Access the full text.
Sign up today, get DeepDyve free for 14 days.
Loading next page...
/lp/ssrn/interim-tax-reporting-accuracy-determinants-and-consequences-vIrpZ4h54x
References
References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.
- Copyright
- Copyright © 2021 Elsevier Inc.
- eISSN
- ARCH-4059
- DOI
- 10.2139/ssrn.3774393
- Publisher site
- See Article on Publisher Site
Abstract
Erik L. Beardsley Mendoza College of Business University of Notre Dame Erik.L.Beardsley.1@nd.edu January 2021 ABSTRACT This study examines the determinants and consequences of tax reporting accuracy. Firms are required to report tax expense each quarter based on their estimated annual effective tax rate (ETR). However, because of both bias and estimation error, these estimates do not always accurately represent annual ETR, creating variation in ETRs and earnings. This study documents several factors that are negatively associated with interim tax reporting accuracy but have no association with bias, suggesting that these factors significantly contribute to estimation error within the tax accounts. Consistent with a monitoring role over financial reporting, analyst following, institutional ownership, and auditor tenure are positively associated with interim tax reporting accuracy through reduced estimation error. In addition, estimated taxable income is more informative to investors for firms that have been more accurate in prior years, and investors respond more positively to beating analysts’ forecasts using a decrease in the tax rate when the firm has a record of accurate tax reporting. Keywords: accounting for income taxes; market pricing; audit quality; analyst coverage; institutional investors; earnings management JEL classification: G10, G14, M40, M41, M42, H20 Acknowledgements: This study stems from my dissertation completed at Texas A&M University. I thank my dissertation committee: Connie Weaver (chair), Sean McGuire, Senyo Tse, and Dudley Poston for their support. I also thank Brad Badertscher, Andy Duxbury, Matt Ege, Jen Glenn, Michelle Hanlon, Stephannie Larocque, Sean McCarthy, Lil Mills, Stevie Neuman, Sarah Rice, David Weber, and the University of Iowa Tax Readings Group for helpful comments. I appreciate suggestions from workshop participants at Texas A&M University, the University of Illinois at Chicago, and the University of Notre Dame. I gratefully acknowledge generous financial support from the AICPA Accounting Doctoral Scholarship. Electronic copy available at: https://ssrn.com/abstract=3774393 1. Introduction The purpose of reported income tax in the financial statements is to accurately portray the current financial performance of the firm (Graham et al. [2012]). However, the tax accounts are one of the most complex areas of financial reporting and are among the most restated accounts (Whalen et al. [2020], Deloitte [2011]), suggesting that taxes are not always accurately reported in GAAP financial statements. We presently know little about the determinants of tax reporting accuracy or whether the market values accuracy when assessing the tax accounts. While academic research has given significant attention to variation in ETRs between firms (e.g., Dyreng et al. [2008], Hanlon and Heitzman [2010], Wilde and Wilson [2018]), an unexplored issue is the causes and consequences of variation in ETRs for a firm within a given year. This study exploits the reporting requirements in APB 28, Interim Financial Reporting, to provide insight into the determinants of interim tax reporting accuracy as well as the effect of tax reporting accuracy on investor use of reported tax expense. I define interim tax reporting accuracy as the extent to which firms’ reported interim GAAP ETRs predict their annual GAAP ETRs. APB 28 requires companies to estimate the annual ETR at each interim period and to allocate tax expense on a pro rata basis, providing a point estimate of the annual ETR. The predominant perception is that interim ETR estimates are biased and firms bias ETR estimates to reach earnings benchmarks (Dhaliwal et al. [2004]). However, the level of estimation error is a distinct and unexplored dimension of interim ETR reporting. Because ETR accuracy is a function of both estimation error and bias (i.e., both contribute to the inaccuracy of ETR estimates), these factors must be separated in order to A notable exception is when firms have “discrete” items that must be recorded fully in the quarter they occur, potentially distorting the role of interim ETRs representing predicted annual ETR. As detailed in Section 5, I re-run the analyses using only firm-quarters without discrete items and obtain similar results. Electronic copy available at: https://ssrn.com/abstract=3774393 examine estimation error directly. That is, the factors associated with ETR accuracy (i.e., the absolute difference between estimated and year-end ETRs) but not ETR bias (i.e., the systematic signed differences between the estimated and year-end ETRs) contribute to ETR estimation error (i.e., unbiased errors in ETR reporting). It is important to examine the determinants of interim tax reporting accuracy for at least three reasons. First, tax expense is substantial for most firms (Dyreng et al. [2008], Graham et al. [2012]) and changes in tax rates can produce significant changes in after-tax earnings. For example, Comprix et al. [2012] report that the standard deviation of changes of estimated ETRs from first to second, second to third, and third to fourth quarters are 6.0 percent, 6.9 percent, and 8.7 percent, respectively, in their post-SOX sample. Using the federal statutory tax rate of 35 percent during their sample period as a baseline, a one standard deviation change in ETR would change after-tax earnings by 9.2 percent, 10.5 percent, and 13.4 percent, respectively. In addition, the GAAP ETR is an important input for investors and analysts’ forecasting and valuation because of its effect on earnings (CFA Institute [2016]). Thus, ETR estimates and subsequent changes can have a significant effect on after-tax earnings and valuation. Second, ETR changes are important to firm managers. Graham et al. [2014] report that 84 percent of tax executives rate the GAAP ETR at least as important as cash taxes paid. Further, the Tax Executive Institute Corporate Tax Department Survey indicates the most common measurement used to evaluate the tax departments is “lack of surprises” (TEI [2011-2012]). My discussions with several Big 4 tax partners indicate that both earnings increasing and earnings decreasing ETR surprises tend to be viewed negatively by investors because they provide signals For example, the first quarter standard deviation of 6.0% changes after-tax earnings by 6.0/(1-0.35) = 9.2%. Electronic copy available at: https://ssrn.com/abstract=3774393 about the credibility of the tax department. For example, even an ETR surprise that has a positive effect on earnings could be viewed as “lucky” and spur distrust from interested parties. To the extent that surprises in year-end ETRs are revealed through inaccurate interim ETR estimates, this study directly relates to one of the most important measurements used to evaluate tax departments. Third, prior ETR reporting accuracy could be indicative of the amount of error in the tax accounts and could have implications for how the market understands and responds to tax information. Prior research suggests that the tax accounts provide incremental information to pretax book income (Hanlon et al. [2005]). However, an open question in the literature is whether the market is inefficient with regard to current period tax information or whether an important pricing factor is being ignored (Graham et al. [2012]). Firms may differ in the reliability of their tax reporting, creating heterogeneity in users’ use of reported tax expense. That is, firms with more accurate tax reporting likely have a stronger reaction to the information in the tax accounts. This study documents firm characteristics that are associated with ETR bias and accuracy. Given the considerable variation in ETRs among firms (Dyreng et al. [2008]), changes in quarterly ETRs (Bauman and Shaw [2005], Comprix et al. [2012]), potential information content in the tax accounts (Lev and Nissim [2004], Hanlon [2005], Ayers et al. [2009]), and the complexity of the tax accounts, it is important to document what specific factors contribute to ETR bias and accuracy. Even the tax expense reported at the end of the year can be considered an estimate because the actual tax return is filed several months later and is subject to possible IRS audit adjustments. Electronic copy available at: https://ssrn.com/abstract=3774393 This study also examines the potential monitoring role of external parties on tax reporting accuracy and bias. Specifically, I examine whether institutional owners, analysts, or auditors influence interim tax reporting via monitoring roles. Because this study examines both accuracy and bias, it provides insight into the role of external parties that may improve interim ETR reporting through more accurate or less biased estimates. Institutional owners can serve as monitors who improve the accuracy of voluntary earnings forecasts (Ajinkya et al. [2005]). However, prior research further suggests that investors pressure managers to meet earnings targets, leading to more biased and less accurate estimates. Likewise, analysts can serve a monitoring role by scrutinizing management behavior (Chung and Jo [1996], Chen et al. [2015]), which affects corporate policy decisions (Derrien and Kecskés [2013], He and Tian [2013], Allen et al. [2016]). However, managers are also pressured by analysts to meet optimistic earnings targets (Habib and Hansen [2008]), potentially leading to manipulation of tax expense in order to meet these targets. Prior research suggests that certain characteristics such as Big 4 auditors, auditor-provided tax services, auditor tenure, office size, and expertise may be associated with higher audit quality (see DeFond and Zhang [2014] for a review). However, because an interim review substantially differs from a year-end audit and taxes are a challenging area for auditors (Deloitte [2011]), it is unclear whether auditor characteristics improve tax reporting accuracy via the interim review process. Results indicate that that geographic complexity, changes in geographic mix of income, discontinued and extraordinary items, deferred tax assets, and R&D activity are negatively associated with interim tax reporting accuracy. However, these factors are not associated with ETR bias, providing evidence that the observed effect on accuracy is due to these factors’ association with estimation error of ETR estimates. Further analyses show that these factors are Electronic copy available at: https://ssrn.com/abstract=3774393 negatively associated with “de-biased” ETR changes, providing direct evidence regarding their association with estimation error. Results also indicate that analyst following, institutional ownership, and auditor tenure are positively associated with interim tax reporting accuracy, consistent with a monitoring role that improves the accuracy of ETR estimates. These findings are new to the literature and differ from prior work that examines ETR estimation bias because I provide evidence that external monitors reduce estimation error of ETR estimates, which is distinct from systematic bias. Importantly, to examine the consequences of tax reporting accuracy, this study investigates whether investor use of current tax information varies predictively with prior tax reporting accuracy. I follow Hanlon et al. [2005] and test whether investor reaction to estimated taxable income varies with prior tax reporting accuracy. The results suggest that the relation between stock returns and changes in estimated taxable income increases with prior tax reporting accuracy, consistent with investors perceiving estimated taxable income as a more reliable performance measure when the firm has a record of accurate tax reporting. Gleason and Mills [2008] use a short-window market test and find that the reward for beating the analysts’ target is significantly discounted for firms that beat the target by decreasing their ETRs from third to fourth quarter. This study examines whether the discount varies with prior tax reporting accuracy by replicating their test using the current sample. I find that the market discount for using a decrease in tax expense to beat analysts’ forecast, consistent with Gleason and Mills [2008]. However, there is no market discount for firms that that have a history of accurate tax reporting. Taken together, these results provide evidence that investor use of reported tax expense varies with prior tax reporting accuracy. Electronic copy available at: https://ssrn.com/abstract=3774393 This study makes several contributions to the literature. First, it documents factors that contribute to ETR accuracy through estimation error, an important dimension that has not been examined in prior research. Second, although the predominant perception is that interim ETR estimates are biased because of pressure to meet market expectations, the findings in this study indicate that analysts and institutional owners serve a monitoring role that improves interim tax reporting accuracy, rather than a pressure role that results in more biased and less accurate tax reporting. Third, this study finds that auditor tenure is associated with improved interim ETR accuracy, consistent with a more effective interim review when the auditor has a continuing relationship with the client. Finally, this study contributes to the literature that examines the pricing of tax information reported in financial statements by providing insight into how the market values the information in the tax accounts. 