Effects of auditor-provided tax services on book-tax differences and on investors' mispricing of book-tax differences

ADIAC-100434; No of Pages 15 Advances in Accounting xxx (xxxx) xxx

Contents lists available at ScienceDirect

Advances in Accounting journal homepage: www.elsevier.com/locate/adiac

Effects of auditor-provided tax services on book-tax differences and on investors' mispricing of book-tax differences Bing Luo Accounting Department, College of Business, San Francisco State University, 1600 Holloway Avenue, San Francisco, CA 94132, USA

a r t i c l e

i n f o

Article history: Received 31 August 2018 Received in revised form 16 August 2019 Accepted 25 August 2019 Available online xxxx Keywords: Non-audit services Auditor-provided tax services Book-tax differences Limited attention Mispricing

a b s t r a c t Over the last decade, the joint provision of audit and non-audit services has been criticized for compromising auditor independence and affecting audit quality. Since 2005, the SEC has enacted rules restricting the types of nonaudit services audit firms can provide clients. While most non-audit services are prohibited, a range of tax services are still allowed. Therefore, if compromises can emerge from allowing non-audit services, permitting tax services could be problematic. This study investigates the effect of auditor-provided tax services (ATS) on firms' levels of book-tax differences and on investors' mispricing of book-tax differences. Using a propensityscore matched sample from 2000 to 2013, I find strong evidence that firms acquiring ATS exhibit a low level of temporary book-tax differences, which in turn mitigate investors' levels of firms' mispricing. These results do not support regulators' claim that the provision of ATS compromises auditor independence. Instead, it suggests that purchasing ATS can improve overall accounting quality through knowledge spillover and thus help investors better price the value of firms. © 2019 Elsevier Ltd. All rights reserved.

1. Introduction The joint provision of audit and non-audit services has been a controversial issue among regulators and academic researchers. Regulators' main concerns have revolved around the claim that providing non-audit services increases the economic bonds between auditors and clients and potentially impairs audit quality. To address these and other concerns, the Sarbanes-Oxley Act (SOX) of 2002 banned most non-audit services. In addition, in 2005, the Securities and Exchange Commission (SEC) further limited the provision of certain types of auditor-provided tax services (ATS) as suggested by the Public Company Accounting Oversight Board (PCAOB). However, despite the regulators' claims, academic evidence on this issue has been inconclusive. In the specific case of ATS, Omer, Bedard, and Falsetta (2006) show that ATS lead to more lenient financial statement audits, and Maydew and Shackelford (2007) show that ATS increase corporate tax avoidance, both supporting regulators' views and demonstrating a negative relation between ATS and auditor independence. Several other studies find contradictory evidence and suggest that ATS do not relate to impaired earnings quality (Choi, Lee, & Jun, 2009; Kinney, Palmrose, & Scholz, 2004; Krishnan & Visvanathan, 2011) or increased tax avoidance (Krishnan & Visvanathan, 2011). These studies argue that ATS facilitate knowledge spillover and,

E-mail address: [email protected].

therefore, enhance audit quality since the insights gained from the tax services can in turn improve financial-statement audits, and vice versa. Since the relationships between ATS and earnings management or between ATS and tax avoidance remain inconclusive, I first investigate the effect of ATS on firms' levels of book-tax differences. Book-tax differences capture the aggregate level of earnings management and tax avoidance. Book income is reported in a firm's financial statement, and taxable income is reported to the tax authorities. There is usually a discrepancy between the two, and the difference is called the booktax difference. Since Blaylock, Shevlin, and Wilson (2012) state that large temporary book-tax differences are likely to arise from upward earnings management and/or tax avoidance, I further examine whether firms purchasing ATS exhibit different levels of temporary book-tax differences. Second, I investigate the effect of ATS on investors' mispricing of book-tax differences. Chi, Pincus, and Teoh (2014) find that investors typically misprice firms with large temporary book-tax differences. They argue that investors need a considerable amount of cognitive and processing resources to estimate taxable income and future cash flow in order to correctly price temporary book-tax differences. Ordinary investors typically only pay close attention to book income and price firms accordingly, that is, they pay relatively high prices for high book income firms and relatively low prices for low book income firms. However, since large temporary book-tax differences are likely to reverse in future periods and produce less persistent book income (Hanlon, 2005), investors are likely to misprice firms. Accordingly, if ATS influence firms' temporary book-tax differences, the effect of ATS

https://doi.org/10.1016/j.adiac.2019.100434 0882-6110/© 2019 Elsevier Ltd. All rights reserved.

Please cite this article as: B. Luo, Effects of auditor-provided tax services on book-tax differences and on investors' mispricing of boo..., Advances in Accounting, https://doi.org/10.1016/j.adiac.2019.100434

2

B. Luo / Advances in Accounting xxx (xxxx) xxx

on firms' temporary book-tax differences may in turn affect investors' levels of firms' mispricing. Using a propensity-score matched sample design to control for potential self-selection bias associated with a firm's decision to purchase ATS, I find that firms that have purchased ATS have significantly lower levels of temporary book-tax differences than firms that have not. To mitigate concerns about reverse causality, I perform a Granger leadlag test and document that the purchase of ATS leads to lower temporary book-tax differences and not vice versa. Last but not the least, I find that investors' mispricing of book-tax differences is reduced among firms purchasing ATS. This study makes the following contributions. First, it informs the debate on whether ATS lead to compromised auditor independence and increased managerial discretion. My results show that the provision of ATS is associated with smaller temporary book-tax difference, in contrast to regulators' concerns about the potential negative consequences of non-audit services. Instead, my findings support the view that providing ATS enhances audit quality and limits firms' management of book and/or taxable income because knowledge spills over. This empirical evidence can potentially help regulation and policy-making. Second, this study extends the existing literature on the implications of book-tax differences. Research on book-tax differences suggests that the difference between book income and taxable income contains information about firms' current and future valuation (Blaylock et al., 2012; Hanlon, 2005). However, the literature on mispricing of book-tax differences suggests that ordinary investors and even some sophisticated analysts do not value book-tax differences efficiently and misjudge the persistence of temporary book-tax differences (Chi et al., 2014; Weber, 2009). My study extends the line of research on this book-tax anomaly by documenting the effect of ATS on the association between book-tax differences and firms' future stock returns. The remainder of this paper proceeds as follows. In Section2, I discuss the background and review relevant literature on ATS and booktax differences. In Section3, I develop my hypotheses. In Section4, I describe my research design and sample-selection procedure. In Section5, I present the empirical results. Finally, I conclude in Section6.

2. Background and literature review 2.1. Auditor-provided non-audit services and audit quality The provision of non-audit services by auditors has been a controversial issue for the past three decades because of its potential to increase the economic bonds between auditors and clients, which could cause auditors to act more favorably toward their clients and impair audit quality. Over the years, despite the sanctions that U.S. regulators have imposed against non-audit services in an effort to minimize the economic dependency between auditors and clients incurred by such services, the Securities and Exchange Commission (SEC) has continued to allow incumbent auditors to perform a range of tax services such as tax compliance, tax planning, and tax advice to audit clients.1 However, the SEC has required firms to disclose the fees billed by their auditing firm for audit services, audit-related services, tax services, and other services in annual proxy statements, and found that tax fees has represented most non-audit revenues for audit firms (Francis, 2006). Specifically, the tax services provided by auditors typically include assessing estimates of tax reserves by reviewing tax returns, work papers, and IRS correspondence to identify areas of tax risk; evaluating managers' own risk analysis; searching for outside legal opinions; and 1 The following types of ATS have been prohibited by SEC since 2005: contingent-fee arrangements, tax marketing, planning, or advice in favor of tax treatments considered confidential or based on an aggressive interpretation of the tax code; tax services to managers who have financial reporting oversight roles at an audit client; and tax services to the immediate family members of such managers (Laffie, 2006).

conducting tax research to assess the probability of loss (Gleason & Mills, 2011). Given the above, if the client's tax services are provided by the auditor's own tax department, the audit personnel can have easy access to a client firm's tax positons. The widespread controversy over non-audit services, specifically tax services, has attracted substantial research interest among academic researchers. Extensive literature has examined whether non-audit services lead to increased earnings management and/or financial statement restatements, which are both potential outcomes of compromised auditor independence (Habib, 2012; Schneider, Church, & Ely, 2006). However, evidence reported by different studies in this research area is inconclusive, with some studies suggesting that auditor independence is impaired by non-audit services and others suggesting the opposite. One of the earliest studies implying impaired auditor independence by non-audit services was conducted by Frankel, Johnson, and Nelson (2002), who document a positive association between nonaudit fees and the magnitude of absolute discretionary accruals. This result is taken as evidence that auditors are more likely to acquiesce to client pressure when the provision of non-audit services generates economic rents. Consistent with the U.S. evidence reported in Frankel et al. (2002), Ferguson, Seow, and Young (2004) also find that, among UK firms, non-audit services are positively related to discretionary accruals and financial statement restatements. Examining the banking industry, Kanagaretnam, Krishnan, and Lobo (2011) find that small commercial banks paying higher abnormal non-audit fees are involved in higher levels of earnings management. Despite the above evidence, another set of studies generally fail to find evidence relating non-audit services to impaired auditor independence. For example, using the same sample as Frankel et al. (2002), Ashbaugh, LaFond, and Mayhew (2003) fail to find a systematic association between non-audit services and earnings management after adjusting discretionary current accruals for company performance. Similarly, after controlling industry-specific effects in the tests in Frankel et al. (2002), Chung and Kallapur (2003) find no association between non-audit fees and abnormal accruals. Antle, Gordon, Narayanamoorthy, and Zhou (2002) find no evidence of a positive relation between non-audit services and abnormal accruals for UK firms. Further, examining the association between non-audit services and financial statement restatement in the United States, Raghunandan, Read, and Whisenant (2003) and Agrawal and Chadha (2005) fail to find a significant relation between restatements and non-audit services. Moreover, many studies document a negative association between non-audit services and earnings management and attribute this negative association to knowledge spillover, explaining that client-specific knowledge gained from non-audit services can, in turn, improve audit quality (Antle, Gordon, Narayanamoorthy, & Zhou, 2006; Arrunada, 1999; Beck, Frecka, & Solomon, 1988; Simunic, 1984). For example, Beck et al. (1988) split non-audit services into recurring and non-recurring services and suggest that recurring non-audit services provide knowledge spillover and reduce the threat to independence. Prior literature investigating the effect of ATS on auditor independence and audit quality also provides mixed results. Omer et al. (2006) find empirical evidence that ATS lead to more lenient financial statement audits, and Maydew and Shackelford (2007) claim that ATS increase corporate tax avoidance. Both studies expose a negative relation between ATS and auditor independence. In contrast, several other studies support the knowledge spillover hypothesis and indicate that ATS are related to less earnings management (Choi et al., 2009), lower likelihood of financial statement restatements (Kinney et al., 2004) and tax-related restatements (Seetharaman, Sun, & Wang, 2011), and lower likelihood of reporting small profits to avoid earnings losses (Krishnan & Visvanathan, 2011). In addition, Krishnan and Visvanathan (2011) provide empirical evidence that ATS do not contribute to corporate tax avoidance, which contradicts Maydew and Shackelford's prediction (2007).

Please cite this article as: B. Luo, Effects of auditor-provided tax services on book-tax differences and on investors' mispricing of boo..., Advances in Accounting, https://doi.org/10.1016/j.adiac.2019.100434

B. Luo / Advances in Accounting xxx (xxxx) xxx

2.2. Book-tax differences Each year, management calculates two types of income: (1) book income and (2) taxable income. While book income is calculated under GAAP for financial reporting purposes, taxable income is subject to tax laws and reported to the tax authorities. There is usually a discrepancy between the two, and the differences are referred to book-tax differences.2 These differences between book income and taxable income can be either permanent or temporary. Permanent book-tax differences are items included in one measure of income but never included in the other and are generally extremely difficult to measure (Hanlon, 2005). Temporary differences between book income and taxable income arise because of differing requirements of the timing of recognition of income and expense items. For book purposes, GAAP provides managers with a substantial amount of discretion in selecting different accounting methods to record revenue and expense items. Such differences do not necessarily affect book and taxable income in the same period. For example, managers may choose between different rates of cost amortization or select different periods to record reserve allowances to achieve short-term earnings goals (Mills & Newberry, 2001). In contrast, the calculation of taxable income allows less discretion. As a result of these managerial decisions, different timing in recording book and taxable income components may create deferred tax expenses or benefits and generate temporary book-tax differences.3 In general, managers have incentives to manage book income upward without affecting taxable income and to manage taxable income downward without affecting book income. Therefore, when a firm has large book-tax differences, its management may have treated one or both of its income measures opportunistically. When looking at the relation between book-tax differences and earnings management, empirical studies tend to support this view and find systematic associations between book-tax differences and proxies of earnings management and earnings quality. For example, Lev and Nissim (2004) show that a higher ratio of taxable-to-book income (i.e., smaller book-tax differences) is associated with higher future earnings growth. Hanlon (2005) reports a negative association between a firm's future earnings persistence and its level of temporary book-tax differences. Extending Hanlon's (2005) study, Blaylock et al. (2012) find that the lower earnings persistence in firms with large temporary book-tax differences is mainly driven by upward earnings management rather than by tax minimization. Phillips, Pincus, and Rego (2003) provide empirical evidence that large positive temporary book-tax differences are more likely to be experienced in firms that just meet or exceed their earnings target, suggesting that firms manage book income upward to please the market. Furthermore, studies have shown that book-tax differences could be used to identify firms in financial distress (Jones & Noga, 2013), detect earnings fraud (Ettredge, Sun, Lee, & Anandarajan, 2008), and predict firms' earnings restatements (Badertscher, Phillips, Pincus, & Rego, 2009).

