Book-Tax Differences: Which Ones Matter to Equity Investors? Jana S. Raedy University of North Carolina Jeri Seidman University of Texas Douglas A. Shackelford University of North Carolina and NBER April 30, 2010 Preliminary and Incomplete Please Do Not Quote Without the Authors’ Permission We appreciate the insights and encouragement of Jim Poterba, the financial support of the UNC Tax Center, and the research assistance of numerous students at the Massachusetts Institute of Technology, the University of North Carolina, and the University of Texas in the collection and organization of the primary data used in this study. 1. Introduction The purpose of this paper is to determine the specific differences between book income and taxable income (book-tax differences or BTDs) that matter to equity investors. Prior studies (e.g., Hanlon, et al., 2005) report that taxable income (estimated from the financial statements) has information content that is incremental to pretax book income, which implies that the market prices BTDs. Unfortunately, the lack of computer-readable information about tax footnotes (the source of detailed BTD information) has prevented these studies from identifying the specific accounts that are incrementally informative. Consequently, it is difficult, if not impossible, to know the specific disclosures that investors find informative in the tax footnotes. For example, are the markets pricing the difference in book and tax depreciation (the largest BTD)? Or are they pricing bad debt reserves, goodwill amortization, foreign tax rate differentials, or a host of other small items? Why should any of these items matter to investors, given that many are disclosed in more and better detail elsewhere in the financial reports? We attempt to shed light on these questions with hand-collected data from the tax footnotes of the Fortune 250 from 1993 to 2007. These data enable us to specify tests of the association between current stock returns and specific BTDs. BTDs have been documented to matter in all sorts of settings. Among other things, extant studies show associations between BTDs and current and future returns, persistence, earnings growth, earnings-to-price ratios, big baths, credit ratings, borrowing costs, restatements, tax shelters, and IRS audits. 1 The fact that BTDs are so well associated is not surprising 1 For a sampling of papers focusing on book-tax differences and accrual quality, see Mills and Newberry (2001), Phillips, et al. (2003), Lev and Nissim (2004), Hanlon (2005), Badertscher, et al. (2009), Blaylock, et al. (2010), among many others. For a sampling linking BTDs and tax avoidance, see Mills (1998), Desai (2003), Desai and Dharmapala (2006, 2009), Wilson (2009), Lisowsky (2009), and Chen, et al. (2010), among many others. For because, as the bridge between the financial statements and the tax returns, BTDs involve almost every element of the firm. Moreover, the standard measure of a firm’s BTDs (the current income tax expense, grossed-up by the statutory tax rate, and subtracted from pretax book income— sometimes with an adjustment for the change in net operating loss carryforwards) draws from every non-conforming (i.e., accounting treatment differs between book and tax) account in the general ledger plus tax credits and foreign and state rate differentials. Therefore, to say that book-tax differences are informative is akin to saying that the financial statements and the tax returns communicate information about a firm. Thus, it is not enough to know that BTDs matter. Furthermore, as BTDs have been shown to be associated with so many things, skepticism has developed about whether they really explain anything. What is needed in the literature now is more understanding of which bits of information imbedded in the BTD hodgepodge of accounts matter, to whom they matter, and why they matter. With disaggregated BTDs from thousands of statements of deferred tax positions and rate reconciliations, we have overcome a data limitation that has prevented researchers from moving forward to address to some of these questions. We begin by revisiting the pricing of BTDs for an important user of financial statements, the equity investor. We first scour the statements of deferred tax positions and rate reconciliations for BTDs that could potentially be incrementally informative. We then drop items from the analysis that pertain to taxes alone rather than to differences in book and taxable income (e.g., state and foreign tax rate differentials). We refer to these items as non-BTDs. We also drop items that cannot provide information incremental to the remainder of the financial statements because the relevant information conveyed in the tax footnote is detailed elsewhere in the financial examples of papers focusing on BTDs and both accrual quality and tax avoidance, see Frank, et al. (2009) and Seidman (2010). For reviews of the entire literature, see Hanlon and Heitzman (2010) and Graham, et al. (2010). 2 statements (e.g., pensions and bad debt reserves). We refer to these items as redundant BTDs. Throughout the study we assume that investors interested in information about pensions, for instance, will look first to disclosures centered on pensions, rather than any indirect information about pensions that might be discernible from the tax footnotes. Note that prior studies limited to Compustat data items have been unable to segregate non-BTDs or redundant BTDs. With access to the actual line items in the statements of deferred tax positions and rate reconciliations, we can purge items that should not be included in tests of BTDs (i.e., the non-BTDs) or tests of the incremental informativeness of BTDs (i.e., the redundant BTDs). After dropping the non-BTDs and the redundant BTDs from the study, we build on Hanlon, et al. (2005) by regressing long-window contemporaneous returns on the remaining BTDs, controlling for pretax book income, total cash flows, year, and industry. We find that both total temporary differences and total permanent differences are incrementally informative, although the sign is opposite our prediction for permanent differences. Drilling down, we find a few relatively small temporary BTDs that appear to provide incremental information about accrual quality. 2 They deal mostly with revenue recognition (e.g., deferred revenue, customer discounts, contract accounting, and deferred gains on sales of assets) and accrued or deferred general business expenses (e.g., frequent flyer plans, direct marketing, policy acquisition, and contingent rent) that are recognized in one period for book and another for tax purposes. Consistent with these BTDs mattering to equity investors, it is easy to see that the underlying accounts could provide opportunity for earnings management and that the manipulation could only be detected through close inspection of the tax footnote disclosures. However, these few accounts provide the most compelling case that we can make in support of BTDs being priced. 