Journal of accounting, auditing finance tập 26, số 02, 2011 4

Accounting Properties of Chinese Family Firms Journal of Accounting, Auditing & Finance 26(4) 623–640 Ó The Author(s) 2011 Reprints and permission: sagepub.com/journalsPermissions.nav DOI: 10.1177/0148558X11409147 http://jaaf.sagepub.com Shujun Ding1, Baozhi Qu2, and Zili Zhuang3 Abstract We posit that family firms in China exhibit accounting properties consistent with the prevalence of Type II agency problems In contrast to the owners of non-family firms, the owners of family firms have more incentives to seek private benefits of control at the expense of minority shareholders and provide lower-quality earnings for self-interested purposes The empirical evidence presented in this study suggests that the accounting earnings of listed Chinese family firms are less informative, and family firms employ less conservative accounting practices than their non-family counterparts We also find that Chinese family firms have higher discretionary accruals compared to non-family firms, which is consistent with the view that family firms engage in more opportunistic reporting behavior Overall, our study suggests that family ownership in China is associated with lower earnings quality, which is in sharp contrast to the findings of prior studies that examine such ownership in the U.S Keyword agency problems, accounting properties, family firms, China This study examines the accounting properties of family firms in China The majority of the world’s firms can be classified as family firms to some extent (Claessens, Djankov, & Lang, 2000; La Porta, Lopez-de-Silanes, & Shleifer, 1999), and such firms thus play a critical role in modern economies Recent studies examining the accounting properties of family firms primarily focus on the United States and offer interesting results Wang (2006), for example, finds that founding family ownership is associated with more informative earnings, more conservative reporting, and lower discretionary accruals Ali, Chen, and Radhakrishnan (2007) analyze the typical agency problems faced by family firms and find University of Ottawa, Ontario, Canada Skolkovo Institute for Emerging Market Studies, China The Chinese University of Hong Kong Corresponding Author: Baozhi Qu, Skolkovo Institute for Emerging Market Studies, Unit 1608, North Star Times Tower, No Beichendong Road, Chaoyang District, Beijing, China, 100101 Email: baozhi_qu@skolkovo.ru Downloaded from jaf.sagepub.com at Taylor's University on December 2, 2012 624 Journal of Accounting, Auditing & Finance that these firms report higher quality earnings These important articles shed light on how family ownership affects accounting properties and disclosure when Type I agency problems, which arise from the separation of ownership and management and thus can be mitigated by family ownership, dominate Type II agency problems, which stem from conflicts between controlling and noncontrolling shareholders Fan and Wong (2002) find listed firms in East Asia to be characterized by less informative accounting earnings This lack of high-quality information disclosure has been said to be responsible, at least in part, for the 1997 Asian financial crisis (Ho & Wong, 2001) Fan and Wong (2002) examine seven East Asian jurisdictions but exclude China However, we believe that examining the interactions among institutional arrangements, family ownership, and accounting properties in China would offer important incremental insights The country’s weak legal system makes it easier for controlling shareholders to expropriate minority shareholders (the Type II agency problem), thus providing us with a good opportunity to investigate whether family firms tend to have different disclosure incentives and hence, exhibit different accounting properties, in an environment in which the Type II agency problem is more pervasive than it is in developed markets such as the United States It is well recognized in the literature that Type II agency problems can lead to the manipulation of accounting earnings by family firms For example, Ali et al (2007) suggest that a variety of incentives arising from these agency problems may lead family firms to manipulate accounting earnings to facilitate private benefit-seeking behavior For instance, these firms may be motivated to manipulate earnings to ‘‘hide the adverse effect of a related party transaction’’ (p 243) Ali et al further point out that family owners usually have a high level of influence over the firm’s board and top management, which is certainly the case in China Hence, they are able to manipulate earnings more easily should they choose to so Although legal institutions that are designed to protect the rights of minority shareholders may help to mitigate the differences in Type II agency problems between family and nonfamily firms, given that such institutions are either nonexistent or ineffective in China (Allen, Qian, & Qian, 2005), family firms are expected to be subject to more severe Type II agency problems and, accordingly, to have poorer earnings quality Using a sample of all listed nonstate firms in China from 2003 to 2006, we first examine the informativeness of accounting earnings and find the reported earnings of family firms to be less informative than those of nonfamily firms.1 We then use a piecewise serial dependence model to test the relationship between family firms and the persistence of the transitory loss components in earnings, which measures accounting conservatism The family firms in our sample are found to use less conservative accounting practices than their nonfamily counterparts Finally, we examine the relationship between discretionary accruals and family firms and find these firms to have a higher level of such accruals Our results remain robust across different model specifications and to the inclusion of different control variables The findings of this study make several contributions to the literature First, we document the importance of Type II agency problems to financial reporting The effect of agency conflicts on disclosure has been extensively investigated (see Healy & Palepu, 2001, for a review), and recent studies have examined the impact of family ownership on corporate disclosure (e.g., Ali et al., 2007; Chen, Chen, & Cheng, 2008; Wang, 2006) However, the aforementioned studies focus on the U.S market, which is characterized by Type I agency problems (e.g., Ali et al., 2007) Our focus on Chinese family firms offers us an opportunity to isolate a setting in which Type II agency problems dominate Our investigation of the accounting properties of these firms thus provides insights that are Downloaded from jaf.sagepub.com at Taylor's University on December 2, 2012 Ding et al 625 complementary to those of previous studies (e.g., Ali et al., 2007; Wang, 2006) examining the way in which agency conflicts affect accounting properties Second, an examination of family firms in China, where privatization took place only recently and the founders of family firms still run their firms directly, enables us to sharpen our tests of the impact of family ownership on financial reporting and disclosure In contrast to Chinese family firms, the majority of those in the United States are entrepreneurial firms The founders of U.S family firms often hire professional managers When these founders retire, their families usually hold only ‘‘marginal ownership’’ (Burkart, Panunzi, & Shleifer, 2003, p 2168) Prior studies that use samples of U.S family firms use either the S&P 500 (e.g., Ali et al., 2007; Wang, 2006) or the S&P 1500 (e.g., Chen et al., 2008), which may raise concerns about the sample (Hutton, 2007) Recent studies have shown that findings involving family firms are ‘‘indeed sensitive’’ to the sample used (Miller, BretonMiller, Lester, & Cannella, 2007, p 831); these authors have discovered that findings based on Fortune 1000 firm data simply cannot be replicated in randomly drawn samples of smaller public companies By no means we suggest that our study uses a noise-free setting, but the early stages of both Chinese family firms and the Chinese stock market may offer a more powerful context for our tests This study also differs from prior studies that examine ownership concentration in East Asia (e.g., Fan & Wong, 2002) and earnings quality in China (e.g., Firth, Fung, & Rui, 2007), none of which investigates the difference between family and nonfamily firms Furthermore, Fan and Wong (2002) did not include China in their sample, and Firth et al (2007) compare earnings quality between the country’s state- and nonstate-owned firms Our study focuses on the impact of family ownership on corporate disclosure and thus adds to this literature The rest of the article proceeds as follows The next section discusses China’s institutional background Section titled ‘‘Relevant Literature and Hypothesis Development’’ reviews the relevant literature and develops our hypothesis Section titled ‘‘Sample and Empirical Analysis’’ discusses our sample and empirical tests, and the last section concludes the article Institutional Background of China China had a centrally planned economy for the three decades following the birth of the People’s Republic of China in 1949 The country’s economic reforms and opening-up policy began in 1978 and initially focused on rural areas In the 1980s, these reforms, which blended central-planning elements with market-based practices, were extended beyond the agricultural sector to state-owned enterprises (SOEs) It was