Change for Change s Sake? Does Mandatory Partner Rotation Improve Audit Quality?

Transcription

1 Change for Change s Sake? Does Mandatory Partner Rotation Improve Audit Quality? ABSTRACT: Opponents of mandatory rotation argue that a change of partner is bad for audit quality as it results in a loss of client-specific knowledge. On the other hand, proponents argue that a change of partner is beneficial as it results in a positive peer review effect and a fresh perspective on the audit. We test the impact of mandatory partner rotation on audit quality using a unique dataset of audit adjustments in China. Our results suggest that mandatory rotation of engagement partners results in higher quality audits. Specifically, we find a significantly higher frequency of audit adjustments during the departing partner s final year of tenure prior to mandatory rotation and during the incoming partner s first year of tenure following mandatory rotation.

2 I. INTRODUCTION Many jurisdictions impose limitations on the length of audit partner tenure but they impose no limitations on the length of audit firm tenure. For example, audit partners are subject to mandatory periodic rotation in Argentina, Australia, Belgium, China, Denmark, France, Germany, Hong Kong, Mexico, Netherlands, New Zealand, Norway, Russia, Taiwan, the United Kingdom, and the United States. However, none of these jurisdictions require the periodic rotation of audit firms. Despite the widespread prevalence of this practice, there is very little evidence on the consequences of mandatory partner rotation in the situation where audit firms do not have to be rotated. This is primarily because most countries do not require partners names to be disclosed and so researchers are unable to identify when partner rotation occurs. When only the audit partner is rotated and the audit firm remains the same, the audit methodology, procedures, and other engagement personnel do not necessarily change (Bamber and Bamber 2009; Chi et al. 2009; Bedard and Johnstone, 2010; Fitzgerald et al. 2012). Therefore, it is an open question whether mandatory partner rotation can really have a substantive impact in terms of improving audit quality in a regime where audit firms are not required to be rotated. This question is important because if mandatory partner rotation does not help to improve audit quality, then regulators are likely to call for alternative and more hard-hitting policies, such as mandatory rotation of the entire audit firm. We are aware of only one published study on the consequences of mandatory partner rotation. Chi et al. (2009) examine mandatory partner rotation in Taiwan, a jurisdiction that requires partners names to be disclosed and does not require mandatory rotation of audit firms. Chi et al. (2009) use abnormal accruals and earnings response coefficients (ERCs) as proxies for audit quality and find little evidence that these variables are affected by mandatory partner rotation. In their discussion of Chi et al. (2009), Bamber and Bamber (2009) point out that these earnings quality metrics may be poor proxies for 1

3 audit quality, which may explain why Chi et al. (2009) obtain insignificant results. Bamber and Bamber (2009: 397) state that: proxies such as abnormal accruals and earnings response coefficients may be among the more popular measures used to date, but they are nonetheless noisy measures of earnings quality, much less audit quality [ ] development of sharper measures of audit quality will provide a major breakthrough. This will require imagination and creativity (p. 399). We respond to this challenge by measuring audit quality using a proprietary dataset of audit adjustments. An audit adjustment occurs when two conditions are met: 1) the client s pre-audit financial statements are misstated, and 2) the auditor detects the misstatement and requires the client to correct the misstatement through an adjustment to the financial statements. The second of these conditions is exactly equivalent to DeAngelo s (1981) conceptual definition of audit quality. That is, audit quality is the probability that an existing misstatement is discovered by the auditor and subsequently corrected. The key to our identification strategy is that mandatory rotation can affect condition #2 (i.e., rotation can affect audit quality) but rotation would not directly affect condition #1. In other words, holding audit quality constant, we have no reason to believe that mandatory rotation would directly affect the quality of the client s pre-audit financial statements. 1 Accordingly, any association between mandatory rotation and audit adjustments is through condition #2 (audit quality). This means that we are able to assess how mandatory rotation affects audit quality by examining how mandatory rotation affects the incidence of audit adjustments. 1 Partner rotation could affect a client s pre-audit financial reporting quality indirectly, through a change to audit quality. For example, suppose that partner rotation causes audit quality to be higher in the incoming partner s first year following mandatory rotation. If a client rationally anticipates this, then the client would know that any misstatements are more likely to be detected and corrected by the new partner, motivating the client to prepare higher quality financial statements in the year following rotation. The key thing to note, however, is that this rotation effect occurs only when there is a change to audit quality. That is, mandatory rotation does not affect the quality of the client s preaudit financial reporting unless there is also a change to audit quality. 2