2 We use the term, “accrual quality” in the spirit of Dechow, et al. (2009), who state that “Higher quality earnings more faithfully represent the features of the firm’s fundamental earnings process that are relevant to a specific decision made by a specific decision-maker. (p.1)” 3 We find no evidence of widespread use of the BTDs to ascertain accrual quality. We also find little evidence that investors find the BTDs that deal primarily with the firm’s tax obligations (e.g., net operating loss carryforwards and tax-exempt income) to be incrementally informative. All in all, investors only seem to care about a few BTDs. There is certainly no overwhelming support for the notion that the BTDs provide important information content to equity investors. The paper proceeds as follows: Section 2 develops the testable hypotheses. Section 3 discusses the sample selection and the research design. Section 4 reports the empirical findings. Closing remarks follow. 2. Hypotheses Development As the connection between reported book and estimated taxable income, book-tax differences potentially can inform investors about the quality of the book numbers (by providing a comparison of the book figures with an alternative measurement system, the tax law) and/or provide additional information (beyond the income tax expense and cash taxes paid) about the firm’s tax liabilities. In this section, we develop hypotheses about both potential roles for BTDs. The first hypothesis concerns accrual quality. Prior studies have concluded that market participants use the BTDs to assess the accrual quality of pretax book income. For example, Hanlon (2005) reports that firms with large, temporary BTDs have less persistent earnings, her indicator of lower accrual quality. She infers that investors interpret large, temporary BTDs as a red flag and lower their expectations about the firm’s future earnings persistence. However, since Hanlon’s (2005) analysis is limited to aggregated temporary BTDs (except for a handful of observations that she hand-collects for robustness checks), she cannot isolate her tests to those 4 BTDs that should matter most to investors and thus verify that her results are driven by BTDs that could reasonably be expected to communicate information about accrual quality. In contrast, this study can identify specific temporary differences that could reasonably be interpreted by the market as indicators of accrual quality. Non-conforming accounts that give rise to accrual quality BTDs are those where sufficient financial accounting discretion exists that firms could manipulate their accruals to advantageously manage their earnings. An example of a deferred tax position potentially related to accrual quality is unearned revenue. When cash prepayments are received, they are recorded as a liability for book purposes, but the cash may be included in taxable income even though the income has not yet been earned. 3 With this account, the tax treatment is determined on the cash basis; however, for book purposes, managers have some leeway in the timing of the income recognition. Thus, managers can use this account to achieve financial accounting goals without adversely affecting their tax liability. If managers use these types of BTDs to assess a firm’s accrual quality, then we predict that unexpected decreases (unexpected increases) in deferred tax assets (deferred tax liabilities) will be associated with negative returns. 4 For example, unexpected decreases in the deferred tax asset for unearned revenue would arise if firms were more aggressively recognizing revenue in the current year than they did in the prior year. Thus, we predict a positive association between stock returns and the unexpected change in the unearned revenue deferred tax asset. 3 The majority of deferred tax positions that could be interpreted as accrual quality are deferred tax asset positions. An example of a deferred tax liability involves long-term construction contracts where some small contractors are allowed to use percentage-of-completion for book purposes (recognizing book income while the work is being conducted) and completed contract for tax purposes (recognizing no taxable income until the project has been finished). Investors could potentially find that the resulting deferred tax liability provides information about the firm’s accrual quality. 4 Our benchmark for expected change in deferred tax position is the prior year change. Because permanent differences are relatively stable and recurring temporary differences are not generally reversing in growth firms, this assumption seems reasonable. 5 In other words, if firms choose to use unearned revenues to manage income, their manipulation will leave footprints in their deferred tax accounts in a predictable manner. Our tests are designed to see if these footprints are informative to investors. Note that this example is consistent with Hanlon’s (2005) finding that positive temporary BTDs (which would arise from decreasing deferred tax assets or increasing deferred tax liabilities) are associated with lower earnings persistence. The first hypothesis states in alternative form: H1: Stock returns are negatively associated with temporary book-tax differences that arise from financial accounting discretion. One reason that we might fail to detect an association between stock returns and BTDs for accrual quality assessment is that earnings management can take many forms, but our hypothesis only relates to an overstatement of current period profits. Under some circumstances (e.g., big baths and the creation of cookie jar reserves), increases (decreases) in deferred tax assets (liabilities) would be consistent with earnings manipulation. Later in the study, we focus on firms that appear to manage their earnings more than other firms and test for cross-sectional differences. Initially, however, we posit a simple hypothesis that investors use the deferred tax accounts to assess the accrual quality of all firms and that a smaller difference between book and taxable income is viewed positively. This is consistent with the prior literature in this area, e.g. Hanlon (2005), which posits that accrual quality is negatively related to BTDs. The second hypothesis concerns tax avoidance. 5 Disclosures in both the statement of deferred tax positions (e.g., NOL carryforwards) and the rate reconciliation (e.g., tax-exempt interest) potentially provides investors with information about current and future tax liabilities that they could not find elsewhere in the financial statements. We anticipate a positive 5 Throughout the study, we follow Dyreng, et al. (2008) in using the term “tax avoidance” to include anything that reduces a company’s taxes, conditional on its pretax book income. Therefore, we do not attempt to distinguish between tax aggressiveness, tax risk, tax evasion, tax planning, tax sheltering, and similar variants. 