 2 did not suggest serious multicollinearity, variance inflation factors (VIF) are calculated after the estimation to rule out the negative potential effects of multicollinearity. As expected, VIF (not reported) are rather low, with a maximum value of 1.68 for variable LEV, thus supporting the initial view that multicollinearity would not affect our results.
The main result in Table 4 (column A) is that partner tenure does not significantly affect the issuance of audit qualifications. This finding had been anticipated by the correlation pattern in Table 2 and later by the univariate analysis in Table 3. Accordingly, H1, stating that the likelihood of a qualified report will decrease with audit partner tenure, is rejected. Since we do not  limited to GCMOs or to financially distressed firms, this result is not fully comparable to those provided by either C&S or Ye et al. (2011), both papers reporting a negative effect of partner tenure on the likelihood of GCMOs in the Australian market. Besides, it should be noted that while C&S and Ye et al. (2011) investigated partner tenure under voluntary partner rotation, our research is carried out under a mandatory partner rotation rule. However, despite the limited comparability between our results with both C&S and Ye et al. (2011), it is somehow surprising that long partner tenures seem to involve negative effects on independence in the high litigation risk Australian audit market but not Spain. Similar to partner tenure, the effects of audit firm tenure on the likelihood audit qualifications are also non-significant, and 0.28 46.14 90 * * * Note: Results from logistic regressions. Parameters estimates and z-values in parentheses. * , * * , * * * Significant at the 10%, 5% and 1% levels, respectively. Dependent variable: OPINION: 1 if company receives an 'unclean' audit report and 0 otherwise. Independent variables: PBANK: probability of bankruptcy; SIZE: natural logarithm of total assets of the company; AGE: natural logarithm of the number of years since listing on the supervisor of the Spanish stock market; LEV: total liabilities divided by total equity; CLEV: change in LEV during the year; LLOSS: 1 if client reported negative net income the previous year and 0 otherwise; INVEST: current assets less debtors and inventories, divided by current liabilities; AUDFIRM: 1 if the company is audited by a Big 4 audit firm and 0 otherwise; FIRMTEN: number of consecutive years audited by the same audit firm; PARTEN: number of consecutive years the partner has been signing the audit report of the company; and PARTEN * LONGFT: the interaction variable of PARTEN and LONGFT (1 if firm tenure ten years or more and 0 otherwise). thus H2 is also rejected. Unlike what occurred with partner tenure, this result was not anticipated by, either the correlation pattern shown in Table 2 or the univariate analysis in Table 3, in both cases suggesting negative impact of firm tenure on the likelihood of audit qualifications.
Results for control variables strongly meet our expectations after the review of prior research and the univariate analysis. Hence, audit qualifications would be more likely for companies with lower solvency (PBANK, LLOSS and INVEST). In Section 4 we had not predicted the sign of the effect for SIZE as, on the one hand, large companies tend to exhibit higher audit quality and negotiating power with the audit firm, but, on the other hand, these companies also involve higher levels of litigation risk for the audit firm. Results would indicate that audit qualifications are more likely for small than for large firms. As predicted, clients of Big 4 audit firms show higher probability of receiving a qualified report. However, we do not report significant results for AGE, LEV or CLEV. In all three cases, the lack of significance had been anticipated by the univariate analysis in Table 3.

The Interaction of Firm and Partner Tenure
In order to control for the potential effects of firm tenure on the actual involvement of the engagement partner with the client, Table 4 (column B) shows the results of the estimations of (2). According to H3 we expect that the negative effect of partner tenure on independence would be stronger under long audit firm tenures. However, the interaction variable PAR-TEN * LONGFT shows no significant effects on the likelihood of audit qualifications, and therefore similarly to H1 and H2, H3 is also rejected. Results regarding FIRMTEN, PARTEN or control variables are qualitative the same as those in Table 4 (column A).
The final analysis regarding the interaction effects of firm and partner tenure consists of the exam of auditor independence in audit engagements characterized by long firm and partner tenures. Although, according to the evidence reported so far in this paper, we do not expect lower independence in these engagements, we consider this a meaningful check, as the prior estimations could not adequately account for potential loss of auditor independence in these engagements (e.g. when long partner tenures interact with extremely long audit firm tenures). Accordingly, we reestimate (2) after substituting PARTEN * LONGFT by the new variable LONGFTLONGPT (long firm and partner tenures defined as 1 for those engagements in which firm tenure is ten years or more and partner tenure is more than four years and 0 otherwise). Results of the new estimation (not reported) show non-significant effects for the new variable. We chose the median value of FIRMTEN as the cutoff point to define long audit firm tenures. However, similar results are observed when long firm tenures are defined as more than 20 years. Thus, even under long tenures with the audit partner and extremely long tenures with the audit firm, we do not report loss of independence.