not until 1992 that the Chinese Communist Party formally announced, at its 14th National Congress, that China was adopting a socialist market-based system A significant chapter in the country’s transition to this economic system was the establishment of the Shanghai and Shenzhen Stock Exchanges in the early 1990s, and its capital markets have experienced unprecedented growth since then The majority of listed firms on the country’s stock market are the result of the corporatization of SOEs Typically, an operational unit of a large SOE was carved out, with its net assets converted to nontradable shares at a certain rate The remaining shares were then issued to the public and can be traded As in other countries in which the government keeps a controlling stake in listed (and partially privatized) SOEs, the Chinese central and local governments remain, either directly or indirectly, the controlling shareholders of these Downloaded from jaf.sagepub.com at Taylor's University on December 2, 2012 626 Journal of Accounting, Auditing & Finance firms (Chen, Firth, Gao, & Rui, 2006) According to Chen et al (2006), 30% of the shares in a typical listed SOE are owned by the government or government-related agencies, and another 30% are held by legal entities that are usually controlled by the state The remaining 40% are owned by individuals (including management and employees), private institutions, and foreign investors In non-SOE listed firms, the controlling shareholder may be a family (discussed below) or some other type of nonstate entity (such as a foreign investor, the firm’s employees, etc.) Hence, the Chinese stock market presents two unique ownership features First, although all shareholders have the same rights, there are six types of shares: state, legal entity, foreign, management, employees, and other individuals (Firth, Fung, & Rui, 2007) The shares held by state and legal entities cannot be traded on the market, whereas those owned by individuals are actively traded Second, ownership is highly concentrated The state, and/or a legal entity shareholder, often controls the listed company, and, typically, there are no other block holders (Chen et al., 2006) Of the aforementioned six types of shares, those held by management, employees, and foreign investors usually account for less than 3% (Firth et al., 2007) Relative to SOEs, private companies, including family firms, are a recent product of the country’s economic reforms and opening-up policy The first group of entrepreneurs generally comprised farmers and workers who had been laid off as a result of the SOE reforms It is estimated that around 140,000 such entrepreneurs started up family businesses in the early 1980s The expansion of economic reform has led to the rapid growth of these family firms, which have had a presence in China’s capital market since its inception In the first 10 years of this market’s establishment, the number of listed family firms increased annually by 83.8% (Zhang & Zhang, 2004) In all, 36% of these firms went public through Initial Public Offerings (IPOs), 3% were listed through management buyouts, and the remainder obtained listing status by acquiring existing listed companies (Zhang & Zhang, 2004) Although Chinese family firms have clearly become increasingly significant, surprisingly, to the best of our knowledge, no one in the literature has studied the impact of family ownership on the accounting properties among the nonstate firms in China Relevant Literature and Hypothesis Development Agency Problems and Family Firms Firms face two types of agency problems, both of which have significant implications for the accounting properties of family firms (e.g., Ali et al., 2007; Wang, 2006) The first type, known as Type I agency problems, results from the separation of ownership and control, and may lead managers to act in their own best interests rather than those of the shareholders (Jensen & Meckling, 1976) Type I agency problems are typical in countries in which ownership is diffuse, such as the United States The second type of agency problem, known as Type II, stems from the conflict between controlling and noncontrolling shareholders (Ali et al., 2007), and is common in regions in which the ownership of listed firms is usually concentrated in the hands of a single shareholder, as is generally the case in East Asia (Fan & Wong, 2002) Both types of agency problems result in incentives and disincentives for accounting transparency and corporate disclosure Compared with their nonfamily counterparts, family firms usually exhibit different patterns in both types of agency problems Several factors influence the Type I agency problems in these firms First, family members usually hold positions among top management Downloaded from jaf.sagepub.com at Taylor's University on December 2, 2012 Ding et al 627 or serve on the board of directors and are sometimes directly involved in the firm’s operations As a result, they usually have better knowledge of the daily operations of the firm, which enables them to monitor managers more effectively and reduce opportunistic behavior on the part of the latter (Anderson & Reeb, 2003) Second, family firms are long-term oriented, and thus, the managers of these firms are less likely to seek short-term benefits by manipulating accounting earnings (e.g., Chen et al., 2008) Third, family firms are more sensitive to negative market events, such as litigation (Chen et al., 2008) For all of these reasons, family firms may be less subject to the severe agency problems that often arise from the separation of ownership and control and more likely to disclose higher quality earnings However, Type II agency problems may lead the controlling owners of these firms to engage in opportunistic activities Family owners may use their controlling positions in the firm to expropriate outside shareholders through various channels, such as related-party transactions (Anderson & Reeb, 2003) and freezing out minority shareholders (Gilson & Gordon, 2003), and they may pursue their own interests at the expense of those of noncontrolling shareholders (Ali et al., 2007) Correspondingly, the controlling shareholders of family firms have more incentives to hide relevant information by disclosing lower quality earnings, as such opacity helps them to expropriate outside shareholders The severity of one type of agency problem over the other determines the quality of the information that firms disclose For example, the U.S market is characterized by diffuse ownership, and Type II agency problems are significantly alleviated due to the wellestablished investor protection mechanisms in that country As a result, Type I agency problems tend to dominate Type II in the United States, which means that the family firms there are more likely to disclose higher quality earnings, as family ownership mitigates Type I problems Ali et al (2007) and Wang (2006) provide evidence to support this argument Our study, in contrast, examines a setting in which Type II agency problems are likely to be dominant, thus enriching our understanding of the impact of such agency problems on accounting properties Hypothesis Development Both types of agency problems exist in China, although, as noted, Type II problems are predominant for several reasons First, the existence of dominant shareholders is a typical feature of listed firms in China Second, the country’s investor protection mechanisms are weak, despite the rapid development of its macro-legal environment The Chinese legal system has been heavily influenced by the civil law tradition La Porta, Lopez-de-Silanes, Shleifer, and Vishny (1998) argue that legal protection for shareholders is weakest in countries with a civil law legal origin Furthermore, Allen et al (2005) provide evidence to show that creditor and shareholder protection in China is even worse than that in other major emerging markets The existence of dominant shareholders, in conjunction with weak protection for investors, renders Type II agency problems more salient, which helps to explain why the China Securities Regulatory Commission, the country’s stock market regulator, has repeatedly asserted that its top priority is strengthening minority shareholder protection The governance mechanisms and incentive structure of family firms differ from those of nonfamily firms in several important aspects, all of which have significant implications for Type II agency problems The key difference lies in their different ultimate controllers Unlike family firms, which are controlled by an individual person and his or her family, Downloaded from jaf.sagepub.com at Taylor's University on December 2, 2012 628 Journal of Accounting, Auditing & Finance nonfamily firms are controlled by a group of legal entities, employees, or private investors, including institutional investors The benefits of expropriation for these controlling shareholders, which are typically institutions, organizations, or a group of private investors that are independent of one another, are thus diluted As a result, the large shareholders of nonfamily firms have fewer incentives to expropriate minority shareholders (Villalonga & Amit, 2006) Family owners in China, in contrast, usually have absolute control over their firm’s board and management and are less constrained by its corporate governance system Such a control structure makes it less costly for them to expropriate minority shareholders Furthermore, the private benefits of control are not diluted, as they all go to the family