2. Background and Hypotheses 2.1 INTERIM TAX REPORTING APB Opinion No. 28, Interim Financial Reporting, requires companies to estimate the annual ETR based on facts and circumstances known at each interim period and to allocate tax expense on a pro rata basis under the integral method. The estimated ETR should reflect anticipated investment tax credits, foreign tax rates, percentage depletion, capital gains rates, and other tax planning alternatives (ASC 740-270-30-8). Appendix A provides an example of this computation. The estimates may not accurately represent the actual annual ETR for a variety of reasons. However, with limited information provided during interim periods, the reported interim The integral method applies to other accounts such as cost of goods sold. However, because the people and processes involved in preparing ETR estimates substantially differ from other accounts (Choudhary et al., 2016), caution should be taken about extending inferences beyond the tax accounts. Electronic copy available at: https://ssrn.com/abstract=3774393 ETR (and subsequent changes to it) could serve as a summary measure regarding the tax position of the firm. If the purpose of reported tax expense is to accurately portray the current financial performance of the firm (Graham et al. [2012]), the accuracy with which firms report interim tax expense may provide information regarding the extent to which reported tax expense serves this purpose. When firms consistently report less accurate ETRs than other firms, their reported tax expense is likely less reliable. Prior research has examined interim tax reporting in different ways. Dhaliwal et al. [2004] report that firms use tax expense to manage earnings by decreasing ETRs from third to fourth quarter to meet analysts’ forecasts. Comprix et al. [2012] document that quarterly ETR estimates are systematically biased upward, creating “slack” that can be used to manage earnings. Schmidt [2006] finds that the initial “tax change component” of earnings is more persistent than the revised tax change component, but the market underweights the forecasting implications of the tax change component. Taken together, these studies suggest that ETR estimates often change significantly but users do not appear to fully utilize the forecasting implications of changes in ETR. Recently, Bratten et al. [2017] examine analysts’ use of management ETR estimates. They find that complexity and discrete items impair management ETR accuracy, and that analysts’ ETR and EPS forecasts are less disperse when management Firms are not required to provide a rate reconciliation during interim periods, and often provide boilerplate statements (e.g., primarily due to mix of foreign income) regarding reasons their ETR deviates from the statutory rate or prior year’s ETR. However, disclosure practices significantly vary across firms. Consistent with this line of research, I find a similar level of bias in my sample. However, while the average ETR decreases during the year, there is a significant portion of ETR increases, suggesting estimation error in addition to bias. Electronic copy available at: https://ssrn.com/abstract=3774393 estimates are more accurate. Following section discuss how external monitors, namely analysts, investors, and auditors, may influence ETR reporting accuracy. 2.2 FINANCIAL ANALYSTS There is significant tension regarding the role that analysts play with regard to the accuracy of tax reporting. On one hand, prior research suggests financial analysts play a monitoring role by scrutinizing management behavior, disseminating information, and improving transparency (Jensen and Meckling [1976], Healy and Palepu [2001], Chen et al. [2015]). Allen et al. [2016] find that analyst coverage is associated with reduced corporate tax avoidance, consistent with a monitoring role. On the other hand, managers are also pressured by analysts to meet optimistic earnings targets (Healy and Wahlen [1999]; Bartov et al. [2002], Graham et al. [2005]), leading to decreases in estimated ETR from third to fourth quarter as a way to manage earnings. Based on the above arguments, analyst coverage could increase interim tax reporting accuracy through increased scrutiny, or decrease interim tax reporting accuracy as a result of pressure to manipulate their ETR to meet earnings targets. However, analysts may have no effect on interim tax reporting accuracy. Although recent research suggests analysts exert effort to understand complex tax situations and identify transitory tax items (Bratten et al. [2017], Beardsley et al. [2020]), analysts also may not incorporate all complex information in their estimates (Plumlee [2003], Shane and Stock [2006] Weber [2009]). If ETR estimates are complex and difficult for management to provide, analysts This study differs from Bratten et al. [2017] in several important ways. First, while they focus on analysts’ use of management’s reported ETR, this study examines investors’ use. Second, this study addresses the factors that affect accuracy, considering both bias and estimation error. Third, this analysis includes both institutional ownership and auditor characteristics as external factors that may impact management reported ETR estimates. Therefore, this study complements theirs and provides additional contributions to enhance our understanding regarding financial reporting of tax expense. Electronic copy available at: https://ssrn.com/abstract=3774393 may not be able to pressure management to provide more accurate estimates. Given these arguments, I state the first hypothesis in null form: H1: Analyst coverage is not associated with interim tax reporting accuracy. 2.3 INSTITUTIONAL OWNERSHIP There is also significant tension with regard to institutional investors’ influence on tax reporting accuracy. On one hand, prior research suggests that institutional investors serve a monitoring role in mitigating agency problems between shareholders and managers (Hartzell and Starks [2003]) and curbing myopic behavior (Bushee [1998]). Ajinkya et al. [2005] note that “institutions consistently probe the company for more specific, unbiased, and accurate information about future earnings.” Consistent with this notion, they find that institutional ownership is positively associated with the likelihood of forecast occurrence, and that these forecasts are more specific and accurate, suggesting that institutional owners are able to exert pressure on management to provide more accurate information. On the other hand, managers may provide their best estimate of annual ETR even without monitoring by institutional owners. In addition, despite demand for more accurate information from institutional owners, managers may not be able to provide more accurate information. Finally, high institutional ownership may place significant pressure on management to meet earnings benchmarks, causing managers to decrease ETRs and/or avoid increases in ETR (Dechow and Skinner [2000], Comprix et al. [2012]). These actions represent the predominant perception of how ETR estimates are reported (i.e., they are biased), resulting in less accurate interim ETR estimates. Given these arguments, I state the second hypothesis in null form: H2: Institutional ownership is not associated with interim tax reporting accuracy. Electronic copy available at: https://ssrn.com/abstract=3774393 2.4 AUDITOR CHARACTERISTICS SEC Regulation S-X (Article 10) requires interim financial statements to be reviewed (but not audited) by an independent accountant, potentially improving their reliability (Manry et al. [2003]). Specifically related to the tax accounts, auditors examine the basis for estimated book-tax differences and inquire about discretionary items as part of their review (Bauman and Shaw [2005]). Because taxes can materially affect reported earnings (Schmidt [2006]), are among the most restated accounts (Plumlee and Yohn [2010], Whalen et al. [2020]), and require significant estimation and judgement (Graham et al. [2012]), the tax accounts are likely to receive auditor attention during the interim financial statement review. Prior research examines auditor characteristics and their association with audit quality. This line of research suggests that larger audit offices, industry expert auditors, longer auditor tenure, and Big N auditors are associated with higher audit quality. Prior research also suggests that auditor-provided tax services (APTS) provide “knowledge spillovers” and are associated with improved financial reporting quality (Kinney et al. [2004], Robinson [2008], Krishnan and Visvanathan [2011]), fewer tax-related restatements (Seetharaman et al. [2011]), and fewer tax- related internal control weaknesses (De Simone et al. [2015]). Gleason and Mills [2011] provide evidence that firms that purchase APTS are fully reserved for IRS disputes, while firms that don’t purchase APTS are not. Thus, prior research suggests that a number of auditor characteristics are associated with improved audit and financial reporting quality. Other review procedures performed by the auditor may include comparing forecasts to actual results, comparing results to entities in the same industry, and examining ratios such as inventory turnover and expenses as a percentage of sales. Specific examples of additional situations where the auditor would ordinarily inquire of management include M&A, revenue recognition, impairment, and derivatives (AU Section 722). See DeFond and Zhang [2014] for an in-depth review of archival audit research. Electronic copy available at: https://ssrn.com/abstract=3774393 However, a review of interim reports substantially differs from an audit. While an audit includes significant testing in accordance with generally accepted auditing standards, a review primarily involves broad analytical procedures and inquiries of management, rather than search and verification procedures (AU 722). Therefore, it is unclear whether these auditor characteristics are associated with improved reviews, and whether this translates to more accurate ETR estimates. Therefore, I state the third hypothesis in null form: H3: Auditor characteristics are not associated with interim tax reporting accuracy. 2.5 USEFULNESS OF REPORTED TAX EXPENSE Financial Accounting Concepts Statement No. 8 states that the fundamental qualitative characteristics of useful financial information are relevance and faithful representation. Information is relevant if it is capable of making a difference to users and has predictive value, confirmatory value, or both. The Financial Accounting Standards Board (FASB) states the following regarding faithful representation: To be useful, financial information not only must represent relevant phenomena, but it also must faithfully represent the phenomena that it purports to represent. To be a perfectly faithful representation, a depiction would have three characteristics. It would be complete, neutral, and free from error. Of course, perfection is seldom, if ever, achievable. The Board’s objective is to maximize those qualities to the extent possible. (FASB 2010, p. 17) Although ETR estimates may be made faithfully by management using the best available information, they are rarely perfect representations of annual ETR (i.e., the “phenomena that it purports to represent”). When users of financial statements consistently observe significant errors in interim ETR estimates, they could infer unreliability of the tax expense reported by the firm. Prior academic research has demonstrated that estimated taxable income, calculated using information in the financial statements, provides information to the market that is incremental to Electronic copy available at: https://ssrn.com/abstract=3774393 pretax book income (Hanlon et al. [2005], Ayers et al. [2009]). However, estimation error in financial statement information creates noise that reduces the beneficial role of the information (Dechow and Dichev [2002]), and this noise reduces the price reaction to the signal (Holthausen and Verrecchia [1988], Kothari [2001], Hanlon et al. [2008]). Therefore, I examine whether the information contained in the tax accounts varies by prior interim tax reporting accuracy, a proxy for potential estimation error, and whether the market differentially impounds the information into stock prices. Consistent with the association of ETR inaccuracy and noise in the tax accounts, greater prior interim tax reporting accuracy may improve the reliability of information in the tax accounts, and this information will be more fully reflected in stock prices. Therefore, the fourth hypothesis is stated as follows: H4: Firms with high prior ETR accuracy have more useful information contained in the tax accounts. 