2 According to Hanlon and Heitzman (2010), book-tax differences capture only nonconforming tax avoidance, which refers to the tax avoidance transactions accounted for differently for book and tax purposes. Conforming tax avoidance, in which financial accounting income is reduced when the taxable income is reduced, is very hard to measure and has not been well captured in the current literature. 3 Two types of temporary book-tax differences exist: future taxable and future deductible amounts (Hanlon, 2005, p.141). Future taxable amounts create deferred tax liabilities and require the recognition of deferred tax expense, while future deductible amounts create deferred tax assets and thus recognize deferred tax benefits. Yet, prior literature on the mispricing of book-tax differences has not documented whether specific types of temporary book-tax differences are more prone to mispricing (Chi et al., 2014; Lev & Nissim, 2004; Weber, 2009). In this regard, my current study closely follows this line of research and is an important first step in investigating the systematic effect of ATS on temporary book-tax differences overall, as well as the investor's mispricing of temporary book-tax differences. It could be an interesting topic for a potential future study to investigate the level of mispricing associated with each type of temporary book-tax difference and then how ATS can influence such mispricing.

3

Prior empirical evidence also supports the positive association between book-tax differences and tax avoidance. For example, Mills (1998) suggests that large positive book-tax differences are positively related to high levels of tax non-compliance. Several subsequent studies further document that the increase in book-tax differences over the 1990s is partly attributable to increased levels of corporate tax sheltering (Manzon & Plesko, 2002; Mills, Newberry, & Novack, 2003; Plesko, 2004). In addition, Wilson (2009) and Frank, Lynch, and Rego (2009) find that firms identified as engaging in tax sheltering activities have relatively large book-tax differences. Consistent with the U.S. evidence reported in Wilson (2009) and Frank et al. (2009), Cho, Wong, and Wong (2006) suggest a similar finding in New Zealand. 2.3. Limited attention and mispricing of book-tax differences In the financial market literature, the efficient-market hypothesis indicates that security prices should fully reflect all available information. Therefore, the expected return on a security should merely be associated with risk factors (e.g., Fama, 1970). Mispricing conceptually refers to associations between security prices/returns and factors other than risk, thus contradicting the efficient-market hypothesis. Specific to book-tax differences, which provide rich information about a firm's future earnings persistence, a series of studies (e.g., Chi et al., 2014; Lev & Nissim, 2004; Weber, 2009) have demonstrated that ordinary investors, and even some sophisticated analysts, fail to correctly interpret or incorporate the information contained in book-tax differences and, therefore, misprice firms' current and future earnings. These studies document a systematic association between firms' book-tax differences and future stock returns (Lev & Nissim, 2004) that exists even after controlling for risk factors (Chi et al., 2014; Weber, 2009) and is most pronounced in firms with weak information environments, where the level of market inefficiency is likely to be highest (Weber, 2009). Hence, the empirical evidence suggests that the association between temporary book-tax differences and returns reflects mispricing rather than risk. In addition, factors related to accrual anomaly cannot fully explain the mispricing (Chi et al., 2014). Chi et al. (2014) attribute mispricing to investors' limited attention. Limited attention underlines the idea that attention is selective in nature and that such selectivity is “a necessary consequence of the vast amount of information available in the environment, and of limits to information processing power” (Kahneman, 1973). In the face of limited cognitive processing resources, people tend to underweigh abstract, statistical, and base-rate information (Kahneman & Tversky, 1973; Nisbett & Ross, 1980) and rely heavily on simple heuristics in making decisions (Libby, Bloomfield, & Nelson, 2002). Therefore, investors with limited attention may fail to estimate book-tax differences because of their lack of cognitive power to estimate taxable incomes from financial statement data and to extract temporary book-tax differences from total book-tax differences (Chi et al., 2014). Instead, investors will typically price firms only according to book income; this is to say, they will tend to pay relatively high prices for firms with high book income and relatively low prices for firms with low book income. However, if firms have high temporary book-tax differences, their high book income will likely reverse in the future and become less persistent, leading to systematic mispricing. 3. Hypotheses development 3.1. H1: auditor-provided tax services and temporary book-tax difference Prior research examining the effect of ATS on audit quality provides contradictory results. On the one hand, the joint provision of audit services and tax services inevitably generates a higher amount of quasirents, thereby increasing the threat to auditor independence and prompting auditors to yield to pressures from clients. This implies that auditors have less incentive to curb earnings management. In addition,

Please cite this article as: B. Luo, Effects of auditor-provided tax services on book-tax differences and on investors' mispricing of boo..., Advances in Accounting, https://doi.org/10.1016/j.adiac.2019.100434

4

B. Luo / Advances in Accounting xxx (xxxx) xxx

auditors who provide tax services are more likely to accept understatement of taxable income. As pointed out by several participants at the PCAOB's roundtable, one of the main concerns of joint provisions is that if an aggressive recommendation comes from the tax department of an audit firm rather than from an external service provider, it is less likely that the auditor will scrutinize the client's tax position closely or call that recommended position into question (PCAOB, 2004). Accordingly, the auditor's clients may not be required to record large contingency reserves related to those positions, resulting in more tax avoidance. Hence, I expect that if ATS impair auditor independence and lead to more lenient auditing, client firms with ATS are more likely to opportunistically manage book income up and/or taxable income down than their counterparts without ATS. Such increased levels of earnings management and/or tax minimization are likely to be captured by large temporary book-tax differences. Given the above, I expect that if ATS lead to impaired auditor independence, firms receiving ATS will have higher levels of temporary book-tax differences than firms not receiving ATS. On the other hand, the knowledge spillover hypothesis posits that joint provision of audit services and tax services can promote information sharing between the audit side and the tax side of the services, and, therefore, knowledge gained from the financial statement audits can improve the quality of tax services and vice versa. It has been suggested that firms with aggressive financial reporting strategies also tend to be aggressive in tax reporting (Frank et al., 2009). Thus, knowledge of aggressive tax planning strategies can inform the auditor about managers' general attitude toward financial reporting. With such preknowledge, the auditor can better focus audit procedures to detect managers' potential earnings management activities, thereby enhancing audit effectiveness. As indicated by prior studies, improved audit quality can lead to lower management of book income (e.g., Choi et al., 2009; Kinney et al., 2004; Krishnan & Visvanathan, 2011). Further, ATS allow auditors to review their clients' quarterly tax statements so that auditors have many opportunities to examine deferred tax expenses and uncover controversial issues (Ettredge et al., 2008; Krishnan & Visvanathan, 2011). This might help reduce firms' tax minimization and further constrain managers' manipulation of earnings through taxes expenses (Christensen, Olson, & Omer, 2013; Dhaliwal, Gleason, & Mills, 2004). Hence, if upward earnings management and/ or tax minimization are reduced by ATS, I expect temporary book-tax differences to be smaller for firms utilizing ATS than for firms not utilizing ATS. Although Krishnan and Visvanathan (2011) find no evidence suggesting an association between ATS and short-term tax avoidance, it is likely that reduced earnings management will still be sufficient to decrease temporary book-tax differences among firms acquiring ATS. Based on the above, whether ATS will have any effect on temporary book-tax differences is an empirical question. Hence, my first hypothesis is posited in a non-directional form:

pay close attention to book income and price firms accordingly. However, large temporary book-tax differences are likely to reverse in future periods and produce less persistent book income (Hanlon, 2005), therefore leading to mispricing of a firm. If firms receiving ATS have higher levels of temporary book-tax differences than firms not receiving ATS, higher levels of temporary book-tax differences would lead to less persistent book income. Therefore, I expect that investors will have increased difficulty in predicting the future earnings of such firms, leading to increased mispricing. In other words, the mispricing of temporary book-tax differences reflects a negative association between temporary book-tax differences and future returns (Chi et al., 2014), and this negative association will worsen with the acquisition of ATS. Alternatively, if firms acquiring ATS have lower levels of temporary book-tax differences than firms not acquiring ATS, lower levels of temporary book-tax differences would lead to more persistent book income. Therefore, I expect that investors will be better able to predict future earnings, leading to reduced mispricing. In other words, the negative association between temporary book-tax differences and future returns will be attenuated with the purchase of ATS. Given the above, my second hypothesis is presented in a nondirectional form:

H1. : All else being equal, auditor-provided tax services are systematically associated with firms' levels of temporary book-tax differences.

ATSit ¼ α0 þ β1 ðTENUREit Þ þ β2 ðINSTOWNit Þ þ β3 ðMERGERit Þ þ β4 ðAUDINDit Þ þ β5 ðLAFit Þ þ β6 ðOPPORit Þ þ β7 ðSIZEit Þ þ β8 ðABACCit Þ þ β9 ðNOLit Þ þ β10 ðΔNOLit Þ þ β11 ðEQINCit Þ þ β12 ðFIit Þ þ β13 ðR&Dit Þ þ β14 ðSLEVit Þ þ β15 ðBTMit Þ þ β16 ðPPEit Þ þ β17 ðROAit Þ þ β18 ðCASHit Þ þ β19 ðDEPit Þ þ β20 ðBIG4it Þ þ β21 ðSECTIERit Þ þ ΣIND þ ΣYEAR þ εit ð1Þ

3.2. H2: auditor-provided tax services and mispricing of book-tax differences Empirical evidence suggests that the association between temporary book-tax differences and returns reflects mispricing rather than risk (Chi et al., 2014; Weber, 2009). Specifically, temporary book-tax differences provide rich information about a firm's future earnings persistence (Hanlon, 2005). However, estimating and correctly pricing temporary book-tax differences are difficult tasks and require a considerable amount of cognitive resources (Chi et al., 2014). Consistent with this notion, investors with limited attention, and also some sophisticated analysts, tend to misprice the earnings information in temporary book-tax differences (Chi et al., 2014; Weber, 2009). They typically

H2. : All else being equal, auditor-provided tax services are systematically associated with investors' mispricing of book-tax differences. 4. Empirical design 4.1. Propensity-score matched sample Prior research suggests that firms purchasing ATS are likely to differ from firms that do not (Krishnan, Visvanathan, & Yu, 2013; Lassila, Omer, Shelley, & Smith, 2010; McGuire, Omer, & Wang, 2012; Omer et al., 2006). As a result, rather than using all non-ATS firms as a control sample, I use a propensity-score matched control sample and compare firms that purchase ATS to this control sample matched on observable firm characteristics shown by prior literature to be associated with the purchase of ATS. This procedure mitigates the concern that observable firm characteristics associated with the choice to purchase ATS drive the differences in the relationship between ATS and firms' level of book-tax differences or between ATS and mispricing of book-tax differences. Following Krishnan et al. (2013), I first estimate the probit model in Eq. (1), which regresses firms' purchase of ATS on the potential determinants of ATS. Because large and complex firms are more likely to purchase ATS, I control for firm size, foreign income, auditor type, leverage, corporate governance, the proportion of shares held by institutions, and other relevant factors.

where: TENURE = Length of the audit firm's tenure with its client; INSTOWN = Percentage of shares owned by institutions at the beginning of the year; MERGER = An indicator variable equal to 1 if a firm participates in any merger activity during the year and 0 otherwise; AUDIND = Auditor independence from the client, measured as nonaudit fees less tax fees divided by total audit fees received from the client; LAF = atural log of audit fees received from the client;

Please cite this article as: B. Luo, Effects of auditor-provided tax services on book-tax differences and on investors' mispricing of boo..., Advances in Accounting, https://doi.org/10.1016/j.adiac.2019.100434