Sensitivity Analysis
As usual in the literature, we perform a series of checks to assess the robustness of the reported results. Firstly, we address the potential effects of influential observations. Hence, after the estimation of (1) we detected 13 influential observations (with Pregibon dbeta higher than 0.2). Results of the reestimation of (1) without these observations (not reported) are qualitatively the same as those in column A, and therefore we conclude that our results are not affected by influential observations. Our second analysis aims to assess potential endogeneity problems associated to those companies changing the audit firm (and thus also the audit partner) as a result of a qualified opinion. Thus, we reestimate (1) after removing from the sample 91 observations with one or two years of audit firm tenure. Results of the new estimation (not reported) are qualitatively the same as those in in column A. Thirdly, similarly to C&S we check the robustness of results to an alternative measure of partner tenure. Hence, we substitute PARTEN for the two dichotomous variables SHORTPARTEN (1 if partner tenure is three years or less and 0 otherwise) and LONGPARTEN (1 if partner tenure is more than four years and 0 otherwise). We chose the median value of PARTEN to delimitate short partner tenures. Results of the new estimation (not reported) do not show any significant effects for either SHORTPAR-TEN or LONGPARTEN. Thus, the non-significant effect of partner tenure on audit qualification does not depend on how partner tenure is measured. In the same line, Davis, Soo, and Trompeter (2009) posed that audit quality could increase in early years through a learning effect, but it would decrease in later years due to a bonding effect. Although the authors studied the effects of audit firm tenure, the explanation could be easily extended to partner tenure. Hence, we estimate a quadratic model with PARTEN and the new variable PARTEN2, defined as the square of PARTEN. Results (not reported) do not support a non-monotonic effect of partner tenure on audit qualifications, as neither PARTEN nor PARTEN2 show any significant effects. The following analysis addresses a potential flaw in our model due to the inclusion of non-going-concern modified opinions (GCMOs) and NGCMOs as a single category of the dependent variable OPINION. It could be argued that, as both types of audit qualifications refer to different dimensions of the audit activity and could also involve different levels of litigation risk, the inclusion of GCMOs and NGCMOs as a single category of the dependent variable might cause misleading results. Hence, similarly to Chi and Chin (2011) we reestimate (1) after the removal from the sample of those observations with GCMOs. As expected, given the low incidence of GCMOs in our data, results (not reported) are very similar to those in column Table 4 (column A), in particular regarding FIRMTEN and PARTEN.
We also address the robustness of our results to situations of superficial partner rotation. Whenever a mandatory partner rotation rule has been implemented, a so-called 'period of grace' has also been established. As an example, the SOX required a five-year time-out period before a partner could re-audit a client. In Spain this period is limited to only two years. Chen et al. (2008) and Bamber and Bamber (2009) posed the importance of differentiating between situations of rotation and those that could be considered as superficial rotation, in which the partner re-audit the same firm again after a short period out the firm. This issue might affect results, since in these situations the measure of tenure as the number of consecutive years auditing the same company will not adequately account for the nature of the relationship between the partner and the client. 6 Therefore, partner tenure measures generally used in the literature, which do not control for superficial rotation, could provide misleading results. Obviously, the potential implications of ignoring situations of superficial rotation will depend on its frequency in the sample. Our dataset comprises 192 partner changes, including eight cases in which the new partner had audited the company before. The eight cases represent 23 observations. While in half of them the partner remained rotated off for three years or less, in the remaining four cases it was out of the firm at least six years. Given the low incidence of superficial rotation in our sample, we do not expect it can seriously affect reported results. Nevertheless, we re-estimate (1), initially without the 23 observations potentially affected by superficial rotation and then reducing the sample only by the four cases (ten observations) in which the partner remained rotated off for three years or less. As expected, results (not reported) do not show any significant differences compared to those reported in Table 4 for the whole sample.
The final analysis performed accounts for the potential effects of economic downturns on the tenure-independence relationship. Periods of economic downturns involve higher litigation risk for the audit firm as the risk of bankruptcy increases. Therefore, we might expect auditors being less willing to impair independence in long-term engagements with either the partner or the audit firm during these periods. To address this issue we split our sample into two subsamples (precrisis and crisis) and afterwards perform separate estimations of (1) for each subsample. Results of the new estimations are shown in Table 5. The most interesting finding would be Accounting in Europe 419 that while PARTEN remains non-significant in both estimations, this is not the case with FIRMTEN, which shows significant results in the estimation conducted with the pre-crisis subsample. The negative sign of the coefficient would indicate loss of independence in longer tenures with the audit firm in the pre-crisis period. The lack of significance of FIRMTEN in the estimation with the crisis subsample and the significance effect with negative sign in the estimation with the pre-crisis subsample would be in line with our expectations discussed above that auditors would be less willing to impair independence during periods of economic downturns.