3. Research Design 3.1 DETERMINANTS OF INTERIM TAX REPORTING ACCURACY AND BIAS I test my first three hypotheses by estimating the following equation: 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 = 𝛽𝛽 +𝛽𝛽 ∑𝐴𝐴𝐴𝐴𝐴𝐴 +𝛽𝛽 𝐼𝐼𝐼𝐼 +𝛽𝛽 ∑𝐴𝐴𝐴𝐴𝐴𝐴 𝐼𝐼𝐴𝐴 𝐼𝐼 𝐴𝐴 + 𝑖𝑖𝑖𝑖 0 1−2 𝑖𝑖𝑖𝑖 3 𝑖𝑖𝑖𝑖 4−8 it 𝛽𝛽 ∑𝐹𝐹 𝐹𝐹𝐼𝐼𝐴𝐴 _𝐹𝐹 𝐴𝐴𝐴𝐴𝐴𝐴𝐼𝐼𝐴𝐴𝐴𝐴 +𝛽𝛽 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐼𝐼𝐴𝐴𝐴𝐴𝐴𝐴 +𝛽𝛽 𝐴𝐴 𝑌𝑌 𝐴𝐴𝐴𝐴 +𝜀𝜀 (1) 9−23 𝑘𝑘 𝑖𝑖 𝑗𝑗 𝑖𝑖 𝑖𝑖𝑖𝑖 The dependent variable is the absolute value of the difference between year-end ETR and the year-to-date ETRs from the first, second, and third quarters (ACCURACY1, ACCURACY2, and ACCURACY3, respectively). This value is multiplied by -1 so larger values indicate greater interim tax reporting accuracy. I also examine factors associated with ETR bias by estimating This calculation is similar to calculations used in research that examines management earnings forecast accuracy (e.g., Hirst et al. [2008], Feng et al. [2009], Bamber et al. [2010], Baik et al. [2011], Billings et al. [2014], Goodman et al. [2014]). Electronic copy available at: https://ssrn.com/abstract=3774393 𝑖𝑖𝑖𝑖 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 Equation (1) after replacing ACCURACY with the signed difference between year-end ETR and the year-to-date ETRs from the first, second, and third quarters (∆Q1Q4, ∆Q2Q4, and ∆Q3Q4, respectively), where positive values indicate an increase in ETR from quarter q to year-end ETR. ANALYST consists of two variables related to analyst coverage, the first of which is the number of analysts (AF) following prior research (He and Tian [2013], Allen et al. [2016]). H1 does not predict a direction on the sign of the AF coefficient estimate; however, a positive coefficient would be consistent with analysts playing a monitoring role, resulting in more accurate ETR estimates. The second analyst variable is an indicator variable, EM, which is equal to 1 if the firm would have missed the annual analysts’ consensus forecast (within 5 cents) without a change in ETR. This variable follows prior research and controls for the incentive to beat analysts’ forecast using tax expense (Dhaliwal et al. [2004]). IO is percent of the firm owned by institutions. H2 does not predict the sign of the IO coefficient estimate; however, a positive coefficient would be consistent with institutional owners playing a monitoring role resulting in more accurate ETR estimates. AUDITOR is a vector of auditor characteristics, including a Big 4 auditor indicator (BIG4), tax fees paid to the auditor (APTS), auditor expertise (EXPERT) , auditor tenure (TENURE), and audit office size (OFFICE_SIZE). A positive coefficient estimate on these variables would be consistent with the notion that improved audit quality extends to the interim review process, resulting in more accurate ETR reporting. Additional analysis examines these factors using a “de-biased” accuracy measure. See Section 4.3.4 for details. In the primary analysis EXPERT is defined as a measure of combined audit and tax expertise, following McGuire et al. [2012]. See Appendix B for details. In untabulated analyses, EXPERT is replaced with TAX_EXPERT or AUDIT_EXPERT following McGuire et al. [2012]. No significant relation between tax-specific or audit-only expertise and ETR accuracy was found. Electronic copy available at: https://ssrn.com/abstract=3774393 FIRM_FACTORS is a vector of firm-specific characteristics that have been associated with the complexity of the tax accounts in prior research. Firm size (SIZE) is included because research suggests larger firms have more opportunity for tax planning (Omer et al. [1993], Rego, [2003]). Larger firms may have more complex tax situations resulting in less accurate ETR estimates, but also may have more effective tax departments that could produce more accurate ETR estimates. Therefore, no prediction is made regarding the sign of the association between SIZE and ACCURACY. To examine various components of tax complexity, I include number of geographic and business segments (GEO_SEGS and BUS_SEGS, respectively), changes in geographic mix of income (∆MIX) , merger and acquisition activity (M&A), research and development activity (R&D), discontinued operations and extraordinary items (DISC_EXTRA), deferred taxes (DTA), and equity compensation (EQUITY_COMP) because prior research suggests these factors are associated with tax rates and therefore could create complexity in the tax function and in estimating interim ETRs (Klassen et al. [1993], Collins et al. [1998], Robinson et al. [2010], Bratten et al., [2017], De Simone et al. [2015]). Sales growth (∆SALES) and prior earnings volatility (EARN_VOL) are included to examine factors related to the difficulty of forecasting earnings (Baik et al. [2011], Goodman et al. [2014]). The analysis also controls for leverage (LEV), profitability (ROA), market-to-book ratio (MTB), firm age (FIRM_AGE) and abnormal accruals (ABACC). I expect the tax complexity and earnings forecast difficulty variables to be negatively associated with interim tax reporting accuracy. However, it is an empirical question as to which factors are associated with both accuracy and bias. Finally, year and industry fixed effects are included to control for cross-sectional and time-series I use absolute value of changes in geographic mix of income because both increases and decreases in ratio of foreign to total income could affect the ability to successfully predict the ETR implications of such a change. Electronic copy available at: https://ssrn.com/abstract=3774393 variation across industries and years, and standard errors are clustered by firm (Petersen [2009], Gow et al. [2010]). All variable definitions are included in Appendix B. 3.2 ESTIMATED TAXABLE INCOME AS A PERFORMANCE MEASURE The primary test of the information content of tax expense examines the slope coefficient relating long-window returns to changes in estimated taxable income and its interaction with prior interim tax reporting accuracy. This test follows prior research (e.g., Francis et al. [2005], Hanlon et al. [2008]) that interprets the slope coefficient as a measure of the informativeness of earnings. Specifically, I examine the informativeness of pretax earnings changes, estimated taxable income changes, and the effect of tax reporting accuracy by estimating the following regression: 𝐴𝐴𝐴𝐴𝑌𝑌𝐴𝐴𝐴𝐴𝐴𝐴 = 𝛽𝛽 +𝛽𝛽 ∆𝑃𝑃 𝑃𝑃𝐴𝐴 𝐼𝐼 +𝛽𝛽 ∆𝐴𝐴𝐼𝐼 +𝛽𝛽 3𝐴𝐴𝐴𝐴 _𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 + 𝑖𝑖𝑖𝑖 0 1 𝑖𝑖𝑖𝑖 2 𝑖𝑖𝑖𝑖 3 𝑖𝑖𝑖𝑖 𝛽𝛽 ∆𝐴𝐴𝐼𝐼𝑇𝑇 3𝐴𝐴𝐴𝐴 _𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 +𝜀𝜀 (2) 4 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 Following Hanlon et al. [2005], RETURN is the 16-month market-adjusted return for firm i starting at beginning of fiscal year t and ending 4 months after the end of fiscal year t. ∆PTBI and ∆TI are year to year changes in pretax and estimated taxable income, respectively. Importantly, taxable income is estimated using amounts in the financial statements, which is available to investors. 3YR_ACCURACY is the quartile rank of interim tax reporting accuracy over the prior 3 years, where higher values indicate greater accuracy. Because tax differences vary across industries (Mills and Newberry [2001]), I rank 3YR_ACCURACY by industry (using two-digit SIC codes) and I scale the ranks to range between -0.5 and 0.5. Appendix B provides specific calculations for all variables. Consistent with prior research (Hanlon et al. [2005], Ayers et al. [2009]), I expect β and β to be positive, indicating that both pretax and taxable income provide information to the Electronic copy available at: https://ssrn.com/abstract=3774393 market. Based on H4, I predict that β will be positive, indicating a stronger relation between returns and changes in estimated taxable income when the firm has reported its tax estimates more accurately in the past. By examining the interaction term while including the main effect for changes in taxable income, this analysis tests how the association between estimated taxable income and returns varies by prior tax reporting accuracy. I also examine the relative information content of changes in estimated taxable income to book income for high and low accuracy firms. By examining the relative information content, I am able to directly test the information content of tax expense (i.e., the numerator of the ETR) relative to pre-tax book income (i.e., the denominator of the ETR) to ensure the effect is driven by tax expense rather than a factor that is associated with both tax expense and pretax income quality. This test follows prior research (Hanlon et al. [2005], Ayers et al. [2009]) to measure the information content of estimated taxable and book income as the adjusted R of regressions of returns on each measure of income individually. Specifically, I estimate the following equations annually for high and low accuracy firms: 𝐴𝐴𝐴𝐴𝑌𝑌𝐴𝐴𝐴𝐴𝐴𝐴 = 𝛽𝛽 +𝛽𝛽 ∆𝐴𝐴𝐼𝐼 +𝜀𝜀 (3) 𝑖𝑖𝑖𝑖 0 1 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝐴𝐴𝐴𝐴𝑌𝑌𝐴𝐴𝐴𝐴𝐴𝐴 = 𝛽𝛽 +𝛽𝛽 ∆𝑃𝑃 𝐴𝐴 𝐼𝐼𝑃𝑃 +𝜀𝜀 (4) 𝑖𝑖𝑖𝑖 0 1 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 Then, I test the average yearly ratio of the adjusted R from Equation (3) and (4) for each group of firms. An advantage of this research design is that it compares the relative information content of the two income measures while holding returns for the firm constant; it does not compare the adjusted R of estimated taxable income across samples, which would be susceptible to possible alternative explanations (Ayers et al. [2009]). That is, differences in the relative information content of taxable income and pre-tax income help to rule out the explanation that prior ETR accuracy is correlated with a firm characteristic that would affect the Electronic copy available at: https://ssrn.com/abstract=3774393 association between both income measures and returns. H4 predicts that the ratio of adjusted R from Equation (3) to adjusted R from Equation (4) will be greater for high accuracy tax reporting firms. 4. Results 4.1 SAMPLE The sample includes all firm-year observations at the intersection of Audit Analytics, Compustat, IBES, and Thomson Reuters Institutional Holdings databases from 2002 to 2015 for which all variables are available. The sample period begins in 2002 because of the limited availability of auditor fee data prior to 2002 as well as the changes in auditing and reporting requirements following SOX. The sample ends in 2015 to avoid issues related to adoption of ASU 2016-09, which introduced volatility and error in firms’ ETRs (Brushwood et al. [2019]). Consistent with prior tax research, financial institutions and utilities (SIC codes 6000-6999 or 4900-4999), firm-years with negative pre-tax income, ETRs greater than 1 or less than zero in any quarter, total assets less than $10 million are excluded. All continuous variables are st th winsorized at the 1 and 99 percentiles to reduce the influence of outliers. These requirements result in a sample of 15,009 firm-year observations. 