Auditors’ Organizational Form, Legal Liability, and Reporting Conservatism: Evidence from China*
Accepted by Michael Willenborg. We thank Associate Editor Michael Willenborg and two anonymous referees for their constructive comments. We also thank K. Hung Chan, Charles Chen, Gongmeng Chen, Helen Choi, Joseph Fan, Jere Francis, Jeong-Bon Kim, Kenny Lin, Agnes Lo, Chung-ki Min, Oliver Rui, William Shafer, T. J. Wong, Donghui Wu, Xi Wu, and George Yang for their helpful comments, discussions, and suggestions on earlier drafts of the paper. In addition, we acknowledge the constructive feedback received from workshop participants at The Chinese University of Hong Kong, City University of Hong Kong, The Hong Kong Polytechnic University, The University of Hong Kong, the China Research Conference 2006 at The Chinese University of Hong Kong, and the American Accounting Association Auditing Section 2006 Midyear Conference, Los Angeles. Michael Firth acknowledges financial support from a grant from the Government of the HKSAR (GRF 340408).
1. Introduction
Limiting the legal liability of auditors can alleviate their exposure to financial loss. However, theoretical studies suggest that limiting auditor liability can also pose a threat to audit quality (Dye 1993, 1995; Chan and Pae 1998). In contrast, many CPA firms argue that limiting their legal liability will not adversely affect audit quality (London Economics 2006; Napier 1998). Given these contrasting views, the issue of whether limiting the legal liability of auditors adversely affects audit quality is an important issue for the accounting profession, the academic community, regulators, and policy makers. Previous research focuses on deriving theoretical models of the association between the limited liability of auditors and audit quality (e.g., Chan and Pae 1998), but the empirical evidence to support these arguments has been restricted to examinations of how different litigation regimes affect auditor behavior (e.g., Choi, Kim, Liu, and Simunic 2008; Francis and Wang 2008).
The objective of this paper is to investigate empirically the association between the organizational form of certified public accountant (CPA) firms and an auditor’s reporting conservatism in China.1 We study the issue in the Chinese setting because of the unique institutional environment there. The disaffiliation exercise for CPA firms that took place in 1998–1999 requires all state-owned CPA firms in China to separate from their government-affiliated parent organizations and form independent CPA firms (Yi 2003). In the restructuring process, subject to approval by the regulatory authorities, CPA firms could choose to take an unlimited liability partnership form of organization or a limited liability form. As all state-owned CPA firms have changed from being government-supported agents to separate legal entities, their risk exposures substantially increase, and their liabilities are no longer covered or protected by the government (Yang, Tang, Kilgore, and Jiang 2001). This change in risk exposure becomes an important concern that has an impact on an auditor’s reporting behavior. Therefore, the reform of the CPA profession in China, which results in samples of both partnerships and limited liability CPA firms, provides an excellent setting to examine empirically how the organizational form of CPA firms affects auditors’ reporting conservatism.
Our paper is the first to study empirically the effect of alternative organizational forms (partnership versus limited liability) of CPA firms on auditor reporting. Although Muzatko, Johnstone, Mayhew, and Rittenberg (2004) investigate the association between IPO underpricing and the organizational form of a CPA firm (general partnership and limited liability partnership (LLP)) in the United States, they do not examine how the reporting behavior of auditors is affected by the limited liability status. While LLPs in the United States limit the liabilities of non-negligent partners, the negligent partners’ personal assets are still at risk. Thus, the negligent partners in an LLP have the same liability exposure as those in a general partnership firm. By contrast, the limited liability CPA firms in China limit the liabilities of both the negligent partners and non-negligent partners. Therefore, our study contributes to the literature by providing clearer evidence on the impact of the liability exposure differences between an unlimited liability partnership firm and a limited liability firm on auditors’ reporting behaviors.
Following Francis and Krishnan 1999, we use an auditor’s propensity to issue a modified audit opinion as a proxy for auditor reporting conservatism. If auditors are more conservative, they should be more likely to issue a modified audit opinion (MAO). While Hope and Langli (2009) use qualified audit reports as a measure of auditor independence, Lim and Tan (2008) and Li (2009) use the propensity to issue going-concern opinions as an indicator of audit quality. The auditor’s reporting decision (clean, nonclean) is therefore frequently used to measure attributes of the audit and auditor.
Audit partners in an unlimited liability partnership firm share liabilities jointly and severally with other partners in the firm, whereas the liabilities of audit partners in a limited liability audit firm are limited to their personal contribution to the capital of the CPA firm. Therefore, auditors in partnership firms have a higher potential risk and liability exposure than auditors in limited liability firms have. We hypothesize that to alleviate this greater uncertainty and liability exposure, auditors in partnership firms are more conservative in their reporting and hence are more likely to issue a MAO. Moreover, there are stronger incentives for the partners to engage in mutual monitoring to ensure a high minimum level of audit quality.
To test our hypothesis, we collect data on 5,007 audits of Chinese listed companies from the period of 2000–2004. Consistent with our hypothesis, the results indicate that auditors in partnership CPA firms are more likely to issue modified audit reports than are auditors in limited liability CPA firms. This positive relation between modified opinions and the organizational form of CPA firms stems from the distressed firms with going-concern opinions, while there is no statistical evidence to support this relation for nondistressed firms. In addition, based on a subsample of CPA firms that were incorporated as limited liability firms during the sample period, we find that after incorporation these auditors are less likely to issue modified reports (to their continuing clients) than they did before the incorporation. Ceteris paribus, this implies that an unlimited liability regime provides a strong motivation for auditors to report conservatively. In summary, the results provide empirical evidence that the organizational form of a CPA firm affects its reporting conservatism, and we attribute this to the threat of liability exposure.
Our results contribute to the broader debate on liability reform for the auditing profession. Currently, in the United States, Europe, and other jurisdictions, there is an intense public policy debate on placing caps on the liability of auditors in civil lawsuits. On the one hand, there is a concern that no one will supply audits because the large potential liabilities may make it not worthwhile to do so (Doralt, Hellgardt, Hopt, Leyens, Roth, and Zimmermann 2008; Napier 1998). Even if there is a supply, unlimited liability may still inhibit the development of auditing. On the other hand, capping liability (either directly through statutory limits on claims and damages or indirectly through incorporation) is seen as a way to evade responsibility and will result in lower audit quality (Plender 1926; Napier 1998). While the Big 4 and other CPA firms are strong and vocal advocates of litigation caps (Wyman 2009), investors take the opposite view. As one example, the International Corporate Governance Network (a global organization of institutional private investors representing assets worth U.S. $15 trillion) recently wrote to the European Union (EU) stating that: “We advise that limiting auditor liability would reduce audit firm accountability, provide a market incentive to take audit shortcuts, and reduce overall audit quality” (Cole 2008). In a consultancy report prepared for the EU Internal Market and Services Commission, institutional investors state that limiting auditor liability will negatively affect their perceptions of audit quality (London Economics 2006). Interestingly, the auditing profession in Germany, which has had liability caps for auditors since the 1930s, recommended raising the caps, as this will signal greater independence (Gietzmann and Quick 1998; Gassen and Skaife 2009).2 Thus, the German auditing profession appears to acknowledge that decreases in liability exposure will be seen in the eyes of investors as impairing audit quality.
In the absence of caps on litigation, many auditing firms have reorganized themselves into LLPs (in the United States) or limited liability companies (LLCs). Incorporation restricts wealth losses from costly litigation to the capital already invested in the firm. Extrapolating our results suggests that capping liability could result in less conservative reporting and potentially lower audit quality.
The paper proceeds as follows. Section 2 describes the recent development of CPA firms in China. Section 3 reviews the literature on the relations between auditors’ liability exposures and audit quality and develops the research hypotheses. Section 4 describes the sample and the research method. We present the empirical results in section 5, and section 6 concludes the paper.
2. Institutional background
Legal liability of auditors in China
Although China may be less litigious than the United States and some other Western countries (Chan, Lin, and Mo 2006), the risk of litigation is not negligible.3 In addition to the normal business risk where the audit firm is unable to repay its creditors, auditors also face litigation risk from investors and their clients’ creditors. Auditors have a direct legal liability to investors for any false and misleading statements or major omissions in an audit report (Johnston and Parker 2007). The first civil litigation against a disaffiliated auditor in China was the case of Jiang vs. PT Hong Guang and Chen Du Su Du CPA Firm in 1998.4 Other cases followed, in which investors sued companies and their auditors. According to a survey of 100 CPA firms carried out in 1999, more than one-third of respondents had been involved in lawsuits (Li and He 2000); the payouts by CPA firms that lost or settled the lawsuits amounted to many millions of RMB each (seven RMB is approximately one U.S. dollar).
Corporate fraud and allegations of auditor negligence continued to grow, and Chinese CPAs fully expected that specific legislation would appear that would make it even easier for clients and investors to sue auditors (Li and He 2000; Zhang 2002). This affected the CPAs’ attitudes towards audit risk. However, it was not until January 2002 that the much-anticipated legislation (the January 15 Notice) was enacted. The January 15 Notice formally documents the circumstances and procedures for investors to initiate litigation against listed companies and their auditors (Supreme Court 2002).5 This notice further opens the door for investors’ claims in China (Zhang 2002: 89).6 Since then, a series of civil litigations against auditors has been initiated.7 International audit firms are also subject to possible shareholder litigation.8 Auditors are well aware of the costs that follow from civil litigation, and this in turn affects their professional judgment and behavior (Li and He 2000; Wen 2002). We contend, however, that this change in behavior is conditioned on the organizational form that the CPA firm has adopted.
