Rotate back or not after mandatory audit partner rotation? Michael A. Firth a, ⇑ , Oliver M. Rui b ,XiWu c a Department of Finance and Insurance, Lingnan University, Hong Kong b China Europe International Business School (CEIBS), Shanghai, China c School of Accountancy, Central University of Finance and Economics, China abstract Many countries have implemented rules that require an audit part- ner to rotate off the audit of a specific client after a certain period of time in the belief that rotation will improve independence and will allow for a fresh look at the audit. The rules are either silent on whether or when a partner can rotate back or else they specify a cooling-off period after which the rotated-off partner can resume the audit. Using archival data from China, a country with a 2-year cooling-off period, this paper explores the determinants of whether the audit partner rotates back or not when the cooling- off period expires, and whether audit quality is weakened by the audit partner rotation-back practice. We find that the audit partner rotation-back practice can be explained by factors relating to switching costs, agency conflicts, client desirability, and the audit partner’s capacity constraint considerations. Interestingly, we find that clients suffering greater audit adjustments immediately prior to the expiration of the cooling-off period are more likely to be associated with subsequent audit partner rotation-back. Further- more, we find that rotation-back partners tend to treat former cli- ents more favorably than non-rotation-back cases using modified audit opinions as our proxy for audit quality. Overall, our findings offer preliminary explanations for and shed light on the conse- quences of rotation-back practice arising from mandatory audit partner rotation requirements and lend support to regulatory con- cerns on rotation-back practice among audit partners. Ó 2012 Elsevier Inc. All rights reserved. 0278-4254/$ - see front matter Ó 2012 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jaccpubpol.2012.05.002 ⇑ Corresponding author. E-mail addresses: mafirth@ln.edu.hk (M.A. Firth), oliver@ceibs.edu (O.M. Rui), wuxi@cufe.edu.cn (X. Wu). J. Account. Public Policy 31 (2012) 356–373 Contents lists available at SciVerse ScienceDirect J. Account. Public Policy journal homepage: www.elsevier.com/locate/jaccpubpol 1. Introduction Mandatory audit partner rotation is now required in many jurisdictions. 1 Rotation is seen as a po- tential means of enhancing auditor independence and audit quality by reducing partner–client familiar- ity and bringing in fresh perspectives. 2 However, the benefits of rotation could be lost if the previously rotated-off audit partner rotates back to the client. In 2003, the U.S. Securities and Exchange Commission (SEC, 2003) expressed its concern about audit partner rotation-back practice in proposing a rule designed to strengthen auditor independence requirements: While the [Sarbanes–Oxley] Act specified that these two partners [the lead and concurring part- ners] were subject to rotation after five years, the Act is silent with regard to the time out period. One approach to the partner rotation rules could have been to preclude the partner from returning to the audit client after he or she rotates off to that engagement. The Commission is adopting rules to require the lead and concurring partners to rotate after five years and, upon rotation, be subject to a five-year ‘‘time out’’ period. Because of the importance of achieving a fresh look to the independence of the audit function, we believe that a five-year time out period is appropriate for these two partners. Prior to the implementation of the SEC, 2003 rule, the American Institute of Certified Public Accountants (AICPA) adopted a 2-year cooling-off professional practice standard (SEC, 2003, footnote 123). After debating various approaches including the permanent barring of audit partners from rotat- ing back to a former client, the SEC settled on a 5-year cooling-off period for lead and concurring audit partners, effective from May 6, 2003. Many professional entities including the AICPA argued for a shorter cooling-off period (SEC, 2003, footnote 126). To strike a balance, the SEC (2003) requires that partners subject to rotation requirements other than the lead and concurring partner rotate after no more than 7 years and be subject to a 2-year time-out. Note that the names of the lead and concurring audit partners are not disclosed. The cooling-off period is shorter than 5 years in some other jurisdic- tions where the mandatory audit partner rotation requirement is in place. For example, a 2-year cool- ing-off period is required in both Australia (Hamilton et al., 2005, p. 3, footnote 4) and China (China Securities Regulatory Commission (CSRC) 2003). Motivated by the regulatory concern about audit partner rotation-back, this study capitalizes on the natural laboratory setting of China, a country where audit partner rotation and rotation back rules apply, to explore what determines whether the audit partner rotates back when the cooling-off period expires and whether audit quality is compromised when the rotated-off partner rotates back. Focusing our investigation on China, a jurisdiction which requires that two certified public accountants sign the published audit report, allows us to identify audit partners and their rotation status for a given client. To examine the determinants of audit partner rotation-back, we test four hypotheses, switching costs, agency conflicts, client desirability, and the audit partner’s capacity constraint, that have been proposed in the literature on auditor selection and the auditor–client relationship. Utilizing data from the mandatory audit partner rotation regime in China, we find that all of these four hypotheses can help