The audit firm s decision to switch from unlimited to. limited liability *

Transcription

1 The audit firm s decision to switch from unlimited to limited liability * Clive Lennox Nanyang Technological University Bing Li Nanyang Technological University Abstract This study investigates the determinants and consequences of the audit firm s decision to become a limited liability partnership (LLP). We find that the likelihood of an audit firm switching from unlimited to limited liability is increasing in its size and its exposure to litigation risk. There is no evidence that audit firms supply lower audit quality, lose market share, or charge lower audit fees after they become LLPs. However, the mix of public and private clients in audit firms portfolios exhibits a significant shift toward riskier publicly traded companies following the adoption of limited liability. * This paper has benefited from the comments of Jong-Hag Choi, Jere Francis, Jeff Pittman, Mike Willenborg, workshop participants at the Hong Kong University of Science & Technology, and participants at the International Symposium of Auditing Research (ISAR, 2010).

2 1. Introduction Litigation is a huge problem for the accounting profession and threatens the very survival of audit firms. Between 2000 and 2008, 314 claims were filed against the Big Four auditors in the United States (US) and an additional 72 claims were against the non-big Four. As of May 2009 the top 50 malpractice settlements totaled $6,644 million, a figure that does not include unresolved lawsuits against the auditors of Tyco International, Parmalat, Fannie Mae, Freddie Mac, Satyam, and others (Audit Analytics, 2009). Litigation is a major problem in other countries too. For example, simulation analysis by London Economics (2006) suggests that a litigation award of million could trigger the collapse of one of the Big Four audit firms in the United Kingdom (UK), while an award of million would cause a smaller audit firm to fold. At the time of the London Economics (2006) study, there were six claims against UK auditors in excess of 250 million, indicating a real threat to their continued survival. Historically, the accountancy profession has lobbied hard for legislation that limits their exposure to litigation. Limited liability partnerships (LLPs) were introduced in the US in the early 1990s. The UK, Germany, India and Singapore have all recently passed legislation allowing their auditors to become LLPs. In contrast, auditors in Ireland, Japan, and Hong Kong are still required to have unlimited liability. These different regulatory approaches co-exist in the absence of any systematic evidence on the implications of allowing auditors to become LLPs. Our study aims to provide such evidence by addressing two research questions. First, which audit firms have the strongest incentives to obtain limited liability and which are more likely to retain unlimited liability? An answer to this question will inform regulators as to which audit firms would be more likely to become LLPs if auditors are permitted to choose between having limited or unlimited liability. 1

3 Second, what are the consequences of allowing audit firms to become LLPs? Regulators who are considering whether it would be desirable to allow audit firms to become LLPs need to understand the likely consequences. For example, is audit quality impaired after audit firms obtain limited liability by becoming LLPs? Do clients perceive that there are adverse consequences of LLP adoption? Does an audit firm s tolerance of client litigation risk change after it obtains limited liability? We investigate these research questions using data from the UK which first permitted audit firms to organize as LLPs in In addition to the recent change in legislation permitting LLPs, we choose this setting because data are publicly available for both private and public company audits. This means that we are able to measure each audit firm s entire portfolio of clients rather than just its publicly traded clients. 1 Moreover, our sample is very large - 392,215 audit engagements - which increases the power of our tests and makes it less likely that any insignificant results are due to a lack of power. With respect to the first research question, we examine how audit firm size and client litigation risk affect the likelihood that the audit firm will become an LLP. For two reasons, we predict that larger audit firms have stronger incentives to become LLPs. First, the negligent actions of just one bad partner threaten the future existence of the entire partnership. This risk is magnified for the larger audit firms that have more partners signing off on audit 1 Johnstone and Bedard (2004) point out that the availability of data on private companies greatly improves the precise measurement of auditors client portfolios. Likewise, DeFond (2004) notes that Choi, Doogar, and Ganguly [CDG] s (2004) reliance on public company data severely limits the paper s ability to observe Big 6 audit firms entire client portfolios. Because CDG can only observe changes in publicly held companies, and because the vast majority of Big 6 clients are privately held, the paper is unable to truly assess client risk in the auditor s total client portfolio. To the extent that auditors balance the riskiness of their public clients with their non-public clients, CDG is likely to mischaracterize the riskiness of auditors' total client portfolios. Consistent with the argument that client portfolios are largely driven by private rather than public companies, we show that 97% of large audit firms clients come from the private sector. 2