6 association between stock returns and temporary or permanent differences that indicate less income than expected will be subject to taxation now or in the future. For example, the rate reconciliation indicates the extent to which tax-exempt municipal bond interest income reduces the effective tax rate (ETR). We predict that stock returns will be rising with increases in the book income that is not subject to tax. Thus, the second hypothesis states in alternative form: H2: Stock returns are positively (negatively) associated with book-tax differences that indicate the firm will pay lower (higher) taxes. That said, we recognize that investors may view some forms of tax avoidance negatively (Hanlon and Slemrod, 2009). For example, involvement in a tax fraud or some forms of corporate tax shelters, even if they reduce current taxes, might bring reputational damage. In later tests, we examine one BTD that might fit this definition, corporate-owned life insurance (COLI). If investors take a dim view of COLI, then we may find that ETR-reducing COLI activity is associated with negative returns. The final hypothesis concerns deferred tax positions and reconciling items with limited descriptions. “Other” is a common label for both temporary and permanent differences. This is because, while all significant reconciling items must be disclosed, SEC Reg S-X Rule 4-08(h)(1) defines significant as 5% of the amount computed by multiplying the income before tax by the applicable statutory federal tax rate. Besides being the account for items that are too small to warrant separate disclosure, some firms could bury sensitive information in “Other.” Whatever the reason, such limited descriptors are presumably insufficient to convey any information. If so, we should find no association between equity prices and these inadequately identified BTDs. This leads to the third hypothesis, stated in null form: H3: Stock returns are not associated with book-tax differences that are termed “other” or are similarly nondescript. 7 3. Empirical Design 3.1. Regression Equation To assess whether individual book-tax differences are informative, conditional on pretax book income, total cash flows, year, and industry, we estimate the following regression equation: Rit = β 0 + β1 ∆PTBI it + β 2 ∆CFit + Σβ 3t YEARt + Σβ 4 j INDit + Σβ 5 k ∆BTDit + ε it , (1) where Rit denotes the market-adjusted return for firm i and is defined as the compound (with dividend) return less the compound return on the value-weighted market portfolio. Returns are calculated over a 16-month period, starting at the beginning of fiscal year t and ending four months after the end of the fiscal year t. ∆PTBI it is the change in our measure of accounting earnings, which is pretax book income (Compustat item PI) less minority interest (MII). ∆CFit is the change in our measure of total cash flows (sum of OANCF, IVNCF, and FINCF). YEARt denotes categorical variables for each year. INDi denotes categorical variables for each of the fourteen industries used in Barth, et al. (1998). ∆BTDit denotes the change in book-tax differences, our key variables of interest in the study. Below we detail our measurement of the k book-tax differences. All explanatory variables, except the indicators, are scaled by the market value of equity at the beginning of the year. Our regression model builds on Hanlon, et al. (2005) who use the same dependent variable and control for pretax accounting earnings in their test of the incremental informativeness of estimated taxable income. Unlike Hanlon, et al. (2005), we add a control for total cash flows to ensure that any informativeness that we find is not arising from the fact that the tax law often has elements of cash-basis accounting. We do not want to erroneously conclude that a BTD is informative simply because our regression model excludes a fundamental 8 measure, such as total cash flows, which investors surely evaluate before delving into the far more complicated tax footnotes. We also add year and industry variables. Hanlon, et al. (2005) test each year separately so they do not need year categorical variables. We include industry indicators since BTDs can vary substantially across industries because GAAP and tax law differ across industries. That said, as discussed below, we find that adding a measure of cash flows and year and industry indicator variables has no impact on our inferences. The other difference between our model and Hanlon, et al. (2005) is the primary contribution of this study, i.e., we substitute individual BTDs hand-collected from the tax footnotes for their estimate of taxable income. Note that the coefficient on their estimate of taxable income (current tax expense, divided by the statutory tax rate, less the change in net operating loss carryforwards) is not directly comparable to the coefficients on our measures of BTDs. The reason is that their estimate of taxable income captures the incremental informativeness of both total BTDs and non-BTDs, such as credits and rate differentials. Hanlon, et al. (2005) recognize this shortcoming and others with their estimate of taxable income, discuss them in detail in their Appendix, and attempt to mitigate any measurement errors. Of course, an advantage of their data is that, by avoiding costly hand-collection, they can evaluate a larger sample of firms. 3.2. Sample Selection As mentioned above, details about specific BTDs are generally unavailable in computerreadable form. Thus, to conduct tests about individual BTDs, we collect data from the tax footnotes in 10-K filings for Fortune 250 firms for the fiscal years 1993-2007. By beginning the sample with fiscal year 1993, the first year when all firms’ financial statements were prepared in 9 accordance with SFAS No. 109, we avoid commingling tax information in the financial statements governed by SFAS No. 109 with tax information in the financial statements governed by its predecessors. We collect data for the entire sample period for any firm in the Fortune 250 in any year between 1995 and 2004. 6 Four hundred and eight firms appear in the Fortune 250 at least once between 1995 and 2004. We do not collect data on 22 firms. 7 Because IPOs, mergers, bankruptcy and going-private transactions eliminate some firms or disclosures, not all firms remain in the sample for all years. To further supplement the sample, we use tax footnote data from Poterba, et al. (2010) for firms involved in corporate control transactions with members of the Fortune 50. Firms acquired by a Fortune 50 firm were collected before the merger and firms divested by a Fortune 50 firm were collected after the spin-off. This process yields an additional 1,090 observations (260 firms), providing us with a total sample of hand-collected data for 5,688 firm-year observations over 646 firms. We match each firm-year observation with Compustat using both firm name and year, and validate the match using total assets and net income. We find that 5,658 firm-year observations have a valid match with Compustat. 