Audit Pricing and Litigation Risk: The Role of Public Equity. Brad Badertscher University of Notre Dame

Transcription

1 Audit Pricing and Litigation Risk: The Role of Public Equity Brad Badertscher University of Notre Dame Bjorn Jorgensen University of Colorado Boulder Sharon Katz Columbia University William Kinney, Jr. * The University of Texas at Austin April 2012 ABSTRACT: Does de facto litigation risk arising from an extant legal environment affect audit pricing, other things equal? The answer is important for the design of efficient legal regimes and for understanding the economics of audit production, but the effects of financial reporting legal environment elements are difficult to isolate. We hold constant the legal regime, issuer disclosure mandates, and auditor responsibilities, but vary de facto litigation risk by analyzing U.S. companies with publicly-traded debt partitioned into those with public equity and otherwise similar companies having private equity. We find that audit fees of public firms are 17% higher than private firms. Results are consistent with higher litigation risk arising from public equity ownership and the availability of equity prices imposing substantial incremental audit costs. Time-series comparisons for firms that change ownership status yield qualitatively similar results for going private firms, but a significantly larger premium for going public firms. * Corresponding Author: Department of Accounting, McCombs School of Business, The University of Texas at Austin, 1 University Station, Office: CBA 3.226, Austin, TX Phone: (512) Fax: (512) william.kinney@mccombs.utexas.edu. Keywords: public firms, private firms, ownership structure, audit fees, litigation risk JEL classification: M41, M42, M44 Data Availability: Data are available from sources identified in the paper. We are grateful for constructive comments by Paul Chaney (discussant), Dain Donelson, Jim Fuehrmeyer, Katherine Gunny, Menon Krishnagopal, and Mike Wilkins, and seminar participants at the 2011 AAA Annual Meeting, the 2012 AAA Midyear Auditing meeting, and University of Colorado Boulder. We would like to thank Ryan Sheenan for research assistance and the Mendoza College of Business for financial support. All errors are our own.

2 I. INTRODUCTION Exposure of audit firms in the United States to private litigation can lead to substantial payments, damage to audit firms reputations, and even declarations of bankruptcy (Seetharaman et al. 2002, DeFond and Francis 2005). Private litigation can arise from audits of companies with publicly or privately traded securities as well from non-audit services. 1 Audit-related private litigation may allege violations of federal and state securities acts, contract law, or tort law, and may involve class actions and punitive damages. Payments made by audit firms to plaintiffs are sizable and claims for damages are several times the likely capitalization of audit firms. According to a recent report by the Center for Audit Quality (CAQ 2008), between 1996 and 2007, private litigation payments by U.S. audit firms for 362 cases totaled $5.66 billion, and 147 claims for damages of $100 million or more for the same time period can be conservatively aggregated at $201 billion. 2 Exposure to litigation risk varies by type of professional service with 78% of damage payments related to audits of public firms, 16% to audits of private firms (including firms with private equity and public debt), and 6% to non-audit services. For potential damages, 94% of the aggregate is attributable to public firm audits, 5% to private firm audits, and 1% to non-audit services. 3 Moreover, prior literature finds that lawsuits are more common and that litigation awards are greater for public firm clients than for private firm clients that are not subject to SEC filing requirements (St. Pierre and Anderson 1984; Lys and Watts 1994). 1 In this paper, the terms public, private, publicly-held, and privately-held refer to whether a firm s equity is publiclytraded on an organized capital market. 2 CAQ summarizes within ranges of values. The conservative summarization uses the lower limit of the range (e.g., the 37 claims between $100 million and $250 million alleging public company audit defects are valued at $100 million each). 3 According the Stanford Law School Securities Class Action Clearinghouse ( which maintains an index of filings of 3,280 issuers that have been named in federal class action securities fraud lawsuits since passage of the Private Securities Litigation Reform Act of 1995, 96% of all federal class action fraud lawsuits were filed against issuers with publicly traded equity. Moreover, similar to firms with public equity, firms with publicly-traded debt and their auditors are subject to federal class action lawsuits (e.g., Stanford securities class action clearinghouse, Nyack Hospital, Docket number: 02- CV-03283). We include a post securities class action lawsuit variable (POST_LS) in our regressions to control for the potential of increased audit fees due to securities class action lawsuits. 1

3 The concentration of realized damage payments and potential damages in audits of companies that are publically traded raises the question of whether the cost of audits reflect the differing litigation risks through either additional audit effort or a risk premium. The answer is important for understanding the economics of audit services by scholars as well as by audit firms, legislators and regulators interested in mechanism design for delivery of assurance services to protect investors and other users such as analysts and those charged with governance. Several prior studies have considered the direct effect of audit litigation risk on audit fees. Simunic and Stein (1996) point out that U.S. empirical evidence is consistent with audit firms increasing audit fees in the face of higher litigation risk, but the evidence is not strong. A possible reason is that prior U.S. studies have primarily analyzed data from firms with publicly-traded equity. 4 As a result, Simunic and Stein (1996) state that the effects of ownership structure, auditor effort, and risk premium, need to be disentangled before it is possible to conclude whether or not audits are properly priced in the face of higher litigation. More recent research that tests the association between audit fees and litigation risk, typically utilizes changes in a legal regime setting that are simultaneously subject to changes in auditor s responsibilities and hence can affect audit fees beyond the effect of litigation risk changes (Venkataraman et al. 2008). Other studies examine audit fees and litigation risk across different legal environments in cross-country settings (Seetharaman et al. 2002, Choi et al. 2009). However, other factors besides litigation risk, such as cross-country differences in audit competition levels, the relative bargaining power of auditors and clients in those markets, and audit quality levels could explain audit pricing differences in an international setting (Sankaraguruswamy and Whisenant 2009). Furthermore, the extent to which the results based on samples consisting of foreign companies can be generalized to U.S. firms is not clear due to differences in these countries reporting regimes (Givoly et al. 2010). 4 An exception is Palmrose (1986). 2