4 engagements. Second, the larger audit partnerships are the potential targets of deep pockets lawsuits (Dye, 1993) and so they have stronger incentives to obtain limited liability. Consistent with both arguments, we find that larger audit firms are more likely to become LLPs compared with smaller auditors. We also predict that audit firms have stronger incentives to become LLPs if their client portfolios expose them to greater risk. We test this prediction using three measures of client litigation risk. First, we use the percentage of the audit firm s portfolio that involves public rather than private companies since public company audits carry greater litigation risk (St. Pierre and Anderson, 1984; Pratt and Stice, 1994; Venkataraman et al., 2008). Second, we examine the percentage of the audit firm s clients that operate in industries with high rather than low litigation risk (Geiger et al., 2006). Third, we use estimates of abnormal audit fees since auditors charge abnormally high fees when they perceive that their clients are more risky (Choi et al., 2009). Using these three measures of client litigation risk we predict that audit firms are more likely to switch to the LLP organizational form when they: (1) have a higher fraction of clients that are publicly traded rather than private, (2) have client portfolios that are more heavily weighted toward industries with high litigation risk, and (3) charge abnormally high audit fees. Our empirical results support all three predictions. Thus, we conclude that the likelihood of an audit firm switching from unlimited to limited liability is increasing in its size and its exposure to litigation risk. Next, we investigate the consequences of LLP adoption for audit quality, the economic value of audits, and risk management. We do so using a difference-in-differences research design that controls for audit firm fixed effects and year fixed effects. In this design, each audit firm is used as a control for itself in order to test whether there is a significant change in audit 3

5 outcomes after the firm becomes an LLP. Our treatment sample comprises the audit firms that become LLPs, while the control sample is the audit firms that retain unlimited liability. Past theoretical studies suggest that the consequences of LLP adoption for audit quality are ambiguous. On the one hand, limited liability could reduce auditors incentives to provide high quality audits by reducing the total cost of litigation (Dye, 1993; Chan and Pae, 1998). On the other hand, the LLP structure shifts the relative cost of litigation away from partners who are not negligent to partners who are negligent (Narayanan, 1994). This shift in incentives occurs because the personal assets of non-negligent partners are protected when the audit firm is an LLP, whereas a non-negligent partner s personal wealth is at risk when another partner acts negligently in a non-llp firm. The protection afforded to non-negligent partners strengthens partners incentives to not be negligent and could therefore lead to an improvement in audit quality after the firm becomes an LLP (Narayanan, 1994). Overall, the effect on audit quality is theoretically ambiguous since there is both a reduction in the total cost of litigation and a shift in the cost of litigation away from nonnegligent partners when a firm becomes an LLP. We provide empirical evidence that helps to address this open question using three proxies for audit quality: (1) the auditor s issuance of modified audit opinions, (2) the magnitude of abnormal accruals (both signed and absolute), and (3) accrual quality. For all three measures we find that LLP adoption has no significant impact on audit quality. The lack of statistical significance is not due to low power since our estimation samples are very large (up to 392,215 audit engagements). Further, we investigate clients perceptions about the impact of LLP adoption on the economic value of audits. If clients perceive that LLP adoption reduces the economic value of auditing, we expect that there would be a fall in demand for an audit firm s services after it 4

6 becomes an LLP. A fall in demand would mean that the audit firm either suffers a drop in its market share or it must offer lower audit fees in order to retain its clients. Therefore, if clients perceive that they are negatively impacted by LLP adoption, we predict that either the market shares of LLP firms or their audit fees would fall relative to the auditors that retain unlimited liability. In fact, the evidence supports neither of these predictions. We therefore conclude that companies do not prefer to be audited by firms that have unlimited liability. A third potential consequence of LLP adoption is on the audit firm s tolerance of client litigation risk. We predict that audit firms would extend their services to riskier types of company after they obtain the protection of limited liability. We find some evidence supporting the prediction. In particular, the mix of public and private clients in the audit firm s portfolio switches toward public companies subsequent to LLP adoption. However, except for the mix of public and private companies, we find no major changes to the riskiness of client portfolios after audit firms become LLPs. In summary, we find no evidence of adverse consequences when auditors voluntarily become LLPs. There is no drop in audit quality and clients apparently do not prefer to be audited by unlimited liability partnerships. These findings raise a further interesting question: why do some audit firms retain unlimited liability? If there are no adverse consequences stemming from LLP adoption, we would expect virtually every audit firm to obtain the benefit of limited liability. In contrast, the majority of audit firms in our sample retain unlimited liability, indicating that there must be costs to becoming an LLP. Two institutional features make it costly for an audit firm to become an LLP in our setting. First, LLP firms are required to file their own financial statements at a central depository where they are made available to the general public. Such disclosure can be costly 5

7 because it forces auditors to disclose proprietary information that they would prefer to remain hidden (Verrecchia, 1983; Feltham and Xie, 1992; Darrough, 1993; Gigler, 1994). In fact, Accountancy Magazine (May 2005) has stated that many audit firms have chosen not to become LLPs because they want their financial statements to be kept private: In discussions with independent practices we discovered that by far the biggest stumbling block is the need for disclosure. The accounts of an LLP are available for anyone to read and there are elements of disclosure of members interests. This [disclosure] reveals the company s profits and the income of its members, and it is this information that some firms are extremely reluctant to reveal. A second cost stems from legislation that requires the LLP s financial statements to be independently verified by another auditor. This implies that LLP adoption is costly because the LLP firm must pay a fee for the audit of its financial statements. Perhaps more important, the LLP firm must allow another auditor access to its proprietary commercial information including its customer base during the course of the audit. We argue that this is potentially costly because the auditor of the LLP firm can approach a client with a lower fee when it finds out that the LLP firm is making high profits on a certain engagement. Such access to the LLP firm s proprietary data makes it easier for the auditor of the LLP firm to grab some of the LLP s clients. Consistent with our prediction about client-grabbing, we find that firms lose clients to their auditors at a higher rate after they become LLPs compared with beforehand. That is, the audit firms that switch to the LLP organizational form suffer significant client losses to the firms that audit their financial statements. 2 The magnitude of this client-grabbing cost is estimated to be 4.1% of annual audit fee revenue for the median audit firm in our sample, which is economically significant. 2 Our results on audit firms market shares exclude any cases in which a company switches between the LLP firm and its auditor. Therefore, there is no inconsistency between our evidence that LLP firms lose clients to their own auditors ( client-grabbing ) but they do not lose market share to other audit firms. 6