To gain further insights into whether auditors in China are concerned about litigation risk, we interviewed partners of several partnership CPA firms and limited liability CPA firms, and the Chinese Institute of Certified Public Accountants (CICPA). The aim of the interviews was to find out more about the various factors that influence CPA firms’ choices of organizational form during the disaffiliation process, such as risk exposures (e.g., engagement risk, litigation risk, or loss of reputation), client portfolio differences, and degrees of government involvement.9 From the interviews, we note that there is a strong consensus that litigation is a major concern for CPA firms and one that they expect will grow in the future. Many partners cite concerns about the spread of the U.S. audit litigation environment to China. Additionally, partners are concerned about regulatory sanctions and loss of reputation that can cause them to go out of business and be personally sued by their creditors.10 However, there are mixed views on how to deal with the threat of litigation (from both investors and creditors). Some auditors say litigation risk has spurred them to incorporate as limited liability firms. Other auditors say that the threat of lawsuits has motivated them to improve audit quality, as they perceive this to be a defense against regulatory sanctions and investor- or creditor-initiated litigation. The interviews reinforce our view that litigation risk is a real concern for auditors in China and one which affects their reporting behavior.
The organizational forms of CPA firms in China
Before 1998, almost all CPA firms in China were state-owned and affiliated with the local or central government, a university, or a government department (DeFond, Wong, and Li 2000; Yi 2003). In 1998, the CICPA implemented a disaffiliation program, which required all state-affiliated CPA firms to be separated from their sponsoring government bodies (the program was completed in 1999). Disaffiliated CPA firms could be registered in the form of an unlimited liability partnership firm or in the form of a limited liability firm subject to the approval of the local regulatory authorities.11 After the disaffiliation exercise, CPA firms became legally independent entities.12
In addition to satisfying the conditions for obtaining a practice license, CPA firms in China are required to have a special license to audit listed companies. Specifically, to be qualified as a special licensed CPA firm, a partnership CPA firm needed to have at least RMB 1 million in capital, while a limited liability CPA firm was required to have a minimum capital of RMB 2 million. The difference in the capital requirement is significant in the Chinese context since before the disaffiliation the auditors were government officials and their salaries and savings were low. Except for the differences in capital and the minimum number of partners, other criteria, such as having at least 20 qualified auditors and RMB 8 million in revenue in the prior year, were the same for the two forms of CPA firms (Ministry of Finance [MOF] 2000). In addition to these specific requirements, the fundamental difference between partnership and limited liability CPA firms is the liability exposure. Specifically, audit partners in a partnership firm are jointly and severally liable for the liabilities of the other partners and the CPA firm. In contrast, if a lawsuit is brought against a limited liability CPA firm, the partners’ legal liabilities are limited to the extent of their capital contribution to the firm.
From our discussions with CPA firms and the regulators (the CSRC and the CICPA), we learned that in some cases the provincial regulatory authorities strongly encouraged the use of a particular organizational form. For example, the Shenzhen authorities required most CPA firms to adopt the partnership form of organization. The political influence of the provincial authorities therefore means that CPA firms did not have a completely free choice of organizational forms at the time of the disaffiliation.13
3. Hypothesis development
Liability exposure and audit quality
Prior U.S. studies generally find that an increased liability exposure induces a higher audit quality level (e.g., Chan and Pae 1998; Dye 1993, 1995; Geiger and Raghunandan 2001; Geiger, Raghunandan, and Rama 2006; Hillegist 1999; Laux and Newman 2010; Liu and Wang 2006; Melumad and Thoman 1990; Schwartz 1997; Venkataraman, Weber, and Willenborg 2008). For example, Melumad and Thoman (1990) find that in an adverse selection setting, the role of auditors is to give information to signal the type of client firm that exists (whether “good” or “bad”). With the threat of litigation, auditors may decide to report truthfully their findings to reduce the probability of paying damages, and thus supply a higher quality audit. Laux and Newman (2010) show that increases in the probabilities of being sued and suffering damage awards result in an increase in both audit quality and the equilibrium audit fee. Venkataraman et al. (2008) examine whether abnormal accruals vary with the legal threat that auditors face in the IPO setting. Consistent with higher audit quality in a high litigation situation, they find that auditors are more conservative when auditing financial statements in an IPO prospectus as the litigation risk exposure is higher than that for a post-IPO audit.
Narayanan (1994) examines analytically the consequences of the Private Securities Litigation Reform Act of 1995, which moved from a joint and several liability regime to a proportionate liability regime in the United States. He shows that under certain conditions a decrease in liability exposure (i.e., proportionate liability regime) could lead to an increase in audit quality. However, his assumption that the plaintiff’s litigation strategy will be the same across regimes seems to be unrealistic in practice (Chan and Pae 1998). By contrast, Chan and Pae (1998) find that the increase in liability protection for auditors may decrease the audit effort and audit quality. They argue that the proportionate liability rule discourages third parties from suing and the disincentive effect on the financial statements users’ litigation decisions can be so severe that it blunts the rule’s potential incentive effect on the auditor’s effort decision. Empirically, auditors (particularly the Big N firms) are found to be less conservative following the adoption of the Reform Act, as they have a lower propensity to issue going-concern modified audit reports (Geiger and Raghunandan 2001; Geiger et al. 2006). Support is also received from several experimental studies (e.g., Dopuch, King, and Schatzberg 1994; Gramling, Schatzberg, Bailey, and Zhang 1998), which conclude that proportionate liability leads to lower audit effort.
The above relation is robust in the cross-country setting that involves different legal regimes and thus auditors’ litigation exposures. Choi et al. (2008) demonstrate that an increase in expected legal liability costs (in different legal regimes) motivates auditors to expend extra effort in the performance of audits, which results in an increased audit fee charged to clients. Francis and Wang (2008) find that in stronger investor protection regimes, the Big 4 auditors are more sensitive to the cost of audit failure and its impact on their reputations, and they therefore allow a lower degree of earnings management by clients.
Based on the analysis of the Chinese market, Mi (2002) argues that limiting an auditor’s liability increases the possibility of fraudulent financial reporting. If the liabilities for individual partners and CPA firms are limited, the benefits from allowing or assisting financial statement manipulation may exceed the costs of being caught. Following the above studies, we expect that increasing an auditor’s liability exposure should motivate the audit partners to have better mutual monitoring and to report more conservatively.
Francis and Krishnan (1999) argue that the reporting conservatism of auditors is a rational mechanism through which auditors can achieve a desired level of audit risk for high-risk clients and lower the probability of being punished for failing to issue a modified report when it is appropriate to do so. Modified audit reports also play a defensive role in protecting auditors from subsequent litigation. Carcello and Palmrose (1994) and Gaeremynck and Willekens (2003) show that modified audit reports that are issued before bankruptcy are more likely to protect auditors from subsequent litigation if bankruptcy eventually occurs.
We argue that a partnership firm is more likely to issue modified audit reports to compensate for the higher risk and liability exposure that is implicitly linked with its organizational form. By contrast, other things being equal, a limited liability firm has less wealth-at-risk (Dye 1995) and thus has a smaller risk and liability exposure. Consequently, a partnership firm is more likely to have a higher threshold for issuing clean audit reports than a limited liability firm. Thus, we hypothesize that the auditors in partnership firms are more likely to issue MAOs than are auditors in limited liability firms due to the intrinsically higher risk and liability exposure that partnerships face. Our hypothesis is as follows.
Hypothesis 1. Ceteris paribus, auditors in a partnership firm are more likely to issue modified audit reports than are auditors in a limited liability firm.
Incorporation of CPA firms and audit quality
Dye (1993, 1995) and Muzatko et al. (2004) demonstrate that once a CPA firm incorporates, the quality of its audits will decrease because the wealth-at-risk is smaller than before. Specifically, Dye (1995) develops a model to show that, in the presence of auditor liability, perceived audit quality is associated with the wealth of auditors. Once auditors have the choice to incorporate, the wealthiest auditors may incorporate only part of their wealth into the firm to reduce their wealth-at-risk relative to what it would be under an unlimited liability regime. Such incorporation induces them to be less conservative. Muzatko et al. (2004) examine empirically the association between the extent of IPO underpricing and the change of organizational form of a CPA firm from a general partnership to a limited liability partnership. They argue that due to reduced intrafirm monitoring, the conversion to a limited liability partnership reduces perceived audit quality, which results in greater ex ante uncertainty for investors and hence greater IPO underpricing. These studies demonstrate that a change in organizational form of CPA firms that increases their liability protection results in lower audit quality.
In this study, we investigate auditors’ reporting behaviors under different organizational liability regimes in China. We expect that when a partnership firm incorporates into a limited liability firm, its risk and liability exposure levels will decrease substantially, and the auditors will be less likely to maintain the threshold of issuing clean opinions. We hypothesize the following.
Hypothesis 2. Ceteris paribus, when a partnership firm incorporates into a limited liability firm, auditors are less likely to issue modified audit reports than before.
4. Research method
Sample collection
Our sample comprises companies that are listed on the Chinese A-share market from 2000 to 2004. We exclude B-share and H-share listed companies because these companies use International Accounting Standards in addition to Chinese Generally Accepted Accounting Principles (GAAP) and have an international auditor. We then exclude companies that are in the first year of audit (IPO firms) and those that are in the utilities and finance sectors, because companies in these industries have different reporting regulations. In addition, we exclude companies that have no auditor information or incomplete financial information and companies whose auditors were engaged in merger exercises in the year of audit to filter out the effects of CPA firms’ mergers and acquisitions. The final sample consists of 5,007 audits of listed companies.
We identify the organizational form of a CPA firm by examining its full name as stated in the listed company’s annual report. If the name of a CPA firm contains the word “limited”, we classify it as a LLC; otherwise, we classify it as a partnership. We confirm the accuracy of our classification of partnerships and limited companies with information provided by the CICPA (the official regulator of the Chinese accounting profession). For example, we check the database of CPA firms maintained by the CICPA which indicates, among other things, whether the CPA firm is a partnership or a LLC. We further confirm the accuracy of our classification by checking with the CPA firms directly.