4 In this study, we analyze the audit fees of U.S.-domiciled firms with publicly-traded debt. 5 Due to their publicly-traded debt, all firms in our sample are subjected to identical reporting regulations and are therefore required to file financial statements with the SEC in accordance with Sections 13 and 15(d) of the Securities Exchange Act of 1934 (see further discussion in Katz 2009 and Givoly et al. 2010). 6 Specifically, we compare private equity and public debt firms that never transition to having public equity (i.e., non-transition firms) with otherwise similar firms having public equity and public debt. Thus, our research design allows us to examine changes in audit fees across different ownership structures and hence exposure to litigation risk, while holding other firm characteristics including the existence of public debt, same legal environment, audit requirements, and the filing of financial reports with the SEC constant across public and private firms. In addition to analyzing fees of private firms relative to public firms in cross-section, we also estimate the effect of inter-temporal change in ownership structure as firms with public debt and private equity become firms with public debt and public equity and as firms with public debt and public equity become firms with public debt and private equity (i.e., transition firms). This difference-in-difference design that holds firm effects constant allows examination of how audit fees change as the firm transitions between public equity and private equity ownership. Higher litigation risk for public firms relative to the private firms may cause auditors to charge higher audit fees to compensate for litigation risk. As compared to private firms, public firm s market liquidity and ownership dispersion increase (Katz 2009), and hence more equity investors may be harmed from misrepresentation of information in the audited financial statements. Indeed, the 2008 CAQ report suggests that it has become increasingly difficult for audit firms to deal with the potential 5 For ease of exposition, we refer to firms with private equity and public debt as private firms while firms with public equity and public debt are referred to as public firms. 6 In particular, [u]nder the 1934 Act auditors must prove they acted in good faith (i.e., not grossly negligent, fraudulent or constructively fraudulent) (Venkataraman et al. 2008, 1317). 3

5 liability from equity market capitalization declines of audit clients. Specifically, potential damages in securities class action typically begin with a decline in market capitalization associated with the disclosure of facts inconsistent with the alleged misrepresentation, and as a result, the larger market capitalization inevitably produces larger claims. Indeed, leading insurers of audit firms (e.g., Aon Corporation) treat the nature of client ownership status (e.g., public vs. private) as an important factor in determining the amount of risk facing the accounting firm (CAQ 2008). As a result, one would expect audit fees to be higher for public firms than for private firms. Contrary to the evidence that suggests public equity firms may have higher audit fees, Bell et al. (2002) and London Economics (2006) document that despite the fact that audit firms consider public status a business risk, audit firms are not likely to be able to pass on the full cost due to pricing competition. Moreover, audit fees could also be higher for private firms relative to public firms. Private firms represent a potentially greater business risk due to increased levels of debt and higher frequency of losses relative to public firms (O Keefe et al. 1994; Simunic and Stein 1996, Seetharaman et al. 2002; Choi et al. 2009). Furthermore, audited accounting information may play a more important monitoring role for private firms because these firms lack a market base measure of firm value, have less vulnerability to takeovers, and less disclosure of non-accounting information, which can increase the monitoring value and importance of auditing in private firms (Lennox 2005). Furthermore, higher ownership concentration among private firms relatively to public firms (e.g., Katz 2009), could lead to higher litigation risk from large and more sophisticated shareholders. Our results, for both non-transition firms and transition firms, provide consistent evidence that relative to public firms, private firms pay significantly lower audit fees. In particular, our results suggest that on average, relative to non-transition private firms, public firms pay statistically 4

6 significantly higher audit fees by 17%. 7 In addition to our cross-sectional pooled regressions, we also examine those firms that transition in ownership status between public and private. We find that, on average, audit fees statistically significantly increase (decrease) by 88% (16%) as firms become public (private). The latter results are important because it rules out low balling as the competing explanation for our findings. Our results are robust to controlling for the Sarbanes-Oxley Act (SOX 2002) and its section 404, and after controlling for firm characteristics such as leverage, occurrence of losses, size, and audit quality. We further control for auditors effort by including control variables such as the length of the 10-K and utilizing each firm as its own control. In addition, we document that audit fees change around both going public and going private transactions. Specifically, we find that audit fees are higher (lower) subsequent to (prior to) the firm going public (private), which is consistent with a greater liability loss component embedded in audit fees for public firms relative to private firms. Finally, we mitigate the role that self-selection might play by conducting both propensity score matching and Heckman s (1979) sample selection correction procedure (in the robustness section). Taken together, these results imply that the ownership status of an audit client significantly affects audit pricing through audit litigation. Our study makes several contributions to the existing literature. First, by examining a unique sample of privately-held firms with publicly-traded debt that are subject to mandatory SEC filings, we can highlight how the role of public equity and litigation risk directly affects audit pricing, controlling for disclosure and filing requirements, the existence of public debt, and audit effort. Second, private firms are important components of the U.S. economy, constituting 99 percent of U.S. companies (Katz 7 In Table 3, PUBLIC is an indicator variable and the dependent variable is the natural logarithm of audit fees, the calculation of the percentage increase in fees for public firms as compared to private firms is e = Therefore, we conclude that public firms pay 17% higher audit fees relative to private firms, all else equal. 5