8 Overall, our results suggest that audit firms become LLPs when the private benefits of obtaining limited liability outweigh the private costs. The large audit firms and the firms with risky clients apparently benefit the most from obtaining limited liability as they are the firms that are most likely to become LLPs. The smaller audit firms with less risky clients prefer to retain unlimited liability because otherwise they would be required to make their financial statements publicly available and those financial statements would need to be audited, exposing the auditor to the risk of client-grabbing. Beyond the private consequences to audit firms, we find no evidence that LLP adoption has adverse consequences for audit quality or that companies prefer to be audited by firms with unlimited liability. The absence of negative externalities is important because it significantly weakens the economic argument for forcing auditors to have unlimited liability. We caution that our inferences about the consequences of LLP adoption apply only to the partnerships that voluntarily choose to obtain limited liability. We do not attempt to draw any conclusions about what the consequences might be if every audit firm were compelled to become an LLP. The auditors that retain unlimited liability may have made this choice because they foresaw that LLP adoption would have undesirable consequences for their quality and their clients. Auditors are permitted to choose the organizational form that they find privately optimal and our findings do not imply that regulators should mandate that every audit firm become an LLP. Instead, we draw inferences about the consequences of audit firms being allowed to become LLPs on a voluntary basis. This is the policy-relevant research question since regulators are considering whether auditors should be permitted to become LLPs, not whether they should be compelled to do so. 7

9 This paper makes three contributions to the literature. First, we are unaware of any other study on the determinants of the decision to become an LLP. This evidence is important because it reveals which firms perceive the greatest net benefits from obtaining limited liability. This in turn will help regulators to predict which firms would be most likely to change their organizational form if auditors are allowed to become LLPs. Second, there is a voluminous literature that examines the implications of client litigation risk and the country s legal environment for various audit outcomes, such as audit quality, audit fees, auditor choice, and the riskiness of auditors client portfolios. 3 Our study is the first to evaluate how these audit outcomes change in response to a change in the audit firm s organizational form. This evidence on consequences is important for regulators who are tasked with deciding whether audit firms should be permitted to become LLPs. Finally, most studies use data from public company audits only and so they cannot test how the mix of public and private clients affects audit firms behavior. Our study is the first to show that audit firms are more likely to become LLPs when their portfolios contain a higher proportion of public companies, and their portfolios become even more heavily weighted toward public companies after they become LLPs. 4 The remainder of this paper is organized as follows. Section 2 discusses the determinants and consequences of the decision to become an LLP. Section 3 explains the research design, while Section 4 describes the sample and presents descriptive statistics. Section 5 reports tests on which audit firms are more likely to become LLPs, while Section 6 examines 3 This literature includes Beatty (1993), Carcello and Palmrose (1994), Krishnan and Krishnan (1997), Morgan and Stocken (1998), Willenborg (1999), Shu (2000), Bell et al. (2001), Seetharaman et al. (2002), Johnstone and Bedard (2003), Johnstone and Bedard (2004), Fargher et al. (2005), Choi et al. (2008), Venkataraman et al. (2008), Landsman et al. (2009), and Choi et al. (2009). 4 Johnstone and Bedard (2004) examine the mix of public and private clients using proprietary data obtained from one audit firm. Our study is different because we have public and private company data for every audit firm and we are therefore able to examine the relation between the public-private mix and the audit firm s organizational form. 8

10 the consequences of LLP adoption. Section 7 concludes by discussing the regulatory implications of our findings. 2. The determinants and consequences of the LLP organizational form 2.1 The benefits to becoming an LLP In an LLP, the personal assets of non-negligent partners are protected in the event of a lawsuit against a negligent partner. In an unlimited liability partnership, a non-negligent partner s personal wealth is at risk when another partner acts negligently. Although the LLP organizational structure protects the personal assets of non-negligent partners, it does not protect the assets of the audit firm or the personal assets of negligent partners. Therefore, LLP firms continue to face a significant threat from liability. 5 The LLP legislation was passed after extensive and costly lobbying by the accounting profession (Cousins et al., 1999), suggesting that there are real economic benefits to becoming an LLP. This is consistent with a study by London Economics (2006), which suggests that litigation can threaten partners personal assets by increasing the risk of bankruptcy. Given that this is so, the following sections develop predictions about which audit firms would perceive the greatest benefits from becoming LLPs. 5 Our study of LLPs is very different from past studies that examine the Private Securities Litigation Reform Act (PSLRA) in the US (Lee and Mande, 2003; Francis and Krishnan, 2003; Choi et al., 2004; Geiger et al., 2006). Whereas the LLP organizational structure protects only the personal assets of nonnegligent partners, the PSLRA protects the assets of the entire audit firm and the personal assets of negligent partners. Therefore, the evidence from past studies of the PSLRA does not readily extend to the LLP setting. Further, the PSLRA was passed in 1995 shortly after US audit firms became LLPs in 1994, which would make it difficult to empirically distinguish between these two liability-reducing events using US data. This problem does not affect our setting because, except for the LLP legislation, there were no changes to the UK s audit liability regime during our period of analysis. 9