owners Family owners thus have stronger incentives to seek private benefits at the expense of minority shareholders and may have more significant Type II agency problems compared with nonfamily firms Consequently, their disclosures tend to be more opaque for self-interested purposes For example, their accounting may be less informative, and their earnings may be managed to bury the wealth effects (transfers) of their expropriation activities Type I agency problems, in contrast, which arise from the separation of ownership and management, are similar for family and nonfamily firms in China because, on average, both have a concentrated ownership structure Large shareholders are likely to monitor management effectively and are sometimes directly involved in management For example, chairman of the board is an executive position in China This chairman, who is presumed to represent the interests of the controlling shareholder(s), is often directly involved in operations (Chen et al., 2006) In addition, chief executive officers (CEOs) and other top managers are usually appointed by the controlling shareholders (or, in some cases, are actually the founders or their family members) In summary, compared with their counterparts in the United States, Chinese family firms constitute a unique sample that is characterized by a higher degree of Type II agency problems Consequently, we posit that Chinese family firms have different disclosure incentives than such firms in the United States, and we thus put forward the following hypothesis Hypothesis: Family firms in China report lower quality accounting earnings than their nonfamily counterparts Sample and Empirical Analysis This section presents our empirical analyses Our sample covers all non-state listed companies in China from 2003 to 2006 We exclude finance firms, although their inclusion has no quantitative effect on our results General accounting data and data on family firms are obtained from the Guotaian (GTA) databases, which are widely used in accounting and finance research using Chinese data (e.g., Haw, Qi, Wu, & Wu, 2005; Sun & Tong, 2003; Wei, Xie, & Zhang, 2005) In this study, a family firm is defined as a firm that is controlled by a private person and his or her family through direct stock ownership or through a pyramid structure In addition, for a firm to be considered a family firm, the ownership stake of the controlling family owner (the largest shareholder) must be greater than or equal to 10% Given the short history of the Chinese stock market and the rarity of mergers and acquisitions among the country’s listed companies, most, if not all, listed Chinese family firms are still controlled by their founders and their families The nonfamily firms in our data set primarily include the following types of companies: dispersedly held companies with no controlling owner (or family), companies that are controlled by a group of Downloaded from jaf.sagepub.com at Taylor's University on December 2, 2012 Ding et al 629 Table Descriptive Statistics Family firm RET EARNINGS DNI ABS_ACC SIZE BETA MB LEVERAGE LOSS SEO CFO ROA RETVOL MVE Nonfamily firm n M Median n M Median 917 829 870 782 922 841 922 922 922 922 869 923 912 922 0.006 0.007 0.009 0.112 20.764 1.117 2.924 0.585 0.180 0.018 0.056 20.012 0.438 1,553.690 20.062 0.001 0.001 0.067 20.729 1.110 2.239 0.535 0.000 0.000 0.049 0.022 0.396 1,016.750 613 606 566 579 619 592 619 619 619 619 589 619 619 619 20.048*** 0.007 0.010 0.101 20.814 1.091 2.854 0.647*** 0.216* 0.019 0.046** 20.028** 0.391*** 1,498.132 20.091** 0.001 0.001 0.068 20.850 1.100 2.090 0.554** 0.000* 0.000 0.044* 0.018** 0.355*** 1,084.356 Note: RET = cumulative market-adjusted returns over the 12-month period from months before the fiscal year-end to months after it (that is, from May of year t to April 30 of year t 1); EARNINGS = the annual change in net income, deflated by the market value of equity at the beginning of the year; DNI the change in net income, calculated as the net income of year t minus that of year t 1, scaled by the book value of equity at the beginning of year t; ABS_ACC = the absolute value of discretionary accruals (performance-matched discretionary accruals calculated following Ali, Chen, & Radhakrishnan, 2007); SIZE = the natural logarithm of the year-end book value of total assets; BETA = the stock beta at year t; MB = the market-to-book ratio, calculated as the year-end share price divided by the book value of equity per share; LEVERAGE = the leverage ratio of the firm at the end of the year, calculated as the year-end book value of total liability divided by total assets; LOSS = a dummy variable that equals one if net income \ and zero otherwise; SEO = a dummy variable that equals one if the company has seasoned equity offerings and zero otherwise; CFO = cash flow from operations scaled by beginning-of-year total assets; ROA = return on assets, measured by net income divided by average total assets; RETVOL = annual stock volatility calculated using monthly stock returns over the year; and MVE = the market value of equity, in millions of Chinese yuan *** indicates that the mean (or median) value of the variable for family firms is significantly different from that for nonfamily firms at the 1% level; ** indicates a significance level of 5%; and * indicates a significance level of 10% investors (who are not from the same family), companies that are controlled by a group of legal entities (such as not-for-profit organizations, township and village organizations, etc.), companies whose shares are held by employees or their unions, foreign-invested companies, and other nonstate firms that are not family controlled Table presents the descriptive statistics of our sample On average, the family firms are smaller and have a higher level of abnormal returns, lower leverage ratios, greater cash flows from operations, greater profitability, and a higher level of return volatility than the nonfamily firms These characteristics suggest that family firms’ Type I agency problems are at least not as severe as that of nonfamily firms even though both groups of firms have similar level of ownership concentration, a result that is consistent with prior studies (e.g., Ali et al., 2007).2 Differences in accounting properties between family and nonfamily firms are thus likely to be driven by Type II agency problems Finally, at the bottom of Table 1, we can see that both groups of firms have a similar degree of market capitalization Table presents the sample distribution for family and nonfamily firms by year and industry The total number of both types of firms increased steadily from 2003 to 2006 Of the 1,542 observations, approximately 59.8% are family firms, about two thirds of which Downloaded from jaf.sagepub.com at Taylor's University on December 2, 2012 630 Journal of Accounting, Auditing & Finance Table Sample Distribution by Year and Industry Number of firms Year 2003 2004 2005 2006 Total observations Family firm Nonfamily firm Total 99 243 271 309 922 198 134 126 162 620 297 377 397 471 1,542 86 57 100 625 54 922 63 28 96 379 54 620 149 85 196 1,004 108 1,542 Industry code 2: Utilities 3: Properties 4: Conglomerates 5: Industrials 6: Commerce Total observations are industrial firms In the overall sample, the firms are from five different industries, with the majority industrial firms (67.8% for family firms and 61.1% for nonfamily firms) We examine three attributes of accounting earnings in Chinese family firms: earnings informativeness; the persistence of transitory loss components in earnings, which is a measure of conservatism; and discretionary accruals These three attributes have also been examined in studies of the earnings quality of U.S family firms (e.g., Ali et al., 2007; Wang, 2006) and that of East Asian firms in general (Fan & Wong, 2002).3 Informativeness of Accounting Earnings We follow the common practice in the literature to measure the informativeness of accounting earnings (e.g., Ali et al., 2007; Collins & Kothari, 1989) and examine whether those of the family firms in our sample are less informative The primary estimation model is given by the following: RET 5a1b1 EARNINGS1b2 EARNINGS3FAMILY 1b3 EARNINGS3SIZE1 b4 EARNINGS3BETA1b5 EARNINGS3MB1b6 EARNINGS3LEVERAGE1 b7 EARNINGS3RETVOL1INDUSTRY EFFECTS1YEAR EFFECTS1error; ð1Þ where RET represents cumulative market-adjusted returns over the 12-month period from months before the fiscal year-end to months after it (from May of year t to April 30 of year t 1), which includes the earnings announcement period; EARNINGS is the annual change in net income,4 deflated by the market value of equity at the beginning of the year; and FAMILY is a dummy variable that equals one if the firm is a family firm and zero otherwise Following the prior literature (e.g., Ali et al., 2007; Collins & Kothari, 1989), we include the following control variables in our regression models: firm size (SIZE, which is Downloaded from jaf.sagepub.com at Taylor's University on December 2, 2012 Ding et al 631 Table Correlations Among Variables RET Family firms RET 1.000 EARNINGS 0.194* SIZE 0.040 BETA 0.027 MB 0.082* LEVERAGE 20.113* LOSS 20.218* SEO 0.093* RETVOL 0.157* Nonfamily firms RET 1.000 EARNINGS 0.294* SIZE 0.115* BETA 0.043 MB 20.001 LEVERAGE 20.148* LOSS 20.275* SEO 0.003 RETVOL 0.016 EARNINGS SIZE BETA MB LEVERAGE LOSS SEO RETVOL 1.000 20.041 0.094* 0.026 0.026 20.382* 0.083* 0.155* 1.000 20.156* 1.000 20.100* 0.048 1.000 20.202* 0.161* 20.116* 1.000 20.230* 0.126* 20.003 0.417* 1.000 