7 2009; AICPA 2004), and making up a significant portion of the market for audit services (Chaney et al. 2004; Minnis 2011). Nevertheless, little is known about financial reporting by private firms (Ball and Shivkumar 2005; Allee and Yohn 2009; Cassar 2009, Badertscher et al. 2011) and on the role of auditing in U.S. private equity firms. 8 Third, we are able to quantify the audit fee premium due to increased litigation as suggested by Simunic and Stein (1996). Finally, we test firms that transition from private to public and those that transition from public to private, and are the first to document that audit fees are significantly altered during the period surrounding going-public and going-private transactions. Such findings further corroborate our main finding that ownership status of an audit client affects audit litigation and auditor pricing. The remainder of this paper is organized as follows. Section II discusses the prior literature, while section III describes the research design. Section IV provides a description of the data collection procedures and descriptive statistics. Section V presents the results. Section VI concludes. II. LITERATURE OVERVIEW AND HYPOTHESIS DEVELOPMENT Auditor s Litigation Risk and Audit Fees Simunic (1980) models the expected total audit fees (E(c)) as cq + E(d)E(ϕ), where c is the per unit cost of audit resources, including a markup for a normal profit and q is the quantity of the resources utilized in performing the audit. E(d) is the expected present value of all potential losses resulting from the period s audit financial statements, and E(ϕ) is the probability that the auditor will actually have to pay for the losses. In other words, an auditors cost function consists of a resource component (cq) and an expected liability loss component (E(d)E(ϕ)). Liability losses typically result 8 Hope and Langli (2010) and Hope et al. (2010) test auditor independence and audit effort in various agency settings for private firms in Norway, which are characterized by low litigation environment, and hence not appropriate to directly test for litigation risk; the results are not easily generalized to the U.S. 6

8 in the form of litigation against the audit firm when managers, shareholders, creditors, or third parties attempt to recover losses attributed to irregularities in the audited financial statements. An auditor s assessment of the expected present value of all potential losses (E(d)) as well as the probability that the auditor will actually have to pay for the losses (E(ϕ)) is likely influenced by a number of factors including the size, riskiness, and ownership status of the firm (Simunic 1980). Behavioral research conducted via survey and questionnaires link audit fees to the marginal cost of auditing plus expected losses from litigation, where higher effort increases the cost of performing audits but decreases the expected litigation and insurance cost. Hence, auditors can either increase effort and hence audit quality in defense of likely litigation, or charge additional insurance premium to cover possible future litigation costs (Palmrose 1986; Simon and Francis 1988; Pratt and Stice 1994; Simunic and Stein 1996). Similarly, prior literature documents that audit effort increases with the assessment of inherent business risk (O Keefe, Simunic, and Stein 1994; Bell et al. 2001). Despite the theoretical and behavior research linking audit fees and litigation risk, Simunic and Stein s (1996) review paper concludes that while the U.S. empirical evidence is consistent with audit firms increasing their audit fees in face of higher litigation risk, this relation is statistically not very strong. They further document that adjustments are made almost exclusively through higher levels of auditor effort, rather than through a pure price premium. Furthermore, with the exception of Bell et al. (2001), prior empirical research has not had the data to distinguish between increased litigation risk premium and increased auditor effort. However, our sample allows us to keep effort similar across ownership phases, and hence, to better capture the litigation risk premium effect on audit fees. More importantly, and unlike prior research, we are able to hold the firm and SEC filing requirements constant across different ownership phases, increasing our ability to control for auditor effort, complexity of the audit, and firm-specific characteristics. 7

9 Recent studies conduct cross-country research to identify the effect of different research regimes on audit fees and effort. Seetharaman et al. (2002) report higher audit fees paid by U.K. firms that are cross-listed in the U.S., consistent with the increased liability and litigation costs in the U.S. Choi et al. (2009) extend this research to 14 countries and document that auditors charge higher fees for firms that are cross-listed in countries with stronger legal regimes than they do for non-cross-listed firms. In addition they find that the cross-listing audit fee premium increases with the differences in the strength of legal regime. Choi et al. (2009) interpret the results as supporting the theory that audit fee premiums are associated with increased legal liability and not with increased audit complexity. However, Skinner and Srinivasan (2010) document that while audit fees in Japan are only a fraction of those in the U.S., Japanese firms cross-listed in the U.S. pay significantly higher audit fees, even higher than those of the largest U.S. firms. They interpret the results as supporting the idea that auditing in Japan is generally lower quality than in the U.S., and further emphasize that the higher audit fees in the U.S. may reflect in part the litigation-driven insurance role of auditors in the U.S. However, they conjecture that this is unlikely to explain the entire difference and suggest that litigation-driven insurance is likely to provide additional incentives for auditors to deliver higherquality audits. As pointed out by Sankaraguruswamy and Whisenant (2009), cross-country research is subject to additional limitations. Specifically, they suggest a number of important factors that could explain initial audit pricing differences in an international setting. These factors include cross-country differences in audit competition levels, the relative bargaining power of auditors and clients in those markets, the level of information available about incumbent auditors costs, and switching costs borne by clients. Furthermore, the extent to which the results based on samples consisting of foreign 8