4.2 DESCRIPTIVE AND UNIVARIATE RESULTS Figure 1 presents histograms of changes in ETRs from the first, second, and third quarters to the fourth quarter in Panels A, B, and C, respectively. Overall, Figure 1 suggests substantial variation in ETR changes wherein a significant number of these changes are large. In fact, 10.1 (16.3) percent of observations have ETR increases (decreases) of more than 5.0 percentage Electronic copy available at: https://ssrn.com/abstract=3774393 points from the first to fourth quarters. Some bias appears to exist in the sample (i.e., the distribution is centered left of zero), consistent with prior research that examines quarterly ETR bias (Comprix et al. [2012]). However, a large portion of observations have positive ETR changes (37.9 percent, 38.1 percent, and 41.7 percent are positive from the first, second, and third quarters, respectively). This results reconciles with prior studies, but also illustrates ETR accuracy is another important dimension to consider with regard to ETR changes in addition to bias. [Insert Figure 1 here] Table 1 presents descriptive statistics for the sample. The mean of ACCURACY1, ACCURACY2, and ACCURACY3 indicates that the absolute value of firms’ ETR changes from the first, second, and third quarter to year-end ETR are 4.9 percent, 3.9 percent, and 2.9 percent on average. This pattern is consistent with the expectation that firms provide more accurate ETR estimates later in the year. ETR changes in the sample exhibit some bias, shown by negative mean and median values for ∆Q1Q2, ∆Q2Q4, and ∆Q3Q4. Approximately 10 percent of the observations in the sample would have missed analysts’ forecasts without a change in ETR from third to fourth quarter within 5 cents (i.e., EM = 1). Therefore, the sample is much broader than studies that specifically examine earnings management through tax expense (e.g., Dhaliwal et al. 2004) because this sample is not restricted to firm-years within 5 cents of analysts’ consensus forecast. Observations with increases (decreases) of more than 5.0 percentage points from the second to fourth, and third to fourth, quarter are 7.7 (12.8) percent and 6.1 (8.4) percent, respectively. A number of variables in the analysis exhibit high correlation (e.g., the correlation between SIZE and AF is 0.66, untabulated). However, an untabulated analysis suggests that multicollinearity is not a major concern in the regressions. The largest variance inflation factor in my analysis is 2.69 on SIZE and prior work suggests a variance inflation factor of 10 as being large enough to indicate a problem (Chatterjee and Price [1991]). Electronic copy available at: https://ssrn.com/abstract=3774393 [Insert Table 1 here] 4.3 MULTIVARIATE RESULTS 4.3.1 Interim Tax Reporting Accuracy Table 2 Panel A presents the results of estimating Equation (1). Columns 1, 2, and 3 present the results with the dependent variables of ACCURACY1, ACCURACY2, and ACCURACY3, respectively. The coefficient on AF is positive in all columns and is significant in columns 2 and 3 (p-values < 0.10 and 0.05, respectively). These results reject H1 and are consistent with analysts serving a monitoring role that improves the accuracy of second and third quarter ETR estimates. Regarding H2, institutional ownership (IO) is positively associated with interim tax reporting accuracy in all three quarters (p-values <0.01). These results reject H2 and provide evidence that institutional investors serve a monitoring role resulting in more accurate ETR estimates. Although analyst following and institutional ownership likely overlap (i.e., they are positively correlated), both coefficients are significant, indicating each serves an incremental monitoring role. Interestingly, firms with the incentive to manage earnings through tax accounts to meet or beat analysts’ expectations (EM =1) have more accurate ETR estimates, on average. A plausible explanation for this result is that to be in this set of firms (i.e., within 5 cents of analysts’ forecast), analysts’ forecast must be relatively accurate, which is more likely when the firm reports more accurate information. Regarding H3, the coefficient on TENURE is positive in all columns and significant in column 3 (p-value < 0.05), consistent with improved auditor-client communication and higher quality reviews when the auditor has been serving the client for a longer period. However, there I further explore the monitoring roles of analyst following and institutional ownership by examining their association with ETR estimation bias (Table 2 Panel B) as well as positive and negative ETR surprises (Table 3). Electronic copy available at: https://ssrn.com/abstract=3774393 is not strong evidence that Big 4 auditors, auditor-provided tax services, audit office size, or expertise is consistently associated with interim tax reporting accuracy. Although prior research generally suggests these characteristics are associated with improved audit quality, these findings do not provide strong evidence that these characteristics improve the interim review process of the tax accounts. As expected, geographic complexity, changes in geographic mix of income, R&D activity, earnings volatility, discontinued operations and extraordinary items, deferred tax assets, and equity compensation are negatively associated with interim tax reporting accuracy. In sum, Table 2 Panel A provides evidence that analyst following and institutional ownership are associated with improved interim tax reporting accuracy, rejecting H1 and H2. Regarding H3, auditor tenure is associated with improved interim tax reporting accuracy, but other auditor characteristics to not appear to be associated with interim tax reporting accuracy. [Insert Table 2 here] 4.3.2 Interim Tax Reporting Bias Importantly, to provide insight into which factors contribute to accuracy through bias or estimation error, this study also examines what factors are associated with signed ETR changes (i.e., if they are systematically biased). Factors associated with accuracy, but not bias, are contributors to estimation error. I therefore re-estimate Equation (1) after replacing the dependent variable with signed changes in ETRs from the first, second, and third quarters to the Several auditor variables are highly correlated, so I also estimate the regressions while including each individually. My inferences remain unchanged, suggesting multicollinearity of the auditor variables is not driving the insignificant results. In untabulated analyses, I also create a composite “auditor score” as the sum of BIG4, EXPERT, and indicator variables equal to one if the firm is above the median for each continuous variable. This composite score is positively associated with tax reporting accuracy, but the result appears driven by auditor tenure. The composite score becomes insignificantly associated with accuracy when tenure is excluded from the composite score. Electronic copy available at: https://ssrn.com/abstract=3774393 fourth quarter (∆Q1Q4, ∆Q2Q4, and ∆Q3Q4), where positive amounts represent an increase in ETR from quarter q to year-end ETR. Table 2 Panel B presents these results. The results provide little evidence of systematic bias associated with changes in geographic mix of income, R&D activity, deferred tax assets, or equity compensation even though these factors are strongly associated with accuracy. These results furthermore suggest that the large estimation error related to these factors is approximately evenly distributed between ETR increases and decreases; that is, there is strong evidence of estimation error but no evidence of bias related to these factors. The results suggest that firms with the incentive to meet or beat analysts’ forecasts (EM = 1) decrease their ETRs on average (p-value < 0.05 from third to fourth quarter), consistent with prior research (e.g., Dhaliwal et al. [2004]). However, analyst following is positively associated with ETR changes (p-values < 0.05 and 0.10 in columns 2 and 3, respectively). This result is consistent with analysts serving a monitoring role by either curbing tax planning later in the year that is “aggressive” or by reducing “slack” in interim ETRs, creating bias. In contrast, institutional ownership is negatively associated with ETR changes. This result is consistent with institutional owners applying pressure on firms regarding ETR changes, either by demanding additional tax planning that reduces ETRs or by penalizing surprise ETR increases, resulting in biased ETR estimates to avoid ETR increases. 4.3.3 Positive and Negative ETR Surprises The bias results related to analysts following and institutional ownership in Table 2 Panel B have two explanations, as discussed in the previous section. For example, in Table 2 Panel B, I also examine these factors using a “de-biased” accuracy measure. See Section 4.3.4 for details. I acknowledge that even if institutional owners do not actually “penalize surprise ETR increases,” firms may still attempt to avoid surprise ETR increases to avoid a potential negative reaction. Electronic copy available at: https://ssrn.com/abstract=3774393 institutional ownership is negatively associated with ETR changes. This relation could be because firms with high institutional ownership tend to tax plan and have greater decreases in ETR throughout the year on average. However, it could also be because firms with high institutional ownership tend to avoid increases in ETR. Both of these explanations are plausible, requiring further examination to provide additional insight into tax reporting accuracy and bias. I investigate this issue further by estimating a multinomial logistic regression with POS_SURPRISE and NEG_SURPRISE as the dependent variables. POS_SURPRISE (NEG_SURPRISE) is an indicator variable equal to 1 if the firm year has an increase (decrease) in ETR from the third to fourth quarters of greater than 5 percent, and zero otherwise. This test examines which factors are associated with substantial changes in ETRs and also allows the changes to be directional, addressing the issue discussed above. For example, if the bias associated with institutional ownership is because of a greater tendency for firms to decrease their ETR, there would be a positive coefficient on IO when NEG_SURPRISE is the dependent variable. In contrast, if the bias associated with institutional ownership is because of a lower tendency for firms to increase ETR, there would be a negative coefficient on IO when POS_SURPRISE is the dependent variable. I include all independent variables from Equation (1) in this analysis and present the results in Table 3. [Insert Table 3 here] The results indicate that analyst following is negatively associated with negative ETR surprises, but not associated with positive ETR surprises. This result is consistent with the result Inferences are unchanged using first to fourth quarters or second to fourth quarters, but ETR surprises from third to fourth quarters are of particular focus because research in this area has focused on “last chance” manipulations of ETR from the third to fourth quarters, and large increases from the third to fourth quarters should be the greatest “surprise” because more information is available at third quarter than first or second quarters. Electronic copy available at: https://ssrn.com/abstract=3774393 in Table 2 Panel B (i.e., analyst following is associated with positive ETR changes). However, this analysis provides an additional insight: the relation is caused by a reduced likelihood of negative ETR changes rather than firms actually increasing ETR at year end (i.e., AF is not associated with POS_SURPRISE). A lower likelihood of negative ETR surprises is consistent with a monitoring role of analysts because this could be caused by reductions in aggressive tax planning at year end or less bias in interim periods that causes an ETR decrease at the end of the year. In addition, institutional ownership is negatively associated with positive ETR surprises but not associated with negative ETR surprises. This result is consistent with Table 2 Panel B. However, the relation is caused by a lower likelihood of positive surprises rather than greater decreases in ETR at year end (i.e., IO is not associated with NEG_SURPRISE). A lower likelihood of positive ETR surprises is consistent with firms avoiding surprises that could have a negative impact on earnings per share which would be viewed negatively by investors. However, because the results indicate no evidence that institutional ownership is associated with negative ETR surprises, the negative association between institutional ownership and ETR changes observed in Table 2 Panel B is caused by a lower likelihood of ETR increases, rather than additional tax planning at year end. 4.3.4 Extracting Bias from Accuracy As an additional approach to separate estimation error and bias, I remove the estimated bias in the estimated ETR to examine the accuracy of a de-biased ETR estimate. Prior research uses a similar approach to extract predictable errors from analysts’ forecasts (e.g., Ali et al. [1992], Larocque [2013]). Specifically, I regress the signed difference between year-end ETR and the year-to-date ETR from the first, second, and third quarters (∆Q1Q4, ∆Q2Q4, and Electronic copy available at: https://ssrn.com/abstract=3774393 ∆Q3Q4, respectively) on all variables included in Equation (1). Next, the bias for each observation is predicted based on the coefficient estimates from these regressions. The predicted bias is removed for each observation by subtracting the predicted bias from the actual change in ETR to calculate de-biased changes in ETR. The absolute value of the de-biased change in ETR is multiplied by -1 to generate ETR accuracy measures with the predicted ETR bias removed (ACC1_debiased, ACC2_debiased, and ACC3_debiased). These variables measure the accuracy of the ETR estimates that are not attributable to bias, and therefore represent