4 The effect of mandatory partner rotation on audit quality is far from obvious. On one hand, a change of partner may worsen audit quality because partners gain more knowledge from longer tenure with the client. Rotation may increase the risk that the incoming partner will be less well informed about the client and therefore less likely to identify financial reporting problems. On the other hand, a change of audit partner may improve audit quality by bringing a fresh perspective to an audit. Indeed, the requirement to periodically rotate audit partners was introduced in the United States for the specific purpose of periodically bringing a fresh perspective to each audit (AICPA 1992, p. 4). Further, it has been argued that a change of partner can provide a powerful peer review effect because the incoming partner assesses the work carried out by the departing partner in the previous year (Seidman 2001; Biggs 2002; Public Oversight Board 2002). This can help incentivize the departing partner to conduct a more thorough audit before the engagement is handed over to the new partner. Accordingly, mandatory rotation can improve audit quality in the year before rotation occurs and in the year of appointment of the new partner. We choose China as our empirical setting for two reasons. First, both the engagement and review partners have to be rotated every five years, just as they are in the United States (US). However, unlike the US, audit reports in China disclose the names of the engagement and review partners. This allows us to identify cases in which one or both of the partners are required to rotate. Second, since 2006, the Ministry of Finance has required audit firms to report to it the pre-audit annual profits of all publicly traded audit clients. The Ministry provided this proprietary data to us for the purposes of academic research. Using this data, we are able to identify which engagements had audit adjustments to reported profits (i.e., an audit adjustment occurs when pre-audit profits post-audit profits). We begin our analysis by examining whether there is a peer review effect that causes the departing partner to conduct a more thorough audit during his final year of tenure before mandatory rotation. When an incumbent partner is scheduled for mandatory rotation 3

5 at the end of the year t audit, he knows for sure that a new partner will take over the audit in year t+1. When planning an audit for the current year, auditors typically look back to the working papers of the previous year (Wright 1988; Tan 1995). If a newly appointed partner finds that the audit in the prior year was unsatisfactory, then this is likely to impair the reputation of the partner who previously worked on the audit. For example, if the incoming partner finds that the previous year s audit was not in accordance with auditing standards, the new partner may inform other partners in the audit firm about the previous partner s lack of competence. We expect that a partner who is scheduled for mandatory rotation would take this into account when deciding how to conduct the audit in his final year of tenure. In particular, the departing partner has a strong incentive to conduct a higher quality audit in his final year (year t) in order to avoid the embarrassment of the audit deficiencies being found by the incoming partner (in year t+1). The new partner is especially likely to detect such deficiencies if he brings a fresh perspective to the audit. Accordingly, when the incumbent partner is in his final year of tenure prior to mandatory rotation, he has a stronger incentive to conduct a high quality audit by detecting and correcting client misstatements. If mandatory rotation has this beneficial peer review effect, we hypothesize that audit adjustments occur more frequently in year t if mandatory partner rotation is scheduled to occur in year t+1 (H1). The proponents of mandatory rotation argue that the new partner may be more independent of the client than the previous incumbent partner because in the first year of tenure the new partner has not had time to develop a close working relationship with management. Moreover, the appointment of a new partner can bring a fresh perspective to an audit. For example, a new partner may spot financial reporting problems that were missed by the previous partner. If mandatory rotation really does lead to higher audit quality, we hypothesize that audit adjustments would occur more often in the replacement 4

6 partner s first year of tenure than in years when partners are unaffected by mandatory rotation (H2). In short, we expect that if mandatory rotation leads to higher audit quality, there would be more audit adjustments during the departing partner s final year of tenure (H1) and during the incoming partner s first year of tenure (H2) than in other years. We test these predictions using a sample of 6,341 audits conducted between 2006 and Our first major finding is that audit adjustments occur more often when the engagement partner is scheduled for mandatory rotation at the end of the year. Consistent with a beneficial peer review effect, this suggests that the departing partner anticipates the arrival of a new partner in year t+1 and this motivates the departing partner to conduct a higher quality audit in year t before the engagement is handed over to the new partner. Our second major finding is that audit adjustments occur more often during the incoming partner s first year of tenure than in other years. This is consistent with mandatory rotation resulting in a fresh eyes effect such that a newly appointed partner is more likely to detect and correct financial reporting problems during his first year of tenure. Overall, our results suggest that mandatory partner rotation has a beneficial effect in the final year of tenure before rotation occurs and in the subsequent year when the new partner is appointed. Interestingly, these results are found to be statistically significant for both material and immaterial audit adjustments. In addition, the results hold for both downward audit adjustments (i.e., where pre-audit profits > post-audit profits) and for upward audit adjustments (i.e., where pre-audit profits < post-audit profits). Further, the results are found to be highly significant for engagement partners but much weaker for review partners. This makes sense given that the engagement partner has a more important role in terms of overseeing the audit fieldwork. Moreover, we provide evidence that the higher frequency of audit adjustments surrounding partner rotation is not attributable to differences of judgment between the departing partner and the new partner. 5