explain the partner rotation-back practice. Specifically, audit partners are more likely to rotate back if there is a greater degree of partner–client familiarity and are less likely to rotate back if the client is larger, more complex, in the high-risk category, or if the client offers a limited tenure for future audit work. More interestingly, we find that if a client suffers greater audit adjustments to pre-audit earnings immediately prior to the expiration of the cooling-off period, the previously rotated-off partner is more likely to rotate back, suggesting an opportunistic switching cost consider- ation by the client. To investigate the consequences of audit partner rotation-back, we follow prior China-related studies (e.g., DeFond et al., 2000; Chen et al., 2010; Chan and Wu, 2011) and use the 1 Based on an international survey by the GAO (2003, Appendix V), these jurisdictions include France, Germany, Japan, Singapore, Spain, the United Kingdom, and the United States. Australia, the Chinese mainland, and Taiwan also require audit partner rotation. 2 There is a growing literature that examines the efficacy of audit partner rotation either from the partner-tenure perspective (Chi and Huang 2005; Carey and Simnett 2006; Chen et al. 2008; Manry et al. 2008; Fargher et al. 2008) or from the perspective of the immediate (i.e., the first-year) effect of mandatory partner rotation (Hamilton et al. 2005; Chi et al. 2009). M.A. Firth et al. /J. Account. Public Policy 31 (2012) 356–373 357 auditor’s propensity to issue modified opinions as a proxy for audit quality. We find that rotation-back partners tend to treat former clients more favorably than non-rotation-back cases in the first post- cooling-off year. This study makes a number of contributions to the auditing literature. First, we are among the first to offer explanations for the audit partner rotation-back/non-rotation-back practice. Recent literature (e.g., Chi et al., 2009) has focused on the immediate effect of mandatory audit partner rotation. We extend the investigation of the rotation policy to a longer window (i.e., when the mandatory cool- ing-off period expires). Second, this study enriches our understanding of the auditor–client relation- ship and (individual-level) auditor selection. Recent studies (Blouin et al., 2007; Chen et al., 2009) have explored why some clients maintain the relationship with former audit partners in a forced audit firm change setting. We add to this line of literature by utilizing another specific setting (i.e., partner rotation-back vs. non-rotation-back), which allows us to extend our understanding of why some cli- ents voluntarily resume the relationships with former audit partners after a mandatory cooling-off period. Third, our findings shed light on the consequences of audit partner rotation-back, and lend support to regulatory concerns about the rotation-back practice among audit partners. Our findings should have implications for the policy debate on the appropriate duration of the cooling-off period. The rest of this paper is structured as follows. Section 2 presents theoretical discussions and hypotheses development of the determinants and the effects of audit partner rotation-back on audit quality. Section 3 describes the institutional background in China and sample selection. Section 4 pre- sents the research design and results of a determinant model on audit partner rotation-back. Section 5 examines the audit quality associated with the rotation-back practice and Section 6 concludes the paper. 2. Hypotheses development Given a specified (mandatory) cooling-off period, why do some audit partners rotate back when the cooling-off period expires? The rotation-back practice is a voluntary manifestation of auditor–client relationships and (individual-level) auditor selection. Such a practice involves three parties: the client, the audit partner, and the audit firm. We base our theoretical analysis on the auditor selection and auditor–client relationship literature. We analyze the issue from both the demand and supply sides. The auditor–client relationship literature suggests three potential costs incurred by the client in auditor selection, i.e., switching costs, agency costs, and implicit insurance costs 3 (Blouin et al., 2007). The switching costs can be defined as start-up costs incurred by the client for a new audit engage- ment (Blouin et al., 2007, p.624), which commonly include costs to educate the auditor about the com- pany’s environment, business, and financial reporting issues. 4 An audit partner rotation invariably leads to switching costs for the client as they have to educate the newly assigned partners. If the rotated off partner retakes charge of the audit after the cooling off period expires, additional switching costs will be incurred. The magnitude of these additional switching costs depends on several factors. If the rotated off partner had developed an extensive knowledge of the client prior to rotation then the switching back costs may be quite low if the partner retains this knowledge and experience. The cost of switching back to the former audit partner has to be compared to any lingering switching costs of the new auditors (i.e., if the new auditors take more than 2 years to gain a thorough understanding of the client). The greater the switching costs the client is likely to incur for its annual audit, the less likely the client is willing to accept a previously rotated-off partner to rotate back. If the rotated-off partner is more experienced and 3 The implicit insurance costs hypothesis suggests that various audit firms provide different values of implicit insurance (Menon and Williams 1994), and audit firm changes may incur costs when switching from a larger audit firm to a smaller firm. However, in our within-firm audit partner analysis, the audit partner rotation itself does not change the implicit insurance the audit firm provides to the client. Therefore, insurance costs are not considered in our subsequent analysis. 