8 Because a key measure in the study, the year- 6 Before 1995, Fortune ranked only manufacturing firms. To avoid including firms that are only in the Fortune 250 because non-manufacturing firms were excluded before 1995, we formed our sample using the Fortune rankings from 1995 to 2004. 7 The 22 excluded firms are nine private companies that do not file publicly-available 10-Ks: AXA Financial, Borden Chemical, Levi Strauss, Premcor, State Farm Insurance, Supermarket General Holdings, TIAA-CREF and the two predecessor firms of TIAA-CREF; two government-sponsored enterprises: Fannie Mae and Freddie Mac; two co-operatives: CHS and Farmland Industries; one publicly traded LP: Plains All American Pipeline; seven mutual companies: Guardian Life of America, Liberty Mutual Insurance Group, Massachusetts Mutual Life Insurance, Nationwide, New York Life Insurance, Northwestern Mutual, and USAA; and one firm for whom we were unable to locate 10-K filings indicating a firm of sufficient size to be included in the Fortune 250: R. H. Donnelley. 8 We collected tax information from the first 10-K or annual report filing for each fiscal year. Restatements can cause differences between the total assets and net income entries in the 10-K and those reported in Compustat. We hand-checked the observations where neither total assets nor net income corresponded to our hand-collected total assets and net income numbers. We dropped 30 firm-years, 23 for which Compustat did not have any data, four 10 to-year change in temporary differences, requires three consecutive years of data for each firm, 1,360 firm-years are lost. Another 181 firms are missing the change in cash flow. We drop 150 observations because at least one month of returns are missing from the required 16-month return beginning at the start of the fiscal year. Finally, we delete 11 firm-years because of a change in fiscal year-end and another 66 firm-years because they did not have a sufficiently large industrygroup in the Compustat universe to calculate the industry-year rank of their cash ETR (a longrun measure of cash taxes paid over pre-tax income), which we use in secondary tests. Our final sample has 3,890 firm-years. 3.3. Book-Tax Difference Variables 3.3.1. Classifications We compile the book-tax differences from tax information in the financial statements that we hand-collected. GAAP requires firms to disclose a schedule of deferred tax positions that provides information about deferred tax assets and liabilities and a rate reconciliation that reconciles reported income tax expense with the amount that would result from applying the domestic federal statutory rate to global pretax income. 9 From these disclosures, we collect the temporary and permanent differences investigated in this study. Companies list more accounts involving more dollars on their statements of deferred tax positions than they do on their rate reconciliations. However, both disclosures are substantial. The number of accounts listed on the statement of deferred tax positions for our sample of Fortune 250 firms ranges from two to 28 with a mean (median) of ten (nine) and a standard where a partial year or merger caused a mismatch between the data we collected and that reported by Compustat, and three with major differences for undeterminable reasons. 9 All significant reconciling items must be disclosed. SEC Reg S-X Rule 4-08(h)(2) defines significant as 5% of the amount computed by multiplying the income before tax by the applicable statutory federal tax rate. 11 deviation of three. The total dollars of deferred tax positions in absolute value ranges from $290,000 to $43 billion with a mean (median) of $867 ($304) million and a standard deviation of $1,975 million. The number of reconciling items ranges from zero to 17 with a mean of five, a median of four and a standard deviation of two. The total dollars of reconciling items in absolute value ranges from zero to $41 billion with a mean (median) of $256 ($81) million and a standard deviation of $906 million. 10 Although most companies follow a stable reporting policy from year to year, disclosures vary substantially across companies in the level of detail and the terms used to describe their accounts. Lacking any standard classifications, we begin by grouping similar items together. We form 22 categories from the statement of deferred tax positions and 19 categories from the rate reconciliation. The Appendix provides details about representative items in each category. 3.3.2. Categories from the Statement of Deferred Tax Positions Table 1 provides descriptive statistics for the 22 categories of temporary differences calculated from the statements of deferred tax position. Temporary differences are calculated as changes in net deferred tax assets (liabilities), stated in millions of dollars of tax. Temporary differences that increase a deferred tax asset or decrease a deferred tax liability (and thereby increase taxable income) are listed as positive values; those that decrease a deferred tax asset or increase a deferred tax liability (decreasing taxable income) are listed as negative values. The first column of figures has the mean for the entire investigation period. The second column 10 In their call for a descriptive study of common permanent differences, which this paper attempts to provide at least in part, Hanlon and Heitzman (2010) speculate that permanent differences are “rare” compared with temporary differences. Based on our comparison of the statements of deferred tax positions and rate reconciliations, “rare” appears a bit strong, but temporary differences are larger and appear more frequently than do permanent differences. 12 shows the percentage of all observations for which that variable is not zero. The remaining columns present the annual mean values from 1995 to 2007. 11 Not surprisingly, the largest category in absolute value is PPE (book-tax differences in property, plant and equipment) with -$47 million, on average, for all years. Three other categories of temporary differences average at least $20 million in net deferred tax position: VA (valuation allowance) at -32 million, INTANG (intangible assets) at $21 million, and NOL (net operating loss and other carryforwards) at $20 million. 