7 Our study makes two main contributions. First, we contribute to the literature on mandatory partner rotation by showing that it has a beneficial impact on audit quality even in the absence of mandatory audit firm rotation. Except for Chi et al. (2009), we are unaware of any published studies examining the consequences of mandatory partner rotation and we are the first to show that mandatory partner rotation can have a positive impact on audit quality. This is important given that many countries currently require partner rotation but few countries require audit firm rotation. Nevertheless, it remains an open question as to whether mandatory audit firm rotation would yield incremental benefits for audit quality beyond those that we detect from mandatory partner rotation. Second, we contribute to the existing literature on audit adjustments (e.g., Kinney and Martin 1994; Wright and Wright 1997; Braun 2001; Keune and Johnstone 2012). In particular, we show that mandatory partner rotation significantly increases the frequency of audit adjustments. We find that the mean predicted probability of an audit adjustment is 66.8% when the engagement partner is not in his final year of tenure prior to mandatory rotation and not in his first year of tenure following mandatory rotation. In comparison, the audit adjustment probability is 77.5% when the engagement partner is in his final year of tenure prior to mandatory rotation and 75.2% when the engagement partner is in his first year of tenure following mandatory rotation. The remainder of the paper is as follows. Section II discusses the prior literature, develops the two hypotheses, and explains the research design. Section III introduces the sample and provides descriptive statistics. Section IV reports the main results and the findings of supplementary analyses. Section V concludes by discussing the policy implications and limitations of our study. 6

8 II. BACKGROUND Audit partner tenure and audit partner rotation In most countries audit partners names are not publicly disclosed, making it difficult for researchers to identify partner rotation events and measure partner tenure. Australia, Taiwan, Germany, and China are four exceptions because these jurisdictions require audit reports to publicly disclose partners names. Carey and Simnett (2006) examine a sample of Australian companies in 1995, a period during which both partner rotation and audit firm rotation were voluntary. Consistent with longer partner tenure leading to lower audit quality, they find that clients are less likely to receive going-concern opinions and are more likely to meet or beat earnings benchmarks when partner tenure exceeds seven years. However, they find no evidence that signed or absolute abnormal accruals are associated with the length of partner tenure. Fargher et al. (2008) find that absolute abnormal accruals in Australia are significantly smaller during the first few years of audit partner tenure than in other years. However, this finding is only obtained for cases in which the new partner is from the same audit firm as the departing partner. Fargher et al. (2008) find that absolute abnormal accruals are significantly larger in the partner s initial years of tenure when the new partner is from a different audit firm. Like Fargher et al. (2008), Chen et al. (2008) jointly examine the effects of both partner tenure and audit firm tenure. Using data from Taiwan during a period in which both partner rotation and audit firm rotation were voluntary, they find that discretionary accruals decrease significantly with audit partner tenure. After controlling for partner tenure, they find that discretionary accruals also decrease with audit firm tenure. Similar to the above studies of Australia and Taiwan, Gold et al. (2012) examine a setting (Germany) in which all audit partner switches are voluntary. Using absolute and signed abnormal accruals, they find evidence of less earnings management as review partner tenure increases, but not as engagement partner tenure increases. 7

9 Audit partners names are not publicly disclosed in the US. To overcome this obstacle, Bedard and Johnstone (2010) use proprietary data from a large US audit firm. They find that planned engagement effort increases following partner rotation, indicating that new partners invest additional effort during their first year of tenure in order to acquire client knowledge. Using a small sample of proprietary data from the US offices of three large audit firms, Manry et al. (2008) find that abnormal accruals are significantly and negatively associated with engagement partner tenure, but only for small clients with partner tenure greater than seven years. Fitzgerald et al. (2012) examine how audit partner rotation and audit firm rotation affect the reporting of internal control deficiencies for a sample of large US not-for-profit organizations. They find a significant increase in reported deficiencies during the first year of audit firm tenure but no significant association between partner rotation and the reporting of deficiencies. However, Fitzgerald et al. (2012) do not distinguish between partner rotations that are voluntary versus those that are mandatory. To our knowledge, the only study to focus specifically on mandatory partner rotation is by Chi et al. (2009). Using data from Taiwan, they find that abnormal accruals during the replacement partner s first year following mandatory rotation are not significantly different from abnormal accruals in a no rotation sample. Moreover, the earnings response coefficient (ERC) during the incoming partner s first year is not significantly different from the ERC in the no rotation sample. In their discussion article, Bamber and Bamber (2009) characterize Chi et al. (2009) as essentially a no-results paper. They note that the lack of significant results could be due to the abnormal accruals and ERC variables being noisy proxies for audit quality or because partner rotation truly has no impact on audit quality. 8