11 2.1.1 Audit firm size We predict that larger audit firms are more likely to become LLPs for two reasons. First, in a partnership that has unlimited liability, the negligent actions of just one partner threaten all the non-negligent partners and this threat is greater for firms that have more partners. For example, in a two partner firm, partner A only has to worry about the degree of care that partner B exercises, whereas in a firm with one hundred partners partner A has to be concerned about the audits of ninety nine other partners. The LLP structure protects the personal wealth of all nonnegligent partners against the actions of a negligent partner. Therefore, the larger partnerships have stronger incentives to become LLPs. 6 Second, Dye (1993) argues that large audit firms are exposed to deep pockets lawsuits in which plaintiffs attempt to recoup some of their losses following corporate failure. Therefore, there is a greater demand from the larger audit firms for protection against such deep pockets lawsuits. On the other hand, it is by no means obvious that the larger audit firms are more likely to become LLPs. Academic research finds that smaller auditors permit more earnings management (Becker et al., 1997; Francis et al., 1998), allow more fraud (Lennox and Pittman, 2010), and are sued more often (Palmrose, 1988). Thus, the smaller auditors may perceive a greater need for protection against lawsuits. In addition, audit firms need to consider the implications of a change in organizational form for their reputations (Mayhew, 2001; Skinner and Srinivasan, 2010). Theory suggests that the larger audit firms have valuable brand name reputations for providing high quality audits (DeAngelo, 1981). LLP adoption could damage these reputations if financial statement users perceive that limited liability would impair audit 6 Consistent with this argument, the President of the Board of Trade proposed the LLP legislation by stating that: The proposals are being made in response to the concerns of many in the professions that, under present partnership law in the UK, the personal assets of the active members of the partnership are at risk from the business decisions of other partners even though, in modern business conditions, it may well be impossible for partners even to know all the other partners (Press Release; February 20, 1997). 10

12 quality. Therefore, the larger reputable firms may retain unlimited liability in order to signal their commitment to maintaining high audit quality. Finally, the large audit firms may be less concerned about litigation risk since their risks can be spread across more highly diversified portfolios of clients (Francis and Krishnan, 2003) Client litigation risk An audit firm has a greater need for legal protection when its clients pose high litigation risk. We therefore expect that, after controlling for its size, an audit firm is more likely to become an LLP if it has a riskier client portfolio. We measure client litigation risk using three proxy variables. The first equals the proportion of audits that involve publicly traded rather than private companies. The litigation risk stemming from an audit of a private company is much smaller than that of a public company because more people rely on the audited financial statements of public companies (St. Pierre and Anderson, 1984; Pratt and Stice, 1994; Venkataraman et al., 2008). Thus, we expect that audit firms whose client portfolios are heavily weighted toward public companies are more likely to become LLPs. 7 Our second measure captures the litigation risk associated with the company s industry. The pharmaceutical, computer, electronic, manufacturing, retail, and service industries pose relatively high litigation risk (Geiger et al., 2006). Therefore, we expect that audit firms have stronger incentives to become LLPs if they have more clients in these industries. Our third risk variable is based on the audit fee charged to the client. When a client 7 Although the majority of publicly traded companies are audited by large firms, it is not the case that the large auditors have the highest values for the ratio of public to private clients. In fact, the public-private mix is just 3% for the five largest auditors because the vast majority of their clients belong to the private sector. The portfolios of several small audit firms contain a higher public-private mix than those of the five largest auditors. Although there is a positive correlation between audit firm size and the fraction of public companies in the auditor s portfolio, the correlation is not large (corr. = 0.264) and we show that the public-private mix variable is empirically distinguishable from the effects of audit firm size. 11

13 poses high risk, the auditor can perform additional audit testing to mitigate its litigation risk or the auditor can charge a risk premium as compensation for bearing the high risk (Beatty 1993; Morgan and Stocken, 1998; Willenborg, 1999; Bell et al., 2001; Seetharaman et al., 2002; Choi et al., 2008; Venkataraman et al., 2008; Choi et al., 2009). In either case, the implication would be that auditors charge higher audit fees to clients that are more risky. Therefore, we expect that such audit firms are more likely to become LLPs. 2.2 The Costs to Becoming an LLP Clients preferences for audit firms that have unlimited liability Companies may prefer their auditors to have unlimited liability for two reasons. First, companies have more to gain when they sue an auditor that has unlimited liability since the LLP structure protects the personal wealth of non-negligent partners. Limited liability could imply a lower level of insurance coverage thereby triggering a fall in demand for the auditor s services after it becomes an LLP. Second, companies may perceive that an LLP firm has weaker incentives to supply high quality audits due to the diminished threat of liability (Dye, 1993; Chan and Pae, 1998). This could also cause a drop in demand for the audit firm s services after it becomes an LLP. If companies prefer to be audited by firms that retain unlimited liability we expect one or both of the following outcomes. Either companies would switch from the LLP firms to auditors that have unlimited liability or the LLP firms would need to offer lower audit fees in order to prevent their clients switching away. Therefore, we test the client preference hypothesis by examining whether there are significant changes in the audit firm s market share and/or changes in audit pricing after the firm becomes an LLP. 12