12 Not surprisingly, the most frequently disclosed category is OTHER-t, which includes all items with such inadequate description, e.g., “other adjustments,” that we could not place that item in any category. Since only material items must be separately reported, 99% of the sample firms have at least one item in their statement of deferred tax position that is included in OTHER-t. Additional categories appearing on at least half of the statements of deferred tax positions are PPE, found 90% of the time, VA listed on 57% of the statements, BENEFITS (employee benefits) on 55%, and CREDITS-t (credit carryforwards) on 53%. The means of the BTD categories vary widely across years. Every category has at least one year in which its mean value is negative and, except for PPE, at least one year in which its mean value is positive. This is consistent with reversing temporary differences, where the book revenue (expense) sometimes exceeds the taxable income (deduction) and sometimes it does not. The mean for ADA (allowance for doubtful accounts) soared in 2007, consistent with increasing expectations of future write-offs, and rising bad debt expense, during the financial crisis. NOL 11 As mentioned above, the regression analysis requires the change in temporary differences, where temporary differences are calculated as the change in each net deferred tax asset (liability). This requirement results in the loss of observations for the first two years of data, eliminating all 1993 and all, but seven, 1994 observations. With only seven observations, we do not report any summary statistics for 1994 in Table 1; however, we do retain these observations in the study’s analyses. 12 The suffix “t” denotes temporary differences from the statement of deferred tax position and an “r” denotes disclosures from the rate reconciliation. 13 jumped in the late 1990s and remained elevated thereafter, consistent with Altshuler, et al. (2008) who find that corporate tax losses soared around the millennium. VA became increasingly negative over the years, apparently reflecting both the recent economic downturn that caused ADA to soar and the longer-term effects boosting NOL. Consistent with that conjecture, the annual means for VA are correlated at -57% with those for ADA (allowance for doubtful accounts) and -51% with those for NOL; however, the ADA means are not significantly correlated with the NOL ones. Several values were somewhat extreme in 2006: PPE, INTANG, SUB-t (subsidiary, partnership and joint ventures) and M&A-t (divestitures, restructurings, and other merger and acquisition-related disclosures) were unusually negative, and BENEFITS and OPEB (pensions and other post-retirement benefits) were unusually positive. Table 2 shows the mean values for each category across 13 industry groups. 13 The largest industries are Retail with 653 firm-year observations, Durables with 623 observations, and Banks and Insurance with 465 firm-years. As expected, the key BTDs vary substantially across industries. For example, the absolute value of PPE is the largest BTD in five industries and ten-fold (eight-fold) the next largest BTD in the Utilities (Extraction) industry, but it is not among the five largest BTDs in Mining, Food, Pharmaceuticals, Durables, and Computers. INTANG is ten-fold the next largest BTD in Chemical and also the largest in Food, Pharmaceuticals and Banks and Insurance. VA is largest in Durables and Computers. INVENT (inventory) reigns in Mining, and ENVIRON (environmental costs) in Textiles. EXP (general business expenses) is strikingly larger for Banks and Insurance than any other industry. Similarly, REV (revenue recognition) is far greater in Computers than any other industry. These 13 Table 2 does not report 50 firm-years for the industry category ‘Other’, which includes four real estate and nine agriculture firm-year observations plus thirty-seven firm-year observations for conglomerates that do not fit neatly into one industry: General Electric, Berkshire Hathaway and Tyco. We do retain all 50 of these observations in the tests in the study. 14 results confirm the importance on our controlling for industries in the regressions. They also suggest that studying BTDs within, as opposed to across, industries might yield fruitful results. 3.3.3. Categories from the Rate Reconciliation Table 3 presents descriptive statistics for the 19 rate reconciliation categories. The format is identical to that used in Table 1, except we add a column that reports the sample mean as a percent of pre-tax book income (since rate reconciliations are typically expressed in reference to the U.S. statutory tax rate of 35%). Reconciling items that increase (decrease) the effective tax rate are listed as positive (negative) values; items are shown in millions of dollars of tax. The two largest categories in absolute value from the rate reconciliations are not differences in book and taxable income. They are non-BTD tax rate differentials. FOR-r (incremental foreign tax rate) is the largest item with an average of -$26 million. Its negative mean value indicates that, on average, U.S. companies enjoy lower taxes on foreign income than they do on domestic income. Immediately behind FOR-r is STATERATE (incremental state tax rate) at $25 million, a positive value consistent with the additional taxes that U.S. companies pay to the states. No other reconciling items average over $20 million. As with the statement of deferred tax position, the most common item in the rate reconciliation is one that cannot be classified into any category because it is insufficiently labeled. An item categorized as OTHER-r has an entry on 92% of the rate reconciliations. Only two other reconciling items appear on half of the reconciliations: STATERATE on 84% of them and FOR-r on half of them. 15 As with the temporary differences in Table 1, there is wide fluctuation in reconciling items over times. However, as expected, the sign of the reconciling items flips less than the sign of the eventually reversing temporary differences reported in Table 1. For example, the annual mean values for AMORT-GW (amortization of goodwill), AMORT-OTH (amortization of other intangibles), and STATERATE are always positive, indicating that these items usually increase the effective tax rate. The annual mean values for CREDITS-r (credits), DIVS (non-taxable dividends, including the dividends-received deduction), ETI (export incentives, such as the foreign sales corporation exemption), OTHER-r, and TEI (tax-exempt interest) are always negative, indicating that these items generally decrease the effective tax rate. ADJTR (adjustments to deferred tax accounts) soared in 2007, consistent with firms being underprovided at the onset of the financial crisis. AMORT-GW spiked in 2002 and again in 2007, likely due to increased write-offs of goodwill following September 11, 2001, and again as the most recent recession began to unfold. AUDIT (audit adjustments) turns more negative in the later years of the study, consistent with taxpayers paying less upon settlement than they had reserved. CREDITS-r reduced the effective tax rate more and more over the years. FOR-r also grew increasingly negative, consistent with firms lowering their