estimation error. I re-estimate Equation (1) after replacing the de-biased accuracy measures as the dependent variable and present the results in Table 4. Consistent with the main analysis, analyst following, institutional ownership, and auditor tenure are associated with more accurate ETR reporting, while a number of factors that contribute to firm complexity (e.g., geographic segments, changes in mix of foreign and domestic income, R&D expenses, earnings volatility, discontinued and extraordinary items, deferred tax assets, and equity compensation) are associated with less accurate ETR reporting using the de-biased accuracy measure. Thus, the main conclusions regarding the effect of these factors on accuracy through estimation error, rather than bias, remain unchanged. [Insert Table 4 here] 4.4 Information Content Results I examine the consequences of interim tax reporting accuracy by testing the association between stock returns and estimated taxable income using the procedures in Ayers et al. [2009]. I collect observations for which all data needed are available at the intersection of Compustat and Electronic copy available at: https://ssrn.com/abstract=3774393 CRSP from 1993 through 2015. I exclude financial institutions and utilities (SIC codes 6000- 6999 or 4900-4999), firm years with fiscal year changes, and firm years with the absolute value of the change in pre-tax book income (∆PTBI) or estimated taxable income (∆TI) greater than 1. Positive tax expense (GAAP_ETR5) and pre-tax book income over the past five years are required. These criteria result in a sample of 33,688 firm-year observations. Descriptive statistics for RETURN, ∆PTBI, ∆TI, and GAAP_ETR5 for the overall sample (untabulated) are similar to those in Ayers et al. [2009]. Untabulated results suggest that RETURN, ∆PTBI, and ∆TI are not significantly different for firms in the top quartile of accuracy based on 3YR_ACCURACY (“high accuracy firms”) compared to all other firms. High accuracy firms have a significantly lower mean five-year GAAP ETRs than all other firms, although the th th 23 opposite is true for the 25 and 50 percentiles. Table 5 presents the results of estimating Equation (3) for the full sample (columns 1 and 2), the pre-SOX sample (columns 3 and 4) and the post-SOX sample (columns 5 and 6). In the odd-numbered columns, I regress returns on changes in pre-tax book income and changes in taxable income. The positive and significant coefficients on both variables (p-values < 0.01 in all columns) are consistent with prior research (Hanlon et al. [2005], Ayers et al. [2009]) and suggest that estimated taxable income provides information to the market that is incremental to changes in pre-tax income. In the even-numbered columns, the coefficient on the interaction term ∆TIx3YR_ACCURACY is positive and significant, providing evidence that the information contained in estimated taxable income increases relating to prior ETR accuracy, even after I include years prior to SOX in my sample for this analysis to align more closely with prior studies that use this research design. However, I also present my results for pre- and post-SOX sample periods. I perform an analysis using a high versus low accuracy sample matched on GAAP_ETR5 in order to rule out the explanation that low ETRs are also more difficult to report accurately. Electronic copy available at: https://ssrn.com/abstract=3774393 controlling for the magnitude of the change in taxable income itself (∆TI). These results provide evidence that potential error in taxable income significantly affects the association between stock returns and estimated taxable income, consistent with H4. [Insert Table 5 here] Table 6 reports the results of the second test of the effect of prior ETR accuracy on the relative information content of estimated taxable income. In Panel A, I split the firms based on the median of 3YR_ACCURACY and compare the ratio of adjusted R values of yearly estimates of Equations (3) to adjusted R values of yearly estimates of Equation (4) for each subsample. A 2 2 higher ratio of R / R indicates a stronger relation between returns and changes in taxable TI PTBI income relative to the relation between returns and changes in pretax income for the same firm. Intuitively, this means that there is relatively more information content in taxable income when this ratio is high. [Insert Table 6 here] In Panel A, the average ratio for all years is 0.532 for high accuracy firms, compared to 0.383 for low accuracy firms, and the difference of 0.148 is statistically significant (p-value < 0.01 using both t-test and Wilcoxon rank sum test). This result suggests the information content of estimated taxable income is relatively greater when the firm has been more accurate with tax estimates in the past, consistent with H4. The relative information content of taxable income for low accuracy firms is only 72% (0.383 / 0.532 = 72%) of the information content of high accuracy firms. Electronic copy available at: https://ssrn.com/abstract=3774393 Notably, Ayers et al. [2009] provide evidence that the information content of taxable income is lower for high tax planning firms (i.e., firms with low GAAP_ETR5). To the extent that high tax planning firms are also inaccurate, the results could be driven by tax planning rather than prior reporting accuracy. To mitigate this alternative explanation, I perform the same analysis on a sample of firms matched on level of tax planning. Specifically, I match each high accuracy firm with a low accuracy firm based on year, two-digit SIC code, and GAAP_ETR5. This matching procedure results in 14,414 firm-year observations for each sample. An untabulated analysis shows no significant difference in level of tax planning (i.e., GAAP_ETR5) between the matched samples (p-value = 0.88), indicating a successful match. Thus, a significant difference in information content of estimated taxable income between the high and low accuracy firms is not attributable to differences in years, industries, or level of tax planning between the samples. Table 5 Panel B presents the results from this analysis. After matching firms based on tax planning, I find that high accuracy firms have relatively more information content in estimated 2 2 taxable income than low accuracy firms. The average R / R for all years in the matched TI PTBI sample is 0.779 for high accuracy firms and 0.462 for low accuracy firms, suggesting the information content for low accuracy firms is only 59 percent of the information content of high accuracy firms on average, even after controlling for level of tax planning. In sum, the results in Tables 5 and 6 provide strong evidence consistent with H4 and suggest that investor use of reported tax expense varies with prior tax reporting accuracy. In untabulated results I replicate this finding using my sample. Electronic copy available at: https://ssrn.com/abstract=3774393 5. Additional Analyses 5.1 MARKET RESPONSE TO BEATING ANALYSTS’ FORECASTS To further investigate the effect of tax reporting accuracy on investor response to reported tax expense, I examine short-window returns surrounding earnings announcements for firms that beat analysts’ forecasts by decreasing their tax rate from third to fourth quarter. Gleason and Mills [2008] show that when firms use tax expense to beat analysts’ forecasts, the market discounts the reward by approximately 86% on average. Because noise reduces the price reaction to accounting information (Holthausen and Verrecchia [1988], Kothari [2001], Hanlon et al. [2008]), to the extent that prior inaccuracies provide a signal regarding the potential noise in current reported tax expense, the market discount may be greater for firms that have been inaccurate in prior years. At the same time, the market might “see through” these earnings management actions more clearly for firms with a history of accuracy, so the discount could be greater for more accurate firms. It is therefore an empirical question whether the market discount is more, less, or the same for firms with greater prior accuracy compared to less accurate firms. First, I replicate the results in Gleason and Mills [2008], presented in column 1 of Table 7. Specifically, I regress cumulative size-adjusted abnormal returns (CAR) for the five-day window around the earnings announcement on Beat_w_Tax, an indicator variable equal to one if the firm used a tax rate decrease to beat analysts’ forecast and zero otherwise, and control variables following Gleason and Mills [2008]. The coefficient estimate on Beat_w_Tax in column 1 is negative (p-value < 0.01), which is consistent with Gleason and Mills [2008] and indicates that the market discounts the reward for beating analysts’ forecasts by using tax expense. Electronic copy available at: https://ssrn.com/abstract=3774393 Next, I split firms that beat analysts’ forecast into two groups based on prior tax reporting accuracy (3YR_ACCURACY). Beat_w_Tax_Accurate is an indicator variable equal to 1 for firms that beat analysts’ target and are above the median of 3YR_ACCURACY, zero otherwise. Beat_w_Tax_Inaccurate is an indicator variable equal to 1 for firms that beat analysts’ target and are below the median of 3YR_ACCURACY, zero otherwise. The results in column 2 of Table 7 indicate a significant market discount for firms that beat the analysts’ forecasts and have been inaccurate in the past. However, a similar market discount is not found for firms that have a record of accurate tax reporting. Further, the coefficient on Beat_w_Tax_Inaccurate is significantly more negative than the coefficient on Beat_w_Tax_Accurate (p-value < 0.05). These results suggest that investor response to beating earnings targets using tax expense varies with prior tax reporting accuracy, a potential signal for the reliability of reported tax expense. [Insert Table 7 here] 5.2 DISCRETE PERIOD ITEMS Accounting standards require that firms recognize certain “discrete” items in the quarter they occur, potentially distorting the ability for the quarterly ETRs to predict annual ETR. To address the concern that the main results are driven by discrete period items, I follow Bratten et al. [2017] and conservatively label quarters as “clean” (i.e., likely free of discrete items) if the reported GAAP ETR is within 0.5% (on either side) from the IBES actual ETR and re-run my accuracy analysis using this clean subsample. This procedure assumes management’s ETR estimate is free of discrete items when it matches the IBES actual ETR because the IBES actual ETR is adjusted for items that require discrete accounting treatment. In untabulated results, the conclusions regarding hypotheses 1, 2, and 3 remain unchanged after removing the potential effect of discrete period items. Electronic copy available at: https://ssrn.com/abstract=3774393 6. Conclusion A significant number of prior studies have focused on the variation in effective tax rates among firms (e.g., Dyreng et al. [2008]). This study examines the causes and consequences of within firm-year GAAP effective tax rate reporting accuracy using the financial reporting requirements under APB 28 to examine the determinants of ETR estimation accuracy as well as the importance of tax reporting accuracy on investor use of information contained in tax expense. In sum, analyst following, institutional ownership, and auditor tenure are positively associated with interim tax reporting accuracy. These results are consistent with a monitoring role over financial reporting of the tax accounts which results in more accurate ETR estimates. The results also suggest that analysts and institutional investors play different monitoring roles: analysts appear to reduce bias in ETR estimates that result in earnings increasing ETR surprises at year end, while institutional owners appear to reduce earnings decreasing ETR surprises at year end. A number of factors such as geographic complexity, change in geographic mix of income, discontinued and extraordinary items, deferred tax assets, and R&D activity are negatively associated with interim tax reporting accuracy. However, these factors are not associated with tax reporting bias, suggesting their association with less accurate reporting is due to estimation error. Market tests suggest a stronger association between stock returns and changes in estimated taxable income when interim tax reporting has been more accurate in the past, providing evidence that market use of tax information varies with prior tax reporting accuracy. In addition, the findings in this study suggest that investors respond more positively to beating analysts’ forecasts using a decrease in tax rate when the firm has a record of accurate tax reporting. These results provide strong evidence that investor use of information in the tax Electronic copy available at: https://ssrn.com/abstract=3774393 accounts varies with prior