Next, we identify CPA firms that incorporated their practices as LLCs during our sample period. For each incorporated CPA firm, we trace all of its clients audited before and after the incorporation during our sample period of 2000–2004. Thus, we look back as far as 2000 and as far forward as 2004. To rule out the possible effects of changes in client mix on the reporting decisions of incorporated CPA firms, we focus our tests on the continuing clients of the CPA firms. The final subsample consists of 14 CPA firms with 435 firm-year observations, where 102 observations are audited before the CPA firms’ incorporation and 333 observations are audited after the incorporation.
We use the full sample of 5,007 audits of listed companies to examine Hypothesis 1. To examine the impact of incorporation (i.e., where a CPA firm changes its organizational form from partnership to limited liability) on the likelihood of issuing a modified report, we use the incorporation subsample of 435 observations to test Hypothesis 2. Table 1, panel A, reports the detailed sample collection procedure.
Panel A: Sample collection | |
---|---|
Total firm-year observations | |
Total A-share listed companies from 2000 to 2004 | 6049 |
Less: Initial public offering companies | (337) |
Utilities and finance companies | (245) |
5467 | |
Less: Companies with no auditor information | (72) |
Companies with no financial statement data | (327) |
Companies’ auditors merged during the sample period | (61) |
Final sample (for testing Hypothesis 1) | 5007 |
Incorporation subsample (14 CPA firms) | |
Number of client companies before the incorporation | 102 |
Number of client companies after the incorporation | 333 |
Total sample (for testing Hypothesis 2) | 435 |
Panel B: Sample client companies | |||
---|---|---|---|
Number (percentage) of observations | |||
Audited by partnership CPA firms | Audited by limited liability CPA firms | Total | |
(1) Total number of client | |||
companies | 843 (16.8%) | 4164 (83.2%) | 5007 (100%) |
(2) Number of client companies | |||
Year 2000 | 217 (27.4%) | 575 (72.6%) | 792 (100%) |
Year 2001 | 162 (17.2%) | 780 (82.8%) | 942 (100%) |
Year 2002 | 156 (15.5%) | 853 (84.5%) | 1009 (100%) |
Year 2003 | 164 (14.8%) | 947 (85.2%) | 1111 (100%) |
Year 2004 | 144 (12.5%) | 1009 (87.5%) | 1153 (100%) |
Panel C: Sample CPA firms | |||
---|---|---|---|
Number (percentage) of observations | |||
Partnership CPA firms | Limited liability CPA firms | Total | |
(1) Number of CPA firms | |||
Year 2000 | 23 (31.1%) | 51 (68.9%) | 74 (100%) |
Year 2001 | 11 (15.9%) | 58 (84.1%) | 69 (100%) |
Year 2002 | 9 (12.9%) | 61 (87.1%) | 70 (100%) |
Year 2003 | 9 (12.7%) | 62 (87.3%) | 71 (100%) |
Year 2004 | 7 (9.9%) | 64 (90.1%) | 71 (100%) |
(2) Average number of clients per CPA firm | |||
Year 2000 | 9.4 | 11.3 | 10.7 |
Year 2001 | 14.7 | 13.4 | 13.7 |
Year 2002 | 17.3 | 14.0 | 14.4 |
Year 2003 | 18.2 | 15.3 | 15.6 |
Year 2004 | 20.6 | 15.8 | 16.2 |
Panel B and panel C of Table 1 document the summary statistics of the client companies and CPA firms, respectively. As shown in panel B, 17 percent (843 audits) of the observations are audited by partnership CPA firms and 83 percent (4,164 audits) of the observations are audited by limited liability CPA firms.14 The number and percentage of audits that are conducted by partnership CPA firms decrease during the sample period. As reported in panel C, the proportion of partnership CPA firms decreases from 31 percent in 2000 to 10 percent in 2004. This decline in partnerships reflects the growing concern over the potential losses in personal wealth from this form of organization. By the end of 2005, there are too few partnership observations to conduct meaningful analysis. Thus, our sample period ends in 2004.
Specification of probit regression models


We use (1) to examine Hypothesis 1 and (2) to examine Hypothesis 2. The dependent variable for the two models is the issuance of a MAO (MAO). MAO has a value of one if the CPA firm issues a MAO and zero otherwise. We classify opinions that are unqualified with an explanatory paragraph, qualified, disclaimer, and adverse as MAOs. Companies that receive a MAO are subject to closer monitoring by the CSRC and may have sanctions being imposed on them (Chen, Su, and Zhao 2000; DeFond et al. 2000). Similar to the United States, unqualified opinions with explanatory paragraphs should only be issued for events and transactions that do not affect the fairness of the financial statements, but are important to financial statement users. However, Chinese auditors often employ unqualified opinions with explanatory paragraphs to avoid issuing qualified opinions because they believe it reduces the probability of losing their clients (Chen et al. 2000). Typical events for this type of opinion include related party transactions, uncertainties about asset values, and financial distress or going-concern issues, while typical events for qualified opinions are scope limitations, uncertainties about asset values, and financial distress or going-concern issues (Chan, Lin, and Mo 2006). Thus, an unqualified opinion with an explanatory paragraph differs in form rather than in substance from a qualified opinion and we include this quasi-qualification as a modified opinion. Previous studies have also included unqualified opinions with explanatory paragraphs as modified opinions (DeFond et al. 2000; Chen, Chen, and Su 2001; Lennox 2005; Chan, Lin, and Mo 2006).
In addition, as a going-concern opinion is a more salient decision for auditors and has been used as a proxy for audit quality in many prior studies (e.g., Geiger and Raghunandan 2001; Lim and Tan 2008; Reynolds and Francis 2001), we further test the impact of the organizational form on an auditor’s reporting decision on issuing going-concern opinions (GOING_CONCERN = 1) and condition our analysis on the presence of financial distress (Carcello, Vanstraelen, and Willenborg 2009; DeFond, Raghunandan, and Subramanyam 2002). Following Carcello et al. 2009, we define financial distress as observations for which either net income is negative, net working capital is negative, or stockholders’ equity is negative.
The main explanatory variable for (1) is PARTNER, which represents the legal form of a CPA firm, and takes the value of 1 if the CPA firm is a partnership firm and 0 otherwise. We expect a positive association between MAO and PARTNER (Hypothesis 1). The test variable for (2), INCORP, captures the effect of an auditor’s incorporation. INCORP has a value of 1 if a client is audited after the CPA firm’s incorporation as a limited liability organization and 0 otherwise. As hypothesized in Hypothesis 2, we expect INCORP to be negatively associated with MAO.
Control variables
We include several control variables that capture various characteristics of auditor and client financial information. LNAUDITOR_SIZE represents the size of an auditor, and we measure it as the natural logarithm of the total assets of all of the CPA firm’s listed clients. We use this variable to control for the relation between auditor size and audit opinion. To the extent that big auditors are of high quality (DeAngelo 1981; Francis and Krishnan 1999; Gassen and Skaife 2009), we expect LNAUDITOR_SIZE to be positively associated with MAO. BIG5 is a dummy variable that equals one if a client firm is audited by one of the Big 5 CPA firms and zero otherwise. Many previous studies indicate that Big N firms are more conservative and have higher audit quality than non–Big N firms (e.g., Fan and Wong 2005; Geiger et al. 2006). However, Francis and Wang (2008) argue that the Big N findings relate to developed countries with strong investor protection; they conclude that the Big N are not quality-differentiated auditors in countries with less investor protection. Thus, we make no directional expectation for BIG5. GOV_DEP represents the economic and political influence from local government on a CPA firm’s reporting behavior, and equals one if the listed company is audited by a local auditor and zero otherwise. Following Chan, Lin, and Mo 2006, we classify an audit firm to be a local firm when the firm is located in the same jurisdiction (province or equivalent in China) as the client, and more than 50 percent of its clients’ total assets come from the same jurisdiction. As government influence on CPA firms and client management was quite common during the economic development phase in China (Aharony, Lee, and Wong 2000; Chan, Lin, and Mo 2006; Chen and Yuan 2004), CPA firms that have a majority of their clients located in the same jurisdiction as the CPA firm are the most vulnerable to political influence from the local government. Thus, we expect GOV_DEP to be negatively associated with MAO (Chan, Lin, and Mo 2006). Reynolds and Francis (2001) argue that large clients may pressure auditors to compromise their independence and report favorably. We thus include ECON_DEP to control for the effects of economic dependence on an individual client (Reynolds and Francis 2001; Khurana and Raman 2006; Li 2009). ECON_DEP is calculated as a client’s total assets divided by the total assets of all of the CPA firm’s listed clients; we expect a negative association between ECON_DEP and MAO.
A significant characteristic of China’s reform process is the uneven distribution of growth and development across the different provinces (Demurger, Sachs, Woo, Bao, Chang, and Mellinger 2002). As the degree of market development of the region where the CPA firm is located could have an impact on that auditor’s reporting behavior, it is important that we control for this in our model. To do this, we construct a composite index of regional market development (LOCATION) from a set of indexes (government decentralization index, legal environment index, market intermediary index, foreign investment index, credit market index, and the indexes showing the relative number of lawyers and CPAs that practice in the province) developed by China’s National Economic Research Institute (Fan and Wang 2004). These market development indexes are used as control variables in other research (e.g., Chen, Firth, Gao, and Rui 2006; Wang, Wong, and Xia 2008). The higher the LOCATION index is, the more developed the province is. We give each client observation a LOCATION score appropriate to the province or city where the CPA firm is located.
