I. INTRODUCTION
Regulators around the world are concerned that longer audit tenure threats auditor independence and lowers audit quality (GAO 2003; European Commission 2010; Humphrey, Kausar, Loft, and Woods 2011). They argue that longer audit tenure may lead to routine and familiarity threat with the client, resulting in excessive reliance on prior years’ results, insufficient audit procedures, and in reporting failures (Brody and Moscove 1998; Arel, Brody, and Pany 2005). Therefore, regulators frequently suggested a mandated system of audit firm rotation (hereafter AFR) as a potential solution to reinforce auditor independence and audit quality. However, opponents of mandatory AFR argue that longer audit tenure is required to develop client-specific knowledge, which is crucial to detect material errors and misstatements (Ghosh and Moon 2005; Jones et al. 2012). A lack of this knowledge will impair auditor independence, and thus audit quality, and may result in more audit failures in the years following an AFR (Geiger and Raghunandan 2002).
Previous studies show contradicting evidence of the actual effect of AFR on audit quality. However, prior researches mainly focus on the effect of mandated system of AFR in the period of normal economic growth (Ewelt-Knauer, Gold, and Pott 2013), this study will focus on the actual effect of voluntary AFR on audit quality during the financial crisis of 2007-2009. The circumstances during the financial crisis are significantly different: the uncertainty is high, risk of going bankrupt increases, and firms have to deal with lower demand which consequently reduces revenue (Paunov 2012). As a result, agency problems increase and firms may have stronger opportunistic incentive for an AFR. First, due to lower profit, firms try to save costs on almost everything (Campello, Graham, and Harvey 2010). Subsequently, it is expected that firms may want to switch audit firms to receive price cuts on audit fees (Ettredge and Scholz 2007). Second, firms are more likely to mask their (potential) poor performance through a more flexible accounting, which also can be achieved through an AFR (Caramanis and Lennox 2008). Finally, firms have higher incentives to shop for an improved audit opinion from a new audit firm, to prevent issuance of a qualified opinion (Krishnan and Stephens 1995). Since this opportunistic incentives for an AFR during the financial crisis may impair auditor independence and increases the extent to which managers are able to report extreme high earnings, it is expected that an AFR during the financial crisis might adversely affect the audit quality. The research question resulting from this is: What is the effect of AFR on audit quality during the financial crisis of 2007-2009?
The sample of this research consist of U.S. companies listed on the New York Stock Exchange (NYSE), during the pre-crisis period (2004-2006) and the crisis period (2007-2009). This study tests the hypothesis with a regression model based on the study of Myers, Myers, and Omer (2003), which used discretionary accruals of Jones’ model (1991) as a proxy for audit quality. Based on the used dataset and this regression model, the effect of the independent variable, AFR during the financial crisis, on the dependent variable, audit quality, is insignificant. Hence, the result of this study does not support that AFR during the financial crisis of 2007-2009 has an adversely effect on audit quality.
The evidence of this study contributes to the literature about the effects of AFR on audit quality. The results of this study are also relevant for practice, especially for regulators, companies who consider an AFR, investors, and other financial statement users. This research informs regulators whether it is reasonable to mandate AFR, in order to reinforce auditor independence and audit quality, to subsequently prevent audit failures and to protect investors. Also, this study informs companies whether AFR is a rational action in case they want to protect their auditor independence and audit quality. Finally, this study alerts investors and other entities who rely on a fair presentation of the financial statement to be more critical toward it.
The structure of this paper is organized as follows. Section II summarizes the existing literature on this topic, leading to a new hypothesis. Section III discusses the methodology, and section IV describes the sample selection. Finally, Section V presents the results, leading to a conclusion in Section VI.
II. LITERATURE REVIEW AND HYPOTHESIS DEVELOPMENT
Quality of audit can be defined as a joint probability that an auditor will both detect and report material misstatements (DeAngelo 1981a). An auditor with more expertise and competence is better able to discover misstatements in client’s accounting system, and a more independent auditor is more likely to report the detected misstatement. Maintaining high audit quality is important to prevent audit failures and protect stakeholders against extreme earnings management (Francis 2004; GAO 2003). Audit failures occur when an audit firm issues an unqualified opinion on financial statement that are actually materially misstated (Kadous 2000). Extreme earnings management can hide firm’s poor performance, and thus mislead investors and other stakeholders who rely on a fair presentation of financial statement. If an auditor would be fully experienced, competent, and independent enough, there would be no audit failure or extreme earnings management. Therefore, regulators frequently suggested a mandated system of AFR, to maintain a high auditor independence and audit quality (GAO 2003; Humphrey et al. 2011). However, research on the actual effects of AFR on AQ resulted in inconclusive evidence, both supporting as contradicting.