14 On the other hand, it is by no means obvious that LLP adoption would lead to a fall in demand for the audit firm s services. In Narayanan s (1994) theoretical model, switching to limited liability has an ambiguous effect on audit quality. He points out that limited liability could cause audit quality to improve because unlimited liability exposes both negligent and non-negligent partners to the same litigation threat and this dulls the incentives of partners to exert effort. Since his model suggests an ambiguous effect on audit quality, the impact on the demand for the audit firm s services is also ambiguous Proprietary information and client-grabbing An audit firm may choose to remain an unlimited liability partnership if it believes that LLP adoption would result in lower audit quality and/or a fall in demand for the firm s services. In addition, there are two other reasons that a firm may choose not to become an LLP. First, an LLP firm is required by law to publicly disclose its financial statements whereas such disclosures are not required for unlimited liability partnerships. Such disclosures can be costly to the extent that they force firms to reveal proprietary information that they would prefer to remain hidden (Verrecchia, 1983; Feltham and Xie, 1992; Darrough, 1993; Gigler, 1994). A survey by The Accountancy Magazine (2005) indicates that audit firms are reluctant to disclose information about their profits and partners incomes and that this is why some audit firms have retained unlimited liability. In addition to the disclosure requirement, LLP firms must have their financial statements independently audited. This imposes a direct cost (i.e., an audit fee) on the LLP firm. Perhaps more important, it means that another audit firm has access to the LLP firm s internal information including its customers. We expect that such inside information makes it easier for 13

15 the auditor of the LLP firm to grab some of the LLP firm s clients. For example, the auditor of the LLP firm can approach a client with a lower bid when it finds out that the LLP firm is making high profits on a certain engagement. We thus predict that LLP firms lose more clients to their auditors during the period they are audited (i.e., after LLP adoption) compared with the period before they are audited. Evidence in support of our client-grabbing prediction would imply that it is costly to obtain limited liability because the LLP firm must share proprietary information about its commercial operations with another auditor. 3. Research design We infer why audit firms become LLPs by examining the ex ante characteristics of auditors that change (do not change) their organizational form and the ex post consequences of this decision. In principle, if the relevant decision makers have rational expectations, the ex ante and ex post evidence should give consistent answers: the motives to become LLPs uncovered on the basis of ex ante evidence should be consistent with the ex post consequences. But in practice the importance of some factors can be assessed only by looking at ex post data. For example, the client-grabbing cost is a factor that can only be gauged by examining whether firms lose clients to their auditors after they become LLPs. Thus, we attack the issue of why audit firms become LLPs by using both ex ante information and ex post evidence on the consequences. 3.1 The ex ante characteristics of audit firms We begin by testing whether larger audit firms and audit firms whose clients pose high litigation risk are more likely to become LLPs. We estimate a Cox proportional hazards model 14

16 that explains the likelihood and speed with which audit firms decide to change their organizational form. h i (t) = h 0 (t) exp (α 1 BIGN i + α 2 AUDITOR_SIZE it-1 + α 3 %PUBLIC it-1 + α 4 %HI_RISK it-1 + α 5 AF_RES it- 1 + α 6 CLIENT_SIZE it-1 + α 7 LEVERAGE it-1 + α 8 ROA it-1 + u it ). (1) Eq. (1) is estimated using the audit firm year as the unit of analysis. The dependent variable (h i (t)) is the hazard rate for audit firm i in year t. The hazard rate is simply the conditional probability that audit firm i will become an LLP at time t given that it had not become an LLP earlier. The t variable is right-censored for the firms that retain unlimited liability because we do not observe whether they will become LLPs after the end of our sample period. We estimate the Cox model, taking into account the econometric issues stemming from this right-censoring. The baseline hazard, h 0 (t), captures the time variation in the LLP adoption probability and therefore controls for any year fixed effects. The independent variables explain the audit firm s decision to become an LLP. The BIGN i variable is an auditor size dummy equal to one if audit firm i is a Big Four firm, and zero otherwise. We also employ a continuous measure of audit firm size (AUDITOR_SIZE it-1 ), equal to the log of the number of audits performed by firm i in year t-1. 8 (We lag all the time-varying independent variables by one year to prevent their values being affected by the decision to become an LLP and take logs to account for skewness in the audit firm size measure.) We expect that larger audit firms are more likely to become LLPs (i.e., α 1 > 0 and α 2 > 0 in eq. (1)). The %PUBLIC it-1, HI_RISK it-1, and AF_RES it-1 variables are our proxies for client litigation risk. The public-private mix (%PUBLIC it-1 ) captures the proportion of i s audits conducted on 8 In untabulated robustness tests we measure audit firm size using the log of clients aggregate assets and the log of aggregate audit fees and obtain very similar results. All untabulated results are available from the authors upon request. 15