4 As elaborated in Blouin et al. (2007), switching costs also include costs incurred by the client in selecting a new auditor, and an increased risk of audit failure. The former costs are more applicable to the case of audit firm switching, but less applicable to the within-firm partner rotation. We argue that the latter costs can arise because of the different knowledge bases of the new and the former audit partner. 358 M.A. Firth et al. / J. Account. Public Policy 31 (2012) 356–373 familiar with the client than the newly introduced auditors, he/she may rotate back if this will save money for the client. Based on these cost considerations, our first hypothesis is as follows: H1. The greater the switching cost, the less likely a previously rotated-off partner is to rotate back when the mandatory cooling-off period expires, ceteris paribus. Agency costs are another major type of cost consideration in auditor selection. Jensen and Meckling (1976) defined agency costs as ‘‘monitoring expenditures by the principal, bonding expenditures by the agent, and loss in welfare experienced by the principal due to the agent not acting in the princi- pal’s best interest’’. A high-quality auditor can reduce agency costs through the competent and inde- pendent monitoring of the agent. Mandatory audit partner rotation has been widely implemented to mitigate agency costs and to improve audit quality (GAO, 2003), whereas the rotation-back practice is deemed by some regulators as compromising auditor independence (SEC, 2003). Therefore, a previ- ously rotated-off partner is likely to be associated with lower perceived audit quality when he/she ro- tates back. If agency costs are a major type of cost consideration in the client’s selection of individual auditors, the client is less likely to accept a rotation-back practice, other things being equal. Jennings et al. (2006) and Kaplan and Mauldin (2008) suggest that the strength of corporate governance plays a role in information users’ perceptions of the effectiveness of mandatory auditor rotation in their experimental studies. Strong corporate governance may place constraints on the close ties between client management and the audit partner, whereas weak corporate governance would do little to cur- tail the negative aspects of auditor–client familiarity. Therefore, we develop our second set of hypoth- eses as follows: H2a. The greater the agency conflict, the less likely a previously rotated-off partner is to rotate back when the mandatory cooling-off period expires, ceteris paribus. H2b. The stronger the corporate governance, the less likely a previously rotated-off partner is to rotate back when the mandatory cooling-off period expires, ceteris paribus. Besides the client (demand-side) considerations in the process of auditor selection, there are also supply-side factors considered by the auditor. Prior literature on auditor-initiated audit firm changes (e.g., Shu, 2000) and on audit firm portfolio management (e.g., Johnstone and Bedard, 2004) suggest that auditors normally avoid audit clients that fall into the high-risk category (see Bedard et al., 2008 for a review). Similarly, when the rotated-off partner considers whether to rotate back after the cooling-off period expires he/she is less likely to do so if the client is deemed as a highly risky one during the cooling-off period. Therefore, we propose our third hypothesis as follows: H3. The more risky the client during the cooling-off period, the less likely a previously rotated-off partner is to rotate back when the mandatory cooling-off period expires, ceteris paribus. Audit firms could also play a role in the audit partner rotation-back practice. As cited in SEC (2003), some respondents expressed a concern about the lack of audit partners, relative to the audit firm’s business capacity, to implement the rotation policy. This concern suggests that an auditor’s capacity constraint can be a deterrent to the rotation-back decision. If an audit firm has fewer qualified prac- titioners relative to the firm’s service jobs, each auditor, including the previously rotated-off auditor, will be fully occupied by the assigned tasks and will have less capacity to rotate back. 5 Therefore, we state our fourth hypothesis as follows: H4. The more business revenues each of an audit firm’s practitioners earns, the less likely a previously rotated-off partner is to rotate back when the mandatory cooling-off period expires, ceteris paribus. 5 Besides the auditor capacity hypothesis, an alternative explanation is related to client desirability from the standpoint of an individual audit partner. That is, when an audit firm provides more (less) business services to its partners, each partner is on average economically less (more) dependent on a particular client, thus leading the rotated-off partner to be less (more) inclined to rotate back. M.A. Firth et al. /J. Account. Public Policy 31 (2012) 356–373 359 Underlying the debate on whether and when the previously rotated-off partners are allowed to ro- tate back (SEC, 2003), the regulator is concerned about an impaired audit quality associated with the audit partner rotation-back. On the one hand, auditors may have incentives to maintain their own rep- utation and audit quality (e.g., Reynolds and Francis, 2001). On the other hand, recent audit failures in Enron and WorldCom suggest that auditors still are subject to client pressures which lead to compro- mised audit quality. It is therefore an empirical question as to whether the audit partner rotation-back is associated with lower audit quality when compared with the non-rotation-back practice. We thus develop our fifth hypothesis in null form as follows: H5. The audit partner rotation-back is associated with the same audit quality as compared to that for the non-rotation-back practice, ceteris paribus. 