Prior studies find that audit opinion type is highly persistent (Dopuch, Holthausen, and Leftwich 1987; Lennox 1999, 2000). We include LagMAO, a dummy variable that equals 1 if the CPA firm issued a modified opinion for the client in the previous year, to control for the effect of repeat audit qualifications to audit clients. We use the natural logarithm of the total assets of the client (LNTA) as a measure of client size, and we expect a smaller client to be more likely to receive a MAO (DeFond et al. 2000; Dopuch et al. 1987). Following Dopuch et al. 1987, DeFond et al. 2000, and Lim and Tan 2008, we include the profitability, leverage, and liquidity measures of client companies, ROE, LEVERAGE, and CURRENT_RATIO, in the regression. ROE represents the return on equity. LEVERAGE represents the financial leverage of the client and is computed as the total long-term liabilities over the total assets of the client. CURRENT_RATIO represents the current ratio of clients and is computed as the total current assets over total current liabilities. Generally, the lower the profitability and liquidity levels are, the more likely it is that an auditor will issue a MAO. We thus expect ROE and CURRENT_RATIO to be negatively correlated with MAO, and LEVERAGE to be positively correlated with MAO.
We include two variables that characterize the complexity of a firm’s operations. INVENTORY and AR measure the year-end inventory and accounts receivable of the client companies (scaled by the total assets of the clients), respectively. In the United States, many financial statement fraud cases involve inventories and accounts receivable manipulation (Feroz, Park, and Pastena 1991). St. Pierre and Anderson (1984) also report a high frequency of lawsuits against auditors because of issues that are related to a client’s inventories and receivables. Consistent with DeFond et al. 2000, we expect a positive association between a firm’s complexity (as captured by AR) and the issuance of MAOs (MAO). Although research studies in the United States find a positive association between INVENTORY and MAO, prior China-based research studies do not (DeFond et al. 2000). In light of this, we do not predict a sign for INVENTORY.
Prior studies in China find that older companies are in relatively poorer financial health and are more likely to engage in earnings management to meet the regulatory profitability requirement, and thus are more likely to receive a modified report (Chen et al. 2001; DeFond et al. 2000). We therefore include AGE, which is the number of years that a client has been listed on the stock market, to control for this effect and expect a positive association between AGE and MAO. In China, listed companies need to meet a target profitability level to raise additional capital by a rights issue, and they will be delisted if they report losses for three consecutive years. These regulatory requirements induce companies to manipulate earnings to meet the various thresholds of ROE for rights issuance and the avoidance of delisting, which in turn leads to auditors’ issuance of modified opinions (Chan, Lin, and Mo 2006; Chen et al. 2001). To control for these regulation effects, we include three dummy variables, CONSEC_LOSS, DELIST, and RIGHTS, in our models. CONSEC_LOSS takes a value of 1 if the client experienced losses in the previous two consecutive years and 0 otherwise. DELIST takes a value of one if the client’s current ROE lies between 0.00 and 0.01, and zero otherwise. The RIGHTS variable is equal to one if the client’s current ROE lies between 0.10 and 0.11, and zero otherwise. If the current ROE is only marginally above the target profitability level, the auditor will be more conservative, as the likelihood of the client having engaged in earnings management is higher. We expect that CONSEC_LOSS, DELIST, and RIGHTS are positively correlated with MAO because these companies are more risky and may have distorted their earnings.
We also include several governance and ownership variables in the model. First, we include the percentage of external directors on the board (EXT_DIR). The external directors are supposed to act as a monitor of the chief executive officer and the executive directors and they may constrain managers’ willful behaviors that lead to MAOs. The greater the proportion of external directors is, the greater the influence they can yield; other China studies have also used the percentage of external directors as an indicator of a firm’s internal governance (Chen, Firth, Gao, and Rui 2006; Firth, Fung, and Rui 2006). Second, we include two ownership variables, STATE and FOREIGN, which represent the percentage of shares owned by the state and foreign ownership, respectively, to control for the ownership effect on audit opinions. ZSCORE is the Altman’s 1983Z-score calculated for each company for each year and is included in the models to capture a client’s financial distress, with higher values indicating a lower probability of receiving a MAO.16
Finally, we include year dummies and industry dummies to control for possible year and industry effects on MAOs for the two models. In particular, the year dummy variables capture the changing regulatory environments that have occurred during our sample period (2000–2004). The 10 industry dummies are based on the CSRC industry classification, which has 11 industries after excluding the finance and utilities sectors. We summarize the variables and their definitions in Table 2.
Dependent variable | Definition |
---|---|
MAO | Indicator variable that equals 1 if the CPA firm issued a modified audit opinion for the client, and 0 otherwise. |
GOING_CONCERN | Indicator variable that equals 1 if the CPA firm issued a going-concern audit opinion for the client, and 0 if a clean opinion is issued. |
Independent Variable | Definition |
---|---|
PARTNER | Indicator variable that equals 1 if the CPA firm is a partnership, and 0 otherwise. |
INCORP | Indicator variable that equals 1 if the client firm is audited after the CPA firm’s incorporation, 0 otherwise. |
AUDITOR_SIZE | Size of the CPA firm measured as the sum of its clients’ total assets. |
LNAUDITOR_SIZE | Natural logarithm of the size of the CPA firm (AUDITOR_SIZE). |
BIG5 | Indicator variable that equals 1 if the auditor is one of the Big 5 CPA firms, and 0 otherwise. |
GOV_DEP | Indicator variable that equals 1 if both the CPA firm and its client are in the same provincial location and more than 50 percent of the CPA firm’s clients total assets come from the same jurisdiction, and 0 otherwise. |
ECON_DEP | This variable is a measure of a CPA firm’s economic dependence on an individual client. It is calculated by dividing the assets of a client by the combined assets of all the listed clients of the CPA firm. |
LOCATION | This variable is the composite index of regional market development from a set of indexes (government decentralization index, legal environment index, market intermediary index, foreign investment index, credit market index, and the indexes showing the relative number of lawyers and CPAs that practice in the province) developed by China’s National Economic Research Institute (Fan and Wang 2004). The higher the LOCATION index, the more developed the province where the CPA firm is located. |
LagMAO | Indicator variable that equals 1 if the CPA firm issued a modified audit opinion for the client in the previous year, and 0 otherwise. |
TA | Total assets of the client company. |
LNTA | Natural logarithm of the total assets of the client company. |
ROE | Return on equity of the client company. |
LEVERAGE | Financial leverage of the client, measured by the ratio of long-term liabilities to total assets of the client company. |
CURRENT_RATIO | Client’s current assets over current liabilities. |
INVENTORY | Client’s inventory level scaled by total assets. |
AR | Client’s accounts receivable level scaled by total assets. |
AGE | Number of years the client has been listed. |
CONSEC_LOSS | Indicator variable that equals 1 if the client experiences losses in each of the previous two years, and 0 otherwise. |
DELIST | Indicator variable that equals 1 if the client’s current ROE lies between 0.00 and 0.01, and 0 otherwise. |
RIGHTS | Indicator variable that equals 1 if the client’s current ROE lies between 0.10 and 0.11, and 0 otherwise. |
EXT_DIR | Percentage of external directors on the board of the client company. |
STATE | Percentage of state ownership of the client company. |
FOREIGN | Percentage of foreign ownership of the client company. |
ZSCORE | Altman’s 1983 Z-score. |
Selection bias issues
In investigating the effects of the organizational form of CPA firms on auditor reporting, there are two possible selection biases caused by a CPA firm’s choice of organizational form and a client firm’s choice of a particular type of audit firm. As mentioned above, there are differences in the licensure requirements and regional government influence between the two types of audit firm. These differences may have an impact on CPA firms’ choice of organizational form. Similarly, as a client company self-selects a particular form of CPA firm as its auditor as well as makes choices on accounting quality, this creates a potential endogeneity problem. For example, if partnership CPA firms are considered to be high quality, then companies may have reasons to select them (e.g., to signal the quality of the company) or to avoid them (e.g., to appoint a more compliant auditor). To deal with these two potential self-selection biases, we use a two-stage approach to estimate (1). Specifically, we first estimate two probit models, one to predict a CPA firm’s choice of organizational form and the other to predict a client firm’s choice of the type of CPA firm (we present details of the models and results in the appendix). Other studies have also used two self-selection coefficients (e.g., Chan, Lo, and Mo 2006).
For the CPA firm’s choice model (panel A of the appendix), factors affecting the choice of the legal form include the characteristics of the CPA firm, its client portfolio, governmental influence and other economic and environmental factors.17 As the unit of analysis in the model is the CPA firm, not the client company, we use the average of a CPA firm’s client characteristics to estimate the probit regression. We obtain the selectivity correction variable, the inverse Mills ratio (λa), for each CPA firm from the probit regression. We then include the auditor-specific λa in the second-stage probit model — that is, (1) — to correct for the potential self-selection problem.
For the client firm’s choice of auditor (panel B of the appendix), we regress the organizational form of CPA firms on various client characteristics (e.g., client size, return on equity, business complexity, shareholding structure). Because clients may switch to a new auditor because of unfavorable audit opinions received in previous years, we also control for the MAOs received in the previous year and the organizational form of the CPA firm in the previous year as additional controls. Our models are guided by prior research (e.g., Kim, Chung, and Firth 2003; Weber and Willenborg 2003; Li 2009). We obtain the selectivity correction variable, the inverse Mills ratio (λc), for each client firm from the probit regression and include this client-specific λc in (1) to account for the self-selection bias.18
5. Empirical results
Descriptive statistics and univariate tests
Table 3 shows the types of audit opinions issued during the sample period. As reported in the table, the average percentage of MAOs issued by CPA firms over the period is 11 percent (panel A), while the average percentage of going-concern opinions issued by CPA firms is 3.6 percent (panel B). The number of MAOs and going-concern opinions fluctuates over time, but there is no overall trend of increase or decrease.