17 public rather than private companies in year t-1. The %HI_RISK it-1 variable measures the fraction of clients that belong to a high-risk industry (i.e., pharmaceuticals, computers, electronics, manufacturing, retail, and services). The AF_RES it-1 variable is the mean audit fee residual for audit firm i in year t-1. Audit fees are abnormally high when auditors perform additional audit testing in response to high litigation risk or when they charge fee premiums to riskier clients. Thus, a positive fee residual suggests that the auditor is relatively concerned about litigation risk. (We measure abnormal audit fees by estimating a fee model and then calculating the difference between actual and predicted audit fees. Descriptive statistics for the variables in the audit fee model are reported in Appendix 1, while the regression results are reported in Appendix 2.) We predict that audit firms are more likely to become LLPs when their client portfolios have larger values for %PUBLIC it-1, %HI_RISK it-1, and AF_RES it-1 (i.e., α 3 > 0, α 4 > 0 and α 5 > 0 in eq. (1)). We control for additional characteristics of the audit firm s client portfolio that could affect its decision to become an LLP. Specifically, CLIENT_SIZE it-1, equals the mean of the natural log of total assets for audit firm i s clients; LEVERAGE it-1 is the mean ratio of liabilities to assets for audit firm i s clients; and ROA it-1 is the mean return on assets. The profitability and leverage variables measure clients financial risk which is different from the risk of a lawsuit (DeFond, 2004), so we do not draw inferences about client litigation risk from these variables. 3.2 The ex post analysis of audit firms client portfolios We examine the ex post consequences for audit firms client portfolios using eq. (2): Auditor_characteristic it = α 0 + α 1 LLP it + u i + d t + e it. (2) 16

18 where u i and d t are auditor-specific and year-specific fixed effects, respectively. The inclusion of auditor fixed effects (u i ) allows us to control for auditor characteristics that remain constant in both the pre- and post-llp periods. The year effects (d t ) control for any time-varying characteristics that affect auditors as a whole. The treatment variable (LLP it ) equals one if audit firm i is an LLP firm in year t, and zero otherwise. The LLP it variable captures the impact of a change in organizational form, while controlling for cross-sectional differences between audit firms (u i ) and year effects (d t ). For the sake of brevity, we tabulate results for the LLP it variable but not for the dummy variables. We are interested in whether companies prefer to be audited by firms that have unlimited rather than limited liability. Thus, we test whether audit firm i loses clients subsequent to becoming an LLP. The first dependent variable (AUDITOR_SIZE it ) equals the log of the number of clients of the audit firm in year t. If clients prefer auditors that have unlimited liability, we would expect that audit firms suffer a fall in the number of clients after they become LLPs. This would imply a negative coefficient on LLP it in eq. (2). In addition to a levels measure, we examine the change in audit firm i s market share. This GROWTH it variable equals the log of (one plus) the number of clients gained minus the log of (one plus) the number of clients lost by audit firm i in year t. (When constructing this variable, we exclude any auditor switches that involve the LLP firm and its auditor in order to distinguish between this test of client preferences and the test of client-grabbing by the LLP s auditor.) If clients prefer to be audited by firms that have unlimited liability, we expect that auditors grow less quickly after they become LLPs, which implies a negative coefficient for the LLP it variable. 17

19 3.3 The ex post analysis of audit fees If companies prefer to be audited by firms that have unlimited liability, the LLP firms would need to offer relatively low audit fees to their clients in order to deter them from switching to other auditors. We test whether LLP adoption has this impact on audit pricing by estimating the model in eq. (3): LAF jt = α 0 + α 1 LLP it + β X + u i + d t + e jit. (3) LAF jt is the log of audit fees paid by company j in year t and X is a vector of control variables. (A full description of the control variables is provided in the results section.) Eq. (3) controls for auditor fixed effects (u i ) and year fixed effects (d t ) in order to implement our difference-indifferences research design. If audit firms receive lower audit fees after they become LLPs we would expect that α 1 < 0 in eq. (3). In summary, we expect that if companies prefer their auditors to have unlimited liability, the LLP firms would either lose clients (eq. (2)) or they would receive lower audit fees (eq. (3)). Evidence that LLP firms neither lose clients nor receive lower fees would be inconsistent with the argument that the LLP structure has adverse consequences for clients. 3.4 The ex post analysis of audit quality We examine the consequences for audit quality using eq. (4): Audit quality jt = α 0 + α 1 LLP it + β X + u i + d t + e jt. (4) The dependent variable measures the quality of the audit supplied to company j in year t. Again, the LLP variable captures the impact of becoming an LLP, after controlling for audit firm fixed effects (u i ) and year fixed effects (d t ). 18