3. Chinese institutional background and sample selection To enhance auditor independence and bring in a fresh perspective, the CSRC and the Ministry of Finance (MOF) of China jointly issued a mandatory audit partner rotation policy on October 8, 2003 which took effect from the 2003 annual audit. Specifically, the CSRC (2003) prohibits a signing auditor from providing audit services for the same listed company for more than five consecutive years and bars the signing auditor from resuming audit services for the entity concerned for at least 2 years after he or she rotates off. 6 Audit reports in China record both the name of the audit firm and the names of the signing audi- tors. 7 This allows us to use the financial auditing history of each listed Chinese company to identify whether a signing auditor has served for five consecutive years or more. For companies that have been listed for less than 5 years, we also check the names of the IPO signing auditors because regulations spec- ify that the tenure for an IPO audit is three consecutive years. 8 Our sample period begins in 2003, which is the earliest year in which the regulations were in force. The MOF made systematic changes to China’s accounting and auditing standards in 2006, and 39 (48) newly enacted or revised accounting (auditing) standards became effective in 2007. Since these two sets of standards are fundamental for the auditors to make accounting judgments and conduct audit procedures, we end our sample period in 2006 to avoid the possible confounding effect of different standard regimes on the auditing process and outcomes. Pre-MR period Mandatory Partner Rotation (MR) period Post-MR period (Partner rotating-back or continuous cooling-off) PreMR(…) PreMR(-1) MR(1) MR(2) PostMR(1) PostMR(…) Fig. 1. The timeline of mandatory partner rotation in China. 6 There is also one exception whereby if both of the signing auditors have provided audit services for the same entity for the same period of five consecutive years, one of them is allowed to extend his or her term as a signing auditor of the entity concerned by a maximum of 1 year. 7 According to the relevant Chinese regulations (e.g., The Ministry of Finance’s Notification on Issues Regarding Certified Public Accountants’ Signing and Stamping on the Audit Report, effective from July 2, 2001), and in normal situations, one of the two signing auditors must serve as the lead auditor in charge of field work and the other must be at least a deputy executive of the CPA firm and serve as the reviewer of the engagement. These two signing auditors are required to assume the same legal liabilities as each other (unless one can prove the contrary). We use the term ‘‘audit partner’’ to describe the signing auditor even if the audit firm is not a partnership (e.g., if it is a limited liability company). 8 The CSRC (2003) requires that for companies that have recently been listed or are awaiting a listing, signing auditors must not provide audit services for the same initial public offering (IPO) entity for more than two consecutive years after the IPO. This 2-year period follows on from the 3 years for which audited financial statements must be included in the IPO applicant’s filing, thus giving partner tenure of 5 years. This requirement is consistent with the (US) SEC Office of the Chief Accountant’s interpretation that the signing auditor’s term under rotation requirements includes the number of years for which audited financial statements are included in the IPO filing (SEC 2004, ‘‘Audit Partner and Partner Rotation’’ section, question 3). 360 M.A. Firth et al. /J. Account. Public Policy 31 (2012) 356–373 To help illustrate our sample construction, we show in Fig. 1 the mandatory rotation timeline, which is partitioned into the pre-rotation period (PreMR), the mandatory partner rotation period (MR), and the post-mandatory-rotation period (PostMR). For partners required to rotate in 2003, the first mandatory rotation or cooling-off year (MR(1)) is 2003, the second cooling-off year (MR(2)) is 2004, and the first post-cooling-off year (PostMR(1)) is 2005. In a similar fashion, partners who are required to rotate in 2004 have a cooling-off period of 2004 and 2005 and the first post-cool- ing-off period is 2006. In the PostMR period, the former audit partner may resume the audit (i.e., part- ner rotation-back) or may not (i.e., continuous cooling-off). Table 1 presents the sample selection process (Panel A) and composition of rotation-back status (Panel B). We first identify 382 A-share listed Chinese companies whose audit partner was subject to mandatory rotation in 2003 or 2004. We eliminate 61 observations where both audit partners should have rotated off in MR(1) but one of them used the exemption article to remain on the audit for one additional year. We then exclude 22 observations where the audit partners should have ro- tated off for 2 years but rotated back in MR(2) (i.e., 1 year earlier), as they represent suspected cases of superficial partner rotation (Bamber and Bamber, 2009). 