10 Audit firm tenure and audit firm rotation There are very few countries that require audit firms to be rotated. Therefore, most studies in the extant literature examine the association between audit firm tenure and audit quality in settings where audit firm switches are purely voluntary (e.g., the United States). Proponents of mandatory rotation argue that short audit firm tenure can strengthen auditor independence due to the avoidance of close personal relationships between the auditors and the client s management. However, the weight of evidence does not support this view. To the contrary, the literature generally finds that shorter audit firm tenure is associated with lower quality reporting. 2 A notable exception is Davis et al. (2009), which finds an increase in the use of discretionary accruals to meet or beat earnings forecasts in both the early and later years of the auditor-client relationship. While the evidence on audit firm tenure helps to inform the debate about the consequences of extended audit firm tenure, it does not directly examine the consequences of mandatory audit firm rotation. Because voluntary rotation is endogenous it is difficult to draw clear inferences from this literature about the likely consequences of mandatory audit firm rotation. For example, companies that engage in earnings management or that shop for clean audit opinions are more likely to voluntarily change their audit firms or experience auditor resignations (DeFond and Subramanyam 1998; Lennox 2000). Thus, the positive association between audit firm tenure and audit quality could reflect the underlying factors that cause voluntary switches rather than the exogenous impact of mandatory rotation. To provide direct evidence on the consequences of mandatory audit firm rotation, recent studies have looked at countries that have implemented this requirement. Kwon et al. (2010) examine the introduction of mandatory audit firm rotation in South Korea. They find 2 In these studies of audit firm tenure, audit quality is measured using abnormal accruals (Johnson et al. 2002; Myers et al. 2003), accounting restatements (Stanley and DeZoort 2007), fraudulent accounting (AICPA 1992; Carcello and Nagy 2004), lawsuits against auditors (St. Pierre and Anderson 1984; Palmrose 1986; Palmrose 1991; Stice 1991), earnings response coefficients (Ghosh and Moon 2005), and the cost of debt financing (Mansi et al. 2004). 9

11 that audit hours and audit fees increased subsequent to mandatory audit firm rotation but they find little evidence of an impact on audit quality. Livne and Pettinicchio (2012) examine earnings quality during Italy s mandatory audit firm rotation regime. They find no significant difference in earnings quality during the replacement audit firm s first year following rotation, compared with the years in which mandatory rotation does not occur. In another study of Italy, Cameran et al. (2012) find that earnings quality is lower following mandatory audit firm rotation than it is in other years. Finally, Harris and Whisenant (2012) find an improvement in earnings quality after mandatory audit firm rotation was introduced into Brazil and South Korea. This finding suggests that the introduction of audit firm rotation helps to improve earnings quality due to an improvement in auditor independence. In addition, Harris and Whisenant (2012) compare earnings quality in the years before and after audit firms are compulsorily rotated in Brazil, Italy, and South Korea. 3 They find that earnings quality is lower in the year immediately after mandatory audit firm rotation than in the year before, suggesting that mandatory rotation can cause lower earnings quality due to a loss of knowledge effect. Our study is different from this literature on mandatory audit firm rotation because we examine the consequences of mandatory partner rotation. As noted by Carey and Simnett (2006) and Bedard and Johnstone (2010), it is problematic to generalize findings on audit firm rotation to individual audit partner rotation because partner rotation generally involves only a change in the identity of the partner assigned to the audit. Therefore, it is important to determine whether mandatory partner rotation can have a positive impact on audit quality in the absence of mandatory audit firm rotation. 3 As Italy introduced mandatory audit firm rotation in 1975, Harris and Whisenant (2012) compare earnings quality under the voluntary and mandatory rotation regimes for Brazil and South Korea alone. 10

12 Audit adjustments During the course of a year-end audit, an auditor may detect potential misstatements and propose adjustments to the financial statements. Then, after negotiations between the auditor and management, the proposed adjustments are either waived (no adjustment) or not waived (the accounts are adjusted). Prior studies largely focus on the decision to waive or not waive a proposed adjustment. For example, Wright and Wright (1997) find that auditors waive approximately 50% of proposed adjustments. Proposed adjustments are more likely to be waived if they require subjective judgment (e.g., accounting estimates) or the adjustments are offsetting (i.e., income-decreasing adjustments are offset by incomedecreasing adjustments). Braun (2001) finds that proposed adjustments are more likely to be waived for clients that pose less risk; e.g., financially healthy companies. Keune and Johnstone (2012) find that proposed adjustments are more likely to be waived when audit fees are lower and when audit committees lack financial expertise. Hatfield et al. (2008) find that auditors sometimes propose and subsequently waive unnecessary adjustments in order to encourage management to agree to more significant adjustments that are not waived. The dependent variable in our study is different from these studies as we examine the incidence of non-waived audit adjustments. A non-waived adjustment occurs when management accepts the need to correct the financial statements. Therefore, a non-waived adjustment is likely to be a more reliable indicator of audit quality than is a waived adjustment. In any case, waived adjustments cannot be examined in our setting because audit firms are only required to disclose the non-waived adjustments to the Ministry of Finance. Kinney and Martin (1994) find that non-waived adjustments occur on between 60-90% of audits in the United States. Moreover, they show that auditors are more likely to require earnings to be adjusted downwards than upwards. As discussed in Section III, we find similar results for our sample of non-waived audit adjustments. 11