B. Luo / Advances in Accounting xxx (xxxx) xxx

OPPOR = Market value of a client divided by the sum of the market value of all clients in the same industry at the same MSA city; SIZE = Natural log of market value of equity at the beginning of the year; NOL = An indicator variable set to 1 if there is a tax loss carryforward during the year and 0 otherwise; EQINC = Equity income scaled by total assets at the beginning of the year; FI = Pre-tax foreign income scaled by total assets at the beginning of the year; R&D = R&D expense scaled by total assets at the beginning of the year; SLEV = Long-term debt-to-asset ratio at the end of the year scaled by total assets at the end of the year; BTM = Book-to-market ratio at the end of the year, measured as book value of equity divided by market value of equity; PPE = Net PPE scaled by total assets at the beginning of the year; ROA = Return on assets, measured as the ratio of income before extraordinary items to the average of total assets for the year; CASH = Cash holding at the end of the year divided by total assets at the beginning of the year; DEP =Depreciation and amortization expense for the year divided by total assets at the beginning of the year; BIG4 = An indicator variable set to 1 if the firm is audited by one of the Big 4 auditors and 0 otherwise; SECTIER = An indicator variable set to 1 if the firm is audited by a second-tier accounting firm, namely Grant Thornton LLP and BDO Seidman LLP, and 0 otherwise. Consistent with prior research, positive coefficients are expected on TENURE, MERGER, LAF, OPPOR, SIZE, NOL, ΔNOL, FI, SLEV, PPE, ROA, DEP, and BIG4, and negative coefficients are expected on INSTOWN, AUDIND, ABACC, EQINC, R&D, BTM, CASH, and SECTIER. Next, I create a matched sample based on the predicted probabilities from the probit regression model. Table A2.1 in Appendix 2 reports the results of the probit regression model. Overall, the probit regression model is significant as assessed by a likelihood ratio test comparing the full model against a reduced model including only the intercept (p b .001; pseudo R2 = 0.1772). For variables with significant coefficients (AUDIND, LAF, OPPOR, SIZE, NOL, PPE, and BIG4), the signs on the coefficients are consistent with prior literature4 (e.g., Krishnan et al., 2013). This indicates that firms that are larger, have a greater market share in the industry, have higher levels of net operating loss carryforwards, have a lower percentage of plant assets compared to total assets, purchase Big 4 audit services, pay relatively high audit fees, or have relatively independent auditors are more likely to retain their auditors for tax services. The matched sample contains a non-ATS control sample whose predicted probabilities of purchasing ATS are highly similar to the ATS firm sample. The means of the predicted probability of purchasing ATS for the ATS firms and matched control (non-ATS) firms are 0.8290778 and 0.8291303, respectively. The mean difference is 0.0000525 and is not statistically significant (t = 0.0175, p = .9860), indicating that the matching procedure is effective. 4.2. Testing H1: impact of auditor-provided tax services on temporary book-tax differences The following regression model (Eq. (2)) is used to test my first hypothesis on whether firms acquiring tax services from their auditors are associated with higher or lower amounts of temporary book-tax 4 The number of significant variables in my model is generally comparable with those reported in prior research as well. For instance, McGuire et al. (2012) employ the same propensity-score matching model and report 9 of the 21 variables as being significant. Similarly, Krishnan et al. (2013) find 11 of the variables to be significant. Both studies have a threshold at the p = .10 level.

5

differences compared with firms not receiving auditor-provided tax services: TEMPit =Ait−1 ¼ α0 þ β1 ðATSit Þ þ β2 ðABACCit Þ þ β3 ðCash3ETRit Þ þ β4 ðTEMPit−1 =Ait−2 Þ þ β5 ðLEVit Þ þ β6 ðGROWTHit Þ þ εit

ð2Þ

where: TEMP = Temporary book-tax differences; A = Total assets; ATS = An indicator variable that is set to 1 if a firm's tax fee is not equal to zero, and 0 otherwise; ABACC = Abnormal accruals; Cash3ETR = Ratio of the sum of cash taxes paid over the previous 3 years to the sum of pretax financial accounting income over the previous 3 years; LEV = Ratio of total liabilities to total assets; GROWTH = Percentage change in sales. 4.2.1. Dependent variable The dependent variable in Eq. (2) is the temporary book-tax differences for firm i in year t. Following Hanlon (2005), I estimate the temporary book-tax differences (TEMP) by grossing up the deferred tax expense using the top corporate tax rate: TEMP ¼

Deferred Tax Expense tx:

The deferred tax expense in the numerator is the sum of deferred federal and foreign tax expenses. If either of these numbers is missing, I calculate the deferred tax expense as the deferred portion of the total income tax expense (Hanlon, 2005). From 1993 onward, tx equals 35%. Finally, following DeSimone and Stomberg (2012), I scale the temporary book-tax differences for each firm-year sample by the total assets at the beginning of the year. 4.2.2. Independent variable The independent variable, ATS, is an indicator variable denoting whether a firm receives joint audit and tax services from its auditor. ATS is set to 1 if the firm procured ATS during the year as indicated by a non-zero tax fee in Audit Analytics and is set to 0 otherwise. 4.2.3. Control variables I include several control variables in Eq. (2). Prior research finds that large temporary book-tax differences (TEMP) likely arise from upward earnings management and/or tax avoidance (e.g., Blaylock et al., 2012). Therefore, I include abnormal accruals (ABACC) and the cash effective tax rate (Cash3ETR) in the model to control for firms' levels of earnings management and tax avoidance, respectively. As firms' abnormal accruals increase, their inflated book income would likely enlarge the temporary book-tax differences. Thus, a positive relation is predicted between abnormal accruals and temporary book-tax differences. Similarly, if firms engage in tax avoidance and thus pay less tax, which results in a decreased cash effective tax rate, the level of temporary book-tax differences would increase. Hence, a negative relation is expected between the cash effective tax rate and temporary book-tax differences. The abnormal accruals for each firm-year sample are calculated via the modified Jones model. Following Dechow, Sloan, and Sweeney (1995), I measure the level of abnormal accruals as deviations from the predicted accruals in the corresponding industry-year regression (Eq. (3)): TAccit =Ait−1 ¼ α0 =Ait−1 þ β1 ðΔ Rit −ΔARit Þ=Ait−1 þ β2 ðPPEit =Ait−1 Þ þ εit

ð3Þ

where:

Please cite this article as: B. Luo, Effects of auditor-provided tax services on book-tax differences and on investors' mispricing of boo..., Advances in Accounting, https://doi.org/10.1016/j.adiac.2019.100434

6

B. Luo / Advances in Accounting xxx (xxxx) xxx

TAccit /Ait-1 = Total accruals for firm i in year t, scaled by total assets at the beginning of year t for firm i; TAcc (total accruals) is measured in Eq. (4) as follows: TAccit ¼ ΔCAit −ΔCLit −ΔCASHit þ ΔSTDit −DEPit

ð4Þ

where: ΔCAit = Change in current assets for firm i from year t-1 to t; ΔCLit = Change in current liabilities for firm i from year t-1 to t; ΔCASHit = Change in cash and cash equivalents for firm i from year t-1 to t; ΔSTDit = Change in short-term debt included in the current liabilities for firm i from year t-1 to t; DEPit = Depreciation expense for firm i in year t; ΔRit = Change in revenue for firm i from year t-1 to t; ΔARit = Change in accounts receivable from operating activities for firm i from year t-1 to t; Ait-1 = Total assets at the beginning of year t for firm i; PPEit / Ait-1 = Gross property, plant, and equipment for firm i in year t, scaled by total assets at the beginning of year t for firm i. For cash effective tax rate (Cash3ETR), I follow Gupta, Mills, and Towery (2014) and compute Cash3ETR as the sum of cash taxes paid over the previous three years divided by the sum of pretax financial accounting income over the same period. Following Goncharov (2009), I also include lagged temporary booktax differences (TEMPit-1) as a control variable because prior studies suggest that the reported temporary differences between book and taxable income is persistent across two consecutive years (Dyreng, Hanlon, & Maydew, 2008; Manzon & Plesko, 2002). Therefore, a positive relation is expected between temporary book-tax differences measured in the current and previous year. Next, I follow Mills and Newberry (2001) and control for firms' leverage (LEV). As suggested in Harrington and Smith (2012), firms engaged in aggressive tax avoidance tend to have higher leverage because they are likely to borrow relatively more debt to maintain lower cash effective tax rates. Thus, I expect a positive relation between leverage and temporary book-tax differences. Furthermore, I incorporate firms' sales growth (GROWTH) in the model as a control variable given that firms with larger temporary book-tax differences have lower prior sales growth (Racca, 2011). Hence, I expect a negative relation between growth and temporary book-tax differences. Finally, I control for firm fixed effects and year fixed effects. The standard errors are clustered by firms. Detailed definitions of the control variables are provided in Appendix 1.

4.3. Testing H2: impact of auditor-provided tax services on mispricing of book-tax differences I use the following regression model (Eq. (5)) to test my second hypothesis, which examines whether firms procuring ATS are associated with increased or reduced mispricing of book-tax differences among investors compared with firms not receiving ATS. Following prior studies examining the mispricing of book-tax differences (Chi et al., 2014; Lev & Nissim, 2004; Weber, 2009), I include firm-specific risk factors as controls (e.g., size, market-to-book, momentum, and beta) in my returns model. I examine the interaction between ATS and book-tax differences, which models the incremental effect of ATS on the association between book-tax differences and future stock returns. Consistent with Weber (2009), independent and control variables in the model are each converted to decile rankings, by year, and scaled to range from zero to one. This scaling allows the interpretation of the variables' estimated coefficients as the returns on a zero investment portfolio with a long position in firm-years in the highest decile and a short position in those in

the lowest decile (Bernard & Thomas, 1990). AdjRETitþ1 ¼ α0 þ β1 ðTEMPit =Ait−1 Þ þ β2 ðATSit Þ þ β3 ðTEMPit =Ait−1  ATSit Þ þ β4 ðSIZEit Þ þ β5 ðMBit Þ þ β6 ðBETAit Þ þ β7 ðEPit Þ þ β8 ðAdjRETit Þ þ β9 ðACCit Þ þ β10 ðCash3ETRit Þ þ β11 ðLEVit Þ þ β12 ðGROWTHit Þ þ εit ð5Þ where: AdjRET = Adjusted returns; TEMP = Temporary book-tax differences; A = Total assets; ATS = An indicator variable that is set to 1 if a firm's tax fee is not equal to zero and 0 otherwise; SIZE = Natural log of market value of equity; MB = Market-to-book ratio; BETA = Systematic risk estimated from regression of monthly raw returns on the value-weighted market portfolio returns (including distribution) over a 60-month return period prior to the adjusted return accumulation period; EP = Earnings-to-price ratio; ACC = Ratio of accruals to average total assets; Cash3ETR = Three-year cash effective tax rate; LEV = Ratio of total liability to total assets; GROWTH = Percentage change in sales. 4.3.1. Dependent variables The dependent variable in my Eq. (5), AdjRET, is the one-year-ahead adjusted buy-and-hold annual return for a given firm-year observation. First, I calculate annual firm-level returns from a firm's adjusted closing prices: Adjusted Priceit −1 Adjusted Priceit−1 ðUnadjusted Priceit þ Dividiendit Þ=Adjustment Factorit −1 ¼ ðUnadjusted Priceit−1 þ Dividiendit−1 Þ=Adjustment Factorit−1 Annual Returnit ¼

Next, I adjust firm-years' annual returns for market returns using the Fama-French portfolios approach (Fama & French, 1995): AdjRETit ¼ Annual Returnit −Market Returnit More specifically, I first use Ken French's data library5 to partition firms into quintiles according to size (market capitalization) and book-to-market, which results in 25 benchmark portfolios by interacting the resulting partitioning of firms based on the size and book-to-market quintiles. I then categorize each of my firm-year samples into one of these 25 portfolios and obtain the annual valueweighted benchmark returns for each of the 25 portfolios from Ken French's data library. Finally, the adjusted returns (AdjRET) for each of my firm-year samples is calculated as the difference between the annual buy-and-hold returns for each sample and the buy-and-hold returns for the portfolio with the same size and book-to-market ratio. 4.3.2. Independent variables The independent variable of interest is the interaction between TEMP and ATS, which models the incremental effect of ATS on the association between TEMP and AdjRET. I also include firms' temporary book-tax differences (TEMP) and ATS to model the main effects. A negative association between TEMP and AdjRET would suggest a systematic mispricing of the book-tax differences. If ATS has an effect on investors' mispricing of book-tax differences, I expect that the association between TEMP and AdjRET will change among firms with ATS, resulting in a significant interaction of TEMP*ATS. 5 Ken French's data library: http://mba.tuck.dartmouth.eduages/faculty/ken.french/ data_library.html.

Please cite this article as: B. Luo, Effects of auditor-provided tax services on book-tax differences and on investors' mispricing of boo..., Advances in Accounting, https://doi.org/10.1016/j.adiac.2019.100434

B. Luo / Advances in Accounting xxx (xxxx) xxx

4.3.3. Control variables I incorporate several variables in Eq. (5) to control for effects of firms' performance and characteristics on future returns. I first include size, market-to-book ratio, return momentum, and beta as firm-specific risk factors. Fama and French (1992) suggest that smaller firms have higher stock returns than larger firms because of the increased risks associated with smaller firms. Hence, I control for firms' size (SIZE) and expect a negative relation between size and returns. I then control for market-to-book ratio (MB), based on evidence from Brown (2001) and Matsumoto (2002) showing that growth firms tend to have fewer negative earnings surprise. Therefore, I expect a negative relation between marketto-book ratio and returns. I also use return from the previous year to control for short-term return momentum (Jegadeesh & Titman, 1993), and I expect a positive relation in returns from two consecutive years. BETA is included to measure systematic risk (Fama & French, 1992), and I expect BETA to be positively related to stock returns. EP is the ratio of earnings before extraordinary items to market value of common equity and is expected to be positively related to returns (Basu, 1977). Again, I include accruals (ACC) to control for the effect of the accrual component of earnings management on returns. A higher level of earnings accruals produces less persistent future earnings (e.g., Hanlon, 2005) and therefore leads to negative returns. I also include 3year cash effective tax rate (Cash3ETR) as a control variable based on evidence from Guenther, Matsunaga, and Williams (2017) that tax risk is positively related to future stock return volatility. I further control for firms' leverage (LEV). As Fama and French (1992) suggested, a firm's required returns can increase as a result of increased leverage because leveraged investments are riskier than unleveraged ones. Thus, I expect a positive relation between firms' leverage and returns. I also control for firms' sales growth (GROWTH), since it has been shown that firms with lower past sales growth exhibit higher returns (Chan & Lakonishok, 2004; Lakonishok, Shleifer, & Vishny, 1994). I therefore expect a negative relation between growth and returns. Finally, I control for firm fixed effects and year fixed effects. The standard errors are clustered by firms. Detailed definitions of the control variables are provided in Appendix 1.