9 Another five observations are excluded be- cause the other audit partner should have rotated off in MR(2) or PostMR(1) due to meeting the 5-year rotation threshold but did not do so. 44 companies changed their audit firms in MR(2) or PostMR(1), and therefore are excluded from the sample. Two companies are further excluded because they were de- listed in PostMR(1). Our final sample consists of 248 companies with an observable status of partner rotation-back or non-rotation-back in the first post-cooling-off year. Panel B shows that for 115 (46.4%) out of 248 companies, the previously rotated-off audit partners rotate back when the 2-year cool- ing-off period expires (RB clients hereafter), whereas for the other 133 companies (53.6%), the previously rotated-off audit partners do not rotate back (NRB clients). Untabulated statistics show that RB clients and NRB clients have similar industry profiles. The frequency in each industry is representative of the overall distribution. Table 1 Sample selection and composition. MR(1) = 2003 MR(1) = 2004 Total N Panel A: Sample selection Companies that underwent mandatory partner rotation in MR(1) 162 220 382 Less Both audit partners should have rotated off in MR(1) but one of them used the exemption article to remain 17 44 61 Audit partners should have rotated off for 2 years but rotated back in MR(2) 9 13 22 The other partner should have rotated off in MR(2) but did not 2 1 3 The other partner should have rotated off in PostMR(1) but did not 0 2 2 Audit firm changed in MR(2) 7 17 24 Audit firm changed in PostMR(1) 7 13 20 The client was delisted in PostMR(1) 1 1 2 Final sample 119 129 248 Rotation-back Continuous cooling-off Total N Number Percent Number Percent Number Percent Panel B: Sample composition based on rotation-back status PostMR(1) 115 46.4 133 53.6 248 100 MR(1) = the first mandatory rotation (cooling-off) year MR(2) = the second mandatory rotation (cooling-off) year PostMR(1) = the first post-mandatory rotation year 9 These companies clearly violated the specified cooling-off requirement and we treat 1-year rotations off as cases of superficial rotation. M.A. Firth et al. /J. Account. Public Policy 31 (2012) 356–373 361 4. Determinants of audit partner rotation back: design and results 4.1. Empirical design To test the four hypotheses (switching costs, agency conflicts, client desirability, and audit partner’s capacity consideration) relating to the possible determinants of audit partner rotation-back vs. non- rotation-back practice, we construct the following probit model: probitðRotBack ¼ 1Þ¼b 0 þ b 1 LTA þ b 2 SqSubs þ b 3 IPOPtn þ b 4 PtnTenr þ b 5 jAudAdjj MR2 þ b 6 StaOwn þ b 7 BODSize þ b 8 IndDir þ b 9 HighRisk þ b 10 Re v nPerCPA þ Yr2006 þ IndDum þ e ð1Þ We set the dependent variable, RotBack, as a dummy variable coded 1 if an observation belongs to the group of clients whose audit partners rotated back when the 2-year cooling-off period expired. As shown in Table 2, we identify a number of experimental variables to test our hypotheses. Simunic (1980) suggests that client size and complexity have the most significant effects on the cost of an audit. Moreover, client size and complexity are also used as proxies for agency concerns, as larger companies are subject to greater reputation costs and more complex companies are more Table 2 Variables used in audit partner rotation-back determinant model: hypotheses and sign predictions. Dep. Var. = RotBack Hypotheses Switching costs Agency conflicts Client desirability Capacity consideration Experimental variables LTA ÀÀ SqSubs ÀÀ IPOPtn + À PtnTenr + À |AudAdj| MR2 + À StaOwn ÀÀ BODSize + IndDir À HighRisk À RevnPerCPA ÀÀ Control variables Yr2006 ?? ? ? IndDum ?? ? ? Variable definitions: RotBack: a dummy variable coded 1 for a rotation-back client and 0 otherwise; LTA: the natural logarithm of total assets (in 10 thousand RMB Yuan); SqSubs: the square root of the number of consolidated subsidiaries; IPOPtn: a dummy variable coded 1 if the rotated-off audit partner served as the initial public offering (IPO) auditor for the client and 0 otherwise; PtnTenr: the number of years for which the audit partner has served as the signing auditor for the client when he or she has to rotate off in line with the mandatory requirement; |AudAdj| MR2 : the absolute value of AudAdj (audit adjustment to pre-audit earnings) made in the second mandatory rotation (cooling-off) year, where AudAdj is equal to (post-audit earnings À pre-audit earnings)/total assets; StaOwn: a dummy variable coded 1 if the client is ultimately controlled by the government or a state-owned enterprise and 0 otherwise; BODSize: the natural log of the number of directors on the board; IndDir: the proportion of independent directors on the board; HighRisk: a dummy variable coded 1 if the client suffers a loss, receives a modified audit opinion, or is subject to a regulatory sanction in any of the 3 years prior to the first post-cooling-off year and 0 otherwise; RevnPerCPA: the natural log (total business revenues of the audit firm divided by the number of certified public accountants of the firm); Yr2006: a dummy variable coded 1 for a 2006 observation and 0 for a 2005 observation; IndDum: industry dummy variables based on the CSRC classification. 362 M.A. Firth et al. / J. Account. Public Policy 31 (2012) 356–373 difficult to monitor (Blouin et al., 2007). Both the switching cost and agency cost considerations pre- dict that client size and complexity are negatively associated with the likelihood of audit partner rota- tion-back. We use LTA, defined as the natural logarithm of total assets, as a proxy for client size and SqSubs, denoted as the square root of the number of consolidated subsidiaries, as a proxy for client complexity. 