10 companies generalize to U.S. firms is not clear due to differences in these countries reporting regimes (Givoly et al. 2010). Lastly, Venkataraman et al. (2008) utilize a litigation regime change during initial public offering (IPO) to examine the relationship between auditor exposure to legal liability, audit quality, and audit fees. They utilize 142 IPOs between 2000 and 2002 to compare the financial and associated audit fees for the two years prior to the IPO ( pre-ipo ), in accordance to prospectus filings, which are subject to the Securities Act of 1933, to the year of IPO ( post-ipo ), in which firms report filings under the Securities Exchange Act Venkataraman et al. (2008) find that auditors provide higher quality (measured by lower abnormal accruals) and receive higher audit fees for the pre-ipo audits under the Securities Act of 1933 due to their exposure to higher litigation risk. The authors acknowledge that auditors responsibilities are higher for IPO engagements and hence can lead to higher fees, but proxy for the portion associated with the litigation risk by utilizing the existence of audit committee, the change in auditors, the S-1 filings, the number of amendments to S-1 registration, and the log of gross proceeds. Our study complements and extends Venkataraman et al. (2008) by examining a sample of firms that have identical reporting requirements across two different ownership structures. 9 Since Venkataraman et al. s (2008) results focus on the period of change in regulation regime, their results cannot easily be generalized to the comparison of private versus public firms, who are subject to identical reporting regulation regimes. Indeed, while Venkataraman et al. (2008) find that pre-ipo firms pay higher fees and enjoy higher audit quality than post-ipo firms, we find that pre-ipo (and 9 Other related studies focus on specific attributes of the IPO process. Beatty (1993), for a sample of 1,191 IPO firms between 1982 and 1984, reports a positive relation between non-underwriting expenses, as a proxy for audit fees, and three ex-post measures of litigation risk: delisting, bankruptcy and lawsuits. Willenborg (1999) shows that IPO audit fees increase with the IPO proceeds, since these proceeds stand for the upper limit on damages, and associated with their insurance role. 9

11 private) firms pay significantly lower audit fees and enjoy higher audit quality than post-ipo (and public) firms. Using questionnaires, Palmrose (1986) examined a sample of 361 audits between 1980 and 1981 where 25% of the firms were privately held and not required to file SEC documents. Palmrose finds that public firms, which are subjected to SEC filing requirements, were associated with higher audit fees. This finding, however, can be also attributed to the different reporting requirements and effort by auditors and is not solely tied to litigation and business risk. We extend this line of research by examining firms across different ownership structures where the SEC filing requirements are held constant to better determine and quantify the impact that ownership structure has on audit fees. The client acceptance decision literature finds that audit firms consider public status as a business risk of auditor reputation loss (Palmrose 1986; Johnstone and Bedard 2003, 2004; Bell et al. 2002). The role of auditors and therefore the pricing of audit fees in private firms may also differ from public firms. For instance, differences in agency conflicts stemming from more concentrated ownership and controlling shareholders may reduce auditors independence in private firms (Hope and Langli 2010; Coffee 2001, 2005) and may further reduce auditor effort and audit fees (Hope et al. 2010). Moreover, Chaney et al. (2004) investigated audit pricing among U.K. private firms and document that private firms choose the lowest-cost auditor available. Further, they found that Big 5 audit firms are not able to charge a premium to private firms. 10 In contrast, lower vulnerability to takeovers and less disclosure of non-accounting information can increase the monitoring value and importance of auditing in private firms (Lennox 2005). Given the prior research, it is not clear whether and to what extent the role of public equity plays in the pricing of audit fees. 10 Both the Norway and the U.K. settings are characterized by much lower litigation environment and hence are less generalizable to the U.S. (Hope and Langli 2010; Hope et al. 2010; Seetharaman et al. 2002). 10