4.4. Sample selection I obtain firms' tax and audit fee data from Audit Analytics. Other relevant financial and stock price information is obtained from Compustat and CRSP. My firm-year samples cover the period from 2000 to 2013. I select 2000 as the beginning year of my sample because tax fee information first became available in Audit Analytics during that year. I use tax fee data to identify firmyears that receive tax services from their auditors. Because firms are only required to disclose tax fees when they purchase auditor-provided tax services, firm-year samples with zero tax fees indicate that no auditor-provided tax services are used. In order to ensure data availability for lagged variables and the computation of the three-year effective tax rate, I further require that relevant data be available during the period between 1998 and 1999. Firm-year observations with any missing data are excluded. Consistent with relevant literature (Krishnan et al., 2013; McGuire et al., 2012), I use the 12 industry categories proposed by Fama and French (1992) and require at least 15 observations per industry-year grouping. Following prior studies, I further exclude firms in utilities and regulated industries (SIC codes between 4900 and 4949) as well as financial services firms (SIC codes between 6000 and 6999) from my sample. Finally, I winsorize all continuous variables at the 1st and 99th percentile. The definitions of all variables used in the models and the empirical analyses are listed in Appendix 1.

7

5. Empirical results 5.1. Descriptive statistics Table 1 provides the summary statistics for the variables used in the main empirical analyses. Panel A presents the descriptive statistics for variables in the regression model used to test H1 (Eq. (2)), whereas Panel B provides the descriptive statistics for variables in Eq. (5) used to test H2. In Panel C and Panel D, I provide the descriptive statistics partitioned by whether the firm-year uses auditor-provided tax services or not. The total number of firm-year observations used to test H1 (the association between ATS and firms' temporary book-tax differences) is 8702. Half of the firm-year sample received ATS, and the other half of the propensity-score matched sample did not receive ATS. The overall mean temporary book-tax differences scaled by total assets (TEMP) is 0.0009 (std. dev. = 0.0323) among all firm-year observations, with firms procuring ATS having significantly lower temporary book-tax differences than firms without ATS. This suggests that ATS might enhance knowledge spillover and limit managerial discretion. The mean level of abnormal accruals (AbAcc) is 0.1862 (std. dev. = 0.0861) for all firmyear observations. However, for firm years with ATS, the mean abnormal accruals is not significantly different from firm-years without ATS. The mean three-year cash effective tax rate (Cash3ETR) for all sample firms is 0.1676 (std. dev. = 0.3256), with firms acquiring ATS having significantly higher three-year cash effective tax rate than firms without ATS. Finally, the mean leverage (LEV) for all firm-year observations is 0.5089 (std. dev. = 0.2393), and the mean growth rate (GROWTH) is 0.0953 (std. dev. = 0.2311). H2 examines whether ATS is systematically associated with investors' mispricing of book-tax differences. The final sample used to test H2 contains 7756 firm-year observations. Half of the firm-year sample received ATS, and the other half of the propensity-score matched sample did not receive ATS. The mean one-year-ahead adjusted returns (AdjRETt+1) among all firm-year samples is 0.0540 (std. dev. = 0.3986). The mean level of temporary book-tax differences scaled by total assets (TEMP) is 0.0008 (std. dev. = 0.0324). For firm-years with ATS, the one-year-ahead adjusted returns and the level of temporary book-tax differences are both significantly lower than for firm-years without ATS. While SIZE, MB, and LEV are comparable between firm-years with and without ATS, other variables including BETA, EP, ACC, Cash3ETR, GROWTH are significantly different between the two groups. These results suggest that firm-years with ATS are different from firm-years without ATS in terms of certain firm characteristics. Hence, it is important to control for such factors in my empirical models while testing for the relationship between ATS and the mispricing of temporary book-tax differences. 5.2. Univariate correlations Table 2 presents pairwise Pearson correlation analysis results between variables in my empirical models. Panel A of Table 2 reports the pairwise correlation coefficients between variables involved in testing H1. The correlation coefficient between temporary book-tax differences (TEMP) and ATS is significantly negative, suggesting that firms procuring ATS are likely to have smaller temporary book-tax differences. TEMP is also inversely correlated with Cash3ETR, suggesting that firms with higher temporary book-tax differences have lower effective tax rates. In addition, TEMPt is positively correlated with TEMPt-1, consistent with prior literature suggesting that temporary book-tax differences tend to be persistent (Dyreng et al., 2008). Panel B of Table 2 presents the correlation coefficients between variables involved in testing H2. The correlation between the oneyear-ahead adjusted returns (AdjRETt+1) and the level of book-tax differences (TEMP) is significantly negative, which is consistent with evidence of mispricing of temporary book-tax difference from prior

Please cite this article as: B. Luo, Effects of auditor-provided tax services on book-tax differences and on investors' mispricing of boo..., Advances in Accounting, https://doi.org/10.1016/j.adiac.2019.100434

8

B. Luo / Advances in Accounting xxx (xxxx) xxx

Table 1 Descriptive statistics. Variable

Std. Dev.

Q1

Median

Q3

Panel A: summary statistics for variables in Eq. (2) (H1): TEMPit /Ait-1 8702 0.0009 ATSit 8702 0.5000 AbAccit 8702 0.1862 Cash3ETRit 8702 0.1676 TEMPit-1/Ait-2 8702 −0.0012 LEVit 8702 0.5089 GROWTHit 8702 0.0953

N

Mean

0.0323 0.5000 0.0861 0.3256 0.0308 0.2393 0.2311

−0.0074 0.0000 0.1526 0.0802 −0.0087 0.3365 −0.0125

0.0000 0.5000 0.1805 0.2035 0.0000 0.5025 0.0725

0.0090 1.000 0.2160 0.2940 0.0096 0.6429 0.1653

Panel B: summary statistics for variables in Eq. (5) (H2): AdjRETit+1 7756 0. 0540 TEMPit /Ait-1 7756 0.0008 ATSit 7756 0.5000 7756 7.5237 SIZEit MBit 7756 3.3522 BETAit 7756 1.3000 EPit 7756 0.0123 AdjRETit 7756 0.0778 ACCit 7756 −0.0419 7756 0.1850 Cash3ETRit LEVit 7756 0.4995 GROWTHit 7756 0.1017

0.3986 0.0324 0.5000 1.7921 4.8463 0.7726 0.1417 0.4133 0.0846 0.2563 0.2290 0.2251

−0.1992 −0.0067 0.000 6.3811 1.5983 0.7593 0.0128 −0.1527 −0.0695 0.0810 0.3321 −0.0036

0.0031 0.0000 0.5000 7.4817 2.5839 1.1692 0.0447 0.0248 −0.0376 0.2106 0.5002 0.0779

0.2155 0.0084 1.0000 8.7222 4.0025 1.6976 0.0636 0.2388 −0.0075 0.2941 0.6280 0.1729

ATS = 0 Mean

ATS = 1 Median

Mean

Difference in Median

Mean (t-stat)

Median (z-stat)

Panel C: Comparison between firm-years with and without ATS for variables in Eq. (2) (H1) Full sample of 8702 firm-years with 4351 ATS firm-years (ATS = 1) and 4351 propensity-score matched non-ATS firm-years(ATS = 0) 0.0030 0.0000 −0.0013 0.0000 0.0043*** (5.4905) TEMPit /Ait-1 AbAccit 0.1875 0.1819 0.1850 0.1795 0.0025 (1.3638) Cash3ETRit 0.1475 0.2013 0.1871 0.2072 −0.0396*** (−5.4464) LEVit 0.5165 0.5059 0.5013 0.4978 0.0152*** (2.9548) GROWTHit 0.1007 0.0797 0.0901 0.0653 0.0106** (2.1417)

0.0000*** (3.635) 0.0024** (2.133) −0.0059** (−2.479) 0.0081** (2.078) 0.0144** (2.104)

Panel D: Comparison between firm-years with and without ATS for variables in Eq. (5) (H2) Full sample of 7756 firm-years with 3878 ATS firm-years (ATS = 1) and 3878 propensity-score matched Non-ATS firm-years(ATS = 0) AdjRETit+1 0.0652 0.0162 0.0431 −0.0139 0.0221** (2.3216) TEMPit /Ait-1 0.0025 0.0000 −0.0009 0.0000 0.0034*** (4.1720) SIZEit 7.5322 7.4568 7.5152 7.5145 0.0170 (0.4185) MBit 3.3123 2.6765 3.3922 2.5100 −0.0799 (−0.7262) BETAit 1.2726 1.1326 1.3273 1.2243 −0.0547*** (−3.1147) EPit 0.0152 0.0462 0.0095 0.0426 0.0057* (1.7820) AdjRETit 0.0865 0.0314 0.0692 0.0172 0.0173* (1.8393) ACCit −0.0460 −0.0378 −0.0379 −0.0354 −0.0081*** (−3.8568) Cash3ETRit 0.1737 0.2107 0.1962 0.2083 −0.0225*** (−3.7266) LEVit 0.5038 0.5204 0.4953 0.4904 0.0085 (1.6256) GROWTHit 0.1109 0.0852 0.0925 0.0688 0.0184*** (3.6011)

0.0301*** (0.280) 0.0000 (0.454) −0.0577 (0.382) 0.1665 (0.987) −0.0917*** (−4.113) 0.0036*** (3.746) 0.0142** (2.172) −0.0024*** (−3.900) 0.0024 (−1.197) 0.0300* (1.887) 0.0164*** (4.058)

See Appendix 1 for variable descriptions. ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively. The sample period is 2000–2013. The numbers in parentheses are t-statistics from t-tests for the differences in the means, and z-statistics from Wilcoxon's rank-sum tests for the differences in the medians. The reason that some of the z-statistics show an inconsistent sign with the median difference is that the Wilcoxon tests are rank sum tests rather than median tests. Basically, the Wilcoxon's rank-sum test ranks all of the observations from both groups, sums the ranks from one of the groups, and then compares the observed rank sum with the expected value. It is possible, although not very common, for a group to have a higher than expected rank sum while having a smaller median than the other group and vice versa. It is also possible for groups to have significantly different rank sums and yet have equal or nearly equal medians.

literature. Further, the one-year-ahead adjusted returns (AdjRETt+1) is negatively correlated with size, earnings-to-price ratio (EP), accruals (ACC), and growth rate (GROWTH); at the same time, one-year-ahead adjusted returns (AdjRETt+1) are positively correlated with three-year cash effective tax rate (Cash3ETR), past price momentum (AdjRETit), and leverage (LEV). 5.3. Results for H1 Table 3 reports the results from the regression analysis on the association between ATS and temporary book-tax differences (TEMP). The coefficient on ATS, β1, is significantly negative (−0.0072; p b .05), indicating that firms receiving ATS have significantly smaller temporary book-tax differences compared with firms not receiving ATS. To assess the economic effects, I multiply the coefficient on ATS (i.e., −0.0072) by average assets to obtain the difference in temporary book-tax

differences between firms that receive ATS and those that do not. The result indicates that, controlling for other factors, the level of temporary book-tax differences in firms that receive ATS is, on average, lower than the temporary book-tax differences in non-ATS firms by $51.60 million. This result is consistent with the notion that ATS enhance audit quality via knowledge spillover and limit client firms' management of book income and/or taxes. The association between abnormal accruals (AbAcc) and temporary book-tax differences is significantly positive (0.0018; p b .05). This is consistent with the findings of Blaylock et al. (2012) that large temporary book-tax differences are likely to be associated with upward earnings management. Also consistent with prior literature (e.g., Hanlon, 2005), the coefficient on the threeyear cash effective tax rate (Cash3ETR) is negative at the 0.01 level, which suggests that firms with large book-tax differences also exhibit low effective tax rates. Furthermore, there is a

Please cite this article as: B. Luo, Effects of auditor-provided tax services on book-tax differences and on investors' mispricing of boo..., Advances in Accounting, https://doi.org/10.1016/j.adiac.2019.100434

B. Luo / Advances in Accounting xxx (xxxx) xxx

9

Table 2 Pearson correlation coefficients. Panel A: variables in Eq. (2) (H1): Variable