10 We expect the coefficients on LTA and SqSubs to be negative. We also use IPOPtn and PtnTenr as proxies for switching costs. IPOPtn is coded 1 if the rotated-off audit partner is the IPO signing auditor and 0 otherwise. PtnTenr is the audit partner tenure since the IPO, i.e., the number of years for which the audit partner has served as the signing auditor for the client when he or she has to rotate off in line with the mandatory requirement. Both variables capture the rotated-off partner’s client-specific experience. Therefore, if the rotated-off partner has served the cli- ent as the IPO signing auditor, or has audited the client for a longer time, the switching costs associ- ated with rotation-back should be lower and we expect a positive sign for both IPOPtn and PtnTenr.On the other hand, IPOPtn and PtnTenr also measure a close partner–client relationship. An IPO audit is associated with the long process prior to a successful IPO in which the audit partner and the client work closely together to float the company on the stock exchange. The financial packaging that is com- mon during Chinese IPOs (Aharony et al., 2000) also suggests that the engagement audit partners may well be familiar with and acquiescent to the client’s earnings management or accounting preferences. As to the post-IPO partner tenure, for the first year of rotation (2003), the affected partners had served for at least 3 years (which may include the 3 years of the IPO period), although some had served for up to 12 years. As familiarity can be a threat to the independence of the previously rotated-off partner, 11 there is a greater demand for a high-quality audit. Therefore, from the concerns about agency conflicts, IPOPtn and PtnTenr are expected to be negatively associated with audit partner rotation-back. Our final measure of switching costs is |AudAdj| MR2 , which is the absolute value of audit adjustments to pre-audit earnings made in the second mandatory rotation (cooling-off) year. AudAdj is defined as (post-audit earnings À pre-audit earnings)/total assets). DeFond and Subramanyam (1998) suggests that clients change auditors due to overly conservative auditing treatment. Following this logic, if the newly assigned auditors treat the client in a conservative manner, the client is more inclined to have the previously rotated-off partner back when the cooling-off period expires. Therefore, we expect the conservatism of auditing treatment in the second cooling-off year is positively associated with audit partner rotation-back. Audit adjustments to pre-audit earnings represent how strict auditors are in curbing clients’ earnings management or misstatements attempts (e.g., Wright and Wright, 1997; Nelson, 2009, p.16) and can thus be used as a direct measure of auditor conservatism. Specifi- cally, pre-audit earnings represent the manager’s reporting preference, whereas the absolute value of audit adjustments (i.e., the divergence between pre- and post-audit earnings) can be viewed as the amount of misstatement corrected by the auditor. The larger the absolute value of audit adjust- ments, the higher the level of the auditor’s effort and independence in terms of withstanding the pres- sure from the manager and making corrections to the manager’s preferred earnings, which include both income-increasing and income-decreasing misstatements. We obtain proprietary data on audit adjust- ments from the Chinese Institute of Certified Public Accountants (CICPA). The CICPA has required audit firms to file pre-audit and post-audit earnings for publicly listed companies since 2001. 12 The coeffi- cient on |AudAdj| MR2 is expected to be positive from the switching costs perspective. However, a greater magnitude of absolute audit adjustments also suggests a higher level of agency conflicts, which may lead to a greater demand for monitoring and a lower likelihood of audit partner rotation-back. Thus the coef- ficient on |AudAdj| MR2 is expected to be negative based on agency cost arguments. We use StaOwn as a proxy for agency conflict. It is coded 1 if the client is controlled by a state- owned entity including a government agency and 0 otherwise. Prior China-related studies (e.g., Wang et al., 2008) suggest agency conflicts between the State shareholders and minority shareholders. Such an agency concern increases the demand for high-quality monitoring and decreases the likelihood of 10 We do not use the number of segments in our research design due to the lack of consistent disclosure practices among most companies in our sample period. A formal Chinese accounting standard on segment reporting took effect from January 1, 2007. 11 The SEC (2003) is concerned that if a lengthy partner-client relationship has been established before the mandatory partner rotation, the rotated-off audit partner may be more likely to rotate back (and the client may request a rotation back). 12 We only have access to the CICPA audit adjustment data for the 2001–2005 annual audits. M.A. Firth et al. /J. Account. Public Policy 31 (2012) 356–373 363 an audit partner rotation-back. The unique institutional background in China also makes the StaOwn variable a useful proxy for client desirability. The government sometimes forces listed companies con- trolled by the state to change auditor to be consistent with the non-listed parent company’s auditor choice (SASAC, 2004). 13 Thus, audit firm tenure (and thus partner tenure) quite often is less assured Table 3 Descriptive statistics of main variables used in audit partner rotation-back determinant model. Continuous variables RB group (RotBack =1) (n = 115) NRB group (RotBack =0) (n = 133) Test of differences Mean (Median) Mean (Median) t-statistic (Mann–Whitney Z) LTA 12.118 12.455 À2.853 *** (12.053) (12.402) (À2.582 *** ) SqSubs 2.588 3.044 À2.144 ** (2.449) (2.828) (À1.734 * ) PtnTenr 5.191 5.380 À0.631 (5) (5) (À0.204) |AudAdj| MR2 # 0.0117 0.0068 2.270 ** (0.0044) (0.0026) (1.998 ** ) BODSize 2.251 2.217 1.270 (2.197) (2.197) (1.516) IndDir 0.332 0.345 À1.689 * (0.333) (0.333) (À1.383) RevnPerCPA ## 3.798 3.961 À2.613 *** (3.752) (3.797) (À1.909 * ) Categorical variables Number (Percent) Number (Percent) Pearson Chi-square IPOPtn 67 65 2.184 (58.3) (48.9) StaOwn 67 101 8.821 *** (58.3) (75.9) HighRisk 25 39 1.853 (21.7) (29.3) Variable definitions: RotBack: a dummy variable coded 1 for a rotation-back client and 0 otherwise; LTA: the natural logarithm of total assets (in 10 thousand RMB Yuan); SqSubs: the square root of the number of consolidated subsidiaries; PtnTenr: the number of years for which the audit partner has served as the signing auditor for the client when he or she has to rotate off in line with the mandatory requirement; |AudAdj| MR2 : the absolute value of AudAdj (audit adjustment to pre-audit earnings) made in the second mandatory rotation (cooling-off) year, where AudAdj is equal to (post-audit earnings À pre-audit earnings)/total assets; BODSize: the natural log of the number of directors on the board; IndDir: the proportion of independent directors on the board; RevnPerCPA: the natural log (total business revenues of the audit firm divided by the number of certified public accountants of the firm); IPOPtn: a dummy variable coded 1 if the rotated-off audit partner served as the initial public offering (IPO) auditor for the client and 0 otherwise; StaOwn: a dummy variable coded 1 if the client is ultimately controlled by the government or a state-owned enterprise and 0 otherwise; HighRisk: a dummy variable coded 1 if the client suffers a loss, receives a modified audit opinion, or is subject to a regulatory sanction in any of the 3 years prior to the first post-cooling-off year and 0 otherwise. * Significance at the 0.10 level (two-tailed). ** Significance at the 0.05 level (two-tailed). *** Significance at the 0.01 level (two-tailed). # The number of observations in RB (NRB) group with available values for |AudAdj| MR2 is 111 (124). ## The number of observations in RB (NRB) group with available values for RevnPerCPA is 111 (127). 13 In 2004, the State-owned Assets Supervision and Administration Commission (SASAC) mandated that a list of major Chinese state-owned enterprises (SOEs) rotate their audit firms with new firms. This policy forced a few listed companies (as subsidiaries of their parent SOEs) to rotate their audit firms. In 2005, the SASAC further required that each SOE under its jurisdiction rotate its audit firm after serving as auditor for five consecutive years. 364 M.A. Firth et al. / J. Account. Public Policy 31 (2012) 356–373 for these companies. This may make rotated-off partners less willing to rotate back as they face the uncertainty of losing the audit again if there is a switch of audit firm due to state ownership. 14 Therefore, both agency conflict and client desirability considerations suggest a negative sign for StaOwn. As discussed earlier, audit partner rotation back could arguably be perceived as one form of very close auditor–client ties. Companies with stronger corporate governance may thus be less willing to accept a switch back to the rotated-off partner. We use BODSize and IndDir as proxies for the corporate governance of a company. BODSize is defined as the natural log of the number of directors on the board, while IndDir is denoted as the proportion of independent directors on the board. Yermack (1996) and Vafeas (2000) suggest that larger boards tend to be less effective than smaller boards. This is possibly due to larger boards being fragmented and less likely to reach agreements among board members. Since 2003, the Chinese regulators have formally required companies to have independent non-executive directors, whose primary aim is to serve the inter- ests of the minority shareholders (Firth et al., 2007). We expect a positive (negative) sign for BOD- Size (IndDir). We use HighRisk as a proxy for client desirability. If a client falls into the high-risk category during the cooling-off period, the rotated-off audit partner is less motivated to rotate back. We classify the client as highly risky (HighRisk = 1) if the client suffers a loss, receives a modified audit opinion, or is subject to a regulatory sanction in any of the 3 years prior to the first post-cooling-off year. We ex- pect the coefficient on HighRisk to be negative. We use RevnPerCPA to capture the potential impact of audit firm characteristics on the capacity of the rotated-off partner to rotate back. 15 It is defined as the natural log (total business revenues of the audit firm divided by the number of certified public accountants of the firm). The more service revenues the audit firm can provide for each of its auditors to earn in the first post-cooling-off year, the more likely the workload capacity of an individual auditor (including the rotated-off auditor) will be saturated. Moreover, as the audit firm has provided each auditor with more economic incentives from other sources, an individual auditor is less incentivized to rotate back. Therefore, from both the view of auditor capacity constraint and that of client desirability, we expect RevnPerCPA to be negatively associated with audit partner rotation-back. Finally, we include Yr2006 and IndDum to control for year and industry fixed effects. Yr2006 is coded 1 for a 2006 observation and 0 for a 2005 observation, while IndDum is industry dummy vari- ables based on the CSRC first-digit industry classification (using the manufacturing industry as the base group). 4.2. Descriptive statistics and univariate analysis Table 3 provides descriptive statistics for the main variables used in model (1) for the RB and NRB clients, respectively. Among the RB (NRB) group, the rotated-off partners serve as the IPO signing audi- tor for 58.3% (48.9%) of the companies. 75.9% of the companies within the NRB group are controlled by the State, whereas a significantly smaller percent (58.3%) within the RB group are State-controlled. 29.3% (21.7%) of companies within the NRB (RB) group fall into the high-risk category. The mean |Au- dAdj| MR2 is 0.0117 for the RB group, which is significantly greater than the magnitude (i.e., 0.0068) for the NRB group (p < 0.05). 