2000 | 2001 | 2002 | 2003 | 2004 | Total | |
---|---|---|---|---|---|---|
Panel A: Types of audit opinions | ||||||
Unqualified with explanations | 82 | 76 | 78 | 41 | 66 | 343 |
Qualified | 46 | 35 | 31 | 16 | 40 | 168 |
Disclaimer | 6 | 8 | 9 | 11 | 13 | 47 |
Adverse | 1 | 0 | 0 | 0 | 0 | 1 |
Modified audit opinions | 135 (17%) | 119 (13%) | 118 (12%) | 68 (6%) | 119 (10%) | 559 (11%) |
Clean opinions | 657 (83%) | 823 (87%) | 891 (88%) | 1043 (94%) | 1034 (90%) | 4448 (89%) |
Total | 792 (100%) | 942 (100%) | 1009 (100%) | 1111 (100%) | 1153 (100%) | 5007 (100%) |
Panel B: Going-concern modified audit opinions | ||||||
---|---|---|---|---|---|---|
Unqualified with explanations | 4 | 11 | 10 | 16 | 36 | 77 |
Qualified | 11 | 11 | 8 | 10 | 11 | 51 |
Disclaimer | 6 | 7 | 6 | 9 | 11 | 39 |
Adverse | 1 | 0 | 0 | 0 | 0 | 1 |
Total going-concern modified opinions | 22 (3.2%) | 29 (3.4%) | 24 (2.6%) | 35 (3.2%) | 58 (5.3%) | 168 (3.6%) |
Clean opinions | 657 (96.8%) | 823 (96.6%) | 891 (97.4%) | 1043 (96.8%) | 1034 (94.7%) | 4448 (96.4%) |
Total | 679 (100%) | 852 (100%) | 915 (100%) | 1078 (100%) | 1092 (100%) | 4616 (100%) |
Table 4 provides descriptive statistics on the selected characteristics of the CPA firms and client companies in the full sample (panel A) and the incorporation subsample (panel B), which is partitioned by the organizational form of the CPA firms and the types of audit opinions. As shown in the table, partnership firms are generally larger (as characterized by AUDITOR_SIZE), less influenced by local government (GOV_DEP), and more likely to be located in the more developed provinces (LOCATION) than limited liability firms are (i.e., column 1 versus column 3 and column 2 versus column 4). These findings suggest that partnership firms have the characteristics of high quality CPA firms in China (Chan, Lin, and Mo 2006), which is consistent with our main argument. Panel A also shows that companies receiving MAOs from partnership CPA firms have a higher liquidity level (CURRENT_RATIO), higher percentage of external directors on the board (EXT_DIR), higher percentage of foreign ownership (FOREIGN), lower ZSCORE and are less complex (AR) than those companies receiving MAOs from limited liability CPA firms (i.e., column 1 versus column 3). For the incorporation subsample (panel B), in which a CPA firm changes its organizational form, client characteristics in the postincorporation period are largely similar to those of the preincorporation period, except that postincorporation clients have a lower leverage and current ratio, a higher Z-score, and a longer listing age. These results are expected because they are continuing clients of the incorporated CPA firms.
Panel A: Full sample (n = 5,007) | ||||||||
---|---|---|---|---|---|---|---|---|
Variable | Companies audited by partnership CPA firm | Companies audited by limited liability CPA firm | Test of mean (median) differences | |||||
MAO = 1 (n = 131) (1) | MAO = 0 (n = 712) (2) | MAO = 1 (n = 428) (3) | MAO = 0 (n = 3736) (4) | t-/(Z-) statistic (1) vs. (2) | t-/(Z-) statistic (1) vs. (3) | t-/(Z-) statistic (2) vs. (4) | t-/(Z-) statistic (3) vs. (4) | |
Mean (Median) | Mean (Median) | Mean (Median) | Mean (Median) | Mean (Median) | Mean (Median) | Mean (Median) | Mean (Median) | |
AUDITOR_SIZE (in billions) | 79.17 (49.24) | 92.63 (54.42) | 56.44 (38.73) | 65.65 (38.73) | −1.010 (−2.402**) | 2.455** (2.585***) | 6.089*** (11.163***) | −1.829* (−0.855) |
BIG5 | 0.076 (0.000) | 0.129 (0.000) | 0.026 (0.000) | 0.052 (0.000) | 2.909* | 7.113*** | 59.980*** | 5.554** |
GOV_DEP | 0.527 (1.000) | 0.483 (0.000) | 0.638 (1.000) | 0.652 (1.000) | 0.841 | 5.216** | 73.028*** | 0.353 |
ECON_DEP | 0.060 (0.020) | 0.059 (0.030) | 0.051 (0.030) | 0.062 (0.030) | 0.024 (−3.517***) | 0.712 (−2.831**) | −0.770 (−4.783***) | −2.369** (−5.008***) |
LOCATION | 30.22 (33.29) | 29.60 (33.29) | 27.48 (29.55) | 27.14 (26.55) | 0.952 (0.172) | 3.631*** (1.215) | 7.677*** (3.564***) | 0.826 (0.646) |
LagMAO | 0.618 (1.000) | 0.081 (0.000) | 0.558 (1.000) | 0.061 (0.000) | 17.848*** | 1.212 | 2.008** | 35.104*** |
TA (in millions) | 1532 (832.5) | 4197 (1429) | 1549 (986.8) | 2180 (1332) | −1.342 (−5.685**) | −0.094 (−1.385) | 5.000*** (2.421**) | −3.073*** (−7.494***) |
ROE | −0.284 (0.020) | −0.132 (0.070) | −0.225 (0.000) | 0.058 (0.060) | −0.272 (−4.743***) | −0.103 (2.970***) | −2.101** (0.539) | −3.407*** (−16.618***) |
LEVERAGE | 0.140 (0.030) | 0.080 (0.040) | 0.060 (0.020) | 0.059 (0.030) | 2.126** (−1.993)** | 2.381** (0.789) | 5.097*** (3.410***) | 0.174 (−2.620***) |
CURRENT_RATIO | 1.808 (0.930) | 1.662 (1.340) | 1.207 (1.015) | 1.675 (1.350) | 0.652 (−6.706***) | 2.087** (−1.286) | −0.218 (−0.049) | −6.024*** (−11.847***) |
INVENTORY | 0.142 (0.110) | 0.190 (0.110) | 0.142 (0.110) | 0.137 (0.110) | −0.433 (0.737) | −0.046 (−0.835) | 2.457** (0.555) | 0.768 (0.138) |
AR | 0.088 (0.060) | 0.109 (0.075) | 0.120 (0.090) | 0.110 (0.090) | −1.793 (−1.911*) | −3.043*** (−3.301***) | −0.246 (−2.543**) | 1.874 (1.340) |
AGE | 6.313 (6.550) | 5.763 (5.455) | 6.435 (6.510) | 5.751 (5.520) | 2.046** (2.257**) | −0.463 (0.369) | 0.097 (−0.080) | 4.670*** (5.203***) |
CONSEC_LOSS | 0.099 (0.000) | 0.024 (0.000) | 0.093 (0.000) | 0.020 (0.000) | 18.308*** | 0.039 | 0.494 | 78.227*** |
DELIST | 0.008 (0.000) | 0.003 (0.000) | 0.005 (0.000) | 0.001 (0.000) | 0.726 | 0.165 | 0.823 | 2.544 |
RIGHTS | 0.008 (0.000) | 0.000 (0.000) | 0.000 (0.000) | 0.001 (0.000) | 5.442** | 3.273 | 0.954 | 0.573 |
EXT_DIR | 0.470 (0.450) | 0.367 (0.370) | 0.395 (0.400) | 0.358 (0.360) | 4.002*** (3.911***) | 2.722*** (2.552**) | 0.898 (0.693) | 2.821*** (2.735***) |
STATE | 0.301 (0.300) | 0.302 (0.310) | 0.281 (0.290) | 0.324 (0.350) | −0.015 (−0.156) | 0.827 (0.787) | −1.999** (−1.936*) | −3.171*** (−3.385***) |
FOREIGN | 0.026 (0.000) | 0.011 (0.000) | 0.011 (0.000) | 0.010 (0.000) | 2.658*** (2.939***) | 2.385** (3.700***) | 0.692 (0.792) | 0.418 (0.975) |
ZSCORE | 0.366 (0.320) | 0.554 (0.480) | 0.473 (0.385) | 0.636 (0.520) | −5.403*** (−6.260***) | −2.679*** (−2.879***) | −4.190*** (−4.158***) | −6.541*** (−9.364***) |
Panel B: Incorporation subsample (n = 435) | ||||||||
---|---|---|---|---|---|---|---|---|
Variable | Companies audited by the incorporated CPA firm before its incorporation (n = 102) | Companies audited by the incorporated CPA firm after its incorporation (n = 333) | Tests of mean (median) differences | |||||
MAO = 1 (n = 18) (1) | MAO = 0 (n = 84) (2) | MAO = 1 (n = 32) (3) | MAO = 0 (n = 301) (4) | t-/(Z-) statistic (1) vs. (2) | t-/(Z-) statistic (1) vs. (3) | t-/(Z-) statistic (2) vs. (4) | t-/(Z-) statistic (3) vs. (4) | |
Mean (Median) | Mean (Median) | Mean (Median) | Mean (Median) | Mean (Median) | Mean (Median) | Mean (Median) | Mean (Median) | |
LagMAO | 0.778 (1.000) | 0.071 (0.000) | 0.500 (0.500) | 0.103 (0.000) | 9.229*** | 1.960* | −0.866 | 6.492*** |
TA (in millions) | 2017 (1490) | 2916 (1677) | 2091 (1362) | 2213 (1306) | −0.898 (−1.018) | −0.076 (0.293) | 1.759* (1.720*) | −0.216 (1.199) |
ROE | −0.175 (0.020) | 0.041 (0.075) | −0.789 (0.020) | 0.002 (0.060) | −2.637*** (−3.305***) | 0.665 (0.385) | 0.567 (1.395) | −3.207*** (−2.430**) |
LEVERAGE | 0.054 (0.050) | 0.085 (0.035) | 0.046 (0.015) | 0.058 (0.030) | −0.910 (0.036) | 0.437 (0.580) | 2.298** (0.020) | −0.814 (−1.081) |
CURRENT_RATIO | 1.149 (1.120) | 1.750 (1.450) | 1.116 (1.035) | 1.434 (1.210) | −2.454** (−2.458**) | 0.220 (0.334) | 2.784*** (3.370***) | −1.981** (−1.780*) |
INVENTORY | 0.145 (0.145) | 0.116 (0.085) | 0.145 (0.120) | 0.129 (0.100) | 1.164 (1.525) | −0.011 (0.051) | −0.924 (−0.502) | 0.743 (1.601) |
AR | 0.094 (0.080) | 0.103 (0.070) | 0.108 (0.070) | 0.092 (0.070) | −0.351 (0.717) | −0.447 (0.598) | 0.899 (1.073) | 0.851 (0.801) |
AGE | 5.073 (4.365) | 4.919 (4.530) | 6.749 (6.730) | 6.920 (6.730) | 0.271 (−0.404) | −3.072*** (−3.123***) | −7.238*** (−6.596***) | −0.415 (−0.340) |
CONSEC_LOSS | 0.111 (0.000) | 0.012 (0.000) | 0.188 (0.000) | 0.033 (0.000) | 5.111** | 0.500 | 1.075 | 15.052*** |
EXT_DIR | 0.411 (0.440) | 0.368 (0.430) | 0.407 (0.415) | 0.335 (0.330) | 0.563 (0.636) | 0.043 (0.101) | 1.011 (0.899) | 1.465 (1.351) |
STATE | 0.309 (0.390) | 0.318 (0.305) | 0.277 (0.305) | 0.308 (0.300) | −0.123 (0.281) | 0.433 (0.559) | 0.314 (0.303) | −0.680 (0.752) |
FOREIGN | 0.014 (0.000) | 0.005 (0.000) | 0.004 (0.000) | 0.008 (0.000) | 0.966 (0.720) | 0.883 (0.446) | −0.692 (−0.572) | −0.553 (−0.176) |
ZSCORE | 0.384 (0.360) | 0.514 (0.490) | 0.416 (0.305) | 0.675 (0.560) | −2.021** (−1.058*) | −0.386 (0.172) | −3.063*** (−2.354**) | −3.104*** (−3.652***) |
- Notes:
- Statistics reported relate to the t-test of difference in means and the Mann-Whitney U-test of difference in medians (in parentheses) of each variable for the sample of partnership versus the sample of limited liability firm.