TEMPit

ATSit

AbAccit

Cash3ETRit

TEMPit-1

LEVit

GROWTHit

TEMPit/Ait-1 ATSit AbAccit Cash3ETRit TEMPit-1/Ait-2 LEVit GROWTHit

1.0000 −0.0683 −0.0631 −0.0879 0.1936 −0.0081 0.0389

1.0000 −0.0165 0.0840 0.0153 −0.0041 −0.0276

1.0000 0.0764 0.0065 −0.0288 0.3708

1.0000 −0.0304 0.0360 −0.0477

1.0000 −0.0066 −0.0029

1.0000 −0.0947

1.0000

ATSit

SIZEit

MBit

BETAit

EPit

AdjRETit

ACCit

Cash3ETRit

LEVit

GROWTHit

1.0000 −0.0324 −0.0278 0.0649 −0.0381 −0.0001 0.0763 0.0410 0.0118 −0.0047

1.0000 0.1799 −0.2138 0.2877 0.0456 0.0896 0.0671 0.1390 0.1188

1.0000 −0.0801 0.0939 0.0444 0.0046 −0.0103 0.1340 0.1065

1.0000 −0.1498 0.0400 −0.0929 −0.2167 0.0127 0.0174

1.0000 0.0537 0.2549 0.0898 −0.1036 0.2087

1.0000 −0.0223 −0.0535 0.0849 −0.0483

1.0000 0.0400 −0.0961 0.1603

1.0000 −0.0132 −0.0827

1.0000 −0.1111

1.0000

Panel B: variables in Eq. (5) (H2): Variable AdjRETit+1 TEMPit AdjRETit+1 1.0000 TEMPit/Ait-1 −0.0589 1.0000 ATSit 0.0015 −0.0469 SIZEit −0.0384 0.0342 MBit −0.0131 −0.0002 BETAit −0.0048 0.0753 EPit −0.1035 0.0294 AdjRETit 0.1047 0.0181 ACCit −0.0974 −0.0089 Cash3ETRit 0.0654 −0.0043 LEVit 0.1645 −0.0163 GROWTHit −0.1133 0.0235

See Appendix 1 for variable descriptions. The numbers in boldface are significant at the 5% level or better.

marginally positive trend between the prior year's level of temporary book-tax differences (TEMPt-1) and the current year's temporary book-tax difference. In addition, highly leveraged (LEV) firms have large temporary book-tax differences, whereas firms with high sales growth rate (GROWTH) have small temporary booktax differences. 5.4. Results for H2 Table 4 reports the results from the regression analysis on the association between ATS and investors' mispricing of book-tax differences. The significantly negative coefficient on TEMP (β1 = −0.0162; p b .05) indicates that higher levels of temporary book-tax differences are associated with lower one-year-ahead adjusted returns (AdjRETt+1). This is consistent with prior literature examining mispricing of temporary book-tax differences (Chi et al., 2014) and implies that investors do not evaluate information in temporary book-tax differences efficiently, leading to systematic pricing errors. Moreover, the coefficient associated with TEMP*ATS is significantly positive (β3 = 0.0117;

Table 3 Regression results from Eq. (2). Variable

Parameter estimate

Robust Std. Err.

z-Value

p-Value

INTERCEPT ATSit AbAccit Cash3ETRit TEMPit-1/ Ait-2 LEVit GROWTHit Year fixed effect Firm fixed effect N = 8702 Adj. R2 = 0.0892

−0.0040 −0.0072 0.0018 −0.0001 0.0982 0.0078 −0.0001 Included Included

0.0039 0.0033 0.0009 3.35e-06 0.0585 0.0051 0.0001

−1.01 −2.13 2.00 −18.05 1.68 1.52 −2.56

0.313 0.033 0.046 0.000 0.093 0.127 0.011

TEMPit /Ait-1 = α0 + β1 (ATSit) + β2 (AbAccit) + β3 (Cash3ETRit) + β4 (TEMPit-1/ Ait-2) + β5 (LEVit) + β6 (GROWTHit) + εit (2) See Appendix 1 for variable descriptions. The standard errors are clustered at the firm level.

p b .01), suggesting that the relationship between book-tax differences and future returns is less negative (i.e., β1 + β3 = −0.0045) in the presence of ATS. The economic effect is essential. Controlling for other factors, the level of book-tax difference mispricing (the association between adjusted returns and TEMP) among firms receiving ATS is, on average, decreased by 0.0117 (β3), or 72.22% (β3/β1), compared with non-ATS firms. This result suggests decreased mispricing of book-tax differences among firms with ATS. Also as expected, one-year-ahead adjusted returns (AdjRETt+1) are positively associated with past price

Table 4 Regression results from Eq. (5). Variable

Parameter estimate

Robust Std. Err.

z-Value

p-Value

INTERCEPT TEMPit /Ait-1 ATSit TEMPit /Ait-1* ATSit SIZEit MBit BETAit EPit AdjRETit ACCit Cash3ETRit LEVit GROWTHit Year fixed effect Firm fixed effect N = 7756 Adj. R2 = 0.1484

0.1303 −0.0162 −0.0174 0.0117 −0.0093 −0.0177 −0.0032 −0.0031 0.0096 −0.0072 0.0090 0.0184 −0.0071 Included Included

0.0849 0.0064 0.0336 0.0045 0.0047 0.0050 0.0041 0.0043 0.0054 0.0042 0.0054 0.0058 0.0036

1.54 −2.52 −0.52 2.59 −2.05 −3.54 −0.79 −0.70 1.76 −1.72 1.65 3.18 −1.98

0.125 0.012 0.603 0.009 0.041 0.000 0.431 0.481 0.078 0.085 0.098 0.001 0.048

AdjRETit+1 = α0 + β1 (TEMPit /Ait-1) + β2 (ATSit) + β3 (TEMPit /Ait-1* ATSit) + β4 (SIZEit) + β5 (MBit) + β6 (BETAit) + β7 (EPit) + β8 (AdjRETit) + β9 (ACCit) + β10 (Cash3ETRit) + β11 (LEVit) + β12 (GROWTHit) + εit (5) See Appendix 1 for variable descriptions. The standard errors are clustered at the firm level. Since TEMP captures both earnings management and tax avoidance, I calculate the Variance Inflation Factor (VIF) in order to detect possible multicollinearity among TEMP, earnings management (ACC), and tax avoidance (Cash3ETR). As a rule of thumb, if the VIFs are b10, then there is no multicollinearity. The results show that the VIFs for TEMP, ACC, and Cash3ETR are 1.94, 1.19, and 1.14, respectively, thus suggesting no multicollinearity among them.

Please cite this article as: B. Luo, Effects of auditor-provided tax services on book-tax differences and on investors' mispricing of boo..., Advances in Accounting, https://doi.org/10.1016/j.adiac.2019.100434

10

B. Luo / Advances in Accounting xxx (xxxx) xxx

momentum (AdjRETit), three-year cash effective tax rate (Cash3ETR), and leverage (LEV) and negatively associated with firm size (SIZE), market-to-book ratio (MB), accruals (ACC),6 and growth rate (GROWTH). 5.5. Robustness tests 5.5.1. Robustness tests for H1 I conduct the following supplemental analyses to test the robustness of the main findings for H1, which states that acquiring ATS leads to lower levels of temporary book-tax differences. 5.5.1.1. Reverse causality. To control for the potential issue of reverse causality where firms with lower temporary book-tax differences selfselect to purchase ATS, I conduct a Granger lead-lag test7 to assess the robustness of my findings. Specifically, I estimate the following two regression models (Eq. (6a) and (6b)). Eq. (6a) tests how ATS impact firms' temporary book-tax differences (TEMPt) with the inclusion of lagged temporary book-tax differences (TEMPt-1); a significant β1 would imply that ATS impact firms' temporary book-tax differences. Eq. (6b) tests how firms' temporary book-tax differences impact the decision to purchase ATS with lagged ATS (ATSt-1) included; a significant β1 would imply that temporary book-tax differences impact the decision to purchase ATS. TEMPit =Ait−1 ¼ α0 þ β1 ðATSit Þ þ β2 ðTEMPit−1 =Ait−2 Þ þ β3 ðAbAccit Þ þ β4 ðCash3ETRit Þ þ β5 ðLEVit Þ þ β6 ðGROWTHit Þ þ εit

ATSit ¼ α0 þ β1 ðTEMPit−1 =Ait−2 Þ þ β2 ðATSit−1 Þ þ β3 ðTEMPit =Ait−1 Þ þ β4 ðTENUREit Þ þ β5 ðINSTOWNit Þ þ β6 ðMERGERit Þ þ β7 ðAUDINDit Þ þ β8 ðLAFit Þ þ β9 ðOPPORit Þ þ β10 ðSIZEit Þ þ β11 ðABACCit Þ þ β12 ðNOLit Þ þ β13 ðΔNOLit Þ þ β14 ðEQINCit Þ þ β15 ðFIit Þ þ β16 ðR&Dit Þ þ β17 ðLEVit Þ þ β18 ðBTMit Þ þ β19 ðPPEit Þ þ β20 ðROAit Þ þ β21 ðCASHit Þ þ β22 ðDEPit Þ þ β23 ðBIG4it Þ þ β24 ðSECTIERit Þ þ εit

ð6bÞ

where: TEMP = Temporary book-tax differences; A = Total assets; ATS = An indicator variable that is set to 1 if a firm's tax fee is not equal to zero and 0 otherwise; ABACC = Abnormal accruals; Cash3ETR = Ratio of the sum of cash taxes paid over the previous 3 years to the sum of pretax financial accounting income over the previous 3 years; LEV = Ratio of total liabilities to total assets; GROWTH = Percentage change in sales; TENURE = Length of the audit firm's tenure with its client; INSTOWN = Percentage of shares owned by institutions at the beginning of the year; MERGER = An indicator variable equal to 1 if a firm participates in any merger activity during the year and 0 otherwise; AUDIND = Auditor independence from the client, measured as nonaudit fees less tax fees divided by total audit fees received from the client; LAF = Natural log of audit fees received from the client; 6 In an untabulated test, I further interact ATS with earnings management (ACC) and tax avoidance (Cash3ETR) in Eq. (5), respectively. The results show that firms' level of earnings management (ACC) is marginally negatively related to adjusted returns, and the coefficient associated with ACC* ATS is positive and marginally significant. This suggests that ATS may have a trend toward reducing accruals mispricing. The results also show that firms' tax avoidance (Cash3ETR) is marginally related to adjusted returns, yet ATS does not appear to have a significant mitigating effect on this relationship. 7 The Granger lead-lag test is commonly used in accounting literature to control for the potential reverse causality issue, e.g., Lennox and Park (2006), Gong, Li, and Xie (2009), and Call, Chen, Miao, and Tong (2014).

OPPOR = Market value of a client divided by the sum of the market value of all clients in the same industry at the same MSA city; SIZE = Natural log of market value of equity at the beginning of the year; NOL = An indicator variable set to 1 if there is a tax loss carryforward during the year and 0 otherwise; EQINC = Equity income scaled by total assets at the beginning of the year; FI = Pre-tax foreign income scaled by total assets at the beginning of the year; R&D = R&D expense scaled by total assets at the beginning of the year; SLEV = Long-term debt-to-asset ratio at the end of the year scaled by total assets at the end of the year; BTM = Book-to-market ratio at the end of the year, measured as book value of equity divided by market value of equity; PPE = Net PPE scaled by total assets at the beginning of the year; ROA = Return on assets, measured as the ratio of income before extraordinary items to the average of total assets for the year; CASH = Cash holding at the end of the year divided by total assets at the beginning of the year; DEP = Depreciation and amortization expense for the year divided by total assets at the beginning of the year; BIG4 = An indicator variable set to 1 if the firm is audited by one of the Big 4 auditors and 0 otherwise; SECTIER = An indicator variable set to 1 if the firm is audited by a second-tier accounting firm, namely Grant Thornton LLP and BDO Seidman LLP, and 0 otherwise. Table 5 reports the main findings from this analysis. The coefficients and t or z statistics for the control variables are not tabulated for simplicity. In Eq. (6a), the coefficient on ATS, β1, is significantly negative, indicating a causal relation running from ATS to firms' temporary book-tax differences. In Eq. (6b), β1, the coefficient on temporary book-tax differences, is not significant. Collectively, I find evidence of causality from ATS to lower temporary book-tax difference, which is consistent with my findings for H1; I do not find evidence of causality from firms' lower temporary book-tax differences to the purchase of ATS. 5.5.1.2. Tax fee measure. In the following model (Eq. 7), I re-estimate Eq. (2) by replacing the indicator variable ATS with firms' actual fees paid to the incumbent auditors for tax services. That is, firm-years will have non-zero tax fees if they purchase ATS in a given year. In order to control for firm size effect, I scale tax fee for each firm-year sample by the total assets. TEMPit =Ait−1 ¼ α0 þ β1 ðTaxFeeit =Ait−1 Þ þ β2 ðAbAccit Þ þ β3 ðCash3ETRit Þ þ β4 ðTEMPit−1 =Ait−2 Þ þ β5 ðLEVit Þ þ β6 ðGROWTHit Þ þ εit

ð7Þ

where: TEMP = Temporary book-tax differences; A = Total assets; TaxFee = Fees paid to the auditor for tax services; ABACC = Abnormal accruals; Cash3ETR = Ratio of the sum of cash taxes paid over the previous 3 years to the sum of pretax financial accounting income over the previous 3 years; LEV = Ratio of total liabilities to total assets; GROWTH = Percentage change in sales. The results from this regression analysis are reported in Panel A of Table 6, and they generally agree with the findings reported in Table 3. Specifically, the coefficient on TaxFee is significantly negative (−0.0017, p b .05), suggesting that firms with ATS have lower levels of temporary book-tax differences. In addition, the negative relation between TaxFee and TEMP suggests that a higher amount of tax fees paid

Please cite this article as: B. Luo, Effects of auditor-provided tax services on book-tax differences and on investors' mispricing of boo..., Advances in Accounting, https://doi.org/10.1016/j.adiac.2019.100434

B. Luo / Advances in Accounting xxx (xxxx) xxx Table 5 Reverse causality test.

Table 7 Mandatory tax fee disclosure.