16 This univariate result is consistent with the client trying to avoid stricter auditing treatments, which we treat as one type of switching cost, but not with the agency concern argu- ment. The mean RevnPerCPA is 3.798 for the RB group, which is significantly smaller (p < 0.01) than the level for the NRB group (3.961). This result is consistent with the audit partner’s capacity and client desirability considerations. Our univariate results also show that NRB clients are significantly larger 14 Feedback from our interviews with CPA firms corroborates that state-controlled companies are more likely to switch audit firms for regulatory reasons. 15 Since 2002, the CICPA began to publicize the total service revenues and the total number of certified public accountants for each of the Chinese top-100 audit firms based on total service revenues. 16 The mean raw value of audit adjustments (AudAdj MR2 )isÀ0.0071 (À0.0056) for the RB (NRB) clients. This suggests that auditors on average make downward audit adjustments to pre-audit earnings in the second cooling-off year, which is consistent with the notion that listed clients generally attempt to overstate earnings. M.A. Firth et al. /J. Account. Public Policy 31 (2012) 356–373 365 [...]... and an audit partner s working capacity) or an auditor’s risk-averse behavior (e.g., client risk profile) However, we also find that audit partners are more likely to rotate back if there is a greater degree of partner- client familiarity or if a client suffers greater audit adjustments to pre -audit earnings These opportunistic incentives associated with audit partner rotation -back, which are also consistent... and audit partner s capacity constraint) that likely explain the audit partner rotation -back or non -rotation -back practice when a 2-year coolingoff period expires We find that all of these four hypotheses can help explain the partner rotation -back practice Specifically, audit partners are more likely to rotate back if there is a greater degree of partner client familiarity and are less likely to rotate. .. number of years for which the audit partner has served as the signing auditor for the client when he or she has to rotate off in line with the mandatory requirement; |AudAdj|MR2: the absolute value of AudAdj (audit adjustment to pre -audit earnings) made in the second mandatory rotation (cooling-off) year, where AudAdj is equal to (post -audit earnings À pre -audit earnings)/total assets; StaOwn: a dummy... prior studies (Blouin et al. , 2007; Chen et al. , 2009) that find that clients with greater magnitudes of earnings management are more likely to follow their former audit partners, should concern the regulator 5 Audit quality associated with audit partner rotation -back 5.1 Research design In this section we further examine whether there is a compromise in audit quality when the rotated-off partner rotates... Carey and Simnett, 2006; Blouin et al. , 2007; Chen et al. , 2008; Chi et al. , 2009) More specifically, we contribute to the auditor tenure and mandatory rotation debate and related literature by employing a wider lens to examine audit partner rotation arrangements within audit firms and showing that partner rotation -back arrangements reinforce the regulatory concern about close partner client relationships... China Auditing: A Journal of Practice & Theory 20 (2), 9–30 Chen, C.J.P., Su, X., Wu, X., 2009 Forced audit firm change, continued partner- client relationship, and financial reporting quality Auditing: A Journal of Practice & Theory 28 (2), 227–246 Chen, C.-Y., Lin, C.-J., Lin, Y.-C., 2008 Audit partner tenure, audit firm tenure, and discretionary accruals: does long auditor tenure impair earnings quality?... companies are more likely to receive modified audit opinions as compared to the year prior to the mandatory rotation, suggesting an immediate positive effect of mandatory partner rotation on audit quality in China Mainland This finding is not consistent with the results in Chi et al (2009) who fail to find a significant improvement in audit quality after mandatory partner rotation in Taiwan However, we note that... with this finding, we also document that partners who rotate back to their former clients treat them more favorably than non-rotating -back cases using modified audit opinions as the proxy for audit quality Overall, our evidence is consistent with the regulatory concerns about the incentives and consequences of audit partner rotation -back Our findings have implications for regulators in China and other... jurisdictions However, regulators have expressed concerns about the practice whereby audit partners rotate back to their former clients A short cooling-off period may be insufficient to sever the close ties between the rotated-off partner and the client Drawing on prior auditor selection and auditor–client relationship literatures, and utilizing data from the mandatory audit partner rotation regime in China,... (two-tailed for variables with bi-directional signs, and one-tailed for variables with one directional sign) ** Significance at the 0.05 level (two-tailed for variables with bi-directional signs, and one-tailed for variables with one directional sign) *** Significance at the 0.01 level (two-tailed for variables with bi-directional signs, and one-tailed for variables with one directional sign) M.A Firth . suggest that auditors normally avoid audit clients that fall into the high-risk category (see Bedard et al. , 2008 for a review). Similarly, when the rotated-off partner considers whether to rotate back after the. explanations for the audit partner rotation -back/ non -rotation -back practice. Recent literature (e.g., Chi et al. , 2009) has focused on the immediate effect of mandatory audit partner rotation. We extend. partner rotation -back or non -rotation -back in the first post-cooling-off year. Panel B shows that for 115 (46.4%) out of 248 companies, the previously rotated-off audit partners rotate back when