tax reporting accuracy, indicating an important pricing factor not examined in prior research. The requirement under APB 28 for companies to estimate annual ETR at each interim period is no easy task. The uncertainty, complex estimation, and substantial judgment create potential for significant estimation error in reported tax expense. This study documents factors associated with interim tax reporting accuracy and demonstrates that accuracy has significant implications regarding investors’ use of tax expense, providing a contribution to the literature regarding the pricing of tax information reported in financial statements. Electronic copy available at: https://ssrn.com/abstract=3774393 Appendix A Interim ETR Estimate Example Suppose a firm has operations in the United States and a foreign subsidiary with statutory tax rates of 35% and 25%, respectively. The firm also has significant R&D expenditures, for which it anticipates a tax credit. For each interim period, the firm should calculate its tax expense using the integral method as follows: In Q1, the firm projects annual pretax income of $100 in both the United States and its foreign subsidiary for a total of $200 annual pretax income. It also projects an R&D tax credit of $2 at the end of the year. Therefore, the firm projects $35 U.S. tax ($100 x 35%), $25 foreign tax ($100 x 25%), and a tax credit of $2 for a total tax expense of $58 ($35 + $25 - $2) for the year. The projected annual ETR as of Q1 is therefore 29% ($58 tax expense / $200 pre- tax income). If the firm reports actual total pre-tax income in Q1 of $75, the firm accrues $21.75 ($75 x 29%) tax expense in Q1. Note that $75 is not proportional to projected annual pre-tax income and $21.75 is not proportional to projected total tax expense; the tax expense accrued should reflect projected ETR at the end of the year, so is applied to each quarter on a pro rata basis. The following table summarizes this calculation: Q1 Projection: Pretax Income Tax Rate Tax U.S. $100 35% $35 Foreign $100 25% $25 Tax Credit $(2) Total Tax Expense $58 Projected ETR 29% Q1 actual pretax income $75 Apply projected ETR 29% Q1 tax expense $21.75 In Q2, the firm projects annual pretax income of $200, however, now $130 is projected in the United States and $70 is projected in the foreign jurisdiction. The projected R&D tax credit remains $2. Therefore, the projected tax rate in Q2 is 30.5% ((($130x35%) + ($70x25%)-$2)/$200). If the firm reports year-to-date pre-tax income in Q2 of $120, year-to-date tax expense should be $36.60 ($120x30.5%) to reflect the projected annual ETR. Because $21.75 of tax expense was accrued in Q1, the Q2 tax expense is $14.85 ($36.60 - $21.75). The following table summarizes this calculation: Q2 Projection: Pretax Income Tax Rate Tax U.S. $130 35% $45.5 Foreign $70 25% $17.5 Tax Credit $(2) Total Tax Expense $61 Projected ETR 30.5% Q2 year-to-date pretax income $120 Apply projected ETR 30.5% Year-to-date tax expense $36.60 Less Q1 tax expense accrued $21.75 Q2 tax expense $14.85 A25 Note that under ASC 740-30-25-17 (formerly APB 23), when firms designate unremitted foreign earnings as permanently or indefinitely reinvested, they are not required to accrue deferred tax expense on those earnings. Electronic copy available at: https://ssrn.com/abstract=3774393 In Q3, the firm does not change its projected U.S. or foreign pre-tax income. However, the firm anticipates an increase in the R&D credit to $4. Therefore, the projected tax rate in Q3 is 29.5%. If the firm reports year-to-date pre-tax income in Q3 of $160, year-to-date tax expense should be $47.20 to reflect the projected annual ETR. Because $36.60 was accrued through Q2, the Q3 tax expense is $10.60. The following table summarizes this calculation: Q3 Projection: Pretax Income Tax Rate Tax U.S. $130 35% $45.5 Foreign $70 25% $17.5 Tax Credit $(4) Total Tax Expense $59 Projected ETR 29.5% Q3 year-to-date pretax income $160 Apply projected ETR 29.5% Year-to-date tax expense $47.20 Less tax expense accrued $36.60 Q3 tax expense $10.60 In Q4, the firm realizes its total annual income of $180, with $110 in the United States and $70 in the foreign jurisdiction. The R&D credit is anticipated to be $4 when the tax return is filed. Therefore, the tax rate at year end is 28.89% and total tax expense is $52. Because $47.2 of tax expense was accrued in Q1-Q3, tax expense for the fourth quarter is $4.80 to reflect annual tax expense. The following table summarizes this calculation: Q4: Pretax Income Tax Rate Tax U.S. $110 35% $38.5 Foreign $70 25% $17.5 Tax Credit $(4) Total Tax Expense $52 ETR 28.89% Annual pretax income $180 Apply projected ETR 28.89% Year-to-date tax expense $52 Less tax expense accrued $47.20 Q4 tax expense $4.80 The interim ETR estimates are summarized as follows: EPS effect (compared to Quarter Estimated ETR Change from Prior Quarter Difference from annual prior ETR estimate) 1 29.00 - +0.11 2 30.50 +1.50 +1.61 -2.11% 3 29.50 -1.00 +0.61 +1.44% 4 28.89 -0.61 - +0.86% The EPS effect is calculated by dividing the change from prior quarter by (1-prior quarter ETR estimate). For example, if the quarter’s pre-tax earnings was $10, after-tax EPS is $7.10 (10x(1-0.290)) using the Q1 ETR estimate; however, using the second quarter ETR estimate, after-tax EPS is $6.95 (10x(1-0.305)), a decrease of 2.11%. In this example, the firm has relatively accurate ETR estimates compared to those in the sample in this study; however, even relatively small changes in ETR have a significant effect on after-tax EPS. Electronic copy available at: https://ssrn.com/abstract=3774393 Appendix B Variable Definitions Primary dependent variables ACCURACY1 Absolute value of the difference between the year-to-date ETR in the first quarter and the year-end ETR, multiplied by -1. ACCURACY2 Absolute value of the difference between the year-to-date ETR in the second quarter and the year-end ETR, multiplied by -1. ACCURACY3 Absolute value of the difference between the year-to-date ETR in the third quarter and the year-end ETR, multiplied by -1. ∆Q1Q4 Year-end ETR minus year-to-date ETR in the first quarter. ∆Q2Q4 Year-end ETR minus year-to-date ETR in the second quarter. ∆Q3Q4 Year-end ETR minus year-to-date ETR in the second quarter. NEG_SURPRISE Indicator variable equal to one if the firm-year has a decrease in ETR from third to fourth quarter of greater than 5%, and zero otherwise POS_SURPRISE Indicator variable equal to one if the firm-year has an increase in ETR from third to fourth quarter of greater than 5%, and zero otherwise External Monitors AF Natural log of the number of analysts following the company in the year t. EM Indicator variable equal to 1 if the company would have missed the last consensus analyst forecast using the previous quarter’s ETR within 5 cents; zero otherwise. IO Percentage of shares owned by institutions at the beginning of the year. APTS Tax fees divided by total fees paid to the auditor BIG4 Indicator variable equal to 1 if audited by a Big 4 firm; zero otherwise. OFFICE_SIZE Natural log of the number of clients audited by the office in year t. EXPERT Indicator variable equal to 1 if an audit firm is both an audit and tax expert; zero otherwise. Audit and tax expertise follows McGuire, Omer, and Wang [2012]. An audit office is defined as an industry audit (tax) expert if its market share in a given MSA (city) and industry (two-digit SIC) is greater than or equal to 30%. Market share is defined as total audit (tax) fees paid to the audit firm divided by total audit (tax) fees paid to all other audit firms in the same industry and MSA. TENURE Natural log of the number of years the audit firm audited the client. Firm characteristics ABACC Abnormal accruals for year t based on the performance-adjusted modified Jones Model (Kothari et al. [2005]). BUS_SEG Natural log of the number of business segments of the company. DISC_EXTRA Indicator variable equal to 1 if the firm has non-zero discontinued operations or extraordinary items during the year, zero otherwise. DTA Total net deferred tax assets at the beginning of the year scaled by lagged total assets. EARN_VOL Standard deviation of earnings per share over the prior 5 years scaled by total assets. EQUITY_COMP Stock compensation expense scaled by total sales FIRM_AGE Natural log of firm age, where firm age is determined based on the number of years since the firm first appeared in Compustat. GEO_SEG Natural log of the number of geographic segments of the company. ∆SALES Sales growth, measured as the absolute value of: sales in year t less sales in year t-1, scaled by sales in year t-1. LEV Long-term-debt-to-asset ratio at the end of year t scaled by total assets at the end of the year. M&A Indicator variable equal to 1 if firm participated in any merger and acquisition activity during year t; zero otherwise. Merger and acquisition activity is determined based on non- zero acquisition expense. ∆MIX Change in geographic mix of income, measured as the absolute value of: foreign pretax income divided by total pretax income in year t minus foreign pretax income divided by total pretax income in year t-1. Electronic copy available at: https://ssrn.com/abstract=3774393 MTB Market-to-book ratio for the end of year t, measured as market value of equity divided by book value of equity. PERM_DIFF Absolute value of year-end GAAP ETR minus 0.35. R&D R&D expense for year t scaled by total assets at the beginning of the year. ROA Return on assets for year t, measured as the ratio of pre-tax income to the average of total assets for the year. SIZE Natural log of market value of equity for the company at the beginning of year t. Variables for market response tests RETURN Buy-and-hold market-adjusted (value-weighted) return for security j over the 16-month return window starting at the beginning of fiscal year t and ending 4 months after the end of fiscal year t. ∆PTBI Change in pretax book income from year t-1 to year t, scaled by market value of equity at beginning of year t. PTBI is computed as pretax book income less minority interest. MVE is computed as Compustat PRCC_F x CSHO. ∆TI Change in estimated taxable income (TI) from year t-1 to year t, scaled by market value of equity (MVE) at beginning of year t. TI is computed as [(FTE + FOTE)/0.35] - ∆NOL, where FTE is current federal income tax expense (Compustat TXFED), FOTE is current foreign tax expense (Compustat TXFO), 0.35 is the top U.S. statutory tax rate for the sample period (1993 and later), and ∆NOL is change in net operating loss carryforwards (Compustat TLCF). MVE is computed as Compustat PRCC_F x CSHO. If federal income tax expense is missing from Compustat, TI is estimated as the difference between total income tax expense (Compustat TXT) and deferred taxes (Compustat TXDI) divided by 0.35, less the change in NOL carryforwards. 3YR_ACCURACY Annual quartile rank by two-digit SIC of prior three year tax reporting accuracy (year t-3 to t-1), scaled to range between -0.5 and -0.5. Prior reporting accuracy is computed as the sum of the absolute value of the differences between year-end ETR and estimated ETR in quarters 1, 2, and 3 over the prior 3 years. There are three differences each year, so cumulative three year tax reporting accuracy is based on nine differences. The sum is multiplied by -1 so higher values indicate greater accuracy, and those values are ranked into quartiles and ranks are then scaled to range between -0.5 and 0.5. GAAP_ETR5 Accumulated GAAP ETR over prior 5 years, calculated as the sum of current tax expense (Compustat TXT less TXDI) over the prior five years (years t-4 to year t) divided by the pretax book income (Compustat PI) over the prior 5 years. CAR Cumulative return for the five-day window around the earnings announcement (day -2 to day +2) minus the cumulative return for an equal-weighted portfolio of firms in the same CRSP size decile. BEAT_W_TAX Indicator variable equal to 1 if the firm would have missed analysts’ forecast without a change in ETR from quarter 3 to quarter 4, zero otherwise. BEAT_W_TAX_ Indicator variable equal to 1 if the firm would have missed analysts’ forecast without a ACCURATE change in ETR from quarter 3 to quarter 4 and the firms’ prior 3YR_ACCURACY is above the median, zero otherwise. BEAT_W_TAX_ Indicator variable equal to 1 if the firm would have missed analysts’ forecast without a INACCURATE change in ETR from quarter 3 to quarter 4 and the firms’ prior 3YR_ACCURACY is below the median, zero otherwise. AFE Actual earnings per share reported by I/B/E/S minus the last I/B/E/S consensus forecast, divided by stock price at the end of the fiscal year. BTM Book-to-market ratio, measured as book value of equity (CEQ) divided by market value of equity (PRCC_F x CSHO). SIZE Natural log of total assets at the end of the year. MOMENTUM Cumulative size-adjusted returns for the 6 months prior to the earnings announcement, ending 3 days before the earnings announcement. Electronic copy available at: https://ssrn.com/abstract=3774393 REFERENCES Accounting Principles Board (APB), 1973. Interim Financial Reporting. APB Opinion 28. Available at: http://clio.lib.olemiss.edu/cdm4/document.php?CISOROOT¼/aicpa&CISO PTR¼522&CISOSHOW=502. Ajinkya, B., Bhorjraj, S., Sengupta, P., 2005. The association between outside directors, institutional investors, and the properties of management earnings forecasts. Journal of Accounting Research 43 (3), 343-76. Ali, A., Klein, A., Rosenfeld, J., 1992. Analysts’ use of information about permanent and transitory earnings components in forecasting annual EPS. The Accounting Review 67 (1), 183-198. Allen, A., Francis, B., Wu, Q., Zhao, Y., 2016. Analyst coverage and corporate tax avoidance. Journal of Banking & Finance 73, 84-98. Ayers, B., Jiang, J., LaPlante, S., 2009. Taxable income as a performance measure: The effects of tax planning and earnings quality. Contemporary Accounting Research 26 (1), 15-54. Baik, B., Farber, D., Lee, S., 2011. CEO ability and management earnings forecasts. Contemporary Accounting Research 28 (5), 1645–68. Bamber, L., Jiang, J., Wang, I., 2010. What’s my style? The influence of top managers on voluntary corporate financial disclosure. The Accounting Review 85 (4), 1131–1162. Bartov, E., Givoly, D., Hayn, C., 2002. The rewards to meeting or beating earnings expectations. Journal of Accounting and Economics 33 (2), 173-204. Bauman, M., Shaw, K., 2005. Disclosure of managers’ private information in interim financial statements: A study of effective tax rate changes. The Journal of the American Taxation Association 27 (2), 57–82. Beardsley, E., M. Mayberry, and S. McGuire. 