13 Hypotheses Audit quality during the departing partner s final year prior to mandatory rotation The proponents of mandatory rotation argue that a change of partner can provide a powerful peer review effect because the incoming partner is likely to examine the work undertaken by the former partner in the previous year (Seidman 2001; Biggs 2002; Public Oversight Board 2002). The reason for this is that auditors often look back to the prior year s working papers when planning the audit for the current year (Wright 1988; Tan 1995). Auditors do not generally go back beyond the most recent year because it is the prior year that is of most relevance to the current year. If an incoming partner finds that the audit in the prior year was unsatisfactory, then this is likely to impair the reputation of the departing partner within the audit firm. For example, if the incoming partner finds that the audit in the prior year was not in accordance with auditing standards, the incoming partner may inform others in the audit firm. A replacement partner is particularly likely to find financial reporting problems if the replacement partner brings a fresh approach to the audit. We expect that a departing partner would take this into account when the partner knows that he is scheduled for mandatory rotation at the end of the audit. In particular, the departing partner has incentives to make necessary adjustments in his final year of tenure because otherwise the replacement partner may find reporting problems (e.g., overstated assets) carrying over from the previous year. The discovery of such problems would reflect badly on the departing partner and tarnish his reputation within the audit firm. We therefore expect that the departing partner conducts a more thorough audit during his final year of tenure prior to mandatory rotation. In contrast, when mandatory rotation is not scheduled for the following year, the incumbent partner does not need to be as careful because it is less likely that a new partner will scrutinize the audit several years later when rotation eventually takes place. 12

14 Therefore, our first hypothesis is that audit adjustments are more likely to occur in year t when a new partner is scheduled to be rotated onto the audit in year t+1: H1: Audit adjustments are more frequent during the departing partner s final year of tenure prior to mandatory partner rotation than in years when audit partners are not affected by mandatory rotation. On the other hand, the data will not support H1 if there is no peer review effect. For example, the replacement partner may adopt an unquestioning approach when assessing how the audit was conducted in the previous year under the former partner. Further, the financial reporting issues that were pertinent in year t may cease to be relevant to the audit in year t+1. Thus, a new partner may not find it necessary to evaluate the work done by the former partner in the previous year. Audit quality during the replacement partner s first year following mandatory rotation Proponents of mandatory rotation argue that a replacement partner brings a fresh perspective to the audit and is therefore more likely to detect and correct financial reporting problems. Consistent with this, Tan (1995) provides experimental evidence that auditors with involvement in the prior year s audit are less likely than newly appointed auditors to pay attention to factual inconsistencies between the current year and the previous year. Similarly, Favere-Marchesi and Emby (2005) show in an experimental study that new partners are more likely than continuing partners to recognize an impairment of goodwill. In sum, both studies suggest that audit partner rotation can result in a beneficial fresh eyes effect. This fresh eyes argument has been a key factor influencing the decision to introduce mandatory partner rotation in many countries. For example, the Cadbury Committee in the UK argued that mandatory partner rotation enhances the rigor of an audit by encouraging a fresh viewpoint (Committee on the Financial Aspects of Corporate Governance 1992). In the 13

15 US, the AICPA introduced mandatory partner rotation every seven years in order to periodically bring a new perspective to audits (AICPA 1992). After the Sarbanes-Oxley Act (2002) reduced the mandatory rotation period from seven years to five years, the AICPA submitted a comment supporting the change, stating that partner rotation provides a periodic fresh look at an issuer s financial statements (AICPA 2003). In addition to the fresh eyes benefit, the proponents of mandatory rotation argue that a newly appointed partner is more independent of the client because the new partner would not have had time to develop close personal relationships with management. In contrast, a partner who has been with a client for several years may be overly trusting of the client s management or unwilling to challenge management s reporting assertions. For example, the Code of Ethics prepared by the International Federation of Accountants (IFAC) claims that using the same lead engagement partner on an audit over a prolonged period may create a familiarity threat (IFAC 2003, para ). These arguments were seemingly influential in China s decision to introduce mandatory partner rotation in Explaining its decision to introduce mandatory partner rotation, the China Securities Regulatory Commission (CSRC) stated that: A number of international and domestic corporate scandals indicate that technical incompetence might not be the reason for audit failure. Rather, extended audit tenure may impair auditor independence Besides, long auditor tenure can also be detrimental to the discovery of new problems. The rotation of lead engagement partners can enhance auditor independence and bring in a fresh perspective. Therefore, the new policy may lead to improved audit quality and better protection of investors. If mandatory rotation does in fact enhance the discovery of financial reporting problems, we would expect the replacement partner to find more audit adjustments during his first year of tenure. Therefore, our second hypothesis is as follows: H2: Audit adjustments are more frequent during the replacement partner s first year of tenure following mandatory partner rotation than in years when audit partners are not affected by mandatory rotation. 14