Proponents of mandatory AFR argue that longer audit tenure may impair auditor independence, and thus audit quality, as a result of an excessive familiarity with the client (Tepalagul and Lin 2015). This tight relationship creates a situation in which an auditor may be over-influenced by the client’s management, and the pressure to retain the client may lead to an eagerness to please the client (Arel et al. 2005). Another argument is the threat of routine, as reflected in excessive reliance on prior-years’ results and static audit programs, resulting in insufficient audit procedures and a lack of attention for detail (Brody and Moscove 1998). AFR may avoid this excessive familiarity with the client and give an opportunity for a fresh perspective to audit, which subsequently improves the audit quality (Ewelt-Knauer et al. 2013).
In contrast, other studies claim that AFR results in a lack of familiarity with the client, which in turn impairs auditor’s expertise, competence, and independence (Geiger and Raghunandan 2002; Myers et al. 2003; Ghosh and Moon 2005). With longer audit tenure, auditor develops client-specific knowledge of items such as firm’s accounting system and internal control structure, which are crucial to detect material misstatements, to subsequently prevent audit failures (Johnson, Khurana, and Reynolds 2002; Jones et al. 2012). Moreover, a lack of firm-specific expertise may impair auditor independence, due to an increased dependence of management and a reliance on managerial estimates when an auditor is new (Solomon, Shields, and Whittington 1999). Myers et al. (2003) used discretionary and current accruals to analyze the effect of audit tenure on earnings quality and found that both accruals decreased with longer audit tenure. This because of auditors placing a greater constraint on extreme management reporting decisions when audit tenure increases.
Even thought, there is a considerable amount of literature available on the actual effect of AFR on audit quality, no study has examined this effect yet under special circumstance, like the global financial crisis of 2007-2009. The circumstances during a financial crisis are significantly different, because of stronger opportunistic incentives of firms to switch audit firms, which subsequently might adversely affect the audit quality.
First of all, during the financial crisis, firms have to cope with lower demands which consequently reduce profits. To survive the crisis, firms try to save costs on almost everything (Campello et al. 2010). Since U.S. listed firms have to be audited, it is expected that firms may want to switch audit firms to receive price cuts on audit fees (Ettredge and Scholz 2007). That is due to a low-balling practice of audit firms, i.e. pricing of initial engagements below costs to attract new clients (Arruñada 2004). It is expected that low-balling competition between audit firms is even higher during the crisis period. Switching audit firms may respectively reduce the audit quality due to impaired auditor independence and shorter auditor-client relationship, which results in lower probability that misstatements will be detected and reported (Jones et al. 2012). Until the low-balling costs are recovered, audit firms may have an incentive to retain clients in the future period and put lower weight on reduced auditor independence (Simon and Francis 1988). Also, audit firms may want to compensate their low fees with less hours and effort spending on the audit work. Lower audit hours and effort are consequently associated with larger abnormal accruals and a higher likeability to which managers are able to manage earnings upwards (Caramanis and Lennox 2008).
Another negative incentive for an AFR during the financial crisis, is that managers may want to hide their less favorable performance, which means that less misstatements are likely to be detected and reported. Hiding poor performance is attractive to let financial statements look good and to create confidence by investors and other stakeholders (DeFond and Park 1997). This can be achieved through earnings management. Myers et al. (2003) argued that longer audit tenure results in auditors placing greater constrains on extreme earnings management. Therefore, firms may want to try manage their earnings upwards through an AFR, which in turn lowers the probability to detect and report errors in the years following an AFR, and thus lower audit quality. Lower audit quality is subsequently associated with more accounting flexibility (Becker, DeFond, Jiambalvo, and Subramanyam 1998). Because of the higher uncertainty about the future and higher risk of going bankrupt, even financially healthy firms might be more interested in increasing their accounting flexibility through an AFR.
Finally, AFR during financial crisis may also impair auditors independence due to an increased incentive of firms to prevent an issuance of qualified opinion, and an incentive to switch to an audit firm that is more likely to approve firm’s preferred reporting behavior. These attempts may be motivated by opinion shopping, i.e. shopping for an improved audit opinion from a new audit firm (Krishnan and Stephens 1995).
Summarizing, the opportunistic incentives for an AFR during the financial crisis to save costs on audit may lower the audit hours and audit effort, low audit effort subsequently increases the extent to which managers are able to report extreme high earnings, and extreme earnings management is attractive for firms who may want to hide their poor performance. Consequently, AFR may impair auditor’s independence and lower the audit quality, because of lower probability that misstatements will be detected and reported. Therefore, the hypothesis of this study is:
H1: AFR negatively affects audit quality during the financial crisis of 2007-2009.