2020. Street vs. GAAP: Which Effective Tax Rate is More Informative? Contemporary Accounting Research, forthcoming. Billings, B., Buslepp, W., Huston, G., 2014. Worth the hype? The relevance of paid-for analyst research for the buy-and-hold investor. The Accounting Review 89 (3), 903-31. Bratten, B., Gleason, C., Larocque, S., Mills, L., 2017. Forecasting taxes: new evidence from analysts. The Accounting Review 92 (3): 1-29. Brushwood, J., D. Johnston, L. Kutcher, and J. Stekelberg. 2019. Did the FASB’s Simplification Initiative Increase Errors in Analysts’ Implied ETR Forecasts? Evidence from Early Adoption of ASU 2016-09. Journal of the American Taxation Association 41 (2): 31-53. Bushee, B., 1998. The influence of institutional investors on myopic R&D investment behavior. The Accounting Review 73 (3), 305-33. CFA Institute. 2016. Letter to Mr. Russell Golden, Financial Accounting Standards Board, RE: Proposed Accounting Standards Update, Income Taxes (Topic 740) – Disclosure Framework – Changes to the Disclosure Requirements for Income Taxes. October 25, Chatterjee, S., Price, B., 1991. Regression Diagnostics, New York: John Wiley. Chen, T., Harford, J., Lin, C., 2015. Do analysts matter for governance? Evidence from natural experiments. Journal of Financial Economics 115 (2), 383-410. Chung, K., Jo, H., 1996. The impact of security analysts’ monitoring and marketing functions on the market value of firms. The Journal of Financial and Quantitative Analysis 31 (4), 493-512. Electronic copy available at: https://ssrn.com/abstract=3774393 Choudhary, P., Koester, A., Shevlin, T., 2016. Measuring income tax accrual quality. Review of Accounting Studies 21 (1), 89-139. Collins, J., Kemsley, D., Lang, M., 1998. Cross-jurisdictional income shifting and earnings valuation. Journal of Accounting Research 36 (2), 209-229. Comprix, J., Mills, L., Schmidt, A., 2012. Bias in quarterly estimates of annual effective tax rates and earnings management. Journal of the American Taxation Association 32 (1), 59-77. Dechow, P., Dichev, I., 2002. The quality of accruals and earnings: the role of accrual estimation errors. The Accounting Review (Supplement), 35-59. Dechow, P., Skinner, D., 2000. Earnings management: Reconciling the views of academics, practitioners, and regulators. Accounting Horizons14 (2), 235-250. DeFond, M., Zhang, J., 2014. A review of archival auditing research. Journal of Accounting and Economics 58 (2), 275-326. Deloitte Development LLC, 2011. Material weaknesses and restatements: Is tax still in the hot seat? Available at http://www.deloitte.com/assets/DcomUnitedStates/Local%20Assets/ Documents/Tax/us_tax_materialweaknesses_012011.pdf Derrien, F., Kecskés, A., 2013. The real effects of financial shocks: evidence from exogenous changes in analyst coverage. Journal of Finance 68 (4), 1407-40. De Simone, L., Ege, M., Stomberg, B., 2015. Internal control quality: The role of auditor- provided tax services. The Accounting Review 90 (4), 1469-1496. Dhaliwal, D, Gleason, C., Mills, L., 2004. Last-chance earnings management: Using the tax expense to meet analysts' forecasts. Contemporary Accounting Research 21 (2), 431-59. Dyreng, S., Hanlon, M., Maydew, E., 2008. Long-run corporate tax avoidance. The Accounting Review 83 (1), 61-82. Feng, M., Li, C., McVay, S., 2009. Internal controls and management guidance. Journal of Accounting and Economics 48 (2), 190–209. Financial Accounting Standards Board (FASB), 2010. Statement of Financial Concepts No. 8. Qualitative Characteristics of Useful Financial Information. Norwalk, CT: FASB. Francis, J., Schipper, K., Vincent, L., 2005. Earnings and dividend informativeness when cash flow rights are separated from voting rights. Journal of Accounting and Economics 39 (2), 329-360. Gleason, C., Mills, L., 2008. Evidence of differing market responses to beating analysts’ targets through tax expense decreases. Review of Accounting Studies 13 (2), 295-318. Gleason, C., Mills, L., 2011. Do auditor-provided tax services improve the estimate of tax reserves? Contemporary Accounting Research 28 (5), 1484-1509. Goodman, T., Neamtiu, M., Shroff, N., White, H., 2014. Management forecast quality and capital investment decisions. The Accounting Review 89 (1), 331-365. Gow, I., Ormazabal, G., Taylor, D., 2010. Correcting for cross-sectional and time-series dependence in accounting research. The Accounting Review 85 (2), 483-512. Graham, J., Hanlon, M., Shevlin, T., Shroff, N., 2014. Incentives for tax planning and avoidance: Evidence from the field. The Accounting Review 89 (3), 991-1023. Graham, J., Harvey, C., Rajgopal, S., 2005. The economic implications of corporate financial reporting. Journal of Accounting and Economics 40 (1-3), 3-73. Graham, J., Raedy, J., Shackelford, D., 2012. Research in accounting for income taxes. Journal of Accounting and Economics 53 (1), 412-34. Electronic copy available at: https://ssrn.com/abstract=3774393 Habib, A., Hansen, J., 2008. Target shooting: review of earnings management around earnings benchmarks. Journal of Accounting Literature 27, 25-70. Hanlon, M., 2005. The persistence and pricing of earnings, accruals, and cash flows when firms have large book-tax differences. The Accounting Review 80 (1), 137-166. Hanlon, M., LaPlante, S., Shevlin, T., 2005. Evidence for the possible information loss of conforming book income and taxable income. Journal of Law and Economics XLVIII, 407-442. Hanlon, M., Maydew, E., Shevlin, T., 2008. An unintended consequence of book-tax conformity: A loss of earnings informativeness. Journal of Accounting and Economics 46 (2-3), 294- Hartzell, J., Starks, L., 2003. Institutional investors and executive compensation. The Journal of Finance 58 (6), 2351-74. He, J., Tian, X., 2013. The dark side of analyst coverage: the case of innovation. Journal of Financial Economics 109 (3), 856-878. Healy, P., Palepu, K., 2001. Information asymmetry, corporate disclosure, and the capital markets: A review of the empirical disclosure literature. Journal of Accounting and Economics 31 (1-3), 405-40. Healy, P., Wahlen, J., 1999. A review of the earnings management literature and its implications for standard setting. Accounting Horizons 13 (4), 365-83. Hirst, D., Koonce, L., Venkataraman, S., 2008. Management earnings forecasts: A review and framework. Accounting Horizons 22 (3), 315–38. Holthausen, R., Verrecchia, R., 1988. The effect of sequential information releases on the variance of price changes in an intertemporal multi-asset market. Journal of Accounting Research 26 (1), 82-106. Jensen, M., Meckling, W., 1976. Theory of the firm: Managerial behavior, agency costs, and ownership structure. Journal of Financial Economics 3 (4), 305-60. 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 (3), 561-88. Klassen, K., Lang, M., Wolfson, M., 1993. Geographic income shifting by multinational corporations in response to tax rate changes. Journal of Accounting Research 31 (Supplement), 141-173. Kothari, S., 2001. Capital markets research in accounting. Journal of Accounting and Economics 31 (1-3), 105-231. Kothari, S., Leone, A., Wasley, C., 2005. Performance matched discretionary accrual measures. Journal of Accounting and Economics 39 (1), 163-197. Krishnan, G., Visvanathan, G., 2011. Is there an association between earnings management and auditor-provided tax services? The Journal of the American Taxation Association 33 (2), 111-35. Larocque, S., 2013. Analysts’ earnings forecast errors and cost of equity capital estimates. Review of Accounting Studies 18 (1), 135-166. Lennox, C., 2016. Did the PCAOB’s restrictions on auditors’ tax services improve audit quality? The Accounting Review 91 (5), 1493-1512. Lev, B., Nissim, D., 2004. Taxable income, future earnings, and equity values. The Accounting Review 79 (4), 1039-1074. Electronic copy available at: https://ssrn.com/abstract=3774393 Manry, D., Tiras, S., Whatley, C., 2003. The influence of interim auditor reviews on the association of returns with earnings. The Accounting Review 78 (1), 251-274. 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., Newberry, K., 2001. The influence of tax and non-tax costs on book-tax reporting differences: Public and private firms. The Journal of the American Taxation Association 23 (1), 1-19. Omer, T., Molloy, K, Ziebart, D., 1993. An investigation of the firm size/effective tax rate relationship in the 1980s. Journal of Accounting, Auditing and Finance 8 (2), 167–182. Petersen, M., 2009. Estimating standard errors in finance panel data sets: Comparing approaches. The Review of Financial Studies 22 (1), 435-480. Plumlee, M., 2003. The effect of information complexity on analysts’ use of that information. The Accounting Review 78 (1), 275-96. Plumlee, M., Yohn, T., 2010. An analysis of the underlying causes attributed to restatements. The Accounting Review 24 (1), 41-64. Rego, S., 2003. Tax avoidance activities of U.S. multinational corporations. Contemporary Accounting Research 20 (4), 805–833. Robinson, D., 2008. Auditor independence and auditor-provided tax service: Evidence from going-concern audit opinions prior to bankruptcy filings. Auditing: A Journal of Practice & Theory 27 (2), 31-54. Robinson, J., Sikes, S., Weaver, C., 2010. Performance measurement of corporate tax departments. The Accounting Review 85 (3), 1035-64. Shane, P., Stock, T., 2006. Security analyst and stock market efficiency in anticipating tax- motivated income shifting. The Accounting Review 81 (1), 227-250. Schmidt, A., 2006. The persistence, forecasting, and valuation implications of the tax change component of earnings. The Accounting Review 81 (3), 589-616. Seetharaman, A., Sun, Y., Wang, W., 2011. Tax-related financial statement restatements and auditor-provided tax services. Journal of Accounting, Auditing & Finance 26 (4), 667-98. Tax Executives Institute, Inc., 2011-2012 Corporate tax department survey. Tax Executives Institute, Inc., Washington, D.C. 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. Whalen, D., Coleman, D., and Tanona, D. 2020. 