16 On the other hand, the data may not support H2 for two reasons. First, a newly appointed partner may have less client-specific knowledge. Therefore, mandatory rotation may be bad for audit quality because the newly appointed partner would be less likely to find financial reporting problems. This would mean a lower likelihood of audit adjustments in the replacement partner s first year of tenure, which is opposite to H2. Second, our first hypothesis is that the departing partner makes more adjustments during his final year of tenure before handing over the audit to a new partner. To the extent that financial reporting problems are resolved before the handover, the new partner may not require more audit adjustments during his first year of tenure. Our institutional setting Audit reports in China disclose the names of both the review partner and the engagement partner. The two partners share the same legal liability (unless there is contrary evidence) and are subject to the same rules on mandatory rotation. 4 Under Articles 3 and 5 issued by the China Securities Regulatory Commission (CSRC) and the Ministry of Finance (October 8, 2003), the review partner and engagement partner have to be rotated every five years or, in the case of newly listed companies, at the end of the second year following the initial public offering (IPO). (The two year rule for newly listed companies reflects that the IPO prospectus contains three years of audited financial statements and so an IPO is counted as three years of continuous service.) 5 These rules are essentially the same as those in the US. In both countries, the review and engagement partners have to be rotated after five years of 4 We use the term audit partner to describe the signing auditor even though audit firms in China are allowed to organize as limited liability companies or as partnerships. 5 Similarly in the US, the SEC Office of the Chief Accountant states that the length of continuous service must include the years for which audited financial statements are included in the company s IPO filing (SEC 2004, Audit Partner and Partner Rotation section, question 3). 15

17 continuous service. 6 Similarly, the 8th European Union directive 2006/43/EC (European Commission, 2006) requires all twenty-seven member states to implement mandatory rotation of engagement and review partners for audit reports dated after June An important advantage of our Chinese setting is that the names of the engagement and review partners are publicly disclosed in audit reports. This allows us to identify all partner rotation events. In addition, the name of the review partner is disclosed in the audit report above the name of the engagement partner in China. 7 This enables us to identify rotation events affecting each type of partner, which is important because engagement partners are expected to have greater influence over audit adjustments than review partners. Research design We test H1 and H2 by estimating the following model of audit adjustments: ADJUST it = α 0 + α 1 MROT_FINAL it + α 2 MROT_FIRST it + CONTROLS + u it (1) The dependent variable (ADJUST it) takes the value one if there is an audit adjustment to company i s profits during year t; ADJUST it equals zero otherwise. The MROT_FINAL it variable equals one if the partner is in his final year of tenure in year t because he is scheduled for mandatory rotation at the end of the audit; MROT_FINAL it equals zero otherwise. Under H1, we predict that audit adjustments are more frequent during the 6 Section 203 of the Sarbanes-Oxley Act (2002) states: It shall be unlawful for a registered public accounting firm to provide audit services to an issuer if the lead (or coordinating) audit partner (having primary responsibility for the audit), or the audit partner responsible for reviewing the audit, has performed audit services for that issuer in each of the five previous fiscal years of that issuer. 7 The signatures of the two partners appear on the audit report with one signature above the other. Based on interviews with senior audit partners, we learned that the top signature is signed by the review partner and the bottom one by the engagement partner. Untabulated tests support what the partners told us. We find that the partners identified in the top signatures are significantly more experienced compared with the partners identified in the bottom signatures. Specifically, the top signing partners are significantly older (t-stat. = 29.14), have significantly more years of service since their CPA qualification (t-stat. = 36.21), and they sign audit reports for more listed clients (t-stat. = 25.50). These results are exactly what we would expect given that the top signatures are by highly experienced review partners whereas the bottom signatures are by less experienced engagement partners. 16

18 partner s final year of tenure prior to mandatory partner rotation, implying a positive coefficient on MROT_FINAL it; i.e., α 1 > 0. The MROT_FIRST it variable equals one if the partner is in his first year of tenure in year t due to mandatory rotation of the former partner at the end of year t-1; MROT_FIRST it equals zero otherwise. Under H2, audit adjustments are more frequent during the replacement partner s first year of tenure, implying a positive coefficient on MROT_FIRST it; i.e., α 2 > 0. Control variables An audit adjustment takes place under two conditions: 1) the client s pre-audit financial statements are misstated, and 2) the auditor detects the misstatement and requires the client to make a correction through an adjustment to the financial statements. Therefore, we control for factors that explain the likelihood that a misstatement exists in a client s pre-audit financial statements (condition #1) and factors that explain the likelihood that an auditor will detect the misstatement and require an adjustment (condition #2). Concerning condition #1 (the likelihood that a misstatement exists), Kinney and McDaniel (1989) find that smaller and less profitable companies are more likely to misstate their financial statements. Therefore, we control for the company s size (Size it = the natural log of sales) and profitability (ROS it = net income divided by sales). Mergers and acquisitions can lead to complicated accounting issues and business integration problems, which increase the incidence of misstatements (Kinney et al. 2004). We control for this using a dummy variable (M&A it) which takes the value one if company i acquires more than 20% of the equity of another company in year t, and zero otherwise. Large corporate groups are also likely to have more complicated accounting issues which could cause a higher probability of misstatement. We control for this using the log of (one plus) the number of consolidated subsidiaries (Subsidiaries it). Finally, we control for Leverage it. Higher leverage can motivate a company to engage in more earnings management (e.g., DeFond and 17