- ***, **, and * indicate two-tailed significance at the 1 percent, 5 percent, and 10 percent level, respectively.
- Notes:
- Statistics reported relate to the t-test of difference in means and the Mann-Whitney U-test of difference in medians (in parentheses) of each variable for the sample before versus after incorporation.
- ***, **, and * indicate two-tailed significance at the 1 percent, 5 percent, and 10 percent level, respectively.
Table 5 reports the Chi-square tests on the association between the organizational form of CPA firms and audit opinions. In panel A, the proportion of MAOs issued by partnership CPA firms is higher than those issued by limited liability CPA firms, and the difference is significant at the 1 percent level. Similarly, for the financially distressed subsample, the proportions of going-concern opinions and other MAOs issued by partnership CPA firms are significantly higher than are the respective proportions issued by limited liability CPA firms (panel B). These results support Hypothesis 1 in that auditors in partnership CPA firms are more likely to issue MAOs than are auditors in limited liability CPA firms.
Panel A: MAO versus clean reports (n = 5,007) | |||
---|---|---|---|
Partnership CPA firms | Limited liability CPA firms | Row total | |
Clean reports | 84.5% (712) | 89.7% (3736) | 88.8% (4448) |
MAOs | 15.5% (131) | 10.3% (428) | 11.2% (559) |
Column total | 100% (843) | 100% (4164) | 100% (5007) |
Panel B: Going-concern reports, non–going-concern MAOs versus clean reports on financially distressed firms (n = 1,668) | |||
---|---|---|---|
Partnership CPA firms | Limited liability CPA firms | Row total | |
Clean reports (1) | 69.6% (206) | 79.2% (1086) | 77.5% (1292) |
Going-concern opinions (2) | 14.2% (42) | 8.6% (118) | 9.6% (160) |
Non–going-concern MAOs (3) | 16.2% (48) | 12.2% (168) | 12.9% (216) |
Column total | 100% (296) | 100% (1372) | 100% (1668) |
Panel C: Incorporation subsample (n = 435) | |||
---|---|---|---|
Before incorporation | After incorporation | Row total | |
Clean reports | 82.4% (84) | 90.4% (301) | 88.5% (385) |
MAOs | 17.6% (18) | 9.6% (32) | 11.5% (50) |
Column total | 100% (102) | 100% (333) | 100% (435) |
- χ2 = 19.566***
- (1) vs. (2) χ 2 = 10.676***
- (1) vs. (3) χ 2 = 5.207**
- (2) vs. (3) χ 2 = 0.819
- χ2 = 4.958**
- Note:
- *** and ** indicate two-tailed significance at the 1 percent and 5 percent level, respectively.
For the incorporation subsample, panel C shows that the proportion of MAOs in the postincorporation period is less than the proportion before the incorporation, and the difference is significant at the 5 percent level. This supports Hypothesis 2 in that, after incorporation, auditors are less likely to issue MAOs.
Table 6 shows the correlation matrix of the dependent and independent variables. The correlations are broadly consistent with the univariate test results. PARTNER is positively correlated with MAO, providing support for Hypothesis 1 in that companies audited by a partnership CPA firm are more likely to receive a MAO than are companies that are audited by a limited liability CPA firm. LagMAO, AGE, and CONSEC_LOSS are also positively associated with MAO, whereas LNAUDITOR_SIZE, BIG5, LNTA, ROE, CURRENT_RATIO, ECON_DEP, STATE, and ZSCORE are negatively associated with MAO. The correlations among the independent variables are modest, so multicollinearity is not a problem in interpreting the results of the regressions.
Regression tests
Table 7 reports the main regression results for (1) (for brevity’s sake, we include the first stage results in the appendix). As shown in the table, the coefficient on PARTNER is significantly positive at the 1 percent level for the full sample (All Years column). The marginal effect for PARTNER in the full sample indicates that moving from a limited liability status to a partnership status will increase the client’s likelihood of receiving a MAO by 5.7 percent. For the yearly analyses, the coefficients on PARTNER for 2001–2004 are significant beyond the 5 percent level. However, the coefficient on PARTNER for the year 2000 is only significant at the 10 percent level. Because the year 2000 is the first year after the disaffiliation exercise, auditors had just detached from the government agencies and might not have fully realized the difference in the risk exposure between the two forms of CPA firms.


For the control variables, consistent with our expectations, clients receiving a modified audit opinion in the previous year, which are smaller in size, and that have a weaker financial condition (e.g., Z-score), are more likely to receive MAOs.
We rerun (1) without λa and λc, and the results (not tabled) show that the coefficient on PARTNER is significantly positive at the 5 percent level for the full sample, and the significance of the control variables remains the same as those shown in Table 7. Thus, our results are robust for the inclusion or exclusion of self-selection corrections.
We then examine the auditors’ reporting behavior for financially distressed and nondistressed firms by running (1) separately for each partition. As defined previously, financially distressed firms are those with negative net income, negative working capital, or negative stockholders’ equity. Table 8 shows that the coefficient on PARTNER is positive and significant at the 5 percent level for the financially distressed sample (panel A), while the coefficient on PARTNER is not significant for the nondistressed firms (panel B). These results indicate that auditors in partnership firms issue more modified opinions to financially distressed firms than auditors in limited liability firms do. However, there is no statistically significant evidence to support the hypothesis for nondistressed firms.

Variables | Exp. sign | Panel A | Panel B | ||
---|---|---|---|---|---|
Financially distressed firms | Non–financially distressed firms | ||||
Coefficient | Z-statistic | Coefficient | Z-statistic | ||
PARTNER | + | 0.486 | 2.538** | 0.308 | 1.528 |
LNAUDITOR_SIZE | + | −0.014 | −0.340 | −0.047 | −1.048 |
BIG5 | ? | 0.171 | 0.808 | −0.212 | −0.902 |
GOV_DEP | − | 0.044 | 0.485 | 0.088 | 0.925 |
ECON_DEP | − | −0.370 | −0.635 | −0.601 | −0.776 |
LOCATION | + | 0.015 | 2.520** | 0.007 | 1.117 |
LagMAO | + | 1.372 | 14.417*** | 1.451 | 14.023*** |
LNTA | − | −0.164 | −3.119*** | 0.093 | 1.254 |
ROE | − | 0.003 | 0.040 | −7.638 | −5.410*** |
LEVERAGE | + | −0.046 | −0.299 | −0.293 | −0.477 |
CURRENT_RATIO | − | −0.008 | −0.361 | 0.006 | 0.204 |
INVENTORY | ? | −0.078 | −2.425** | 0.314 | 0.836 |
AR | + | 0.861 | 1.751* | 0.178 | 0.413 |
AGE | + | −0.009 | −0.526 | 0.000 | 0.018 |
CONSEC_LOSS | + | 0.086 | 0.494 | 0.315 | 1.228 |
DELIST | + | 1.366 | 2.341** | 0.233 | 0.486 |
RIGHTS | + | −7.200 | −25.741*** | 0.630 | 0.772 |
EXT_DIR | − | −0.050 | −0.289 | −0.006 | −0.041 |
STATE | ? | −0.101 | −0.615 | 0.181 | 1.076 |
FOREIGN | ? | −0.193 | −0.297 | 1.421 | 1.857* |
ZSCORE | − | −0.423 | −3.130*** | −0.281 | −2.140** |
Year2000 | ? | −0.313 | −1.885* | 0.615 | 3.937*** |
Year2001 | ? | −0.285 | −2.121** | 0.360 | 2.375** |
Year2002 | ? | −0.076 | −0.646 | 0.157 | 1.047 |
Year2003 | ? | −0.660 | −5.250*** | −0.221 | −1.257 |
Industry dummies | ? | included | included | ||
λc | ? | 0.112 | 1.008 | −0.050 | −0.380 |
λa | ? | −0.164 | −1.228 | −0.179 | −1.441 |
Intercept | ? | 2.554 | 2.194** | −2.523 | −1.665* |
N | 1668 | 3339 | |||
χ2 | 481.220*** | 420.077*** | |||
Pseudo R2 | 0.286 | 0.139 |
- Notes:
- Dependent variable is MAO. The main variable of interest is PARTNER. λc = inverse Mills ratio obtained from the first stage probit client-choice model. λa = inverse Mills ratio obtained from the first stage probit auditor-choice model. All other variables are defined in Table 2. Z-statistics are based on standard errors adjusted for clustering on client firms.