12 III. RESEARCH DESIGN Modeling the Impact of Litigation risk on Audit Fees To investigate whether litigation risk impacts audit fees we first compare private equity and public debt firms that never transition to having public equity, with otherwise similar firms having public equity and public debt. In particular we compare our sample of non-transition private firms to all public firms with public debt as well as to a matched sample of firms with publicly-traded debt and equity. For the matched sample, we use a propensity score match, which is based on industry, year, leverage, abnormal accruals, sales, assets, and foreign operations. This research design enables us to hold constant SEC filings and disclosure requirements, audit requirements, and the audit legal regime, but vary the ownership structure of the audit client, and hence exposure to the audit firms litigation risk. Second, in addition to analyzing fees of private firms relative to public firms in cross-section, we also estimate the effect of inter-temporal change in ownership structure as private firms transition to public firms (private-to-public transactions), and as public firms transition to private firms (public-toprivate transactions). This difference-in-difference design holds firm effects constant and allows examination of how audit fees change as the firm transitions between public equity and private equity ownership or vice versa. Our basic regression specification is the following: FEES i = α 0 + α 1 PUBLIC i + α 2 INVREC i + α 3 ASSETS_GR i + α 4 BIGAUD i + α 5 SEG i + α 6 MNC i + α 7 LIT i + α 8 LOSS i + α 9 LEV i + α 10 ASSETS i + α 11 ABNACC i + α 12 POST_LS i + α 13 AUD_TO i + α 14 WORDCT i + α 15 ACCEL_FILER i + α 16 SOX404 i + α 17 COUNT_WEAK i + α j Σ i YEAR i + α k Σ k INDUS i + ε i, (1) where the dependent variable, FEES, represents the natural logarithm of audit fees (AUDIT), audit fees plus audit-related fees (AUDIT + AUDIT_RELATED), or total fees (TOTAL) (see Venkataraman et al for a similar use of the natural logarithm). Although the main focus of our study is on audit fees, we also examine audit fees plus audit-related fees, and total fees to provide additional evidence on the role that ownership structure and litigation risk has on different types of 11

13 fees paid to the audit firm. We include an indicator variable, PUBLIC, which equals one for public equity and public debt firms, and zero otherwise. If public firms are subject to higher audit fees, then the coefficient on PUBLIC should be positive. See the Appendix for a detailed definition of each variable included in equation (1). We also control for factors that affect audit fees, as documented by prior research, which include both characteristics of the client firms (e.g. business risk, size, audit complexity, and asset structure) as well as characteristics of the audit firm (e.g. Simunic 1980; Palmrose 1986; Simon and Francis 1988; Beatty 1993; Francis et al. 1994; Craswell et al. 1995; Craswell and Francis 1999; Menon and Williams 2001; Seetharaman et al. 2002; Venkataraman et al. 2008; Choi et al 2009; Hribar et al. 2010). Growth in assets (ASSET_GR) is associated with litigation risk due to significant changes in transaction cycles and overburden on client s financial control. Auditor type (BIGAUD) can reflect greater expertise and hence command higher fees. We include a number of control variables for the firm s complexity, including the fraction of assets in inventories and receivables (INVREC), the number of operating business segments (SEG), and indicator variable for foreign operations (MNC). Our final complexity measure is the number of words in the 10-K (WORDCT). Prior research by Loughran and McDonald (2011) suggests that the number of words in a 10-K proxies for the complexity of the firm. Following the aftermath of the WorldCom and Enron failures, Congress enacted the Sarbanes- Oxley Act of 2002 (SOX). SOX requires a management report on internal controls related to financial reporting (section 404a) and can require that the management report be attested by an audit firm (section 404b). Given the documented audit costs of section 404b (Iliev 2010) and to ensure that section 404b requirements are not driving our results, we include an indicator variable (SOX404) if the audit firm was required to attest the internal control report. Prior research also suggests that firms with 12

14 material weaknesses in internal control have complex operations, higher accounting risk, financial stress, and poorer accounting quality (Ashbaugh-Skaife et al. 2007; Doyle et al. 2007). Hence, we further control for the count of material weaknesses identified in the assessment of internal controls (COUNT_WEAK). In addition to controlling for material weaknesses, we also control for whether the firm was an accelerated filer. Loss (LOSS), and leverage (LEV) are proxies for the financial strength of the client firms and client-specific litigation risks borne by the auditors. The litigious industry-based indicator variable (LIT), as in Francis et al. (1994), controls for the premium auditors charge in industries that are more litigious. Similarly, we control for firms that were involved in a class-action lawsuit in year t-1, according to the Stanford class action clearinghouse database (POST_LS). The natural logarithm of total assets (ASSETS) is used as a proxy for the client firm size. We further include year (YEAR) and industry (INDUS) fixed-effects, which have not been tabulated, to control for fundamental differences in audit fees that may exist across years and industries. In addition to the variables included in equation (1), we also include indicator variables for the year of IPO (YEAR_IPO) and the year of going private (FIRSTYEAR_PRV) when examining firms that transition between private equity and public equity ownership. These additional variables capture specific corporate events that increase the litigation risk and effort required by the auditors. Heninger (2001) reports that the risk of auditor litigation is positively associated with abnormal accruals, and similarly, Hribar et al. (2010) expect auditors to charge higher fees to firms with lower quality accounting. In addition, Venkataraman et al. (2008), who use abnormal accruals as a proxy for audit quality, 11 document that an increase in litigation risk increase both the audit fees and the audit quality. Therefore, to the extent which our firms exhibit different financial reporting quality between 11 Discretion in earnings was commonly used in prior literature as a proxy for audit quality (Becker et al. 1998; Francis Krishnan 1999; Frankel et al. 2002; Myers et al. 2003; Sankaraguruswamy and Whisenant 2009). 13