0.153* 20.024 0.010 20.037 20.064 1.000 20.263* 0.182* 0.083* 0.123* 0.070* 20.068* 1.000 1.000 20.050 0.143* 0.001 0.080* 20.328* 0.027 0.062 1.000 20.147* 1.000 20.142* 20.073 1.000 20.288* 0.162* 20.199* 1.000 20.297* 0.122* 20.002 0.471* 1.000 0.098* 20.031 20.015 20.074 20.074 20.250* 0.026 0.026 0.293* 0.308* 1.000 0.006 1.000 Note: RET = cumulative market-adjusted returns over the 12-month period from months before the fiscal yearend to months after it (that is, from May of year t to April 30 of year t 1); EARNINGS = the annual change in net income, deflated by the market value of equity at the beginning of the year; SIZE = the natural logarithm of the year-end book value of total assets; BETA = the stock beta at year t; MB = the market to book ratio, calculated as the year-end share price divided by the book value of equity per share; LEVERAGE = the leverage ratio of the firm at the end of the year, calculated as the year-end book value of total liability divided by total assets; LOSS = a dummy variable that equals one if net income \ and zero otherwise; SEO = a dummy variable that equals one if the company has seasoned equity offerings and zero otherwise; RETVOL = annual stock volatility calculated using monthly stock returns over the year * indicates that the correlation is significant at the 5% level or better the natural logarithm of the year-end book value of total assets); stock beta (BETA); growth potential (MB, which is the market-to-book ratio, calculated as the year-end share price divided by the book value of equity per share); risk of bankruptcy (LEVERAGE, which is the leverage ratio of the firm at the end of the year, calculated as the year-end book value of total liability divided by total assets); and return volatility (RETVOL, which represents annual stock volatility and is calculated using monthly stock returns over the year) Finally, INDUSTRY EFFECTS and YEAR EFFECTS are dummy variables that are included to control for industry and time-fixed effects, respectively To mitigate the undue influence of outliers, the continuous variables used in our analysis are winsorized at the 1st and 99th percentiles.5 Simple correlation analysis (see Table 3) reveals that cumulative market-adjusted returns and earnings are more positively correlated for nonfamily firms than they are for family firms (both correlation coefficients are significant at the 5% level), which is consistent with the supposition that family firms are characterized by less informative earnings Table presents our estimation results Robust standard errors adjusted for clustering and heteroscedasticity are reported in parentheses for all of the coefficient estimates (Petersen, 2009) Regression (1) in Table is conducted to determine whether firm Downloaded from jaf.sagepub.com at Taylor's University on December 2, 2012 684 Journal of Accounting, Auditing & Finance Table Sample Selection Panel A: All Restatements Based on Audit Analytics Number of observations Number of restated company-years with audit fee data during 2003-2005 based on Audit Analytics Less: Restated company-years without necessary COMPUSTAT data for the restated year Final sample 5,908 (2,020) 3,888 Panel B: Tax-Related Restatements Based on Proprietary Data From a Big-Four Accounting Firm Number of observations Number of restated company-years during 2003-2005 for 183 tax-related restatements announced between 1/1/2005 and 3/ 31/2007 Less: (1) Restated company-years without necessary COMPUSTAT data for the restated year (2) Restated company-years without audit fee data on Audit Analytics for the restated year Final sample 307 (40) (8) 259 misapplications of GAAP in the tax area are coded and classified separately This separate file was made available to us To ensure the reliability of this database, we examined a random sample of the database by reading relevant SEC filings Specifically, we randomly chose 16 companies (approximately 10% of the tax-related restatement sample of 150 unique companies) and reviewed their SEC filings (8-k, 10-k, 10-k/a, 10-q, 10q/a, etc.) to determine the restatement causes and dates We detected no errors in either the incidence of restatements or the attribution of the causes of the restatements to the tax area Accordingly, we used this proprietary sample of tax-related restatements in our study From this proprietary sample, we identified 307 tax-related restated company-years during 2003 to 2005 It should be noted that restatements of interim results is coded as a restatement year For example, if a company restates numbers for the first quarter but not the numbers for the fiscal year (because the error was corrected before fiscal year-end), it is still coded as a restatement year for the company We regard restatements of both interim and fiscal year results as reporting failures We removed 48 observations due to missing data on COMPUSTAT and AA Our final tax-related restatement sample contains 259 company-years for 150 distinct companies (Table 2, Panel B) Appendix A depicts examples of disclosures made by such restating companies along with the names of the associated auditors Research Design and Results We first examine the association between auditor-provided tax services and the likelihood of general restatements (H1) We match each restated company-year (sample) with a nonrestated company-year (control) for the same year, from the same two-digit Standard Downloaded from jaf.sagepub.com at Taylor's University on December 2, 2012 Seetharaman et al 685 Industrial Classification (SIC) industry, and with the closest company size (total assets) The control company does not restate its financial statements for any year during the period 2003 to 2005 To test H1, we estimate the following logistic regression: ProbRestate ịit ẳ b0 1b1 APTSit ðorTax Totalit Þ 1b2 Specilaistit 1b3 LNTAit 1b4 MAit 1b5 LNNBSit ð1Þ 1b6 LNGBSit 1b7 LEVit 1b8 BIG4it 1b9 Tenureit 1eit : Restate is an indicator variable that takes a value of one if company i’s financial statements for year t are restated, and zero otherwise We measure our main variable of interest, auditor-provided tax services, in two ways First, we use an indicator variable, APTS, that takes a value of one if company i’s auditor provides tax services to that company in year t and zero otherwise Second, we use the ratio of tax fees paid to the auditor to total fees paid to that auditor by company i in year t, Tax_Total.14 If the provision of tax services to an audit client enables the auditor to gain valuable knowledge and improve financial reporting quality via a knowledge spillover effect, there would be a lower likelihood for restatement and the coefficient b1 on APTS or Tax_Total would be negative In contrast, if the likelihood of a restatement increases with APTS, the coefficient b1 on APTS or Tax_Total would be positive and consistent with an impaired auditor independence explanation In regression (1), we include several control variables that may affect the likelihood of restatements First, Romanus, Maher, and Fleming (2008) find that auditor industry specialization is negatively associated with the likelihood of accounting restatements Accordingly, our model controls for auditor industry specialization by means of the variable Specialist We define company i’s auditor as an industry specialist (Specialist) if the auditor has the largest number of clients in the company’s industry (two-digit SIC) in year t Second, although our matching procedure requires restating companies to be matched as closely as possible with control companies according to total assets, we find that in some cases, the differences in total assets between the restating and control companies are quite large So, we include the natural logarithm of company i’s total assets at the end of year t, LNTA, as an independent variable to control for the possible effect of company size, not already controlled for by the matching procedure, on the likelihood of restatements Third, we include an indicator variable for merger or acquisition activity in year t,15 MA, because either activity may complicate accounting issues and increase the likelihood of restatements (Kinney et al., 2004) Fourth, because multiple business or geographical segments may increase complexity and affect the likelihood of restatements, we include the natural logarithm of company i’s number of business segments (LNNBS) and geographic segments (LNNGS) in year t Logarithms are used because business segments and geographic segments are right skewed (see, for example, Choi, Kim, Liu, & Simunic, 2009) Fifth, highly levered companies may make accounting choices to avoid violating accounting-based debt covenants (e.g., DeFond & Jiambalvo, 1994; Sweeny, 1994) Alternatively, to reduce the agency costs of debt, companies with greater leverage may choose more conservative accounting practices (Frankel & Roychowdhury, 2008; Khan & Watts, 2009) To control for the possible effect of leverage on restatements, we include company i’s leverage (LEV), measured as total debt divided by total assets at the end of year t Next, because audit quality may affect the likelihood of restatements, we include an indicator variable for the Big-Four auditors (BIG4) Finally, DeZoort and Stanley (2007) report a negative Downloaded from jaf.sagepub.com at Taylor's University on December 2, 2012 686 Journal of Accounting, Auditing & Finance Table Variable Definitions Dependent variable Restate Is an indicator variable that equals one if company i’s financial statements for year t are