20 We use audit quality proxies that are commonly found in past studies. DeAngelo (1981) defines audit quality as the joint probability that a breach in financial reporting is discovered and reported by the auditor. Following prior research (e.g., DeFond et al., 2002), we use audit opinions to surrogate for the auditors propensity to both detect and report problems related to the client. Thus, our first proxy for audit quality is the auditor s issuance of a modified or qualified opinion. 9 If auditors have weaker incentives to supply high quality audits after they become LLPs, we predict that they would more often issue clean unqualified opinions (α 1 < 0 in eq. (4)). The opposite would be expected if the LLP structure increases audit quality. We also examine the consequences of LLP adoption for earnings management (Venkataraman et al., 2008), which is measured using the company s signed abnormal accruals. The primary advantage of using signed rather than absolute accruals is that auditors and clients tend to disagree about income-increasing rather than income-decreasing earnings management and auditors generally require their clients to adjust earnings downwards rather than upwards (DeFond and Jiambalvo, 1993; Kinney and Martin, 1994; Nelson et al., 2002). However, earnings can be managed either upward or downward depending on the manager s objectives, so we also use absolute abnormal accruals as an alternative measure of audit quality (Chen et al., 2008). Finally, we use the accrual quality measure of Dechow and Dichev (2002) because managers can use accruals to signal future cash flows rather than for opportunistic reasons. 9 The FAME database includes a data field indicating whether the audit report contains an unqualified clean opinion or the report is a departure from this usual type of opinion. Unlike in the US, audit opinions are not modified for changes in accounting standards or accounting policy and the vast majority of modified opinions in the UK are issued due to going-concern problems. A qualified opinion is issued if the auditor discloses other problems with the financial statements such as violation of accounting standards or limitations on audit scope. The FAME database does not allow us to distinguish between modified and qualified opinions, but qualified opinions are known to be extremely rare. Thus, the vast majority of non-clean opinions in our sample are issued due to going-concern problems. 19

21 3.5 The ex post analysis of client grabbing An audit firm that becomes an LLP needs to have its own financial statements independently audited. We posit that this is costly because the LLP s auditor is uniquely positioned to grab clients from the LLP firm. We test this by comparing how many clients the firm loses to its auditor before and after it becomes an LLP. L#CLIENTS_LOST it = α 0 + α 1 LLP it + α 2 LLP_SIZE it + α 3 AUDITOR_SIZE it + u i + d t + e it (5) In eq. (5), the LLP it variable equals one during the years in which the LLP firm is audited, and zero during the same firm s pre-llp years. For example, Deloitte LLP is audited by Grant Thornton from 2004 to 2008, so LLP it takes the value one for Deloitte LLP between 2004 and 2008, and zero between 1999 and The dependent variable (L#CLIENTS_LOST it ) equals the log of (one plus) the number of clients lost by LLP firm i to its auditor in year t. For example, in the case of Deloitte LLP and its auditor Grant Thornton, we count the number of clients that switch from Deloitte to Grant Thornton in each year from 1999 to (We take logs to reduce the impact of skewness and outliers in the raw variable.) Under the client-grabbing hypothesis, we expect that firms lose more clients to their auditors after LLP adoption than they do before LLP adoption, which would imply a positive coefficient on LLP it in eq. (5). The firms that remain unlimited liability partnerships are not audited so eq. (5) is estimated on just the firms that do become LLPs. Eq. (5) controls for the size of the LLP firm (LLP_SIZE it ) and the size of the LLP firm s auditor (AUDITOR_SIZE it ) because larger audit firms have more clients to lose. We also include fixed effects for each audit firm (u i ) and the year fixed effects (d t ) take care of any yearly variation in the rate of client switching. 20

22 4. Sample and descriptive statistics 4.1 Sample The Limited Liability Partnerships Act permitted audit firms to become LLPs as of April 1 st Our tests of consequences involve comparing a firm s pre-llp period with its post-llp period, so we collect data prior to as well as after the legislation took effect. Financial statement information comes from the FAME database which provides data for the most recent ten years (1999 to 2008). FAME covers over one million UK companies but the data have to be downloaded a few thousand companies per time and, to obtain the data within a reasonable time frame, we restrict the sample by requiring that each audit firm undertakes at least 100 audit engagements in one or more of the ten sample years. This yields an initial sample of 95 audit firms. Next we drop Andersen and KPMG leaving a final sample of 93 audit firms. We exclude Andersen because the fallout from the Enron scandal caused it to disappear shortly after the new LLP legislation came into effect in We drop KPMG because it was already a limited liability company before the introduction of the LLP Act. 10 (Untabulated tests reveal that all of the main results are unchanged if we include Andersen and KPMG in the sample.) Every company is required to file its financial statements with Companies House, where the accounts are available to the general public. This is very different from the US where only SEC registrants are required to make their financial statements publicly available. The rationale 10 KPMG became incorporated following the Companies Act of 1989 which allowed audit firms to organize as companies. With the exception of KPMG, most audit firms preferred to remain as partnerships because the tax rules are much less favorable for companies (Cousins et al., 1999). Companies are subject to double taxation (once at the corporate level and again at the shareholder level via either dividends or capital gains) whereas partnerships face taxation only at the partner level. The tax rules for unlimited liability partnerships are the same as for LLPs so tax is not a factor in the partnership s decision to become an LLP whereas it is a major factor in the decision to incorporate. Unfortunately, we are unable to test the effects of KPMG s incorporation because data for its pre-incorporation period are unavailable as FAME provides data for the most recent ten years only. We therefore drop KPMG from the sample. 21