- ***, **, and * indicate two-tailed significance at the 1 percent, 5 percent, and 10 percent level, respectively.
In Table 9, we focus our analysis on the association between PARTNER and the specific type of MAO for the financially distressed sample. Similar to Butler, Leone, and Willenborg 2004, we separate the MAOs into going-concern and other non–going-concern opinions. Panel A presents the regression results for going-concern opinions versus clean opinions, whereas panel B shows the results for non–going-concern MAOs versus clean opinions. As shown in the table, the coefficient on PARTNER is significantly positive for the going-concern subsample at the 1 percent level but only marginally significant for the non–going-concern MAO subsample at the 10 percent level.19 These results suggest that, ceteris paribus, auditors in partnership firms are more likely to issue going-concern opinions to financially distressed firms than are auditors in limited liability firms.

Variables | Exp. sign | Panel A | Panel B | ||
---|---|---|---|---|---|
Going-concern vs. Clean | Non–going-concern MAO vs. Clean | ||||
Coefficient | Z-statistic | Coefficient | Z-statistic | ||
PARTNER | + | 0.742 | 2.973*** | 0.415 | 1.905* |
LNAUDITOR_SIZE | + | −0.005 | −0.106 | −0.059 | −0.979 |
BIG5 | ? | 0.039 | 0.120 | 0.022 | 0.092 |
GOV_DEP | − | −0.001 | −0.008 | −0.023 | −0.223 |
ECON_DEP | − | 0.120 | 0.183 | −1.949 | −1.932* |
LOCATION | + | 0.009 | 1.106 | 0.015 | 2.125** |
LagMAO | + | 1.478 | 12.391*** | 1.132 | 9.857*** |
LNTA | − | −0.252 | −3.742*** | 0.050 | 0.698 |
ROE | − | 0.002 | 0.285 | −0.000 | −0.043 |
LEVERAGE | + | 0.072 | 0.437 | −0.664 | −1.801* |
CURRENT_RATIO | − | −0.177 | −1.279 | 0.021 | 0.817 |
INVENTORY | ? | −0.097 | −2.489** | −0.058 | −1.748* |
AR | + | 1.226 | 1.790* | 0.730 | 1.344 |
AGE | + | −0.008 | −0.394 | −0.015 | −0.798 |
CONSEC_LOSS | + | 0.114 | 0.539 | −0.031 | −0.136 |
DELIST | + | 1.535 | 2.560** | 1.622 | 2.239** |
RIGHTS | + | −6.975 | −21.373*** | −6.572 | −19.298*** |
EXT_DIR | − | 0.031 | 0.128 | −0.074 | −0.380 |
STATE | ? | −0.090 | −0.410 | −0.086 | −0.476 |
FOREIGN | ? | −0.814 | −0.934 | 0.149 | 0.200 |
ZSCORE | − | −0.406 | −1.886* | −0.353 | −2.307** |
Year2000 | ? | −0.703 | −2.910*** | 0.057 | 0.311 |
Year2001 | ? | −0.418 | −2.407** | −0.004 | −0.026 |
Year2002 | ? | −0.438 | −2.625*** | 0.197 | 1.475 |
Year2003 | ? | −0.477 | −3.174*** | −0.675 | −4.102*** |
Industry dummies | ? | included | included | ||
λc | ? | 0.043 | 0.291 | 0.141 | 1.174 |
λa | ? | −0.111 | −0.630 | −0.135 | −0.910 |
Intercept | ? | 3.910 | 2.538** | −0.990 | −0.694 |
N | 1452 | 1508 | |||
χ2 | 383.702*** | 264.145*** | |||
Pseudo R2 | 0.279 | 0.179 |
- Notes:
- Dependent variable is MAO. The main variable of interest is PARTNER. λc = inverse Mills ratio obtained from the first stage probit client-choice model. λa = inverse Mills ratio obtained from the first stage probit auditor-choice model. All other variables are defined in Table 2. Z-statistics are based on standard errors adjusted for clustering on client firms.
- ***, **, and * indicate two-tailed significance at the 1 percent, 5 percent, and 10 percent level, respectively.
We further test whether there is any difference in auditor going-concern reporting across different years. Table 10 reports the yearly regression results. We find that, except for the year 2000, the coefficients on PARTNER are significant at the 5 percent level. The results are similar to the yearly analyses for the full sample.

Variables | Exp. sign | Year 2000 | Year 2001 | Year 2002 | Year 2003 | Year 2004 | |||||
---|---|---|---|---|---|---|---|---|---|---|---|
Coefficient | Z-statistic | Coefficient | Z-statistic | Coefficient | Z-statistic | Coefficient | Z-statistic | Coefficient | Z-statistic | ||
PARTNER | + | 0.199 | 0.218 | 1.875 | 2.431** | 1.799 | 2.100** | 2.350 | 2.317** | 1.466 | 1.978** |
LNAUDITOR_SIZE | + | −0.000 | −0.001 | 0.288 | 0.925 | −0.291 | −1.186 | 0.452 | 1.191 | 0.000 | 0.002 |
BIG5 | ? | −7.021 | −0.000 | −1.552 | −1.602 | 1.014 | 0.949 | 0.540 | 0.668 | −1.102 | −1.094 |
GOV_DEP | − | −0.508 | −0.681 | −0.332 | −0.795 | 0.185 | 0.464 | 0.296 | 0.715 | 0.084 | 0.300 |
ECON_DEP | − | −0.991 | −0.227 | 3.743 | 1.373 | 0.869 | 0.344 | 2.572 | 1.344 | −2.539 | −1.487 |
LOCATION | + | 0.067 | 1.379 | 0.001 | 0.021 | 0.013 | 0.509 | −0.007 | −0.304 | 0.019 | 0.983 |
LagGC | + | 0.683 | 1.009 | 1.835 | 4.302*** | 1.501 | 3.594*** | 1.889 | 4.825*** | 1.791 | 5.387*** |
LNTA | − | −0.200 | −0.587 | −0.636 | −2.449** | −0.258 | −1.021 | −0.456 | −2.019** | −0.142 | −0.857 |
ROE | − | −0.547 | −1.015 | −0.398 | −3.408*** | −0.172 | −1.436 | −0.211 | −1.580 | 0.021 | 1.169 |
LEVERAGE | + | 0.552 | 0.596 | 1.972 | 1.129 | −0.981 | −1.128 | 0.244 | 0.161 | −0.573 | −0.894 |
CURRENT_RATIO | − | 0.507 | 0.972 | −0.216 | −0.647 | 0.210 | 1.008 | −1.899 | −3.066*** | −2.194 | −4.137*** |
INVENTORY | ? | 1.504 | 0.992 | −1.837 | −0.710 | 0.046 | 0.277 | 2.602 | 1.002 | 2.382 | 1.505 |
AR | + | 2.620 | 1.437 | 2.589 | 1.081 | −1.991 | −0.741 | 1.196 | 0.938 | 0.767 | 0.427 |
AGE | + | 0.357 | 1.924* | −0.045 | −0.464 | −0.051 | −0.585 | −0.032 | −0.433 | 0.060 | 1.278 |
CONSEC_LOSS | + | 0.328 | 0.348 | −0.456 | −0.613 | −0.062 | −0.086 | 0.006 | 0.015 | 0.609 | 1.340 |
DELIST | + | N/A | N/A | N/A | N/A | N/A | N/A | −5.522 | −0.000 | N/A | N/A |
RIGHTS | + | N/A | N/A | N/A | N/A | N/A | N/A | −7.047 | −0.000 | N/A | N/A |
EXT_DIR | − | −0.245 | −0.062 | −0.969 | −1.348 | 0.412 | 0.547 | −0.711 | −0.904 | −0.169 | −0.300 |
STATE | ? | 0.260 | 0.230 | −0.094 | −0.124 | 0.131 | 0.165 | 0.350 | 0.500 | −0.894 | −1.762* |
FOREIGN | ? | 0.559 | 0.136 | −1.004 | −0.228 | 0.428 | 0.068 | −12.683 | −0.660 | 1.387 | 0.949 |
ZSCORE | − | −3.719 | −2.017** | 0.128 | 0.209 | −1.610 | −1.953* | −0.798 | −1.445 | 0.180 | 0.493 |
Industry dummies | ? | included | included | included | included | included | |||||
λc | ? | 0.582 | 1.045 | 0.060 | 0.126 | 0.647 | 1.385 | −0.101 | −0.220 | −0.313 | −0.864 |
λa | ? | 0.708 | 1.159 | 0.209 | 0.402 | −1.581 | −2.178** | −1.054 | −1.401 | −0.291 | −0.637 |
Intercept | ? | −1.083 | −0.118 | 4.637 | 0.658 | 10.908 | 1.710* | −2.489 | −0.310 | 1.057 | 0.249 |
N | 142 | 240 | 280 | 371 | 419 | ||||||
χ2 | 66.185*** | 92.328*** | 88.082*** | 129.210*** | 165.678*** | ||||||
Pseudo R2 | 0.506 | 0.423 | 0.363 | 0.402 | 0.424 |
- Notes:
- Dependent variable is GOING_CONCERN. The main variable of interest is PARTNER. λc = inverse Mills ratio obtained from the first stage probit client-choice model. λa = inverse Mills ratio obtained from the first stage probit auditor-choice model. All other variables are defined in Table 2. Z-statistics are based on standard errors adjusted for clustering on client firms.