15 ownership phases, we need to control for financial reporting quality. Ball and Shivakumar (2005) and Givoly et al. (2010) find that public firms recognize losses in a more timely manner and engage in more earnings management than private firms. Thus, we control for both timely loss recognition and earnings quality by including ABNACC. ABNACC is the amount of abnormal accruals after controlling for conservatism in our abnormal accruals calculation (see Ball and Shivakumar 2006 and the Appendix for detailed discussions of the computation of ABNACC). Finally, as a result of prior research by Maher et al. (1992) how document that audit fees increase when a firm changes auditors we include an auditor turnover indicator variable for the firms that change auditors (AUD_TO). 12,13 IV. SAMPLE SELECTION Our initial sample consists of private firms that have publicly-traded debt. Because their debt is public, these firms must file financial statements with the SEC, even though their equity is privatelyheld. We follow Katz (2009) and Givoly et al. (2010) and select all firm-year observations on Compustat in any of the ten years from 2000 through 2009 that satisfy the following criteria: (1) the firm s stock price at fiscal year-end is unavailable; (2) the firm has total debt as well as total annual revenues exceeding $1 million; (3) the firm is a domestic company; (4) the firm is not a subsidiary of another public firm; and (5) the firm is not a financial institution or in a regulated industry (SIC codes and ). 12 Eighteen of our private transition firms (27%) changed auditors during our sample period. Of those 13 changed from a Big 4 to another Big 4, two changed from non-big 4 to another non-big 4 and three changed from Big 4 to non-big 4. As a robustness check we removed the 18 firms from our analysis and our results are quantitatively similar to those presented. For the non-transition sample we have 615 audit turnovers, with 77% switching from big 4 to another big 4 and 23% switching from either non-big 4 to big 4 or vice versa. As a robustness check I removed the 615 firms (and a separate analysis by removing only the 144 firms) from our analysis and our results are quantitatively similar to those presented. 13 In an untabulated analysis, following Berkovitch et al (2005) we further control for sales growth, profitability (ROA), volatility of sales, volatility of ROA, tangible assets, R&D intensity, CAPEX, and lag count week, while results remain qualitatively similar the benefit to R-square is marginal. Despite larger number of controls variables we have no unacceptable variance inflation factors. 14

16 To ensure that the sample includes only private firms with public debt, we examine each firm and remove public firm observations, firms lacking required financial statement data, firms involved in bankruptcy proceedings, and foreign firms. We further omit firms that do not have the required audit fees data on Audit Analytics. Details are provided in Table 1. The resulting sample consists of 229 private firms (774 private firm-years) with public debt between 2000 and For the first set of analyses we limit our sample to firms that remain private during the sample period (until mid 2011). This limitation generates a sample of 628 private firms with 162 firm year observations. We compare this sample of private firms to a sample of public firms with public debt and a matched sample of public firms with public debt. Following Givoly et al. (2010), we establish the presence of public debt in a given year based on one of the following indications: (1) the availability of S&P senior debt ranking (Compustat item SPLTICRM), (2) the existence of debt debentures (DD), or (3) the issuance of public debt (according to the Mergent Fixed Income Securities database (FISD) and Thomson Reuter SDC database, prior to the observation year with a maturity date beyond the observation year). Applying the above criteria generates a final sample of distinct public equity firms with publicly-traded debt (11,322 firm-year observations). To mitigate concerns that some characteristics that affect private versus public ownership choice also affect audit fees, we implement a propensity score for the matched sample. In particular, we first match public firms in the same industry (three-digit SIC code) and fiscal year as the private firms. We then create propensity scores that are derived from a PROBIT model which include control variables that significantly differ between the two groups: leverage (LEV), abnormal accruals (ABNACC), size (both total sales and ASSETS), and foreign operations (MNC). We then match each public firm, one-to-one, with a private firm with the closest propensity score, without replacement (for further discussion see Angrist and Pischke 2009). This matching procedure ensures that our results are 15

17 not driven by such differences as growth, size, earnings quality or leverage between public and private firms. As a result of this matching process, both samples include 427 firm-year observations. [PLACE TABLE 1 HERE] For the second set of analyses we limit the sample to firms that transition between private equity and public equity ownership and hence have both private and public ownership phases. This limitation generates a sample of 67 firms with 391 firm-year observations, of which 245 are public and 144 are private. In order to maximize subsample size, when Audit Analytics required data is not available, we hand-collected the audit fees data from proxy statements as well as 10-K filings (if not disclosed in proxy statements). As indicated in Table 1, the final sample consists of 45 firms (152 firm-years) that transition from public to private and 22 firms (237 firm-years) that transition from private to public. V. DESCRIPTIVE STATISTICS AND REGRESSION RESULTS Descriptive Statistics for Public and Private Firm-Years Table 2, Panel A presents descriptive statistics for our sample of private firms with public debt, that remain private, as compared to all public equity firms with pubic debt. 14 Both means and medians of AUDIT (AUDIT_S) are statistically smaller for private firms as compared to public firm. Where AUDIT is the natural log of audit fees and AUDIT_S is the audit fees converted into millions of dollars and scaled by total assets. The AUDIT_S results indicate that private firm exhibit significantly lower audit fees relative to public firm by 33% ( / ) on average. We find no statistical difference in audit-related fees between public and private firms but we do find evidence that the mean amount of total fees is higher for private firms. 14 We winsorize all continuous variables included in the regressions at the 1st and 99th percentiles. We further adjusted the t- stats to control for the clustering by year and multiple firm observations. 16