worldwide effective tax rate by sourcing more taxable income in countries with tax rates less than the U.S. statutory rate of 35%. Meanwhile, STATERATE was increasing, consistent with both increasing state tax levies and states becoming more aggressive in collecting tax revenue. REPAT (repatriations effect) jumped around the tax holiday for repatriations under the American Jobs Creation Act of 2004. Lastly, TEI rose sharply toward the end of the investigation period. Table 4 (as with Table 2) shows the mean values for each category across 13 industry groups. As with the temporary differences, the reconciling items vary considerably across 16 industries. However, the two tax rate differentials dominate most industries. FOR-r has the largest mean value in absolute terms in Food, Chemical, Pharmaceuticals, Extraction, Computers, and Other. In these industries, FOR-r is always negative and far exceeds any other items, indicating how important foreign taxes are to companies in those industries. STATERATE is the largest item in the Textile, Retail and Services industries and barely exceeded by CREDITS-r among Utilities. ADJTR is the largest item in Mining and Durables, consistent with both industries being underprovided. AMORT-GW dominates the Transport industry. Not surprisingly, TEI more than doubles any other item for Banks and Insurance. 3.3.4. Items that do not represent differences between book income and taxable income As discussed above, not all of the items in the schedule of deferred tax positions or the rate reconciliation concern BTDs. For example, common items include CREDITS-t (credit carryforwards) in the schedule of deferred tax positions and STATERATE (incremental tax rate attributable to state income taxes) in rate reconciliations. Both of those items relate to differences in the computation of the tax but do not relate to differences in the income measures for book and tax. Since this study addresses differences between book and taxable income, we drop two categories of deferred tax positions (CREDITS-t and VA) and seven categories of reconciling items (ADJTR, AUDIT, CREDITS-r, FOR-r, ETI, REPAT, and STATERATE) that represent a tax difference, rather than a difference between book income and taxable income. As discussed above, studies that are forced to estimate BTDs by taking the difference between current tax expense (grossed-up) and pre-tax book income erroneously include these non-BTD items in their estimate of BTDs. Because we hand-collect data from the statement of deferred 17 tax positions and rate reconciliations, this study is able to exclude these non-BTDs from the analyses below. 3.3.5. Redundant Information Next, we drop from the analysis those temporary differences from the schedule of deferred tax positions and those permanent differences from the rate reconciliation that cannot provide incremental information to investors because the information in the BTDs is disclosed elsewhere in the financial statements, usually in more and better detail. For example, because the reserve method is required for book purposes, but the cash method is required for tax purposes, the pension account often creates a deferred tax asset. This DTA may reveal information about the financial situation of the company’s pension plans and its book and tax treatment. However, since GAAP requires extensive detail about pensions in other sections of the financial statements, including the expense for books and the cash funding provided (which equals the deduction for tax purposes), the pension BTD itself, which nets the expense and deduction, cannot possibly provide anything to investors that is incremental to information that they can find elsewhere in the financial statements. Thus, since the purpose of this study is to isolate those BTDs that provide incremental information to the markets, we exclude pension and other post-employment BTDs, which are found in the OPEB category. Information is similarly redundant for the following six BTD categories, which we also drop: ADA, MTM (mark-tomarket adjustments), WARRANTY (warranties), both forms of amortization (AMORT-GW and AMORT-OTH), and MININT (minority interests). 14 14 We rely on the following pronouncements in our assertions that any information in these BTDs is redundant: ADA: ASC 310-10-50-6 through 10, SEC Reg S-X Rule 12-09; MTM: ASC 320-10-50-2, ASC 820-10-50, ASC 825-10-50-10 through 19; OPEB: ASC 715-30-50-1; WARRANTY: ASC 460-10-50-8; AMORT-GW: ASC 350-2050-1 and 2; AMORT-OTH: ASC 350-30-50-1 through 3; MININT a required disclosure on the income statement. 18 Thus, the rest of the paper focuses on the remaining 16 temporary and nine permanent book-tax differences that have the potential to provide investors with incremental information that they could not find elsewhere in the financial statements. The 16 temporary difference categories are: BENEFITS (employee benefits), ENVIRON (environmental costs), EXP (general business expenses), FOR-t (differences related to foreign income), INTANG (intangible property differences), INVENT (inventory differences), LEASE (leased property differences), M&A-t (differences arising from mergers, acquisitions, divestitures or restructuring), NOL (net loss carryforwards), OTHER-t (other adjustments), PPE (differences related to owned tangible property), REG-t (items unique to the regulated industries), REV (differences in revenue recognition), STATE (state and local taxable income differences), SUB-t (differences related to subsidiaries), and UNUSUAL-t (items significant enough to warrant separate line item disclosure for at least one firm but too unusual to form a separate category). The nine permanent difference categories are: COLI (corporate-owned life insurance), DIVS (non-taxable dividends), DMD (domestic manufacturing deduction), M&A-r (tax-free merger costs, non-deductible divestiture or restructuring charges), OTHER-r (other items), REG-r (items unique to regulated industries), SUB-r (subsidiary differences), TEI (tax-exempt income), and UNUSUAL-r (material, but infrequent items). To summarize, these 25 categories include the temporary and permanent differences between book and taxable income that potentially provide investors with information that is incremental to that found in the rest of the financial reports. Note that in all cases we erred on the side of retaining a category in the study when we were uncertain about its redundancy. For some of these categories, ample information is available outside the tax footnotes and estimating the BTD is probably a straightforward exercise even without the tax footnote. However, if any 19 information appeared to be only found in the tax footnote, we retained that category for the study. For example, many investors can likely estimate with some precision the PPE BTD, which is largely the difference in book and tax depreciation. However, since the tax depreciation figure is not found anywhere in the financial statements, we retained PPE BTD in the analysis, even though we recognize that some investors likely can compute a rough estimate of the tax depreciation without the tax footnotes. 