19 Jiambalvo 1994) and, at the same time, higher leverage can incentivize auditors not to waive any proposed audit adjustments (Braun 2001). Therefore, there may be a positive relation between leverage and non-waived adjustments. On the other hand, lenders are more willing to supply finance to companies that are known to be trustworthy (e.g., Datta et al. 1999), so the relation might instead be a negative. Concerning condition #2 (the likelihood that an auditor detects and corrects an existing misstatement), we include three controls for other determinants of audit quality. First, we control for audit firm size as there is evidence that larger audit firms supply higher quality audits (e.g., Francis 2004). Our audit firm size measure is a dummy variable, Big4 it, which equals one if company i is audited by a Big Four firm in year t, and zero otherwise. If the Big Four provide higher quality audits they would be more likely to detect any misstatements in the pre-audit financial statements. However, the association could go in the opposite direction as the clients of Big Four audit firms are likely to have higher quality pre-audit financial statements compared with the clients of non-big Four firms. In this case, we may find a negative association between Big4 it and audit adjustments. Second, Chan and Wu (2010) show that audit firm mergers result in higher audit quality in China. This is potentially important as a lot of companies in our sample are audited by a firm involved in a merger. Our auditor merger variable, AUD_M&A it, takes the value one if company i s audit firm merges with another audit firm in year t, and zero otherwise. Third, we control for the audit opinion because the financial reporting problems that lead to non-clean audit opinions may contribute to explaining the frequency of audit adjustments. The Unclean Opinion it variable equals one if company i receives a qualified or modified audit opinion in year t, and zero otherwise. Finally, we include indicator variables for each industry and each year. 18

20 III. SAMPLE AND DESCRIPTIVE STATISTICS Sample Starting in 2006, an audit firm is required to report privately to the Ministry of Finance the company s pre-audit earnings number after the audit firm signs off on the company s audited financial statements. This is required on every public company audit in China. Using the pre-audit earnings number, the Inspection Bureau of the Ministry of Finance is able to identify whether a company had an audit adjustment to its earnings number. The Inspection Bureau obtains this information from audit firms in order to target the audit engagements that it feels warrant close scrutiny in its regular inspections of audit firms. (Similarly, the PCAOB looks for signs of potential misreporting when it decides which audit engagements should be sampled in its inspections of US audit firms.) During an inspection, the Bureau selects a sample of audit engagements for examination. Any inconsistency between the adjustments reported to the Ministry and the adjustments recorded in an audit firm s working papers would result in further investigation and possible regulatory action. Further, an audit firm is not given advance notice of the working papers that will be examined by the inspectors. Consequently, it would be risky for an audit firm to report the audit adjustments to the Ministry of Finance in a strategic or dishonest way. Consistent with this, we understand from the Ministry that no audit firm has so far had to be disciplined for misreporting the pre-audit earnings of its clients. We obtained access to the audit adjustments data through one of the study s authors, who had previously worked for the Chinese Institute of Certified Public Accountants (CICPA) under the Ministry of Finance from The data were provided to us on the understanding that it would be used for academic research purposes. 9 8 This person was responsible for drafting Chinese auditing standards and other technical and policy matters relating to the Chinese public accounting profession. He has had an extensive working relationship with regulatory agencies within and outside the Ministry of Finance. After joining 19

21 We start with 8,087 company-year observations held in the Bureau s audit adjustments database for the period We require the partners names for years t and t-1 in order to identify whether the partner is new to the audit in year t (i.e., the partner is new if he did not conduct the audit in year t-1). We also require the partners names for years t and t+1 in order to identify whether the partner in year t will be rotated in the following year. This means that we require partners names for at least three consecutive years (t-1, t, and t+1). Further, for those cases in which there is a change of audit partner, we trace partners names for up to five years prior to the change in order to determine whether the change is mandatory or voluntary. We lose 651 observations as a result of these data requirements. We also lose 338 observations where data are missing for one or more variables in year t. Because our objective is to examine the consequences of a change in audit partner, holding constant the identity of the client s audit firm, we drop 621 observations where there is a change of audit firm and 136 observations where a client s financial statements are audited by two audit firms. 10 This leaves a final sample of 6,341 company-year observations. In all regressions, the standard errors are clustered on each company as there are multiple yearly observations for each company. Mandatory audit partner rotation Table 1 presents descriptive information on the mandatory rotation of engagement partners and review partners. Panel A reports the number of partners who are in their final year of academia, he has remained actively involved in auditing-related professional activities and research programs initiated by the CICPA and the Inspection Bureau, among others. 9 We are not allowed to use the audit adjustments data for any non-academic purpose and we must not disclose any information about specific audit firms or specific clients. 10 Between April 1, 2002 and March 8, 2007, dual audits are required when a company in China makes an Initial Public Offering or a Seasoned Equity Offering. Dual audits are also common among companies that offer B or H shares to foreign investors. The B and H share companies are required to prepare one set of financial statements for domestic investors under Chinese accounting standards and another set of financial statements for foreign investors under international accounting standards and these two sets of financial statements are often audited separately by two audit firms. 20