restated, and zero otherwise Tax_Restate Is an indicator variable that equals one if company i’s financial statements for year t are restated for tax reasons, and zero otherwise Primary independent variable (auditor-provided tax services) APTS Is an indicator variable that equals one if company i’s auditor provided tax services in year t, and zero otherwise Tax_Total Is tax fees of company i’s auditor in year t divided by the total fees that the auditor received from company i in year t Other variables Specialist Is an indicator variable that equals one if company i’s auditor has the largest number of clients in the two-digit SIC industry that company i belongs to in year t according to COMPUSTAT LNTA Is the natural logarithm of company i’s total assets (COMPUSTAT #6) at the end of year t MA Is an indicator variable that equals one if company i engaged in merger or acquisition activities in year t as defined by COMPUSTAT (Footnote 1), and zero otherwise LEV Is the leverage of company i at the end of year t, defined as the sum of long-term debt (COMPUSTAT #9) and short-term debt (COMPUSTAT #34) divided by total assets (COMPUSTAT #6) LNNBS Is the natural logarithm of the number of business segments for company i at the end of year t reported on COMPUSTAT LNNGS Is the natural logarithm of the number of geographic segments for company i at the end of year t reported on COMPUSTAT BIG4 Is an indicator variable that equals one if company i’s auditor in year t is one of the Big-Four accounting firms, and zero otherwise TENURE Is the number of years in which company i’s auditor has continuously served as the auditor for company i according to COMPUSTAT Note: SIC = Standard Industrial Classification relationship between auditor tenure and the likelihood of restatements We define auditor tenure (Tenure) as the length of the auditor–client relationship and include Tenure in regression (1) All of these variables are defined in Table 3.16 Descriptive statistics for the general restatement sample are reported in Table 4, Panel A As discussed earlier, there are 3,888 restated company-years in the sample Compared with control company-years that have no restatements, company-years with restatements have a greater likelihood to engage in merger or acquisition activities (MA), a larger number of geographic segments (LNNGS), greater leverage (LEV), a smaller likelihood of using a Big-Four auditor (BIG4), and shorter auditor tenure (TENURE) Panel B of Table reports the Pearson correlations between the variables Consistent with Panel A, Restate is positively correlated with the indicator for merger or acquisition activities (MA), the number of geographic segments (LNNGS), and leverage (LEV) but negatively correlated with the indicator for Big-Four auditor (BIG4) and the auditor tenure (TENURE) We report the estimation results of Equation in Table using 3,888 restated companyyears based on AA data plus 3,888 matched company-years that not have restatements Our main variables of interest are auditor-provided tax services, APTS, an indicator Downloaded from jaf.sagepub.com at Taylor's University on December 2, 2012 687 Downloaded from jaf.sagepub.com at Taylor's University on December 2, 2012 0.7523 0.1108 0.3027 5.4763 0.1870 1.0083 1.0440 0.3150 0.6651 8.1229 APTS Tax_Total Specialist LNTA MA LNNBS LNNGS LEV BIG4 Tenure 0.4317 0.1265 0.4595 2.6245 0.3900 0.4350 0.4657 0.4878 0.4720 7.2968 SD 1.0000 0.0010 0.0000 3.7834 0.0000 0.6931 0.6931 0.0314 0.0000 3.0000 P25 1.0000 0.0679 0.0000 5.6829 0.0000 0.6931 0.6931 0.2002 1.0000 6.0000 Median 1.0000 0.1742 1.0000 7.2180 0.0000 1.3863 1.3863 0.4014 1.0000 11.0000 P75 0.7587 0.1148 0.2883 5.4637 0.1628 0.9926 1.0141 0.2664 0.6867 8.4514 M 0.4279 0.1266 0.4530 2.6122 0.3692 0.4225 0.4450 0.4470 0.4639 7.6984 SD 1.0000 0.0027 0.0000 3.7843 0.0000 0.6931 0.6931 0.0146 0.0000 3.0000 P25 1.0000 0.0769 0.0000 5.6878 0.0000 0.6931 0.6931 0.1632 1.0000 6.0000 Median Control company-years 1.0000 0.1777 1.0000 7.1676 0.0000 1.3863 1.3863 0.3396 1.0000 11.0000 P75 20.0064 20.0040 0.0144 0.0126 0.0242*** 0.0156 0.0300*** 0.0486*** 20.0216** 20.3284* Comparison of M Difference Tax_Total Specialist LNTA MA LNNBS LNNGS LEV BIG4 1.0000 20.0075 1.0000 20.0158 0.5073*** 1.0000 0.0158 20.0401*** 20.0352*** 1.0000 0.0024 0.3000*** 0.1730*** 20.0647*** 1.0000 0.0318*** 0.0232** 0.0012 20.0519*** 0.0683*** 1.0000 0.0183 0.0804*** 0.0292** 20.0744*** 0.2548*** 0.0828*** 1.0000 0.0329*** 0.1619*** 0.1580*** 20.1200*** 0.2111*** 0.1088*** 0.1808*** 1.0000 0.0519*** 20.1045*** 20.0642*** 0.0392*** 20.2438*** 20.0343*** 20.0401*** 20.1223*** 1.0000 20.0231** 0.2615*** 0.1949*** 20.2043*** 0.5399*** 0.0671*** 0.1819*** 0.2310*** 20.1833*** 1.0000 20.0219* 0.1347*** 0.1005*** 20.0203* 0.1480*** 20.0260** 0.1766*** 0.1302*** 20.0349*** 0.1855*** APTS Note: This table presents Pearson correlations between variables for all restatements based on Audit Analytics data and their matched observations ***, **, and * indicate statistical significance at the 1%, the 5%, and the 10% confidence levels, respectively (two-tailed) Variables are defined in Table Restate APTS Tax_Total Specialist LNTA MA LNNBS LNNGS LEV BIG4 Tenure Restate Panel B: Correlation Matrix (Audit Analytics Restatements and Their Matched Observations) 1.0000 Tenure Note: This table presents mean and median of each variable used in empirical analysis and compares these variables between sample company-years and their matched observations Each sample (restated) company-year is matched with a non-restated observation for the same year, from the same two-digit Standard Industrial Classification (SIC) industry, and with the closest company size (total assets) Test for differences in means are based on t statistics (two sided) Variables are defined in Table ***, **, and * indicate statistical significance at the 1%, the 5%, and the 10% confidence levels, respectively (two-tailed) M Variable Restated company-years Panel A: Comparison Between Audit Analytics Restated Company-Years and Control Company-Years Table Descriptive Statistics 688 Journal of Accounting, Auditing & Finance Table Association Between Auditor-Provided Tax Services and the Likelihood of Restatement (Based on Audit Analytics Data): Logistic Models ProbTax Restateịit ẳ b0 1b1 APTSit 1b2 Specilaistit 1b3 LNTAit 1b4 MAit 1b5 LNNBSit 1b6 LNGBSit 1b7 LEVit 1b8 BIG4it 1b9 Tenureit 1eit : Dependent variable = Restate Independent variables Intercept APTS (1) 20.3223*** (13.00) 20.0261 (0.21) Tax_Total Specialist LNTA MA LNNBS LNNGS LEV BIG4 Tenure n Likelihood ratio chi-square Pseudo R2 0.0694 (1.83) 0.0187* (2.95) 0.1517** (6.26) 0.0746 (1.77) 0.1848*** (12.20) 0.2487*** (22.91) 20.1340** (4.95) 20.0065** (4.25) 7,776 55.79 010 (2) 20.3230*** (13.86) 20.2370 (1.63) 0.0696 (1.84) 0.0188* (3.06) 0.1503** (6.14) 0.0722 (1.65) 0.1903*** (12.87) 0.2487*** (22.91) 20.1291** (4.59) 20.0064** (4.10) 7,776 57.21 010 Note: This panel uses all restated company-years during 2003 to 2005 and a matched sample of non-restating company-years based on year, two-digit SIC industry, and size The primary independent variable, auditor-provided tax services, is defined as APTS in column (1) and TAX_TOTAL in column (2) Coefficients and Wald chi-squares (in parentheses) are reported ***, **, and * indicate statistical significance at the 1%, the 5%, and the 10% confidence levels, respectively (two-tailed) Variables are defined in Table variable, and the ratio of tax fees to total fees paid to the auditors, Tax_Total The coefficients on APTS and Tax_Total are not statistically significant, that is, we find no association between auditor provided tax services and the likelihood of restatements.17 The results for control variables are similar to those of the univariate tests in Table The likelihood of restatements increases with firm size (LNTA), the indicator for merger or acquisition activities (MA), the number of geographic segments (LNNGS), and leverage (LEV) but decreases with the indicator for Big-Four auditor (BIG4) and the auditor tenure (TENURE).18 Next, we examine the association between auditor-provided tax services and the likelihood of tax-related restatements (H2) using Equation Downloaded from jaf.sagepub.com at Taylor's University on December 2, 2012 Seetharaman et al 689 Table Comparison Between Company-Years With Tax-Related Restatements and Control Company-Years Restated company-years Variable M SD P25 APTS Tax_Total Specialist LNTA MA LNNBS LNNGS LEV BIG4 Tenure 0.8108 0.1211 0.1776 6.4942 0.2124 1.0945 1.2207 0.2559 0.8571 9.2510 0.3924 0.1365 0.3297 1.9178 0.4098 0.4534 0.5051 0.2603 0.3506 7.9662 1.0000 0.0087 0.0000 5.3803 0.0000 0.6931 0.6931 0.0148 1.0000 3.0000 Median Control company-years P75 1.0000 1.0000 0.0690 0.1851 0.0000 0.0000 6.4001 7.6348 0.0000 0.0000 0.6931 1.6094 1.0986 1.6094 0.1970 0.4083 1.0000 1.0000 7.0000 14.0000 M SD P25 0.8610 0.1443 0.2432 6.4792 0.2664 1.0483 1.1666 0.2263 0.8533 9.2587 0.3506 0.1379 0.3862 1.9093 0.4429 0.4464 0.4857 0.2521 0.3545 7.2742 1.0000 0.0295 0.0000 5.3684 0.0000 0.6931 0.6931 0.0145 1.0000 4.0000 Median P75 Comparison of M Difference 1.0000 1.0000 0.1059 0.2222 0.0000 0.0000 6.3805 7.6245 0.0000 1.0000 0.6931 1.3863 1.0986 1.6094 0.1842 0.3207 1.0000 1.0000 7.0000 11.0000 20.0502 20.0232* 20.0656* 0.0150 20.0540 0.0462 0.0541 0.0296 0.0038 20.0077 Note: This table presents mean and median of each variable used in empirical analysis and compares these variables between sample company-years and their matched observations Each sample (restated) company-year