3.3.6. Accrual Quality The remainder of the paper focuses on the 25 categories of temporary and permanent differences not excluded in Sections 3.3.4 and 3.3.5. As discussed above, one reason that BTDs might be incrementally informative is that they signal information to investors about the quality of the firm’s accruals. To test this proposition, we identify among the remaining temporary BTDs those involving accounts with enough financial accounting discretion that investors potentially could learn about the firm’s accrual quality through evaluating changes in the BTD. The nine BTDs that could potentially provide investors with information about the firm’s accrual quality (AQ) are: BENEFITS, ENVIRON, EXP, INTANG, INVENT, LEASE, PPE, REV, and UNUSUAL-t. Below we explain why each potentially serves as an indicator of accrual quality. The BENEFITS BTD relates to common year-end compensation and benefit accruals, such as accrued vacation pay and bonuses and deferred compensation. In most cases these accounts arise because accrued costs are expensed for book purposes in one period and deducted for tax purposes in the following period. Since managers probably have some flexibility to accelerate or delay expensing, investors might study this account to garner some information about the firm’s accrual quality. To our knowledge, this information is not necessarily provided 20 elsewhere in the financials. The EXP BTD is similar to the BENEFITS BTD in that it also includes common accruals around year end that are not disclosed elsewhere. In contrast to the recurring accruals found in BENEFITS and EXP BTDs, the ENVIRON BTD mostly captures non-recurring expenses, e.g., environmental costs that are not deductible until paid. Again, these expenses are not reported elsewhere and managers likely have some discretion in booking these charges. The INTANG BTD relates to book-tax differences for intangible assets, such as the difference between book and tax amortization (or book impairments and tax amortization) for goodwill acquired in a taxable merger. The INVENT BTD relates to differences in book and tax bases, such as LIFO methods, uniform capitalization (UNICAP), or lower-of-cost-or-market (LCM). To the extent the basis difference arises from tax law, e.g., UNICAP, then investors likely will not gain any insights into accrual quality. However, to the extent the bases differences arise from accounting choices, such as application of LCM, then investors might use the INVENT BTD to assess accrual quality. The LEASE BTD relates to difference arising from leased assets, including the difference between book and tax depreciation on leased assets and differences between the treatment of a lease—capital or operating—for book and tax. The PPE BTD concerns differences related to owned tangible property, primarily the difference between tax and book depreciation but also asset impairments and other property differences. While detailed information is disclosed regarding book expenses for owned tangible property, tax information generally must be inferred from the PPE BTD. 15 The REV BTD concerns differences in revenue recognition for book and tax. Examples include unearned revenue (which may trigger taxes when the cash is received upfront and create 15 For a discussion of firms managing the book depreciation in a manner that would give rise to changes in the PPE BTD, see Keating and Zimmerman (2000). As a counter to those who might doubt that investors price differences in book and tax depreciation, they document that a de facto decoupling of book and tax depreciation, arising from the enactment of statutorily determined tax depreciation schedules in the Economic Recovery Act of 1981, increased the frequency of book income-increasing depreciation estimate revisions. 21 a deferred tax asset until income is recognized later for book purposes) and installment sales (where book income is recorded at the time of the sale but a deferred tax liability exists until cash, which triggers taxation, is received). Managerial discretion might be observable through changes in the BTD. Finally, because UNUSUAL-t includes a hodgepodge of accounts, it is difficult to make a prediction about its coefficient. However, since these items are material for the firms that list them, analysts and other market participants may study them closely. Furthermore, our review of the items in this category suggests that many involve managerial discretion. Thus, we have chosen to include UNUSUAL-t among the AQ BTDs. For purposes of constructing regression variables, the temporary difference due to each category of DTA/(DTL) disclosure is calculated by subtracting the prior year’s balance in the deferred tax position from the current year balance; these “levels” of BTDs are presented in Tables 1 and 2. Our regression variable is calculated as the change in each temporary difference (itself already a change variable), divided by the market value of equity at the beginning of the year. 16 A positive coefficient on the change in the AQ temporary differences will be interpreted as evidence that investors find the BTD incrementally informative. For example, the EXP BTD declines when the firm deducts more than it expenses; our regression variable, EXP, is negative when the change in the BTD this year is less than the change in the BTD next year, regardless of whether the BTD actually increased or decreased. Such behavior might raise concerns that the firm has understated its expenses and thus overstated its accounting earnings. If so, the drop in 16 For example, the revenue variable, REV, was $2,598 for Hewlett Packard in 2007; this includes disclosures about deferred revenue and intercompany profits. In 2006 and 2005, this variable was $2,063 and $1,220, respectively. The 2007 temporary difference is calculated as ($2,598-$2,063) = $535; the 2006 temporary difference is calculated as ($2,063-$1,220) = $843. The unscaled REV variable for 2007 is the change in the temporary difference, so ($535 - $843) = -$308. This implicitly assumes that we expected an increase in the REV BTD of $843 and instead Hewlett Packard reported an increase of only $535 so our “surprise” is $308 less change in REV than we expected. The actual REV regression variable for 2007 is equal to 0.29% of beginning MVE. 22 the EXP BTD would be associated with negative returns, resulting in a positive coefficient. We will interpret such a finding as evidence that investors use the EXP BTD to assess the quality of the firm’s accruals. Conversely, an increase in the EXP BTD occurs when a firm accrues expenses in excess of their deductions. Such behavior might allay fears that the book earnings are not sustainable. If so, the increase in the EXP BTD would be associated with positive returns, once again resulting in a positive coefficient. We would interpret this result as evidence that the market is using the EXP BTD to assess the quality of the firm’s accruals. 