18 The statistics for leverage indicate that private firm have significantly higher leverage ratios (mean LEV is 0.70 vs. 0.31) than public firm, and private firms are more likely to report net current year losses (LOSS). Private firms exhibit significantly lower total assets, number of business segments, and foreign operations (ASSET, SEG, MNC) compared to public firms. Private firms also have lower word count on the 10-K filings (WORDCT). Fewer private firms are operated in litigious industries (LIT), are audited by big 4 auditors (91% and 93% for private firms and public firms, respectively), and were involved in class-lawsuit action (POST_LS), but more private firms change auditors (AUD_TO). Fewer private firms are subject to the internal control audit requirements of section 404 of SOX (4% versus 53%), report internal control weaknesses (COUNT_WEAK), and are accelerated filers (ACCEL_FILER). Finally, Panel A suggests that private firms exhibit lower abnormal accruals (ABNACC) relative to public firms, which is consistent with results in Givoly et al. (2010). In sum, the results in Table 2, Panel A suggest the importance of controlling for client and audit firm characteristics, as well as for other factors that may affect litigation risk when comparing the audit fees of public and private firms. [PLACE TABLE 2 HERE] Table 2, Panel B presents the Pearson and Spearman correlation tables for our variables of interest. Consistent with public firms having higher audit fees, both the Pearson and Spearman correlations between an indicator variable for a public firm-year (PUBLIC) and AUDIT are positive and significant. In addition, PUBLIC is positively related to TOTAL fees, multinational corporations (MNC), number of operating business segments (SEG), fewer loss years (LOSS), lower leverage (LEV), higher assets (ASSETS), abnormal accruals (ABNACC), involvement in class actions suits (POST_LS), greater number of words in their 10K (WORDCT), more likely to be an accelerated filer 17

19 (ACCEL_FILER), lower auditor turnover (AUD_TO), and more likely to be subjected to section 404 of SOX (SOX404). Results for Comparison of Private Firms and Public Firm-Years We present the regression results for tests of audit fees among public and private firms in Table 3. The first column of Table 3 examines the impact that litigation risk and ownership structure has on audit fees (AUDIT). The results indicate that public firms exhibit higher audit fees than private firms, after controlling for important client firm and audit firm characteristics. The coefficient on PUBLIC is and significant (t-statistic = 5.207), consistent with audit fees being 17% higher audit fees due to the higher litigation risk. 15 Similarly, AUDIT + AUDIT_RELATED fees and TOTAL fees are significantly higher (0.113; t-statistic = and 0.171; t-statistic = 5.683) for the public years. Most control variables are also significant. The coefficient on INVREC is significantly positive indicating that the higher the fraction of assets in inventories and receivables, and hence the firm s complexity, the higher the audit fees. As expected, the coefficients on client firm size (ASSET), foreign operations (MNC), number of operating business segments (SEG), and word count (WORDCT), which proxy for required effort by the audit firm, are significantly positive. In contrast to our expectations, asset growth (ASSET_GR) is significantly negative. 16 The coefficient on leverage (LEV) is significantly positive, indicating that financial weakness of the client firm and higher litigation risk could lead to higher audit fees. Similarly, losses (LOSS) are positively correlated with audit fees. Increase in litigation risk as capture by class actions suits (POST_LS) and weakness in internal control (COUNT_WEAK) are positively correlated with audit fees. As further expected, big auditors charge higher audit fees (BIGAUD), however, change in auditors (AUD_TO) is negatively associated with audit fees. Consistent with the literature documenting the increase in audit fees due to 15 Since PUBLIC is an indicator variable and the dependent variable is the natural logarithm of audit fees, the calculation of the percentage increase in fees for public firms as compared to private firms is e = Similar results are documented in Hope and Langli (2010) and Hope et al. (2010), but in a lower litigation setting. 18

20 section 404 of SOX (Iliev 2010), the coefficient on SOX404 is positive and significant. Finally abnormal accruals (ABNACC) are significantly negative, and may indicate higher audit quality (Venkataraman et al. 2008). [PLACE TABLE 3 HERE] Private Firms Compared to Propensity Matched Public Firms To mitigate concerns that some characteristics that affect private versus public ownership choice also affect audit fees, we implement a propensity score for the matched sample. In particular we compare the private firms to a sample of publicly-traded companies with public debt in the same industry (three-digit SIC code) and fiscal year. We then generate a single, matched public firm-year observation for each private firm-year by selecting the public firm-year with the closest propensity match score. The univariate results for the private firms relative to the matched sample of publicly-traded firms with public debt can be found in Table 4, Panel A. The results indicate that private firms have lower audit fees and total fees but higher audit-related fees. By construction of the matching process, private firms significantly differ from public firms across fewer firms characteristics, all with lower economic significant, including leverage (LEV, 56% and 48%, respectively). In the matched sample, private firms are audited by more big 4 auditors than public firms (BIGAUD, 92% and 84%, respectively) which would mitigate our findings, subject to less audit turnover (AUD_TO), class action litigation (POST_LS), accelerated filings (ACCEL_FILER), section 404 (SOX404), and report fewer material internal weaknesses (COUNT_WEAK). After controlling for these differences, the regression results found in Table 4, Panel B indicate that public firms have higher audit fees as well as higher AUDIT + AUDIT_RELATED fees relative to a sample of private firms. In addition, the estimated coefficient on PUBLIC for TOTAL fees is significantly and economically positive. 19