Variables

Dependent variable TEMPit /Ait-1 (6a) −0.0028*** (−2.23) 0.0713*** (5.63)

ATSit TEMPit-1/ Ait-2 ATSit-1 Control variables, Ind. And Year Dummies N Adj. R2

Included 7866 0.0748

0.4312 (1.02) 2.0119*** (30.99) Included 7848 0.4704

to the incumbent auditor is associated with lower levels of temporary book-tax differences among firms.

5.5.1.3. Positive tax fee measure. For this analysis, I further drop firm-year observations with zero tax fees (TaxFeeit /Ait-1 = 0) and re-estimate Eq. (7). The results from this analysis are shown in Table 6, Panel B. Consistent with the results in Table 3 and Panel A of Table 6, the coefficient on TaxFee is again significantly negative (−0.0002, p b .05), confirming that a higher amount of tax fees paid to the auditor is related to lower temporary book-tax differences.

Table 6 Regression Results from Eq. (7). Parameter Estimate

Robust Std. Err.

z-Value

Panel A: Re-estimation of Eq. (2) by replacing ATSit with TaxFeeit INTERCEPT −0.0040 0.0038 −1.07 TaxFeeit −0.0017 0.0007 −2.31 AbAccit 0.0018 0.0009 2.01 Cash3ETRit −0.0001 3.51e-06 −16.96 TEMPit-1/ Ait-2 0.0965 0.0578 1.67 LEVit 0.0079 0.0051 1.54 GROWTHit −0.0001 0.0001 −2.54 Year fixed effect Included Firm fixed effect Included N = 8702 Adj. R2 = 0.0804

Variable

Parameter Estimate

Robust Std. Err.

z-Value

p-Value

INTERCEPT TEMPit /Ait-1 ATSit TEMPit /Ait-1* ATSit SIZEit MBit BETAit EPit AdjRETit ACCit Cash3ETRit LEVit GROWTHit Year fixed effect Firm fixed effect N = 7432 Adj. R2 = 0.1805

0.1622 −0.0178 −0.0442 0.0103 −0.0151 −0.0092 0.0017 0.0029 0.0081 −0.0077 0.0151 0.0236 −0.0078 Included Included

0.0929 0.0067 0.0292 0.0046 0.0071 0.0069 0.0044 0.0056 0.0038 0.0036 0.0055 0.0057 0.0043

1.75 −2.64 −1.51 2.22 −2.13 −1.32 0.40 0.52 2.11 −2.13 2.71 4.09 −1.80

0.081 0.008 0.131 0.027 0.033 0.186 0.689 0.604 0.035 0.033 0.007 0.000 0.072

ATSit (6b)

TEMPit /Ait-1 = α0 + β1 (ATSit) + β2 (TEMPit-1/ Ait-2) + β3 (AbAccit) + β4 (Cash3ETRit) + β5 (LEVit) + β6 (GROWTHit) + εit (6a) ATSit = α0 + β1 (TEMPit-1/ Ait-2) + β2 (ATSit-1) + β3 (TEMPit /Ait-1) + β4 (TENUREit) + β5 (INSTOWNit) + β6 (MERGERit) + β7 (AUDINDit) + β8 (LAFit) + β9 (OPPORit) + β10 (SIZEit) + β11 (ABACCit) + β12 (NOLit) + β13 (ΔNOLit) + β14 (EQINCit) + β15 (FIit) + β16 (R&Dit) + β17 (LEVit) + β18 (BTMit) + β19 (PPEit) + β20 (ROAit) + β21 (CASHit) + β22 (DEPit) + β23 (BIG4it) + β24 (SECTIERit) + εit (6b) See Appendix 1 for variable descriptions. The values in parentheses are t-values for Eq. (6a) and z-values for Eq. (6b). The standard errors are clustered at the firm level. ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively.

Variable

11

p-Value

0.284 0.021 0.044 0.000 0.095 0.123 0.011

Panel B: Re-estimation of Eq. (2) by replacing ATSit with TaxFeeit and retaining only firm-years with TaxFeeit N 0 INTERCEPT −0.0024 0.0014 −1.69 0.092 TaxFeeit −0.0002 0.0006 −2.34 0.019 AbAccit 0.0007 0.0003 2.39 0.017 Cash3ETRit −0.0001 1.30e-06 −20.49 0.000 TEMPit-1/ Ait-2 −0.0042 0.0089 −0.46 0.643 LEVit 0.0030 0.0019 1.53 0.126 GROWTHit −0.0001 0.0001 −2.13 0.033 Year fixed effect Included Firm fixed effect Included N = 4351 Adj. R2 = 0.0601 TEMPit /Ait-1 = α0 + β1 (TaxFeeit /Ait-1) + β2 (AbAccit) + β3 (Cash3ETRit) + β4 (TEMPit-1/ Ait-2) + β5 (LEVit) + β6 (GROWTHit) + εit (7) See Appendix 1 for variable definitions (TaxFeeit/Ait-1 is in thousands of dollars). The standard errors are clustered at the firm level.

AdjRETit+1 = α0 + β1 (TEMPit /Ait-1) + β2 (ATSit) + β3 (TEMPit /Ait-1* ATSit) + β4 (SIZEit) + β5 (MBit) + β6 (BETAit) + β7 (EPit) + β8 (AdjRETit) + β9 (ACCit) + β10 (Cash3ETRit) + β11 (LEVit) + β12 (GROWTHit) + εit (5) See Appendix 1 for variable descriptions. The standard errors are clustered at the firm level.

5.5.2. Robustness tests for H2 I conduct the following supplemental analyses to test the robustness of the main findings for H2, which examines the association between ATS and mispricing of book-tax differences. 5.5.2.1. Mandatory tax fee disclosure. I re-estimate Eq. (5) after limiting the sample period to post-2003, when the disclosure of tax fees became mandatory. The results presented in Table 7 show that the coefficient on TEMP is significantly negative (−0.0178, p b .01), whereas the coefficient for the interaction term TEMP*ATS is significantly positive (0.0103, p b .05). These results are consistent with the main results from the full sample (Table 4), suggesting that the incremental effect of ATS on the negative association between TEMP and adjusted returns is not sensitive to the inclusion of data between years 2000 and 2002. 5.5.2.2. Effects of SEC regulation. In 2005, the SEC set new regulations for non-audit services and prohibited certain types of ATS in favor of tax treatments, such as tax marketing, planning, or advice. To control for the potential effects of these regulatory differences, I estimate Eq. (5) separately for the pre-2005 period (2000–2004) and the post-2005 period (2005–2013) and test the difference in the coefficient estimate of the interaction term TEMP*ATS. As shown in Panel A and Panel B of Table 8, respectively, the coefficients on TEMP*ATS are significantly positive in both the pre-2005 period (0.0169, p b .05) and the post-2005 period8 (0.0133, p b .05). These findings provide further confirmation that ATS help reduce the mispricing of book-tax differences and that the effect of ATS on mispricing did not change following the regulatory changes of 2005. 5.5.2.3. Initiation/continuation/termination of ATS and mispricing of booktax differences. I next test the impact of a firm's initiation, continuation, and termination of ATS on investors' pricing of temporary book-tax differences to corroborate the results in Table 4. Based on the main results from Eq. (5) that ATS mitigate investors' mispricing of book-tax differences, I expect that the mitigating effect will be apparent when firms initiate and continue ATS. On the contrary, when a firm terminates 8 When ATS is present in the post-2005 period, the coefficient on TEMP is positive (i.e., β1 + β3 = 0.0037) but not significantly different from zero (z = 0.67; p = .502) in the post-2005 period. This suggests that the mispricing of temporary book-tax differences has been reduced to a level that is not distinguishable from zero when ATS is present during the post-2005 period.

Please cite this article as: B. Luo, Effects of auditor-provided tax services on book-tax differences and on investors' mispricing of boo..., Advances in Accounting, https://doi.org/10.1016/j.adiac.2019.100434

12

B. Luo / Advances in Accounting xxx (xxxx) xxx

ATS, I expect that investors' mispricing of temporary book-tax differences will increase following the termination. I modify Eq. (5) by replacing the indicator variable ATS with three indicator variables – START, CONT, and END – to identify the instances when firms initiate, continue to use, or terminate ATS (Eq. 8). Specifically, START equals 1 during the first year when ATS appear for a given firm and 0 otherwise. CONT equals 1 during years when a firm continues to acquire ATS after the initiation and 0 otherwise. END equals 1 for the year immediately after the last year ATS appear for a given firm and 0 otherwise. To ensure that START picks up the impact of initiating ATS, I require that the firm in question does not have ATS in the year prior to initiation (STARTt = 1 if ATSt-1 = 0 and ATSt = 1). Similarly, to ensure that END picks up the impact of terminating ATS, I require that the firm in question have ATS in the year prior to termination but not in the current year (ENDt = 1 if ATSt-1 = 1 and ATSt = 0). For CONT, I require that the firm have ATS in the current year and in the prior year (CONTt = 1 if ATSt-1 = 1 and ATSt = 1). Accordingly, I expect the coefficients on TEMPit/Ait-1 * STARTit and TEMPit/Ait-1 * CONTit to be positive, and the coefficient on TEMPit/Ait-1* ENDit to be negative. I use the following modified regression model to test the impact of a firm's initiation, continuation, and termination of ATS on investors' pricing of temporary book-tax differences: AdjRETitþ1 ¼ α0 þ β1 ðTEMPit =Ait−1 Þ þ β2 ðSTARTit Þ þ β3 ðTEMPit =Ait−1  STARTit Þ þ β4 ðCONTit Þ þ β5 ðTEMPit =Ait−1  CONTit Þ þ β6 ðENDit Þ þ β7 ðTEMPit =Ait−1  ENDit Þ þ β8 ðSIZEit Þ þ β9 ðMBit Þ þ β10 ðBETAit Þ þ β11 ðEPit Þ þ β12 ðAdjRETit Þ þ β13 ðACCit Þ þ β14 ðCash3ETRit Þ þ β15 ðLEVit Þ þ β16 ðGROWTHit Þ þ εit ð8Þ where: AdjRET = Adjusted returns; TEMP = Temporary book-tax differences; A = Total assets; START = An indicator variable that is set to 1 if a specific year is the first year that ATS appeared for a firm and 0 otherwise; CONT = An indicator variable that is set to 1 if ATS appears for a firm in both a given year and a prior year and 0 otherwise; END = An indicator variable that is set to 1 if a specific year is the year immediately after the last ATS is provided for a firm, and 0 otherwise; SIZE = Natural log of market value of equity; MB = Market-to-book ratio; BETA = Systematic risk estimated from regression of monthly raw returns on the value-weighted market portfolio returns (including distribution) over a 60-month return period prior to the adjusted return accumulation period; EP = Earnings-to-price ratio; ACC = Ratio of accruals to average total assets; Cash3ETR = Three-year cash effective tax rate; LEV = Ratio of total liability to total assets; GROWTH = Percentage change in sales. The basic descriptive statistics regarding the variables START, CONT, and END are provided in Table A2.2 in Appendix 2. Among the full sample of 7756 firms, 344 firms initiate or switch to ATS (START = 1); 3506 firms continue to use ATS for at least 2 consecutive years (CONT = 1); 1359 firms end ATS (END = 1); and the remaining 2547 firms continue to not use ATS for at least two consecutive years during the sample period. With respect to the variables of interest, the mean of the temporary book-tax differences (TEMPit /Ait-1) for firm-years that continue to use ATS (CONT = 1) is decreased compared with firm-years that start to use ATS (START = 1; 0.0003 vs. 0.0032, t = −3.9233, p b .0001). This suggests that the continued use of ATS reduces firms' temporary book-tax differences. In addition, the mean of the one-year-ahead adjusted returns (AdjRETit+1) among firm-years that continue to use

Table 8 Effects of the SEC regulation. Variable

Parameter estimate

Robust Std. Err.

z-Value

p-Value

Panel A: Re-estimation of Eq. (5) with pre-2005 data (N = 2576; Adj. R2 = 0.1991) INTERCEPT −0.3304 0.1933 −1.71 0.087 TEMPit /Ait-1 −0.0298 0.0908 −3.29 0.001 ATSit −0.0345 −0.0345 −0.55 0.583 TEMPit /Ait-1* ATSit 0.0169 0.0070 2.39 0.017 SIZEit −0.0204 0.0114 −1.78 0.074 MBit −0.0235 0.0119 −1.98 0.048 BETAit 0.0066 0.0097 0.68 0.493 EPit −0.0017 0.0104 −0.17 0.868 AdjRETit 0.0182 0.0094 1.92 0.055 ACCit −0.0185 0.0106 −1.74 0.082 Cash3ETRit 0.0198 0.0111 1.78 0.075 LEVit 0.0182 0.0090 2.02 0.043 GROWTHit 0.0019 0.0076 0.26 0.798 Panel B: Re-estimation of Eq. (5) with post-2005 data (N = 5180; Adj. R2 = 0.1817) INTERCEPT 0.2285 0.0993 2.30 0.021 TEMPit /Ait-1 −0.0096 0.0056 −1.70 0.089 0.0023 0.0304 2.00 0.046 ATSit TEMPit /Ait-1* ATSit 0.0133 0.0067 −1.99 0.047 SIZEit −0.0167 0.0059 −2.84 0.005 MBit −0.0193 0.0066 −2.90 0.004 BETAit 0.0004 0.0048 0.09 0.931 EPit −0.0056 0.0077 −0.73 0.465 AdjRETit 0.0067 0.0039 1.72 0.085 ACCit −0.0119 0.0063 −1.87 0.061 Cash3ETRit 0.0091 0.0064 1.43 0.153 LEVit 0.0203 0.0052 3.84 0.000 GROWTHit −0.0064 0.0048 −1.34 0.179 AdjRETit+1 = α0 + β1 (TEMPit /Ait-1) + β2 (ATSit) + β3 (TEMPit /Ait-1* ATSit) + β4 (SIZEit) + β5 (MBit) + β6 (BETAit) + β7 (EPit) + β8 (AdjRETit) + β9 (ACCit) + β10 (Cash3ETRit) + β11 (LEVit) + β12 (GROWTHit) + εit (5) See Appendix 1 for variable descriptions. The year fixed and firm fixed effects are included in the models. The standard errors are clustered at the firm level.