III. METHOD
To empirically test the hypothesis of this study, the following regression model will be used:
AQi, t = ꞵ0 + ꞵ1 AFRi, t + ꞵ2 CRISISi, t + ꞵ3 AFR_CRISISi, t + ꞵ4 SIZEi, t + ꞵ5
SALESGROWTHi, t + ꞵ6 CFOi, t + ꞵ7 LEVERAGEi, t + ꞵ8 BIG4i, t + ꜫi, t
Table 6 in the Appendix details the definition of each variable.
The regression model is based on the study of Myers et al. (2003), which used discretionary accruals of Jones’ model as a proxy for audit quality (AQ), with an independent variable, audit firm tenure. Due to a small sample, this study replaces independent variable Tenure by independent variable AFR, as also was done in the study of Ruiz-Barbadillo, Gómez-Aguilar, and Carrera (2009). The model of this study will test the association between AQ and AFR during the financial crisis of 2007-2009 (CRISIS). To test this, three independent variables are included in the model. The independent variables, AFR and CRISIS, will subsequently investigate whether they affect the dependent variable, AQ. Combining these two independent variables results in an interaction term, AFR_CRISIS, which will test the hypothesis of this study. If the hypothesis of this study is correct, then I expect the coefficient on AFR_CRISIS to be positive, concluding that during the financial crisis, firms are more likely to increase their discretionary accruals through an AFR. These higher absolute discretionary accruals are in turn associated with lower audit quality (i.e. negative relation).
Several independent control variables are added to the regression model to reduce the possibility that AFR and CRISIS proxy for other determinants of accruals. First, control variable SIZE is included because larger firms tend to record more stable accruals (Dechow and Dichev 2002). SALESGROWTH is included because accruals are likely to be positively correlated with company’s growth opportunities (Carey and Simnett 2006). CFO is used because firms with higher cash flows from operations are likely to perform better (Frankel et al. 2002), and because of the negative relationship between accruals and cash flows (Dechow 1994). The variable LEVERAGE is used as a proxy for the possibility of debt covenant violations that may generate an incentive to increase earnings through higher accruals (Cameran, Prencipe, and Trombetta 2016). Finally, control variable BIG4 is included because the conservatism of larger audit firms tends to limit extreme accruals (Becker et al. 1998). The control variable AGE is excluded from the model of Myers et al. (2003) because this data is difficult to gather, and because the analyses of Myers et al. (2003) reveal no significant impact of AGE on AQ.
IV. SAMPLE SELECTION
The sample of this study is collected from the database COMPUSTAT North America from the period of 2004-2009. The primary interest of this research is the audit firm rotation during the global financial crisis of 2007-2009. The three years before the financial crisis, 2004-2006, are used as a control sample period. The sample is counted by company’s unique CIK number and focusses on the U.S. non-financial companies listed on the New York Stock Exchange (NYSE), which is the largest stock exchange in the world. Looking at the availability of the data, the companies without a CIK number, companies that are unaudited, or for which an auditor is not defined, are excluded from the sample. Also, other observations with missing values are excluded. This resulted in a final dataset of 7,758 firm-year observations, represented by 1,293 companies. Further, this sample selection procedure yielded to a sample of 315 audit firm rotations, with 147 rotations during the financial crisis, and 168 in the pre-crisis period.
Descriptive statistics
Table 1 shows the descriptive statistics of the continuous variables of the regression model for the full sample. The standard deviation of AQ (0,781) shows that values are not concentrated around the mean (0,621), what indicates a large difference in AQ between companies. The independent variables AFR and AFR_CRISIS show no differences in frequency of AFR in the pre-crisis period and in the crisis period (both 2%). So, companies have no extraordinary incentives for an AFR during bad financial times. Further, 94% of companies that are used in the sample, is audited by Big 4 audit firms.
Univariate analyses
Table 2 shows the results of an independent-samples T-test to compare the means of AQ. This study investigates whether AFR during the financial crisis has an effect on audit quality. Therefore, the sample is split into groups based on the dummy variables AFR and CRISIS. During the financial crisis, the results show higher mean of AQ after an AFR (0,021 difference). During the pre-crisis period, the mean of AQ after an AFR is even higher than during the crisis period (0,679 and 0,644 respectively). In general, the results show higher mean of AQ after an AFR (0,044 difference). There is no large difference between the means of AQ in the pre-crisis period compared to the crisis period (0,618 and 0,624 respectively). While the differences between all the means are insignificant, no evidence can be provided to state that firms are more likely to report higher discretionary accruals after an AFR during the financial crisis, and thus lower AQ.