21 [PLACE TABLE 4 HERE] Audit Fees around Going-Private and Going-Public Events Prior research documents that organizational changes such as going public and going private can affect earnings quality in the periods surrounding those transactions. For example, Ball and Shivakumar (2008) and Katz (2009) provide evidence that private firms alter their earnings quality several years before public stock exchange listings, in anticipation of increased scrutiny by regulators and public investors. Thus, we examine whether audit firms adjust their fees in the periods surrounding going-private and going-public transactions. Such changes in audit fees could be the consequence of anticipated changes in stock ownership concentration, financial reporting pressures, reliance on more or less debt financing, earnings quality, and overall litigation risk. Indeed, as documented by Venkataraman et al. (2008), in the years prior to the IPO, when firms submit prospectus filings and are subject to the Securities Act of 1933, auditors provide higher audit quality, and receive higher audit fees for their additional responsibilities and higher litigation risk, relative to the IPO year, when firms file reports subject to the Securities Exchange Act of These finding may imply that pre-ipo private firms pay higher audit fees than public firms during the IPO year. In contrast, we expect the changes in audit fees in the periods surrounding the going-private and goingpublic transactions to reflect the broader differences in audit fees that we document for public and private firms. However, we have no predictions for how quickly the changes will occur. Thus, we examine two years before and two years after the going-private and going-public transactions. Figure 1 presents the mean values for the different fees in the two years preceding and the two years after a public firm is taken private. The plots of mean AUDIT fees suggest a substantial decrease in audit fees in the year after the going-private transaction. In contrast, mean TOTAL fees (which are affected by non-audit fees as well) increase in the year preceding the going-private transaction and 20

22 then decrease in both the year of going private and the first year after the going-private transaction. These results are consistent with the decline in audit fees during the private ownership phase, and with an increased need for non-audit services before such going-private transaction. Figure 1 also presents the audit and total fees for a matched sample of public firms that do not transition to private equity ownership. The match is conducted in year t-1 and the matched sample remains constant throughout the time period. The match is based on the same fiscal year, industry (3- digit SIC code), and closest propensity score match, where propensity score match is based on leverage, abnormal accruals, sales, assets, and foreign operations. The average percentage change in audit fees from year t-1 to t+1 for the transition sample is -30.3% while the average change in audit fees over the sample period for the non-transition matched sample is 15.3%. The difference in percentage change in audit fees is -45.6% which is statistically significant at the 1% level (t-stat = -4.15). The results provide additional support that public equity ownership significantly impacts audit fees. [PLACE FIGURE 1 HERE] Figure 2 presents similar charts for our measures of fees; however, this figure is centered on the two years preceding and the two years after a private firm goes public. The plots of mean AUDIT fees suggest a substantial and monotonic increase in audit fees beginning one year before the IPO until one year after the IPO. These results are consistent with the increase in audit fees during the public ownership phase. 17 In contrast, mean TOTAL fees increase in the year preceding the going-public transaction and starts to decline to the pre-ipo level in the year after the IPO, consistent with an 17 These results, however, are inconsistent with the findings of Venkataraman et al. (2008), who document that the audit fees in the two years prior to the IPO are higher than the fees in the year of IPO. While Venkataraman et al. (2008) focus on pure private firms that go public through an IPO without prior SEC filings; private firms in our sample already submitted audited financial reports with the SEC prior to the IPO. Therefore, we can expect lower auditor effort and exposure to litigation risk in our setting were private firms previously submitted audited SEC filings. 21

23 increased need for non-audit-services (e.g. compliance with SOX) as well as tax-fees consultation before such going-public transactions (Badertscher et al. 2010). Analogous to Figure 1, Figure 2 presents the results for a matched sample of private firms that do not transition to public equity ownership (see Figure 2 for a description of the matching procedure). Interestingly, the average percentage change in audit fees from year t-1 to t+1 for the transition sample is 75.4% while the average change in audit fees for the non-transition matched sample is 14.9%. The difference in percentage change is 60.5% which is statistically significant at the 1% level (t-stat = 5.88). Taken together, the patterns of audit fees exhibited in Figures 1 and 2 are consistent with an increase in litigation risk during the public phase, with systematically increasing audit fees charged by audit firms. [PLACE FIGURE 2 HERE] Results for Firms that Transition between Public and Private Equity Table 5, Panel A, reports the descriptive statistics for the firms that transition from public to private ownership. The results for Table 5, Panel A are qualitative similar to those reported in Tables 2 and 4. While the difference between the AUDIT fees for public and private years is not as large relative to the full sample in Table 2, Panel A, both the mean and median amounts are statistically lower for the private firm years. Also consistent with the results in Table 2, we find no difference in AUDIT + AUDIT_RELATED fees for the two samples. In contrast, the mean for the total fees no longer significantly differ between the two samples. In addition, private firm-years exhibit significant lower inventory and receivables to total assets, as well as asset growth as compared to public firmyears. Furthermore, losses, leverage, change of auditors (LOSS, LEV, and AUD_TO) are significantly higher for private firm-years. 22