is matched with a non-restated observation for the same year, from the same two-digit SIC industry, and with the closest company size (total assets) Test for differences in means are based on t statistics (two-sided) Variables are defined in Table ***, **, and * indicate statistical significance at the 1%, the 5%, and the 10% confidence levels, respectively (twotailed) ProbTax Restate ịit ẳ b0 1b1 APTSit ðorTax Totalit Þ 1b2 Specilaistit 1b3 LNTAit 1b4 MAit 1b5 LNNBSit ð2Þ 1b6 LNGBSit 1b7 LEVit 1b8 BIG4it 1b9 Tenureit 1eit : The only difference between Equations and is the dependent variable Tax_Restate is an indicator variable that takes a value of one if company i’s financial statements for year t are restated for errors or irregularities in the tax area, as defined earlier, and zero otherwise To estimate the coefficients of Equation 2, our sample consists of 259 tax-related restated company-years from 150 distinct companies (see Table 2, Panel B) We match each restated company-year (sample) with a non-restated company-year (control) for the same year from the same two-digit SIC industry and with the closest company size as measured by total assets The descriptive statistics for the dependent and independent variables of regression (2) are reported in Table We find that the ratio of company i’s tax fees paid to the auditor to total fees paid to that auditor (Tax_Total) is significantly higher for the non-restating, control observations than for the restating sample We also find that the non-restating, control observations are more likely to use industry specialist (Specialist) as the auditor In an unreported correlation matrix, we find that the occurrence of tax-related restatements (Tax_Restate) is significantly negatively associated with Tax_Total and Specialist Next, we examine these relationships in a multivariate setting The results of estimating Equation are reported in Table using 259 restated company-years for tax-related reasons and the matched non-restated control observations The coefficients on APTS and Tax_Total are significantly negative.19 Thus, companies that purchase tax services from their auditors are less likely to have tax-related restatements Downloaded from jaf.sagepub.com at Taylor's University on December 2, 2012 690 Journal of Accounting, Auditing & Finance Table Association Between Auditor-Provided Tax Services and the Likelihood of Tax-Related Restatement: Logistic Models ProbTax Restateịit ẳ b0 1b1 APTSit 1b2 Specilaistit 1b3 LNTAit 1b4 MAit 1b5 LNNBSit 1b6 LNGBSit 1b7 LEVit 1b8 BIG4it 1b9 Tenureit 1eit : Dependent variable = Tax_Restate Independent variables Intercept APTS (1) (2) 20.0569 (0.02) 20.6062** (5.53) 20.2369 (0.33) Tax_Total Specialist LNTA MA LNNBS LNNGS LEV BIG4 Tenure n Likelihood ratio chi-square Pseudo R2 20.3995* (2.93) 0.0094 (0.03) 20.2126 (0.94) 0.1335 (0.43) 0.3854** (3.98) 0.4676 (1.72) 20.0078 (0.00) 20.0082 (0.46) 518 14.48 037 21.2298* (3.3) 20.4100* (3.10) 20.0001 (0.00) 20.2262 (1.06) 0.1243 (0.37) 0.3553* (3.45) 0.4288 (1.44) 20.0571 (0.04) 20.0093 (0.59) 518 12.18 031 Note: This panel uses all restated company-years during 2003 to 2005 where the restatements are tax-related plus a matched sample of non-restating company-years based on year, two-digit SIC industry, and size The primary independent variable, auditor-provided tax services, is defined as APTS in column (1) and TAX_TOTAL in column (2) Coefficients and Wald chi-squares (in parentheses) are reported ***, **, and * indicate statistical significance at the 1%, the 5%, and the 10% confidence levels, respectively (twotailed) Variables are defined in Table than those that not In addition, as the ratio of tax fees paid to auditors to total fees paid to auditors increases, the likelihood of tax-related restatements decreases The odds ratio for the coefficient on APTS is 0.55, indicating that the odds of tax-related restatement for firms with auditor-provided tax services is 0.55 times the odds of tax-related restatement for firms without auditor-provided tax services For the control variables, we find that the likelihood of tax-related restatements is negatively associated with the auditor’s industry expertise (Specialist), which is consistent with Romanus et al (2008) We also find a positive coefficient on the natural logarithm of the number of geographic segments (LNNGS) As the number of geographical segments Downloaded from jaf.sagepub.com at Taylor's University on December 2, 2012 Seetharaman et al 691 increases, there would be corresponding increases in multijurisdictional tax, and perhaps transfer pricing, issues Results indicate that these factors increase the likelihood of taxrelated financial statement errors that are subsequently corrected.20 The coefficients on the other control variables are not significant, probably due to loss of statistical power from the much smaller sample size Summary and Conclusions In this study, we examine the association between auditor-provided NATS and overall financial reporting quality using a sample of public companies in a post-SOX environment We also examine the association between such NATS and financial reporting quality in the tax area Previous research documents a positive association between auditor-provided NATS and financial reporting quality in a pre-SOX environment However, a confluence of regulatory and voluntary initiatives has limited the scope and nature of auditor-provided NATS and may have served to limit or sever knowledge spillovers Simultaneously, the ban on other auditor-provided NAS may have shifted the source of independence concerns to NATS, further diluting the potential for improved financial reporting quality Our study is designed to test this tension between the intended (enhanced financial reporting quality) and the potentially unintended consequences of SOX We measure financial reporting quality by means of appropriately screened financial statement restatements We not find a statistically significant association between the auditor’s provision of NATS and restatements in general Thus, the proposition that the ban on all other NAS shifts the source of independence problems to NATS is not empirically supported Similarly, the proposition that NATS improve overall financial reporting quality via knowledge spillovers is also not supported However, we find a significantly negative association between the auditor’s provision of NATS and tax-related restatements, suggesting that in a post-SOX environment, the benefits of NATS by the auditor manifest themselves in terms of a lower likelihood of tax-related financial statement restatements Downloaded from jaf.sagepub.com at Taylor's University on December 2, 2012 692 Downloaded from jaf.sagepub.com at Taylor's University on December 2, 2012 1/11/2006 (2004) 1/17/2006 (2004) 1/18/2006 (2005) AGILENT TECHNOLOGIES (Auditor) FLOW INTERNATIONAL (Auditor) Date reported (fiscal years restated) AEP INDUSTRIES (Internal) Company (tax service provider) Price water house Coopers Price water house Coopers KPMG Auditor Appendix A Examples of Disclosures by Companies Making Tax-Related Restatements On January 6, 2006, the Company discovered that in determining the value of its deferred tax liability at October 31, 2004, it used an amount for the net tax bases of its fixed assets that exceeded the amount that should have been used for such purposes This error resulted in an understatement of the deferred tax liability related to the depreciation of these fixed assets and the provision for income taxes from continuing operations as of and for the year ended October 31, 2004 by approximately US$1.0 million The restatement reflects the company’s identification of an error related to the accounting for income taxes The company miscalculated and understated its estimate of U.S jurisdictional loss during the preparation of its consolidated financial statements for the fiscal year ended October 31, 2004 The Company improperly excluded the tax provision for Flow Autoclave Systems, Inc., a partially owned consolidated entity, in its calculation of its consolidated tax provision As a result of this error, income tax expense was misstated for the periods referred above Income tax accounting deficiency/details disclosed by the restating company Seetharaman et al 693 Appendix B On the basis of a review of the literature and discussions with tax partners of several national and Big-Four accounting firms, examples of nonaudit tax services (NATS) that many accounting firms routinely provided for their audit clients prior to Sarbanes-Oxley Act of 2002 (SOX) that are either banned by SOX or unlikely to be preapproved (or be preapproved inconsistently) by audit committees include the following:  Tax compliance services (e.g., payroll, sales, property, state income, federal income, and others) that become part of the financial statements through the recording of assets or liabilities  Traditional tax planning services such as analysis of research and development costs, lease versus buy decisions, and so on, where the client uses the auditor’s analysis to prepare and record the appropriate financial accounting entries  Appraisal services undertaken for tax-compliance reasons (e.g., assigning values to intangible assets under Section 197 of the Internal Revenue Code, calculating gains on distributions of assets to shareholders under Section 311, valuing assets and liabilities for a Section 338 election, implementing mark-to-market values under Section 475, and allocating purchase prices under Section 1060) when