Correlation matrix
The Pearson correlation matrix in table 3 shows that the dependent variable AQ is positively correlated with the independent variables AFR (0,011), CRISIS (0,004), and AFR_CRISIS (0,004), as was expected. However, no evidence can be provided about this positive correlation because this correlation is insignificant. The control variable LEVERAGE is insignificantly positively correlated to the dependent variable AQ and to the independent variable AFR_CRISIS (0,008 and 0,010 respectively), but significantly positively correlated to the independent variable CRISIS (0,045). Further, both control variables SIZE and BIG4 are significantly negatively correlated to AQ (-0,165 and -0,068 respectively), as was expected. So it can be stated that smaller companies and companies audited by non-Big 4 audit firms, report higher discretionary accruals, thus have lower audit quality, as also was argued in the study of Becker et al. 1998. Moreover, the control variable BIG4 is significantly negatively correlated to the dependent variables AFR, CRISIS, and AFR_CRISIS (-0,166, -0,025, and -0,094 respectively). Hence, it can be argued that companies are likely to switch to a non-Big 4 firm when they change audit firms.
Since all the values in Table 3 are below 0,6, there are no multicollinearity problems.
Multivariate results
Table 4 shows results of the regression analyses. Based on the F-statistics, the model is significant at the 1% level. The adjusted R2 is only 0,050, which is close to the adjusted R2 of 0,083 reported in the similar study of Johnson, Khurana, and Reynolds (2002). The relationship between the interaction term AFR_CRISIS and the dependent variable AQ is negative (-0,040), which is not consistent with H1. While this relationship is insignificant, it cannot be concluded that AFR during the financial crisis has any effect on the audit quality. Therefore, the hypothesis of this study is rejected. This insignificant effect can be explained by the fact that audit firms might be more concerned about their reputation (DeAngelo 1981a), or that during the financial crisis, firms might be more under pressure to present a fair financial statement, in order to avoid major audit failures like at the beginning of the 21st century (GAO 2003). The results of all other control variables have a significant effect on AQ. Moreover, all control variables, except the variables SALESGROWTH and CFO, are in line with the expectations of this study. However, negative coefficient of SALESGROWTH is in line with the studies of Myers et al. (2003) and Cameran et al. (2016).
Although considerable amount of research has been done on the topic actual impact of mandatory AFR on audit quality, the resulted outcomes are divided, both supporting as well as contradicting. The purpose of this study is to investigate whether an AFR during the financial crisis of 2007-2009 has a negative effect on audit quality. For this study, data of U.S. listed companies during the financial crisis period is used, because it is expected that during bad financial times an AFR will negatively affect the audit quality. During the financial crisis, firms have to deal with higher uncertainty, higher risk of going bankrupt, and lower demand which consequently reduces revenue. As a result, firms may have increased incentives to cut their audit fees, to increase their accounting flexibility, or to shop for an improved audit opinion from a new auditor, through an AFR. These negative incentives for an AFR may consequently impair auditor independence, which results in lower ability of an auditor to detect and report misstatements, thus lower audit quality.
Based on the sample of U.S listed companies on NYSE, it cannot be concluded that AFR during the financial crisis has a negative effect on audit quality. Therefore, the hypothesis of this study is rejected. This can be explained by the fact that audit firms might be more concerned about their reputation during the crisis period, in order to avoid major audit failures like at the beginning of 21st century. Another reason could be that that during the financial crisis, firms might be more under pressure to present a fair financial statement, to protect creditors, investors, and others who rely on a fair presentation of financial statement.
This study contributes to the literature on the potential effects of AFR on AQ, by providing more information about the opportunistic incentives of firms to switch audit firms during the crisis period, which may negatively affect the audit quality. As the results of this study are insignificant, no evidence can be provided that regulators should mandate or should put a restriction on AFR.
Several limitations of this study should be mentioned. Firstly, the used sample contains only U.S. firms that are listed on the NYSE. Therefore, the results cannot be generalized to smaller firms or other countries. Secondly, not all U.S. companies listed on the NYSE are used, due to limited data availability. Because of this small sample, the number of AFR (316) compared to total firm-years observations (7,758), is relatively small, which limits the extent to which the results of this study are valuable enough to base conclusions on.
Further research can include more data, with a larger number of companies listed on other stock markets, which may lead to more reliable results. Besides that, as there are several models to measure audit quality (DeFond and Zhang 2014), further research could empirically test whether different proxies for audit quality lead to different results.