22 tenure in year t. We find that 327 engagement partners are in their final year of tenure because they are scheduled for mandatory rotation at the end of the year (MROT_FINAL it = 1). There are another 416 review partners who are in the same situation. Panel B reports the number of partners who are in their first year of tenure following mandatory rotation of the previous partner. There are 291 (319) engagement partners (review partners) who are in this situation (MROT_FIRST it = 1). [INSERT TABLE 1 HERE] Audit adjustments Table 2 presents descriptive statistics on the adjustments to reported profits. As shown in Panel A, there are: 1) 2,858 audits with net downward adjustments (45.07%), 2) 2,087 audits with no adjustments (32.91%), and 3) 1,396 audits with net upward adjustments (22.02%). This is consistent with Kinney and Martin (1994) who find that adjustments are required in 60-90% of audits. An untabulated test finds the frequency of downward adjustments to be significantly higher than the frequency of upward adjustments (i.e., 45.07% > 22.02%; p- value < 0.001). This is consistent with US research which finds that auditors are more likely to require profits to be adjusted downwards than upwards (Kinney and Martin 1994; Nelson et al. 2002). [INSERT TABLE 2 HERE] Panel B of Table 2 reports descriptive statistics on the magnitudes of audit adjustments. The adjustment magnitude is defined as the dollar amount of the adjustment scaled by the absolute dollar amount of the pre-audit profit. As shown in Panel B, the mean downward adjustment reduces reported profits by 16.2% whereas the mean upward adjustment increases reported profits by 9.8%. An untabulated test shows that the downward adjustments are significantly larger than the upward adjustments (i.e., 9.8% is smaller than 16.2%; p-value < 0.001). Therefore, not only are downward adjustments more 21

23 frequent than upward adjustments, but also the downward adjustments are larger in magnitude. Overall, it is clear that audit adjustments generally have the effect of reducing rather than increasing reported profits. 11 Univariate tests of H1 and H2 Table 3 presents univariate tests for H1 and H2. Under H1, we expect the audit adjustment frequency to be significantly higher during the partner s final year of tenure than in other years. Panel A of Table 3 shows that the univariate results for H1 are highly significant for engagement partners. The audit adjustment frequency is 77.06% during the engagement partner s final year of tenure prior to mandatory rotation, compared with 66.54% in all other years. The difference (77.06% versus 66.54%) is both large and statistically significant (pvalue < 0.001). In contrast, Panel B finds that the results for H1 are insignificant for review partners. The audit adjustment frequency is 66.83% during the review partner s final year of tenure prior to mandatory rotation, compared with 67.11% in other years. The difference (66.83% versus 67.11%) is small and statistically insignificant. [INSERT TABLE 3 HERE] Panel C of Table 3 shows that the results for H2 are highly significant for engagement partners. The adjustment frequency is 74.91% during the engagement partner s first year of tenure following mandatory rotation, compared with 66.71% in other years. Thus, audit adjustments are more frequent when the new engagement partner takes over the audit as a result of mandatory rotation. The difference (74.91% versus 66.71%) is statistically significant (p-value < 0.001). The univariate tests of H2 are also significant for review 11 Although an adjustment to reported profits may be immaterial when measured on a net basis, this does not mean that the adjustments to individual components of the income statement are immaterial. For example, a material downward adjustment to reported sales together with a material downward adjustment to the reported cost of sales can result in a small net adjustment to bottom line profits. Auditors are required to judge materiality with respect to individual components of the income statement as well as to bottom-line profits. 22

24 partners. In Panel D, the audit adjustment frequency is 71.79% during the review partner s first year of tenure following mandatory rotation, compared with 66.84% in other years. The difference (71.79% versus 66.84%) is statistically significant (p-value = 0.067) but it is smaller than for engagement partners. Overall, the univariate results in Table 3 suggest that audit quality is higher during the engagement partner s final year of tenure prior to mandatory rotation and during the engagement partner s first year of tenure following mandatory rotation. In addition, we find some (weaker) evidence that audit quality is higher during the review partner s first year of tenure following mandatory rotation. However, it is important to control for other client and auditor characteristics that potentially affect the incidence of audit adjustments. Control variables Table 4 provides descriptive statistics for the control variables. The mean value of the log of sales (Sales it) is and ranges from to The mean return on sales (ROS it) is 6.8% and the mean value of leverage (Leverage it) is 53.0%. We find that 24.3% of clients are involved in a merger or acquisition transaction during the year (M&A it). The natural log of (one plus) the total number of subsidiaries (Subsidiaries it) ranges from to Only 6.7% of companies are audited by one of the Big Four firms (Big4 it) which is consistent with the low frequency of Big Four audits reported in other China studies (e.g., Chen et al. 2011). We find that 20.0% of clients are audited by an audit firm that merges with another audit firm (AUD_M&A it), while 4.8% of companies receive qualified or modified audit opinions (Unclean Opinion it). [INSERT TABLE 4 HERE] 23