- ***, **, and * indicate two-tailed significance at the 1 percent, 5 percent, and 10 percent level, respectively.
Table 11 reports the multivariate test results for the auditor-incorporation subsample. The main test variable is INCORP. Consistent with the univariate test, we find that the coefficient on INCORP is significantly negative at the 1 percent level and the marginal effect for INCORP is 8.4 percent. This result supports Hypothesis 2 and suggests that auditors are less likely to issue MAOs after incorporation than before.

Variables | Expected sign | Coefficient | Z-statistic |
---|---|---|---|
INCORP | − | −0.713 | −2.928*** |
LNAUDITOR_SIZE | + | −0.025 | −0.609 |
BIG5 | ? | −0.540 | −0.959 |
GOV_DEP | − | −0.014 | −0.054 |
ECON_DEP | − | −2.618 | −0.576 |
LOCATION | + | 0.359 | 1.457 |
LagMAO | + | 1.195 | 5.266*** |
LNTA | − | 0.189 | 0.919 |
ROE | − | −0.129 | −2.071** |
LEVERAGE | + | −2.378 | −1.773* |
CURRENT_RATIO | − | −0.309 | −2.388** |
INVENTORY | ? | 1.144 | 1.350 |
AR | + | 0.932 | 0.837 |
AGE | + | 0.089 | 1.919* |
CONSEC_LOSS | + | 0.921 | 2.347** |
EXT_DIR | − | −0.093 | −0.232 |
STATE | ? | 0.433 | 0.998 |
FOREIGN | ? | 0.117 | 0.046 |
ZSCORE | − | −0.918 | −2.684*** |
Year2000 | ? | 0.876 | 1.617 |
Year2001 | ? | 0.646 | 2.533** |
Year2002 | ? | 0.033 | 0.123 |
Year2003 | ? | −0.691 | −1.816* |
Industry dummies | ? | included | |
Intercept | ? | −2.594 | −0.773 |
χ2 | 110.550*** | ||
Pseudo R2 | 0.269 |
- Notes:
- Dependent variable is MAO. The main variable of interest is INCORP. All other variables are defined in Table 2. Z-statistics are based on standard errors adjusted for clustering on client firms.
- ***, **, and * indicate two-tailed significance at the 1 percent, 5 percent, and 10 percent level, respectively.
We also perform supplemental tests to check the robustness of our main results. As suggested by Reynolds and Francis 2001, it is possible that large auditors behave more conservatively to protect their reputation rather than because of perceived liability risk. To alleviate the concern of reputation effects on auditors’ reporting behaviors, we partition our samples based on auditor size and rerun the regressions for (1) and (2).20 Overall, the supplementary tests show that our main results are not driven by auditor size; rather, it is the organizational form that a CPA practice takes that drives auditor conservatism.
6. Conclusion
This paper examines the association between the organizational form of CPA firms and the reporting conservatism of auditors in China. As risk and liability exposures have become important concerns for Chinese auditors, we specifically investigate how these factors affect the reporting behavior of auditors that use different organizational forms. Our results indicate that auditors in a partnership CPA firm report more conservatively, particularly in issuing more going-concern opinions to financially distressed firms, than auditors in a limited liability CPA firm do. In addition, we find that the incorporation of a CPA firm decreases the auditor’s threshold to issue clean opinions. Specifically, a client that is audited after the CPA firm’s incorporation is less likely to receive a modified audit report than before.
Our paper examines how the difference in intrinsic liability exposure affects an auditor’s propensity to issue MAOs in China. Our findings indicate that the unlimited liability form of a CPA firm induces behavior that is more conservative; therefore, partnerships are more likely to issue MAOs. By contrast, the limited liability regime provides better protection for CPA firms and audit partners from lawsuits and thus auditors are less likely to issue modified opinions. These results provide empirical support to the perceptions that limiting auditor liability would reduce audit quality (Cole 2008; London Economics 2006). Our results should be useful for regulators and the accounting profession as they consider ways to implement a limited liability regime that do not reduce audit quality or auditor independence. While our analysis pertains to the legal organizational form of CPA firms in China, our results and conclusions can possibly be generalized to countries with a similar legal environment and should have resonance for the wider debate on liability reform for the auditing profession.
Footnotes
Appendix

Panel A: Auditor choice of the legal form of CPA firms | |||
---|---|---|---|
(DV: PARTNER) 355 auditor-year observations | |||
Variables | Exp sign | Coefficient | Z-statistic |
LNAUDITOR_SIZE | − | −0.040 | −0.835 |
GOV_DEP | ? | −1.828 | −4.402*** |
LNTA # | ? | 1.332 | 3.037*** |
ROE # | + | 0.019 | 0.723 |
LEVERAGE # | + | 2.499 | 0.668 |
CURRENT_RATIO # | + | 0.101 | 0.436 |
INVENTORY # | − | −0.196 | −0.138 |
AR # | − | −9.861 | −2.486** |
AGE # | ? | −0.081 | −0.536 |
EXT_DIR # | ? | 0.406 | 0.310 |
STATE # | ? | 2.678 | 1.831* |
FOREIGN # | ? | 20.858 | 2.809*** |
GDI | ? | −0.641 | −3.290*** |
LEI | + | 1.390 | 3.720*** |
MII | − | −0.489 | −3.829*** |
FII | ? | −0.284 | −2.293** |
CMI | ? | −0.170 | −2.127** |
LAWYER_INDEX | ? | 0.642 | 4.075*** |
CPA_INDEX | ? | −0.713 | −4.008*** |
CLIENT_NO | ? | 0.035 | 2.065** |
CPA_NO | ? | −0.003 | −1.024 |
Year2000 | ? | 2.096 | 3.302*** |
Year2001 | ? | 0.704 | 1.178 |
Year2002 | ? | 0.335 | 0.757 |
Year2003 | ? | 0.817 | 2.012** |
Intercept | ? | −29.841 | −3.138*** |
χ2 | 156.685*** | ||
Pseudo R2 | 0.453 |
- Notes:
- The variables are defined as follows: GDI = Government decentralization index, which measures the extent of central government involvement in the local economy. A high GDI indicates lower central government involvement in the local economy. LEI = Legal environment index, which measures the development of the legal environment of each province. A higher LEI indicates a more developed legal environment. MII = Market intermediary index, which measures the development of market intermediaries in each province. A higher MII indicates a more developed system of market intermediaries. FII = Foreign investment index, which measures the extent of foreign investment in each province. A higher FII indicates a greater extent of foreign investment. CMI = Credit market index, which measures the development of the local credit market. A higher CMI indicates a more developed local credit market. LAWYER_INDEX = Relative index of the number of lawyers that practice in the province. CPA_INDEX = Relative index of the number of CPAs that practice in the province. CLIENT_NO = Number of clients that a CPA firm has in the year of observation. CPA_NO = Number of CPAs that a CPA firm has in the year of observation. Other client characteristic variables are defined in Table 2 except that we take the average of the client characteristics of a CPA firm (denoted by #), instead of the characteristics of the individual client company. Z-statistics are based on standard errors adjusted for clustering on audit firms.
- ***, **, and * indicate two-tailed significance at the 1 percent, 5 percent, and 10 percent level, respectively.

Panel B: Client choice of the legal form of CPA firms | |||
---|---|---|---|
(DV: PARTNER) 5007 client-year observations | |||
Variables | Expected sign | Coefficient | Z-statistic |
LNTA | ? | 0.029 | 0.920 |
ROE | + | −0.005 | −1.408 |
LEVERAGE | + | 0.595 | 2.908*** |
CURRENT_RATIO | + | 0.014 | 1.175 |
INVENTORY | − | 0.037 | 1.409 |
AR | − | −0.180 | −0.680 |
AGE | ? | 0.006 | 0.495 |
CONSEC_LOSS | ? | 0.289 | 1.511 |
DELIST | ? | −0.222 | −0.368 |
RIGHTS | ? | 0.007 | 0.018 |
EXT_DIR | ? | 0.157 | 1.350 |
STATE | ? | −0.103 | −0.900 |
FOREIGN | ? | −0.016 | −0.032 |
LagMAO | ? | 0.081 | 0.937 |
LagPARTNER | + | 2.462 | 36.520*** |
Year2000 | ? | −0.188 | −1.842* |
Year2001 | ? | −0.296 | −2.878** |
Year2002 | ? | −0.042 | −0.457 |
Year2003 | ? | 0.013 | 0.150 |
Industry dummies | ? | included | |
Intercept | ? | −2.613 | −3.786*** |
χ2 | 2378.698*** | ||
Pseudo R2 | 0.464 |
- Notes:
- The variables are defined as follows: LagPARTNER = 1 if the client is audited by a partnership CPA firm in the previous year; 0 otherwise. All other variables are defined in Table 2. Z-statistics are based on standard errors adjusted for clustering on client firms.
- ***, **, and * indicate two-tailed significance at the 1 percent, 5 percent, and 10 percent level, respectively.