23 underlying the UK s disclosure requirements is that both private and public companies enjoy limited liability as a result of their incorporation and, in return, they must publicly disclose their accounts. These disclosure requirements allow us to obtain data for every annual audit performed by the audit firms, including both public and private company engagements. The 93 audit firms in our sample conduct a total of 392,215 audits between 1999 and Only 10,459 (2.67%) audits involve publicly traded companies while 381,756 are on private companies. The completeness of the data allows us to accurately measure each audit firm s entire client portfolio (DeFond, 2004; Johnstone and Bedard, 2004). It also allows us to examine how the public-private mix co-varies with the audit firm s organizational form. A handful of audit firms enter or exit the sample over the ten year period of analysis so there are 892 rather than 930 (= 10 x 93) audit firm years. (Our conclusions are unchanged if the sample is instead restricted to the audit firms that have data in every year.) 4.2 Descriptive statistics Table 1 provides descriptive statistics for each of the 93 audit firms. The largest auditor is PricewaterhouseCoopers LLP which conducted 64,733 audits and earned audit fees of 5.0 billion. The next largest auditors are Deloitte LLP ( 2.59 billion), Ernst & Young LLP ( 2.23 billion), Grant Thornton LLP ( 426 million), and BDO Stoy Hayward LLP ( 421 million). The ten largest audit firms all became LLPs, which is consistent with our prediction that the larger firms have stronger incentives to obtain limited liability. The smallest auditor (Ford Bull Watkins & Co, ranked #93) earned 0.2 million in fees, conducted 725 private company audits (none on public companies), and did not become an LLP. [INSERT TABLE 1 NEAR HERE] 22

24 In the portfolios of the five largest audit firms, the proportion of public companies ranges from just 3.10% to 3.89%, reflecting that most audits are conducted on private companies. Surprisingly, the largest auditors do not have the largest values for the public-private mix (%PUBLIC it ) even though they audit the majority of publicly traded companies. As shown in Table 1, the largest value of %PUBLIC it is 13.27% for Jeffreys Henry LLP (ranked #56), followed by 5.99% for RSM Robson Rhodes LLP (ranked # 11), and 5.39% for UHY Hacker Young LLP (ranked #16). These three firms with high exposure to publicly traded clients all became LLPs. Table 2 provides summary statistics for the 93 audit firms. Panel A shows that 36 firms (38.7%) become LLPs while 57 auditors do not. Between 1999 and 2008, the 36 LLP firms conduct 201,365 audits and the 57 non-llp firms conduct 190,850 audits. Panel B summarizes the year of LLP adoption. Just one firm (Ernst & Young) became an LLP in Four firms became LLPs in 2002, seven in 2003, and eight in both 2004 and There are three adopters in 2006, four in 2007, and just one in 2008, suggesting that most firms intending to become LLPs had done so by the end of [INSERT TABLE 2 NEAR HERE] 5. The audit firm s decision to become an LLP 5.1 Univariate results We model the decision to become an LLP by estimating the Cox model in eq. (1). The sample period is from 2001 to 2008 because firms were not allowed to become LLPs before The model is estimated at the level of the audit firm year because the decision to become an LLP is made by the audit firm rather than the audit client. The treatment sample comprises the 36 years in which audit firms become LLPs (N = 36), while the control sample comprises the pre- 23

25 adoption years of these 36 firms plus the years of the 57 auditors that do not become LLPs (N = 550). The post-adoption years of the 36 LLP firms are not used in the Cox model, but they are used in the ex post analysis where we examine the consequences of becoming an LLP. Table 3 reports univariate results for the treatment and control samples. The results for BIGN i and AUDITOR_SIZE it-1 reveal a significant positive association between audit firm size and the likelihood of becoming an LLP. The LLP adopters also have significantly greater exposure to companies that are publicly traded (%PUBLIC it-1 ) and in high-risk industries (%HI_RISK it-1 ) and these auditors charge abnormally high fees (AF_RES it-1 ). These univariate results are consistent with our arguments about auditor size and client litigation risk, but caution should be exercised since we have not controlled for other factors that influence the decision to become an LLP. For example, larger audit firms tend to audit bigger companies and such companies are more likely to be publicly traded. Table 4 reports the pair-wise correlations. The largest correlation (0.392) is between the two auditor size variables (BIGN i and AUDITOR_SIZE it-1 ). (Given this high correlation we do not include these two variables in the same regression.) Large audit firms have portfolios that are more heavily weighted to public companies (the correlation between AUDITOR_SIZE it-1 and %PUBLIC it-1 is 0.264) and there is a positive relation between client size and the proportion of public companies in the auditor s portfolio (the correlation between CLIENT_SIZE it-1 and %PUBLIC it-1 is 0.191). [INSERT TABLES 3 & 4 NEAR HERE] 5.2 Regression results Table 5 presents the regression results for eq. (1). Consistent with the univariate results, the BIGN i and AUDITOR_SIZE it-1 coefficients are positive and highly significant, indicating that the 24