the company uses the derived values in part for financial statement purposes  Tax-consulting engagements that examine, for example, the efficiency of internal tax departments, procedures used to protest state and local property tax valuations, or state income tax studies with recommendations for redesign  ‘‘Loaning’’ tax staff or supervisors to audit clients  Designing or commenting on the tax aspects of a compensation package—for example, reviewing the applicability of antidiscrimination provisions and the reasonable compensation and incentive compensation provisions of Section 162(m)  Representing the audit client in Internal Revenue Service (IRS) exams, sales tax proceedings, state income tax audits, payroll tax audits, local government property tax proceedings, and the like  Representing the audit client in an administrative dispute with the IRS  Performing expert services for the audit client to the extent such services involve advocacy  Performing transfer pricing studies and other tax-only valuations if they have financial statement implications (e.g., if the transfer pricing study were a consideration in determining the company’s tax reserves)  Preparing or assisting in preparing the tax provisions appearing in the financial statements  Assistance and consultations in connection with the defense of field audits by taxing authorities, administrative proceedings/notices, and appeals These include research, tax computations, and complex interest computations on principal tax liabilities and refunds resulting from income tax filings and audit adjustments  Research, consultations, planning, and report preparation with respect to transfer pricing matters Includes preparation of reports used by the company to comply with taxing authority documentation requirements regarding intercompany transactions  Tax opinions of proposed tax treatments with the threshold of at least ‘‘more likely than not’’ (.50% chance of success)  Global tax consulting and planning on general tax opportunities and challenges, Downloaded from jaf.sagepub.com at Taylor's University on December 2, 2012 694 Journal of Accounting, Auditing & Finance mergers, acquisitions, divestitures, restructurings, other business transactions, supply chain enhancement, tax attribute planning, employee benefit plans and compensation arrangements, and special tax projects/planning/documentation to mitigate the company’s tax burden  Preparation and/or review of required tax return compliance filings on a global basis, including administrative coordination services, and addressing many different forms of taxation such as income, indirect, property, customs, and so on Consultations regarding applicable handling of items for tax returns, required disclosures, elections, and filing positions available to the company.21 Declaration of Conflicting Interests The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/ or publication of this article Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The authors acknowledge the financial support from the John Cook School of Business at Saint Louis University Notes Specifically, Section 201(a) of Sarbanes–Oxley Act of 2002 (SOX) adds new Section 10A(g) to the Securities Exchange Act of 1934 Section 10A(g) prohibits a registered public accounting firm from providing any nonaudit services (NAS), including the following nine categories of NAS, contemporaneously with an audit: (a) bookkeeping or other services related to the accounting records or financial statements of the audit client; (b) financial information systems design and implementation; (c) appraisal or valuation services, fairness opinions, or contribution-in-kind reports; (d) actuarial services; (e) internal audit outsourcing services; (f) management functions or human resources; (g) broker or dealer, investment adviser, or investment banking services; (h) legal services and expert services unrelated to the audit; and (i) any other service that may be determined to be impermissible Final rules adopted by the Securities and Exchange Commission (SEC) to carry out the provisions of SOX define nonaudit tax services (NATS) as all services performed by professional staff in the independent accountant’s tax division except those services related to the audit Typically, it would include tax compliance, tax planning, and tax advice (SEC, Release No 33-8183) Specifically, Section 201 of SOX amends Section 10A(h) of the Securities Exchange Act of 1934 to include the requirement that NATS by the auditing firm must be ‘‘approved in advance’’ by the company’s audit committee; such approval is generally required on a case-by-case basis A restatement occurs when a company, either voluntarily or prompted by auditors or regulators, revises public financial information that was previously reported (Government Accountability Office [GAO], 2002) We assume that such restatements are indications of lower quality financial reporting because the financial statements were not fairly presented in accordance with U.S generally accepted accounting principles (GAAP) We exclude restatements caused by mandated retrospective application of GAAP or by changes in the accounting entity SEC Final Rule S7-13-00, effective for proxy statements filed after February 5, 2001, requires the public disclosure of fees for audit and tax services paid to auditors See Ethics And Independence Rules Concerning Independence, Tax Services And Contingent Fees, Public Company Accounting Oversight Board (PCAOB) Release No 2005-014 (July 26, 2005), and Sections 304, 308, 802, 804, 806, 807, 1105, 1106, and 1107 of SOX Downloaded from jaf.sagepub.com at Taylor's University on December 2, 2012 Seetharaman et al 695 For example, when a tax specialist in the audit firm responds to a question by management about 10 11 12 13 14 15 a complicated tax return position and its financial reporting implications and offers some advice, the question of whether the activity is an audit service, a NATS, or a prohibited service that even preapproval cannot sanction is a difficult one to answer unequivocally Online at http://www.internationaltaxreview.com/Article/2608944/Search/How-Sarbanes-Oxleyis-changing-tax-services.html?Keywords=annual1survey&PartialFields=(CATEGORYIDS%3a10 329)&Brand=ITR For example, one of the respondents in the 2005 survey of Tax Directors in North America cited earlier said, ‘‘Seeking audit committee approval every time we request advice from our statutory auditors has in essence caused us to cease using their services.’’ In related research, Cook, Huston, and Omer (2008) consider whether SOX affects companies’ incentives to manage earnings via changes in effective tax rates (ETR) when auditors provide tax services Their mixed finding is that, for companies that would miss consensus earnings forecasts in the absence of ETR changes, higher tax fees are associated with greater third-to-fourth quarter reductions in ETRs; however, they find this phenomenon both before and after the passage of SOX Direct evidence of legislative concern and interest in this topic is provided by two U.S Government Accountability Office studies on post-SOX restatements, in 2002 and 2006, commissioned by the Committee on Banking, Housing, and Urban Affairs of the U.S Senate Furthermore, Section 408(a) of SOX specifically directs the SEC to periodically review issuer’s financial statements whereas Section 408(b)(1) directs the Commission to use financial statement restatements as one of six criteria for scheduling the reviews required by Subsection (a) In addition, the details of audit and NATS fees disclosed by companies under the threat of expanded criminal and civil penalties for SOX violations, which we use, are probably more accurate than the details of audit and NAS fees obtained via surveys in Kinney, Palmrose, and Scholz (2004) and other similar pre-SOX studies For example, companies may incur costs in locating a new tax services provider and developing a business relationship with that vendor Alternatively, companies may incur costs in setting up or developing their own tax departments In addition, companies may incur costs from not being able to retain their preferred provider of NAS, if that preferred provider is their audit firm The difference in value between a preferred provider and a second choice may be substantial, particularly if the preferred provider has relatively rare service offerings or service offerings that are particularly well suited to the needs of the company Specifically, we read the SEC filings of a random sample of Audit Analytics (AA) restatements that were attributed to the tax area We found that such restatements often include those that were caused primarily by nontax reasons For example, CECO ENVIRONMENTAL CORP’s restatement was primarily due to revenue recognition issues that had a collateral effect on deferred taxes; however, it is classified by AA as a tax-related restatement http://sec.gov/ Archives/edgar/data/3197/000118143105007912/rrd67425.htm Our matched pair research design matches companies by two-digit Standard Industrial Classification (SIC) codes and total assets, and our logistic regression model includes additional variables to control for size and complexity Accordingly, we assume that the level of required tax services is also matched Thus, after controlling for the level of required tax services in the manner described, as the ratio of tax fees paid to auditors to total fees paid to auditors decreases, more of the required tax services is done by third-party vendors or