ATS is also decreased compared with those starting to use ATS (0.0322 vs. 0.0501), although the difference in the mean fails to reach statistical significance (t = −1.2314, p = .2182). After firm-years cease to use ATS

Table 9 Regression results from Eq. (8). Variable

Parameter estimate

Robust Std. Err.

z-Value

p-Value

INTERCEPT TEMPit /Ait-1 STARTit TEMPit/Ait-1 * STARTit CONTit TEMPit/Ait-1 * CONTit ENDit TEMPit/Ait-1* ENDit SIZEit MBit BETAit EPit AdjRETit ACCit Cash3ETRit LEVit GROWTHit Year fixed effect Firm fixed effect N = 7756 Adj. R2 = 0.1342

0.1602 −0.0228 −0.0661 0.0187 −0.0272 0.0146 0.0601 0.0023 −0.0128 −0.0133 0.0058 −0.0073 0.0787 −0.0082 0.0035 0.0164 −0.0086 Included Included

0.0888 0.0062 0.0830 0.0106 0.0419 0.0040 0.1354 0.0205 0.0045 0.0055 0.0038 0.0058 0.0455 0.0049 0.0042 0.0055 0.0041

1.80 −3.64 −0.80 1.77 −0.65 3.59 0.44 0.12 −2.81 −2.41 1.49 −1.25 1.73 −1.65 0.83 2.98 −2.08

0.071 0.000 0.426 0.077 0.516 0.000 0.657 0.908 0.005 0.016 0.135 0.211 0.084 0.098 0.407 0.003 0.037

AdjRETit+1 = α0 + β1 (TEMPit/Ait-1) + β2 (STARTit) + β3 (TEMPit/Ait-1 * STARTit) + β4 (CONTit) + β5 (TEMPit/Ait-1 * CONTit) + β6 (ENDit) + β7 (TEMPit/Ait-1* ENDit) + β8 (SIZEit) + β9 (MBit) + β10 (BETAit) + β11 (EPit) + β12 (AdjRETit) + β13 (ACCit) + β14 (Cash3ETRit) + β15 (LEVit) + β16 (GROWTHit) + εit (8) See Appendix 1 for variable descriptions. The standard errors are clustered at the firm level.

Please cite this article as: B. Luo, Effects of auditor-provided tax services on book-tax differences and on investors' mispricing of boo..., Advances in Accounting, https://doi.org/10.1016/j.adiac.2019.100434

B. Luo / Advances in Accounting xxx (xxxx) xxx

(END = 1), the mean of the temporary book-tax differences (TEMPit / Ait-1) increases significantly compared with firm-years that continue to use ATS (0.0014 vs. 0.0003; t = 2.8392, p = .0045). The mean of the adjusted returns (AdjRETit+1) declines significantly compared with those that continue to use ATS (−0.0106 vs. 0.0322; t = −5.4062, p b .0001). It should be noted that the above descriptive statistics do not consider any firm- or year- fixed effect; therefore, they may not correctly reflect the effect of the start, continuation, and termination of ATS on firms' level of book-tax differences or the mispricing of booktax differences. Table 9 shows the results from the regression analysis of Eq.(8). As expected, the coefficient on TEMP is negative and significant at the 0.01 level, which is consistent with the mispricing of book-tax differences (Chi et al., 2014; Lev & Nissim, 2004). The coefficient on TEMP*CONT (β5 = 0.0146) is significantly positive at the p b .01 level, which suggests a significantly decreased mispricing of booktax differences during years when firms continue to acquire ATS. However, the coefficients on the interaction term TEMP*START and TEMP*END are both insignificant at the p b .05 level, indicating that ATS may not have an immediate impact on investors' mispricing of book-tax differences when firms start or cease to use ATS. Collectively, these results imply that investors are most likely to better price a given firm when the firm acquires ATS continuously, possibly because of the improved earnings quality. 6. Conclusion

13

mispricing of book-tax differences. The joint provision of audit and non-audit services has been a controversial issue for decades. SOX of 2002 prohibited most non-audit services, but ATS is one of the few non-audit services still allowed in the post-SOX period. In 2005, the SEC further limited the provision of certain types of ATS. Despite the claim by regulators that non-audit services, including ATS, may be a threat to auditor independence, academic evidence of the effects of ATS on audit quality and auditor independence has been inconclusive. The present study provides valuable empirical evidence supporting the knowledge spillover hypothesis and suggests that the provision of ATS is associated with lower levels of temporary book-tax differences among firms and decreased mispricing of book-tax differences among investors. The empirical results obtained in this study are not consistent with the concern that ATS lead to impaired auditor independence as a result of the increased economic bonds between auditors and client firms. These empirical results provide new insights to the ongoing debate on whether ATS should be continued. Overall, this study contributes to the research on the impact of book-tax differences on a firm's overall accounting quality and the implications of book-tax differences for investors. Acknowledgments I am grateful for the guidance from my dissertation committee members, Teresa Lightner (chair), Eric Rapley, and Robert Pavur. I also appreciate the helpful comments and suggestions from Roger Graham (editor) and anonymous referees.

This study examines the association between ATS and firms' levels of book-tax differences and the association between ATS and investors' Appendix 1. Definition of variables Variables

Description

A ABACC

Total assets Abnormal accruals, measured as deviations from the predicted values from the corresponding industry-year regression TAccit /Ait-1 = α0 /Ait-1 + β1 (ΔRit - ΔARit) /Ait-1 + β2 (PPEit / Ait-1) + εit. A is the total asset, R is revenue, AR is accounts receivables, and PPE is gross property, plant, and equipment ACC Ratio of accruals to average total assets, where accruals are (ΔCurrent Assets - ΔCash) - (ΔCurrent Liabilities - ΔDebt included in current liabilities) - ΔDeferred Tax Liability - ΔDepreciation AdjRET Adjusted Returns, measured as annual returns adjusted for market returns with the same size and book-to-market. Return accumulation begins in the fourth month after the fiscal year-end to allow the dissemination of financial reports. Annual returnsit = (Priceit + Dividendit) / Adjustmentit) / ((Priceit-1 + Dividendit-1) / Adjustmentit-1) - 1. Price is the unadjusted closing price for the year, Dividend is dividend per share, and Adjustment is the cumulative adjustment factor ATS An indicator variable that is set to 1 if a firm's tax fee is not equal to zero and 0 otherwise AUDIND Auditor independence from the client, measured as non-audit fees less tax fees divided by total audit fees received from the client BETA Systematic risk estimated from regression of monthly raw returns on the value-weighted market portfolio returns (including distribution) over a 60-month return period prior to the adjusted return accumulation period. Fifteen months of return data are required to calculate beta BIG4 An indicator variable set to 1 if the firm is audited by one of the Big 4 auditors and 0 otherwise BTM Book-to-market ratio at the end of the year, measured as book value of equity divided by market value of equity CASH Cash holding at the end of the year divided by total assets at the beginning of the year Cash3ETR Three-year cash effective tax rate = the sum of cash tax payments over the past 3 years / the sum of pre-tax income less special items over the past 3 years CONT An indicator variable that is set to 1 if ATS appears for a firm in both a given year and a prior year and 0 otherwise DEP Depreciation and amortization expense for the year divided by total assets at the beginning of the year END An indicator variable that is set to 1 if a specific year is the year immediately after the last ATS is provided for a firm and 0 otherwise EP Earnings-to-price ratio EQINC Equity income scaled by total assets at the beginning of the year FI Pre-tax foreign income scaled by total assets at the beginning of the year GROWTH Percentage change in sales INSTOWN Percentage of shares owned by institutions at the beginning of the year LAF Natural log of audit fees received from the client LEV Ratio of total liabilities to total assets MB Market-to-book ratio at the end of the year, measured as market value of equity divided by book value of equity MERGER An indicator variable equal to 1 if a firm participates in any merger activity during the year and 0 otherwise NOL An indicator variable set to 1 if there is a tax loss carryforward during the year and 0 otherwise OPPOR Market value of a client divided by the sum of the market value of all clients in the same industry at the same MSA city PPE Net PPE scaled by total assets at the beginning of the year R&D R&D expense scaled by total assets at the beginning of the year ROA Return on assets, measured as the ratio of income before extraordinary items to the average of total assets for the year SECTIER An indicator variable set to 1 if the firm is audited by a second-tier accounting firm, namely Grant Thornton LLP and BDO Seidman LLP, and 0 otherwise SIZE Natural log of market value of equity SLEV Long-termdebt-to-asset ratio at the end of the year scaled by total assets at the end of the year START An indicator variable that is set to 1 if a specific year is the first year that ATS appeared for a firm and 0 otherwise (continued on next page)

Please cite this article as: B. Luo, Effects of auditor-provided tax services on book-tax differences and on investors' mispricing of boo..., Advances in Accounting, https://doi.org/10.1016/j.adiac.2019.100434

14

B. Luo / Advances in Accounting xxx (xxxx) xxx

(continued) Variables

Description

TAcc

Total accruals = ΔCA - ΔCL - ΔCASH + ΔSTD - DEP. CA is current asset, CL is current liabilities, CASH is cash and cash equivalent, STD is short-term debt included in the current liabilities, and DEP is depreciation expense Fees paid to the auditor for tax services Temporary book-tax differences = (Deferred federal tax expenses + Deferred foreign tax expenses) / 0.35. If either deferred federal tax expenses or deferred foreign tax expense is missing, temporary book-tax differences = Total deferred taxes expenses / 0.35 Length of the audit firm's tenure with its client.

TAXFEE TEMP TENURE

Appendix 2. Appendix Table A2.1 First stage probit regression results from Eq. (1). Variable

Parameter Estimate

Standard Error

z-Value

p-Value

INTERCEPT TENUREit INSTOWNit MERGERit AUDINDit LAFit OPPORit SIZEit ABACCit NOLit ΔNOLit EQINCit FIit R&Dit LEVit BTMit PPEit ROAit CASHit DEPit BIG4it SECTIERit Industry fixed effect Year fixed effect N = 36,622 Log likelihood = −2425.9815 Pseudo R2 = 0.1772

−3.8662 0.0037 −0.0033 −0.0095 −0.0960 0.2536 0.2326 0.0395 0.0055 0.1487 −0.0729 0.4664 0.2880 −0.2352 0.1068 0.0017 −0.3696 −0.0105 0.0560 0.6672 0.7081 0.3549 Included Included

0.4404 0.0028 0.0033 0.0610 0.0353 0.0366 0.0682 0.0212 0.0119 0.0736 0.0649 2.6940 0.3606 0.2822 0.1300 0.0020 0.1983 0.1802 0.1035 1.1194 0.0708 0.2822

−8.78 1.31 −1.00 −0.16 −2.72 6.91 3.41 1.86 0.46 2.02 −1.12 0.17 0.80 −0.83 0.82 0.82 −1.86 −0.06 0.54 0.60 9.99 1.26

0.000 0.191 0.320 0.876 0.007 0.000 0.001 0.063 0.644 0.043 0.262 0.863 0.424 0.404 0.411 0.415 0.062 0.953 0.588 0.551 0.000 0.209

ATSit = α0 + β1 (TENUREit) + β2 (INSTOWNit) + β3 (MERGERit) + β4 (AUDINDit) + β5 (LAFit) + β6 (OPPORit) + β7 (SIZEit) + β8 (ABACCit) + β9 (NOLit) + β10 (ΔNOLit) + β11 (EQINCit) + β12 (FIit) + β13 (R&Dit) + β14 (LEVit) + β15 (BTMit) + β16 (PPEit) + β17 (ROAit) + β18 (CASHit) + β19 (DEPit) + β20 (BIG4it) + β21 (SECTIERit) + εit (1) See Appendix 1 for variable descriptions. The standard errors are clustered at the firm level. Table A2.2 Descriptive statistics for Eq. (8). START = 1

AdjRETit+1 TEMPit /Ait-1 SIZEit MBit BETAit EPit AdjRETit ACCit Cash3ETRit LEVit GROWTHit

CONT = 1

END = 1

Mean

Std. Dev.