Concluding Remarks
The main finding of our study would be the lack of significant effects of partner tenure on the likelihood of audit qualifications. Thus, auditor independence does not seem to be Note: Results from logistic regressions. Parameters estimates and z-values in parentheses. * , * * , * * * Significant at the 10%, 5% and 1% levels, respectively. Dependent variable: OPINION: 1 if company receives an 'unclean' audit report and 0 otherwise. Independent variables: PBANK: probability of bankruptcy; SIZE: natural logarithm of total assets of the company; AGE: natural logarithm of the number of years since listing on the supervisor of the Spanish stock market; LEV: total liabilities divided by total equity; CLEV: change in LEV during the year; LLOSS: 1 if client reported negative net income the previous year and 0 otherwise; INVEST: current assets less debtors and inventories, divided by current liabilities; AUDFIRM: 1 if the company is audited by a Big 4 audit firm and 0 otherwise; FIRMTEN: number of consecutive years audited by the same audit firm; PARTEN: number of consecutive years the partner has been signing the audit report of the company. compromised in long-term engagements with the audit partner. This result is robust to a series of checks. Moreover, we do not find different implications of partner tenure on auditor independence depending on the tenure with the audit firm. Evidence reported by C&S and Ye et al. (2011) for Australia showed negative effects of partner tenure on the likelihood of audit qualifications. Although important differences in the methodology and the regulatory context would make difficult the comparability of our results with those in C&S and Ye et al. (2011), it is somewhat surprising that long partner tenures seem to compromise independence in the high litigation Australian market but not it in the Spanish low litigation setting. However, it should be noted that the loss of independence reported by C&S for the Australian market was rather weak and limited to non-Big 4 firms. If compared with Australia, the pronounced concentration of the Spanish market by Big 4 firms could explain these contradictory results. Moreover, C&S and Ye et al. (2011) examined partner tenure under voluntary partner rotation while the institutional context of this research is characterized by mandatory partner rotation. Precisely, the mandatory rotation of partners was imposed in many countries to preserve auditor independence.
Focusing on audit firm tenure, our results do not support loss of independence in long tenures with the audit firm for the whole research period. However, we report a negative effect in the precrisis period, but not during the economic downturn.
These results might have some interesting implications for regulators and policy makers. Firstly, as we do not observe lower independence in longer tenures with the audit partner, the accelerated rotation from seven to five years established by the 2015 Spanish Audit Law, would lack of empirical support. Secondly, the inconclusive results for firm tenure might suggest that auditors could be willing to impair independence in long tenures with the audit firm, except during periods of economic downturns when the risk of litigation increases. Therefore, further research is needed to assess the convenience of imposing the mandatory rotation of the audit firm.
Our study is subject to at least two limitations which need to be considered when interpreting the reported results. The most important would be the ability of the opinion of the audit report to effectively measure audit quality. As posed by DeFond and Zhang (2014), audit quality is a difficult concept. Secondly, the extreme level of concentration of the Spanish audit market by Big 4 audit firms, would make difficult to draw sound conclusions regarding the effects of audit firm type.
There are at least two meaningful extensions of this investigation. Firstly, the scarce research carried out at the audit partner level, jointly with the surprising finding that partner tenure seems to impair independence in the high litigation Australian market but not in Spain would welcome further research examining the issue in countries with different legal traditions. Secondly, it would be interesting to check whether the lack of significant effects of tenure on audit quality reported in this paper is robust to other measures of audit quality.
Notes 1 Quick (2012) provides a detailed discussion of the proposals contained in the Green Paper. 2 Although they did not find a significant effect of tenure on either working capital accruals or abnormal working capital accruals, they reported that a lower proportion of clients misses breakeven for long partner tenure observations, suggesting a greater ability to manage earnings. 3 According to the authors, when the audit firm rotates, the new firm brings a new team, applies its own methodology and new client procedures, while in partner rotation, in most cases, all that changes is one audit partner. 4 For example, the average audit firm tenure is 5.7 years in Chi and Huang (2005); 3.6 years in Knechel and Vanstraelen (2007); 8.6 years in Gul, Jaggi, and Krishnan (2007); 6.9 years in Chen et al. (2008); and 6.9 years in Lim and Tan (2010).