2019 financial restatements: a nineteen year comparison. Sutton, MA: Audit Analytics. Electronic copy available at: https://ssrn.com/abstract=3774393 Figure 1 Panel A: ETR Changes from Quarter 1 to Quarter 4 -.1 -.05 0 .05 .1 Change in ETR from Q1 to Q4 Panel B: ETR Changes from Quarter 2 to Quarter 4 -.1 -.05 0 .05 .1 Change in ETR from Q2 to Q4 Panel C: ETR Changes from Quarter 3 to Quarter 4 -.1 -.05 0 .05 .1 Change in ETR from Q3 to Q4 FIG. 1. This figure presents histograms for changes in ETR from quarters 1 to 4, 2 to 4, and 3 to 4 in Panels A, B, and C, respectively, for the full sample (n=15,009). Electronic copy available at: https://ssrn.com/abstract=3774393 0 10 20 30 40 50 0 10 20 30 40 50 0 10 20 30 40 50 TABLE 1 Descriptive statistics th th th Mean Std Dev 25 Pctl 50 Pctl 75 Pctl Dependent variables ACCURACY1 -0.049 0.081 -0.053 -0.020 -0.006 ACCURACY2 -0.039 0.069 -0.041 -0.015 -0.005 ACCURACY3 -0.029 0.059 -0.027 -0.010 -0.003 ∆Q1Q4 -0.011 0.093 -0.030 -0.005 0.009 ∆Q2Q4 -0.008 0.079 -0.023 -0.004 0.007 ∆Q3Q4 -0.004 0.065 -0.014 -0.001 0.006 External monitors AF 1.769 0.963 1.099 1.946 2.485 EM 0.102 0.302 0.000 0.000 0.000 IO 0.644 0.298 0.444 0.714 0.875 BIG4 0.865 0.341 1.000 1.000 1.000 APTS 0.107 0.123 0.001 0.062 0.170 EXPERT 0.424 0.494 0.000 0.000 1.000 TENURE 2.076 0.886 1.609 2.197 2.708 OFFICE_SIZE 2.705 1.098 2.079 2.773 3.555 Firm characteristics SIZE 7.233 1.735 6.025 7.115 8.343 FIRM_AGE 2.886 0.715 2.398 2.890 3.466 GEO_SEGS 0.978 0.725 0.000 1.099 1.609 BUS_SEGS 0.777 0.731 0.000 1.099 1.386 ∆MIX 0.125 0.244 0.000 0.014 0.116 M&A 0.193 0.394 0.000 0.000 0.000 R&D 0.035 0.057 0.000 0.005 0.050 ∆SALES 0.142 0.287 0.019 0.095 0.203 EARN_VOL 1.248 3.483 0.319 0.573 1.093 DISC_EXTRA 0.184 0.388 0.000 0.000 0.000 DTA 0.000 0.060 -0.023 0.000 0.028 EQUITY_COMP 0.010 0.029 0.001 0.004 0.011 LEV 0.155 0.165 0.001 0.119 0.249 ROA 0.081 0.058 0.041 0.068 0.106 MTB 3.159 3.346 1.570 2.396 3.745 ABACC 0.029 0.504 -0.063 0.002 0.085 This table presents the descriptive statistics for the primary sample (n = 13,509). Appendix B provides variable definitions. All continuous variables are winsorized at the 1% and 99% levels. Electronic copy available at: https://ssrn.com/abstract=3774393 TABLE 2 ETR Accuracy and Bias Panel A: Determinants of interim tax reporting accuracy. (1) (2) (3) ACCURACY1 ACCURACY2 ACCURACY3 External monitors AF 0.002 0.002* 0.002** (0.001) (0.001) (0.001) EM 0.004** 0.004*** 0.004*** (0.002) (0.002) (0.001) IO 0.011*** 0.008*** 0.007*** (0.003) (0.003) (0.002) BIG4 -0.004* -0.001 -0.001 (0.002) (0.002) (0.002) APTS 0.008 0.003 0.001 (0.006) (0.005) (0.004) EXPERT 0.001 0.000 -0.000 (0.002) (0.001) (0.001) TENURE 0.001 0.001 0.002** (0.001) (0.001) (0.001) OFFICE_SIZE 0.001 0.001 0.001 (0.001) (0.001) (0.001) Firm characteristics SIZE 0.003*** 0.001** 0.002*** (0.001) (0.001) (0.001) FIRM_AGE 0.002 0.001 0.002* (0.001) (0.001) (0.001) GEO_SEGS -0.006*** -0.004*** -0.003*** (0.001) (0.001) (0.001) BUS_SEGS 0.001 0.001 0.000 (0.001) (0.001) (0.001) ∆MIX -0.050*** -0.044*** -0.032*** (0.004) (0.004) (0.003) M&A -0.001 -0.001 -0.001 (0.002) (0.002) (0.001) R&D -0.131*** -0.102*** -0.081*** (0.019) (0.015) (0.013) ∆SALES 0.000 0.000 0.000 (0.000) (0.000) (0.000) EARN_VOL -0.001** -0.000* -0.000 (0.000) (0.000) (0.000) DISC_EXTRA -0.006*** -0.004** -0.003** (0.002) (0.002) (0.001) DTA -0.060*** -0.050*** -0.047*** (0.014) (0.012) (0.011) EQUITY_COMP -0.112*** -0.074** -0.026 (0.040) (0.032) (0.024) LEV -0.007 0.001 0.001 (0.005) (0.004) (0.004) ROA 0.294*** 0.275*** 0.249*** (0.017) (0.015) (0.013) MTB -0.001** -0.001** -0.001*** (0.000) (0.000) (0.000) ABACC -0.002 -0.001 -0.001 (0.001) (0.001) (0.001) Constant -0.062*** -0.053*** -0.051*** (0.006) (0.005) (0.005) Year Fixed Effects Yes Yes Yes Industry Fixed Effects Yes Yes Yes R 0.124 0.123 0.117 N 15,009 15,009 15,009 This table presents the results of estimating Equation (1). In Column 1 – 3, the dependent variable is ACCURACY1, ACCURACY2, and ACCURACY3, respectively. *, **, and *** indicate statistical significance at the 0.10, 0.05, and 0.01 levels, respectively. p-values are two-tailed and are calculated based on standard errors that are clustered by firm (Petersen, 2009; Gow et al., 2010). Appendix B provides variable definitions. Electronic copy available at: https://ssrn.com/abstract=3774393 Table 2, continued Panel B: Determinants of interim tax reporting bias. (1) (2) (3) ∆Q1Q4 ∆Q2Q4 ∆Q3Q4 External monitors AF 0.002 0.003** 0.002* (0.001) (0.001) (0.001) EM 0.000 -0.003 -0.004** (0.003) (0.002) (0.002) IO -0.006 -0.009*** -0.005** (0.004) (0.003) (0.002) BIG4 -0.002 0.001 0.000 (0.003) (0.002) (0.002) APTS 0.005 -0.009 -0.005 (0.007) (0.006) (0.005) EXPERT -0.002 -0.001 -0.001 (0.002) (0.001) (0.001) TENURE 0.001 0.001* 0.001 (0.001) (0.001) (0.001) OFFICE_SIZE -0.001 -0.001 -0.001 (0.001) (0.001) (0.001) Firm characteristics SIZE 0.000 -0.001 -0.001 (0.001) (0.001) (0.001) FIRM_AGE -0.000 -0.002 -0.000 (0.001) (0.001) (0.001) GEO_SEGS -0.003* -0.002 -0.001 (0.002) (0.001) (0.001) BUS_SEGS -0.000 -0.001 -0.000 (0.001) (0.001) (0.001) ∆MIX -0.004 0.005 0.006* (0.005) (0.005) (0.004) M&A 0.000 -0.001 -0.001 (0.002) (0.002) (0.002) R&D -0.039* -0.017 -0.027* (0.022) (0.018) (0.015) ∆SALES 0.000* 0.000 -0.000 (0.000) (0.000) (0.000) EARN_VOL -0.000 -0.000 -0.000 (0.000) (0.000) (0.000) DISC_EXTRA -0.006*** -0.003 -0.001 (0.002) (0.002) (0.002) DTA -0.022 -0.025* -0.022* (0.016) (0.013) (0.012) EQUITY_COMP 0.139*** 0.052** 0.005 (0.052) (0.024) (0.019) LEV -0.007 -0.001 0.001 (0.006) (0.005) (0.004) ROA -0.080*** -0.064*** -0.055*** (0.019) (0.016) (0.014) MTB 0.001*** 0.001** 0.001*** (0.000) (0.000) (0.000) ABACC -0.002 0.001 0.001 (0.002) (0.001) (0.001) Constant 0.010 0.012** 0.010* (0.007) (0.006) (0.005) Year Fixed Effects Yes Yes Yes Industry Fixed Effects Yes Yes Yes R 0.011 0.009 0.009 N 15,009 15,009 15,009 This table presents the results of estimating Equation (1). In Column 1 – 3, the dependent variable is ∆Q1Q4, ∆Q2Q4, ∆Q2Q4, respectively. *, **, and *** indicate statistical significance at the 0.10, 0.05, and 0.01 levels, respectively. p- values are two-tailed and are calculated based on standard errors that are clustered by firm (Petersen, 2009; Gow et al., 2010). Appendix B provides variable definitions. Electronic copy available at: https://ssrn.com/abstract=3774393 TABLE 3 Positive and Negative ETR Surprises (1) (2) Variables POS_SURPRISE NEG_SURPRISE External monitors AF -0.019 -0.116** (0.061) (0.053) EM -0.304** -0.060 (0.122) (0.101) IO -0.524*** -0.157 (0.151) (0.140) BIG4 -0.005 0.039 (0.132) (0.119) APTS -0.120 0.248 (0.342) (0.296) EXPERT 0.019 0.030 (0.083) (0.072) TENURE -0.028 -0.016 (0.051) (0.042) OFFICE_SIZE -0.044 -0.034 (0.039) (0.035) Firm characteristics SIZE -0.117*** -0.044 (0.036) (0.032) FIRM_AGE -0.020 -0.111* (0.068) (0.060) GEO_SEGS 0.300*** 0.242*** (0.068) (0.058) BUS_SEGS -0.056 0.029 (0.065) (0.053) ∆MIX 0.850*** 0.884*** (0.126) (0.114) M&A 0.031 0.089 (0.093) (0.083) R&D 2.866*** 3.732*** (0.851) (0.622) ∆SALES -0.005 0.010 (0.015) (0.017) EARN_VOL 0.004 0.014** (0.010) (0.007) DISC_EXTRA 0.157 0.265*** (0.096) (0.083) DTA 0.879 3.557*** (0.721) (0.617) EQUITY_COMP 0.085 0.681 (1.533) (0.762) LEV -0.122 -0.235 (0.284) (0.232) ROA -31.562*** -14.196*** (2.580) (1.246) MTB 0.056*** 0.006 (0.014) (0.013) ABACC 0.043 -0.021 (0.079) (0.064) Constant -0.117*** -0.044 (0.036) (0.032) Year Fixed Effects Yes Industry Fixed Effects Yes < 0.01 Wald Χ (p-value) N 15,009 This table presents the results of estimating a multinomial logistic regression with POS_SURPRISE and NEG_SURPRISE as the outcomes. *, **, and *** indicate statistical significance at the 0.10, 0.05, and 0.01 levels, respectively. p-values are two-tailed and are calculated based on standard errors that are clustered by firm (Petersen, 2009; Gow et al., 2010). Appendix B provides variable definitions. Electronic copy available at: https://ssrn.com/abstract=3774393 TABLE 4 De-Biased Accuracy Measure (1) (2) (3) Variables ACC1_debiased ACC2_debiased ACC3_debiased External monitors AF 0.001 0.002** 0.002*** (0.001) (0.001) (0.001) EM 0.004** 0.004** 0.004** (0.002) (0.002) (0.001) IO 0.012*** 0.007*** 0.006*** (0.003) (0.002) (0.002) BIG4 -0.003 0.000 -0.000 (0.002) (0.002) (0.002) APTS 0.007 0.002 -0.000 (0.005) (0.005) (0.004) EXPERT 0.001 0.001 -0.000 (0.001) (0.001) (0.001) TENURE 0.001 0.001* 0.002*** (0.001) (0.001) (0.001) OFFICE_SIZE 0.001 0.001 0.001 (0.001) (0.001) (0.000) Firm characteristics SIZE 0.003*** 0.001** 0.001*** (0.001) (0.001) (0.000) FIRM_AGE 0.002* 0.001 0.002** (0.001) (0.001) (0.001) GEO_SEGS -0.006*** -0.004*** -0.003*** (0.001) (0.001) (0.001) BUS_SEGS 0.001 0.001 0.000 (0.001) (0.001) (0.001) ∆MIX -0.050*** -0.044*** -0.032*** (0.003) (0.002) (0.002) M&A -0.001 -0.001 -0.001 (0.002) (0.001) (0.001) R&D -0.130*** -0.100*** -0.086*** (0.013) (0.011) (0.010) ∆SALES 0.000 -0.000 0.000 (0.000) (0.000) (0.000) EARN_VOL -0.001*** -0.000** -0.000* (0.000) (0.000) (0.000) DISC_EXTRA -0.006*** -0.004*** -0.003** (0.002) (0.001) (0.001) DTA -0.055*** -0.049*** -0.046*** (0.011) (0.010) (0.008) EQUITY_COMP -0.103*** -0.076*** -0.031* (0.023) (0.019) (0.017) LEV -0.007* -0.000 -0.000 (0.004) (0.004) (0.003) ROA 0.273*** 0.251*** 0.227*** (0.012) (0.011) (0.009) MTB -0.001*** -0.000** -0.001*** (0.000) (0.000) (0.000) ABACC -0.003** -0.001 -0.000 (0.001) (0.001) (0.001) Constant -0.065*** -0.054*** -0.050*** (0.006) (0.005) (0.004) Year Fixed Effects Yes Yes Yes Industry Fixed Effects Yes Yes Yes R 0.118 0.114 0.108 N 15,009 15,009 15,009 This table presents the results of estimating Equation (1) after replacing the dependent variable with de-biased accuracy as the dependent variable. *, **, and *** indicate statistical significance at the 0.10, 0.05, and 0.01 levels, respectively. p- values are two-tailed and are calculated based on standard errors that are clustered by firm (Petersen, 2009; Gow et al., 2010). Appendix B provides variable definitions. Electronic copy available at: https://ssrn.com/abstract=3774393 TABLE 5 Information Content of Book and Estimated Taxable Income Dependent variable = RETURN Full Sample Pre-SOX Post-SOX (1) (2) (3) (4) (5) (6) ∆PTBI 1.784*** 1.764*** 1.791*** 1.761*** 1.723*** 1.709*** (0.087) (0.087) (0.080) (0.080) (0.045) (0.045) ∆TI 0.371*** 0.493*** 0.626*** 0.846*** 0.256*** 0.328*** (0.073) (0.082) (0.084) (0.093) (0.038) (0.044) 3YR_ACCURACY -0.012 -0.000 -0.021* (0.010) (0.021) (0.011) ∆TIx3YR_ACCURACY 0.601*** 1.085*** 0.358*** (0.168) (0.204) (0.105) Constant 0.090*** 0.089*** 0.080*** 0.078*** 0.096*** 0.095*** (0.004) (0.004) (0.008) (0.008) (0.004) (0.004) N 33,688 33,688 12,519 12,519 21,169 21,169 R 0.087 0.088 0.084 0.086 0.092 0.092 This table presents the results of Equation (2). *, **, and *** indicate statistical significance at the 0.10, 0.05, and 0.01 levels, respectively. p- values are two-tailed and are calculated based on standard errors that are clustered by firm (Petersen, 2009; Gow et al., 2010). Appendix B provides variable definitions. Electronic copy available at: https://ssrn.com/abstract=3774393 TABLE 6 Relative Information Content of Estimated Taxable Income and Book Income Panel A: Relative information content of estimated taxable income to book income for high and low tax accuracy firms (split at the median of 3YR_ACCURACY) High Accuracy Firms (Median) Low Accuracy Firms (Median) Year Difference in 2 2 2 2 2 2 2 2 2 2 Obs ∆PTBI R ∆TI R R /R Obs ∆PTBI R ∆TI R R /R R /R PTBI TI TI PTBI PTBI TI TI PTBI TI PTBI Avg_Pre02 713 3.160 0.164 2.519 0.111 0.670 682 2.927 0.119 1.557 0.069 0.575 0.095 Avg_Post02 776 2.297 0.119 1.174 0.050 0.425 742 2.187 0.137 0.886 0.032 0.236 0.189*** Avg_Total 749 2.672 0.139 1.759 0.076 0.532 716 2.116 0.130 1.166 0.048 0.383 0.148*** Panel B: Relative information content of estimated taxable income to book income for high and low tax accuracy firms (split at the median of 3YR_ACCURACY) for a sample of firms matched on level of tax planning (based on GAAP_ETR5) High Accuracy Firms (Median) Low Accuracy Firms (Median) Year Difference in 2 2 2 2 2 2 2 2 2 2 Obs ∆PTBI R ∆TI R R /R Obs ∆PTBI R ∆TI R R /R R /R PTBI TI TI PTBI PTBI TI TI PTBI TI PTBI Avg_Pre02 610 3.319 0.172 2.933 0.139 0.778 610 2.212 0.129 1.674 0.075 0.610 0.167* Avg_Post02 638 2.446 0.123 1.390 0.058 0.780 638 1.831 0.119 0.796 0.038 0.347 0.433** Avg_Total 626 2.826 0.144 2.061 0.093 0.779 626 1.997 0.123 1.178 0.054 0.462 0.318*** This table presents the results of estimating Equations (3) and (4). *, **, and *** indicate statistical significance at the 0.10, 0.05, and 0.01 levels, respectively. p-values are based on one-tailed t-tests and computed using the yearly values of 2 2 R /R for high accuracy firms versus low accuracy firms. Inferences are unchanged if using Wilcoxon rank sum tests to TI PTBI calculate p-values. Appendix B provides variable definitions. Electronic copy available at: https://ssrn.com/abstract=3774393 TABLE 7 Earnings Announcement Abnormal Returns for Firms that Beat Analysts’ Forecast DV = CAR (1) (2) Intercept 0.007*** 0.007*** (0.003) (0.003) Beat_w_Tax -0.004*** (0.001) Beat_w_Tax_Accurate 0.000 (0.002) Beat_w_Tax_Inaccurate -0.006*** (0.002) AFE 9.204*** 9.194*** (0.448) (0.448) BTM 0.005** 0.005** (0.002) (0.002) SIZE -0.000 -0.000 (0.000) (0.000) MOMENTUM -0.080*** -0.080*** (0.002) (0.006) N 18,331 18,331 R 0.033 0.033 This table presents the results of estimating Equation (4). *, **, and *** indicate statistical significance at the 0.10, 0.05, and 0.01 levels, respectively. p-values are two-tailed and are calculated based on standard errors that are clustered by firm (Petersen, 2009; Gow et al., 2010). Appendix B provides variable definitions. Electronic copy available at: https://ssrn.com/abstract=3774393
Journal
ARN Conferences & Meetings – SSRN
Published: Jan 27, 2021
Keywords: accounting for income taxes, market pricing, audit quality, analyst coverage, institutional investors, earnings management