Mean

Std. Dev.

Mean

Std. Dev.

0.0501 0.0032 7.0945 2.9280 1.2158 0.0250 0.0622 −94.44 0.1775 0.4770 0.0858

0.2868 0.0139 1.4264 1.8271 0.6240 0.0521 0.2908 181.55 0.1344 0.1899 0.1306

0.0322 0.0003 7.5752 2.9565 1.2844 0.0332 0.0407 −145.35 0.1891 0.4892 0.0814

0.2542 0.0130 1.4435 1.7069 0.5890 0.0466 0.2554 258.11 0.1228 0.1814 0.1276

−0.0106 0.0014 7.6465 3.0360 1.2349 0.0331 0.0388 −182.22 0.1740 0.4915 0.1136

0.2303 0.0095 1.2656 1.5901 0.65451 0.0476 0.2489 270.26 0.1226 0.1673 0.1230

See Appendix 1 for variable descriptions.

References Agrawal, A., & Chadha, S. (2005). Corporate governance and accounting scandals. Journal of Law and Economics, 48(2), 371–406. Antle, R., Gordon, E., Narayanamoorthy, G., & Zhou, L. (2006). The joint determination of audit fees, nonaudit fees, and abnormal accruals. Review of Quantitative Finance and Accounting, 27, 235–266.

Antle, R., Gordon, E. A., Narayanamoorthy, G., & Zhou, L. (2002). The joint determination of audit fees, non-audit fees, and abnormal accruals. Working paper Yale University. Arrunada, B. (1999). The provision of non-audit services by auditors: Let the market evolve and decide. International Review of Law and Economics, 19(4), 513–531. Ashbaugh, H., LaFond, R., & Mayhew, B. (2003). Do nonaudit services compromise auditor independence? Further evidence. The Accounting Review, 78, 611–639.

Please cite this article as: B. Luo, Effects of auditor-provided tax services on book-tax differences and on investors' mispricing of boo..., Advances in Accounting, https://doi.org/10.1016/j.adiac.2019.100434

B. Luo / Advances in Accounting xxx (xxxx) xxx Badertscher, B., Phillips, J. D., Pincus, M., & Rego, S. O. (2009). Earnings management strategies and the trade-off between tax benefits and detection risk: To conform or not to conform? The Accounting Review, 84(1), 83–97. Basu, S. (1977). Investment performance of common stocks in relation to their priceearnings ratios: A test of the efficient market hypothesis. Journal of Finance, 32(3), 663–682. Beck, P. J., Frecka, T. J., & Solomon, I. (1988). A model of the market for NAS and audit services: Knowledge spillovers and auditor-auditee bonding. Journal of Accounting Literature, 7(1), 50–64. Bernard, V., & Thomas, J. (1990). Evidence that stock prices do not fully reflect the implications of current earnings for future earnings. Journal of Accounting and Economics, 13(4), 305–340. Blaylock, B., Shevlin, T., & Wilson, R. J. (2012). Tax avoidance, large positive temporary book-tax differences, and earnings persistence. The Accounting Review, 87(1), 91–120. Brown, L. (2001). A temporal analysis of earnings surprises: Profits and losses. Journal of Accounting Research, 39(2), 221–241. Call, A. C., Chen, S., Miao, B., & Tong, Y. H. (2014). Short-term earnings guidance and accrual-based earnings management. Review of Accounting Studies, 19(2), 955–987. Chan, L. K. C., & Lakonishok, J. (2004). Value and growth investing: Review and update. Financial Analysts Journal, 60(1), 71–86. Chi, S., Pincus, M., & Teoh, S. H. (2014). Mispricing of book-tax differences and the trading behavior of short sellers and insiders. The Accounting Review, 89(2), 511–543. Cho, J., Wong, J., & Wong, N. (2006). Book-tax differences and inland revenue audit adjustments in New Zealand. Journal of Business Finance and Accounting, 33, 1650–1667. Choi, W., Lee, H., & Jun, B. W. (2009). The provision of tax services by incumbent auditors and earnings management: Evidence from Korea. Journal of International Financial Management & Accounting, 20(1), 79–103. Christensen, B. E., Olson, T., & Omer, T. (2013). Auditor-provided tax services and earnings management through tax accounts. Working paper Texas A&M University. Chung, H., & Kallapur, S. (2003). Client importance, nonaudit services, and abnormal accruals. The Accounting Review, 78, 931–955. Dechow, P., Sloan, R., & Sweeney, A. (1995). Detecting earnings management. The Accounting Review, 70, 193–225. DeSimone, L., & Stomberg, B. (2012). Do investors differentially value tax avoidance of income mobile firms? Working paper Stanford Graduate School of Business, and University of Georgia. Dhaliwal, D., Gleason, C., & Mills, L. (2004). Last-chance earnings management: Using the tax expense to meet analysts' forecasts. Contemporary Accounting Research, 21, 431–459. Dyreng, S., Hanlon, M., & Maydew, E. (2008). Long-run corporate tax avoidance. The Accounting Review, 83(1), 61–82. Ettredge, M., Sun, L., Lee, P., & Anandarajan, A. (2008). Is earnings fraud associated with high deferred tax and/or book minus tax levels? Auditing: A Journal of Practice & Theory, 27, 1–33. Fama, E. (1970). Efficient capital markets: A review of theory and empirical work. Journal of Finance, 25, 383–417. Fama, E., & French, K. R. (1992). The cross-section of expected stock returns. The Journal of Finance, 47, 465–3427. Fama, E., & French, K. R. (1995). Size and book-to-market factors in earnings and returns. The Journal of Finance L(1), 131–155. Ferguson, M. J., Seow, G. S., & Young, D. (2004). Nonaudit services and earnings management: U.K. evidence. Contemporary Accounting Research, 21(4), 813–841. Francis, J. R. (2006). Are auditors compromised by nonaudit services? Assessing the evidence. Contemporary Accounting Research, 23(3), 747–760. Frank, M., Lynch, L. J., & Rego, S. (2009). Tax reporting aggressiveness and its relation to aggressive financial reporting. The Accounting Review, 84, 467–496. Frankel, R. M., Johnson, M. F., & Nelson, K. K. (2002). The relation between auditors' fees for nonaudit services and earnings management. The Accounting Review, 77, 71–105. Gleason, C. A., & Mills, L. F. (2011). Do auditor-provided tax services improve the estimate of tax reserves? Contemporary Accounting Research, 28(5), 1484–1509. Goncharov, I. (2009). Does reporting timeliness affect book-tax differences? working paper University of Amsterdam. Gong, G., Li, L. Y., & Xie, H. (2009). The association between management earnings forecast errors and accruals. The Accounting Review, 84(2), 497–530. Guenther, D. A., Matsunaga, S. R., & Williams, B. M. (2017). Is tax avoidance related to firm risk? The Accounting Review, 92(1), 115–136. Gupta, S., Mills, L., & Towery, E. (2014). The effect of mandatory financial statement disclosures of tax uncertainty on tax reporting and collections: The case of FIN 48 and multistate tax avoidance. The Journal of American Taxation Association, 36(2), 203–229. Habib, A. (2012). Non-audit service fees and financial reporting quality: A meta-analysis. Abacus, 48(2), 214–248. Hanlon, M. (2005). The persistence and pricing of earnings, accruals, and cash flows when firms have large book-tax differences. The Accounting Review, 80, 137–166. Hanlon, M., & Heitzman, S. (2010). A review of tax research. Journal of Accounting and Economics, 50, 127–178. Harrington, C., & Smith, W. (2012). Tax avoidance and corporate capital structure. Journal of Finance and Accountancy, 11, 144–164.

15

Jegadeesh, N., & Titman, S. (1993). Returns to buying winners and selling losers: Implications for stock market efficiency. The Journal of Finance, 48(1), 65–91. Jones, A., & Noga, T. (2013). Book-tax differences as an indicator of financial distress. Accounting Horizons, 27, 469–489. Kahneman, D. (1973). Attention and effort. Englewood Cliffs, NJ: Prentice-Hall. Kahneman, D., & Tversky, A. (1973). On the psychology of prediction. Psychological Review, 80, 237–251. Kanagaretnam, K., Krishnan, G., & Lobo, G. (2011). An empirical analysis of auditor independence in the banking industry. The Accounting Review, 85, 2011–2046. Kinney, W., Palmrose, Z., & Scholz, S. (2004). Auditor independence, non-audit services, and restatements: Was the U.S. government right? Journal of Accounting Research, 42, 561–588. Krishnan, G. V., & Visvanathan, G. (2011). Is there an association between earnings management and auditor provided tax services? The Journal of the American Taxation Association, 33, 111–135. Krishnan, G. V., Visvanathan, G., & Yu, W. (2013). Do auditor-provided tax services enhance or impair the value relevance of earnings? The Journal of the American Taxation Association, 35(1), 1–19. Laffie, L. (2006). SEC approves PCAOB rules for tax services. The Tax Adviser, 37, 444–445. Lakonishok, J., Shleifer, A., & Vishny, R. (1994). Contrarian investment, extrapolation and risk. Journal of Finance, 49, 1541–1578. Lassila, D. R., Omer, T. C., Shelley, M. K., & Smith, L. M. (2010). Do complexity, governance, and auditor independence influence whether firms retain their auditors for tax services? The Journal of the American Taxation Association, 32, 1–23. Lennox, C. S., & Park, C. W. (2006). The informativeness of earnings and management's issuance of earnings forecasts. Journal of Accounting and Economics, 42(3), 439–458. Lev, B., & Nissim, D. (2004). Taxable income, future earnings, and equity values. The Accounting Review, 79, 1039–1074. Libby, R., Bloomfield, R., & Nelson, M. (2002). Experimental research in financial accounting. Accounting, Organizations and Society, 27(8), 775–810. Manzon, G., & Plesko, G. (2002). The relation between financial and tax reporting measures of income. Tax Law Review, 55, 175–214. Matsumoto, D. (2002). Management's incentives to avoid negative earnings surprises. The Accounting Review, 77(3), 483–514. Maydew, E. L., & Shackelford, D. A. (2007). The changing role of auditors in corporate tax planning. Taxing corporate income in the 21st century. Cambridge University Press. McGuire, S., Omer, T., & Wang, D. (2012). Tax avoidance: Does tax-specific industry expertise make a difference? The Accounting Review, 87(3), 975–1003. Mills, L. (1998). Book-tax differences and internal revenue service adjustments. Journal of Accounting Research, 36(2), 343–356. Mills, L., & Newberry, K. (2001). The influence of tax and non-tax costs on book-tax reporting differences: Public and private firms. The Journal of American Taxation Association, 23, 1–19. Mills, L., Newberry, K., & Novack, G. (2003). How well do Compustat NOL data identify firms with U.S. tax return loss carryovers? Journal of the American Taxation Association, 25(2), 1–17. Nisbett, R., & Ross, L. (1980). Human inference: Strategies and shortcomings of social judgment. Englewood Cliffs, NJ: Prentice-Hall. Omer, T. C., Bedard, J. C., & Falsetta, D. (2006). Auditor-provided tax services: The effects of a changing regulatory environment. The Accounting Review, 81(5), 1095–1117. Phillips, J., Pincus, M., & Rego, S. O. (2003). Earnings management: New evidence based on deferred tax expense. The Accounting Review(2), 491–521. Plesko, G. (2004). Corporate tax avoidance and the properties of corporate earnings. National Tax Journal, 57(3), 729–737. Public Company Accounting Oversight Board (PCAOB) (2004). Auditor independence tax services roundtable. Retrieved from http://pcaobus.org/standards/standards_and_ related_rules/2004074_roundtable_transcript.pdf. Racca, J. (2011). Stable book-tax differences, prior earnings and earnings persistence. Working paper University of North Texas. Raghunandan, K., Read, W. R., & Whisenant, J. S. (2003). Initial evidence on the association between nonaudit fees and restated financial statements. Accounting Horizons, 17(3), 223–224. Schneider, A., Church, B. K., & Ely, K. M. (2006). Non-audit services and auditor independence: A review of the literature. Journal of Accounting Literature, 25, 169–211. Seetharaman, A., Sun, Y., & Wang, W. (2011). Tax-related financial statement restatements and auditor provided tax services. Journal of Accounting, Auditing and Finance, 26(4), 677–698. Simunic, D. (1984). Auditing, consulting, and auditor independence. Journal of Accounting Research, 22(2), 679–702. Weber, D. (2009). Do analysts and investors fully appreciate the implications of book-tax differences for future earnings? Contemporary Accounting Research, 26(4), 1175–1206. Wilson, R. (2009). An examination of corporate tax shelter participants. The Accounting Review, 84(3), 969–999.

Please cite this article as: B. Luo, Effects of auditor-provided tax services on book-tax differences and on investors' mispricing of boo..., Advances in Accounting, https://doi.org/10.1016/j.adiac.2019.100434