in-house by the tax department Note that unlike the regulatory mandate to disclose tax fees paid to auditors, companies need not disclose the cost of operating their tax departments Hence, the level of required tax services cannot be directly observed As mergers and acquisitions can impact the likelihood of restatements beyond the current year, we alternatively define merger or acquisition activity (MA) over a 3-year horizon, that is, MA Downloaded from jaf.sagepub.com at Taylor's University on December 2, 2012 696 16 17 18 19 20 21 Journal of Accounting, Auditing & Finance equals one if the company has merger and acquisition activity during year t – to year t and zero otherwise The results are similar and not reported To the extent good corporate governance constrains earnings management (e.g., Agrawal & Chadha, 2005; Klein, 2002), corporate governance will be negatively associated with the likelihood of accounting restatements because less managed earnings are less likely to be restated Another way governance could impact the likelihood of restatement is through its influence on the choices of auditors and tax service providers, which in turn influence accounting quality For example, Lassila, Omer, Shelley, and Smith (2009) find a positive association between the strength of corporate governance and the likelihood of retaining auditors as tax services providers, which in turn could have a positive (or negative) impact on audit quality However, we not add governance variables (e.g., board and audit committee characteristics) and acknowledge this limitation because our sample size would be severely restricted after adding these variables Our tax restatement sample will be reduced by more than 70% and have only 72 restated company-years left (plus 72 matched firm-years), that is, too few degrees of freedom to allow meaningful statistical inferences In analysis, we not tabulate; we reestimate the regression using only the first restated year for each restating company plus its corresponding control observation For example, if a company announced in 2006 that it was restating the financial restatements for both 2003 and 2004, only 2003 is included in our estimation The reason for conducting this additional test is that a subsequent year restatement (as opposed to the 1st year restatement) may not be a reflection of poor reporting quality of that year but rather a collateral effect from the previous year’s errors We obtain qualitatively similar results A total of 54 restated company-years in our propriety tax–related restatement data set were not identified as restated observations by AA Adding these 54 observations to the 3,888 restated company-years based on AA and reestimating Equation produces qualitatively similar results In analysis, we not tabulate; we use only the 1st restated year for each restating company during our sample period and get qualitatively similar results: the likelihood of tax-related restatements is negatively associated with auditor-provided tax services, APTS, and the ratio of tax fees to total auditor fees, Tax_Total As an additional validity test, we also examine the economic significance of APTS by measuring market reactions to restatement announcements We find that the average abnormal returns over the days around tax-related restatement announcements for restating companies that have APTS is less negative than the abnormal returns for companies that not have APTS It is likely that restatements by companies that have APTS are less severe than restatements by companies that not have APTS Another possible explanation is that restatements by companies with APTS are perceived by the market as less severe, which is consistent with the notion that the market values auditor-provided tax services Without a reliable measure for the ‘‘severity’’ of the restatement, we cannot distinguish between the two explanations We find no difference in market reactions for companies with APTS versus companies without APTS in the general restatement sample The last seven items (bullet points) were specifically made available to us by a Big-Four accounting firm as examples of services that the firm no longer provides to its audit clients Note that some of these items are not specifically prohibited by SOX References Agrawal, A., & Chadha, S (2005) Corporate governance and accounting scandals Journal of Law and Economics, 48, 371-406 Ainsworth, R T (2004) Sarbanes-Oxley and tax services Sarbanes-Oxley Compliance Journal Retrieved from http://www.s-ox.com/dsp_getFeaturesDetails.cfm?CID=197 Ashbaugh, H., Lafond, R., & Mayhew, B (2003) Do non-audit services compromise auditor independence? Further evidence Accounting Review, 78, 611-639 Downloaded from jaf.sagepub.com at Taylor's University on December 2, 2012 Seetharaman et al 697 Choi, J., Kim, J., Liu, X., & Simunic, D (2009) Cross-listing audit fee premiums: Theory and evidence Accounting Review, 84, 1429-1463 Chung, H., & Kallapur, S (2003) Client importance, non-audit fees, and abnormal accruals Accounting Review, 78, 931-955 Cook, K., Huston, G., & Omer, T (2008) Earnings management through effective tax rates: The effects of tax planning investment and the Sarbanes-Oxley Act of 2002 Contemporary Accounting Research, 25, 447-471 Cripe, B., & McAllister, B (2009) Determinants of audit/tax separation decisions American Journal of Business, 24, 47-56 DeFond, M., & Jiambalvo, J (1994) Debt covenant violation and manipulation of accruals Journal of Accounting and Economics, 17, 145-176 DeZoort, F T., & Stanley, J (2007) Audit firm tenure and financial restatements: An analysis of industry specialization and fee effects Journal of Accounting and Public Policy, 26, 131-159 Durst, M C., & Gibson, T H (2003) Audit vs non-audit tax services under Sarbanes-Oxley Tax Executive, 55, 474-477 Elder, R., Harris, D., & Zhou, J (2008) Tax consulting and reported weaknesses in internal control (Working paper) Syracuse, NY: Syracuse University Frankel, R., Johnson, M., & Nelson, K (2002) The relation between auditors’ fees for non-audit services and earnings management Accounting Review, 77, 71-105 Frankel, R., & Roychowdhury, S (2008) Are all special items equally special? The predictive role of conservatism (Working paper, Washington University) St Louis MO: MIT Press Gleason, C., & Mills, L (2007) Do auditor-provided tax services improve the estimate of tax expense? (McCombs Research Paper Series No ACC-01–07) Retrieved from http://ssrn.com/ abstract=977973 Government Accountability Office (2002) Financial statement restatements: Trends, market impacts, regulatory responses and remaining challenges (GAO 03-138) Retrieved from http://www.gao.gov/products/GAO-03-138 Government Accountability Office (2006) Financial statement restatements: Update of public company trends, market impacts, and regulatory enforcement activities (GAO-06-678) Retrieved from http://www.gao.gov/products/GAO-06-678 Huang, H., Mishra, S., & Raghunandan, K (2007) Types of nonaudit fees and financial reporting quality Auditing: A Journal of Practice and Theory, 26, 133-145 Khan, M., & Watts, R (2009) Estimation and validation of a firm-year measure of conservatism Journal of Accounting and Economics, 48, 132-150 Kinney, W J., Jr., Palmrose, Z.-V., & Scholz, S (2004) Auditor independence, non-audit services, and restatements: Was the U.S government right? Journal of Accounting Research, 42, 561-588 Klein, A (2002) Audit committee, board of director characteristics, and earnings management Journal of Accounting and Economics, 33, 375-400 Krishnan, G., & Visvanathan, G (2008, August 3-6) The association between auditor provided tax services and the use of deferred tax expense to avoid reporting losses (Working Paper, American Accounting Association Annual Meeting) Anaheim, CA Retrieved form http://aaahq.org/ AM2008/abstract.cfm?submissionID=834 Larcker, D., & Richardson, S (2004) Fees paid to audit firms, accrual choices and corporate governance Journal of Accounting Research, 42, 625-658 Lassila, D R., Omer, T., Shelley, M., & Smith, L (2009) Do complexity, independence, and governance influence whether firms retain their auditors for tax services? (Working paper) College Station: Texas A&M University Maydew, E., & Shackelford, D (2006) The changing role of auditors in corporate tax planning (Working paper) Chapel Hill: University of North Carolina Omer, T., Bedard, J., & Falsetta, D (2006) Tax fees paid to auditors: The effects of a changing regulatory environment Accounting Review, 81, 1095-1117 Downloaded from jaf.sagepub.com at Taylor's University on December 2, 2012 698 Journal of Accounting, Auditing & Finance Purcell, T J., & Lifson, D (2003) Tax services after Sarbanes-Oxley: How to cope with an uncertain present Journal of Accountancy, 196, 32-37 Reynolds, K., Deis, D., & Francis, J (2004) Professional service fees and auditor objectivity Auditing: A Journal of Practice and Theory, 23, 29-52 Romanus, R., Maher, J., & Fleming, D (2008) Auditor industry specialization, auditor changes, and accounting restatements Accounting Horizons, 22, 389-413 Scholz, S (2008) The changing nature and consequences of public company financial restatements: 1997-2006 (The U.S Department of the Treasury) Retrieved from http://www.treasury.gov/presscenter/press-releases/documents/financialrestatements_1997_2006.pdf Sweeny, A (1994) Debt covenant violations and managers’ accounting responses Journal of Accounting and Economics, 17, 281-308 Downloaded from jaf.sagepub.com at Taylor's University on December 2, 2012