17 3.3.7. Tax Avoidance—Temporary Differences In addition to using the disclosures in the schedule of deferred tax positions to assess the firm’s accrual quality, we expect that some investors may find five of the remaining temporary BTDs informative about the firm’s future tax outlays. The first is NOL, which includes all net operating loss and other carryforwards that will be available to offset taxable income in the future but will not affect future book income. Unlike the AQ BTDs, the NOL BTD does not communicate information about the manager’s discretion in recording book income and expenses. It solely communicates information about future tax liabilities. The other four BTD categories involve similarly complex, technical tax decisions, but little, if any, book discretion: FOR-t, which concerns foreign income issues; M&A-t, which concerns expenses related to taxable mergers, acquisitions, divestitures, or restructuring events; STATE, which concerns state 17 The same logic applies to BTDs for both deferred tax assets and deferred tax liabilities. However, for deferred tax liabilities, we flip the sign of the year-to-year change so that we can continue to speak of a positive coefficient as an indicator of accrual quality informativeness. To see why this is necessary, consider installment sales, which create deferred tax liabilities because they are recognized as revenue for books when sold but for tax when paid. One reason that the installment sales deferred tax liability might increase is that the firm is aggressively accelerating revenue recognition for book purposes. If so, we would anticipate increases in the installment sales deferred tax liability being associated with negative returns. This relation would exhibit itself as a negative coefficient in the regression. Therefore, to ensure that accrual quality relevance is only indicated by positive coefficients, we flip the signs for all deferred tax liabilities. By changing the signs, what would be a negative coefficient for an installment sales deferred tax liability becomes a positive coefficient. 23 and local tax issues; and REG-t, which concerns special tax issues for regulated industries. Each of these five temporary “TAX BTD-t” variables is coded such that a positive coefficient will be interpreted as evidence that investors find these BTDs to be incrementally informative about the firm’s future tax outlays. 3.3.8. Tax Avoidance—Permanent Differences We also anticipate that investors use seven categories of reconciling items to better understand the extent to which book income is taxed. For five of the seven (DIVS, DMD, TEI, M&A-r and REG-r), the prediction is straightforward. We anticipate that a change in these “TAX BTD-r” variables that increases (decreases) the effective tax rate will be associated with negative (positive) stock returns. For example, the rate reconciliation indicates the extent to which tax-exempt municipal bond interest income (TEI) reduces the ETR. Thus, we predict that stock returns are increasing in tax-exempt municipal bond interest income because learning that some of the firm’s book income is not subject to tax should be good news to the market. Similarly, positive returns should be associated with the exclusion for dividends from subsidiaries (DIVS), lower taxes on domestic manufacturing (DMD), minimizing non-deductible costs associated with tax-free mergers or with divestitures and restructuring (M&A-r), and any special taxes on regulated firms (REG-r). Predictions are less straightforward for UNUSUAL-r and COLI. For UNUSUAL-r, we take the same approach that we did with UNUSUAL-t above. That is, this category includes a hodgepodge of unusual reconciling items that are material enough to warrant separate line item disclosure. Therefore, it is probably important enough that investors examine it closely. Thus, 24 in keeping with the analysis above, we also predict a negative coefficient on the UNUSUAL-r coefficient. For COLI, we also anticipate that investors will view lower ETRs as good news. However, as mentioned above, some investors may view certain forms of tax avoidance negatively (Hanlon and Slemrod, 2009). For example, involvement in a tax fraud or a corporate tax shelter, even if it reduces ETRs, might cause reputational damage. The IRS considers some forms of corporate-owned life insurance, specifically those that are leveraged and cover many non-executives, to be tax shelters. It has won numerous court cases involving these policies. Although broad-based, leveraged corporate-owned, life insurance is less common now and deductibility of such policies is subject to considerable legislative restriction, some of the corporate-owned life insurance in the rate reconciliations that we study over the investigation period may be of the tax shelter variety. If so, and if investors take a dim view of these policies, then we may find that increased ETR-reducing corporate-owned life insurance is associated with negative returns. Recognizing that the sign on the COLI coefficient could be positive or negative, we make no prediction about its sign. The effect of a rate reconciliation disclosure, including the effect of a permanent taxrelated BTD, is calculated as the impact (in % change in ETR) of a reconciling item times pretax income. These “levels” of permanent BTDs are shown in Tables 3 and 4. To compute the TAX BTD-r regression variables, we subtract the impact (in dollars) of a reconciling item last year from the impact this year and divide the change in the permanent difference by market value of equity at the beginning of the year. 18 18 For example, Hewlett Packard disclosed that the effect of state taxes in 2007 was 0.5%. In 2006, state taxes decreased the ETR by 0.1%. Pretax income in 2007 and 2006 was $9,177 and $7,191, respectively. The 2007 rate reconciliation disclosure for state taxes equals (0.005 x $9177) = $45.89; the 2006 disclosure equals (-0.001*$7191) = -$7.19. The unscaled STATERATE variable for 2007 is the change in the dollar effect of state taxes: ($45.89 less - 25 3.3.9. No information The final category includes BTDs from which we do not expect investors can extract any information. These “NO INFO” BTDs include both BTDs categories that have the items without sufficient description to understand them (OTHER-t and OTHER-r). Despite their high frequency (99% and 92% of the firms report some BTD in the OTHER-t and OTHER-r categories, respectively), we predict that the coefficient on these two BTD variables will not be statistically significant. We also predict non-significance on the two SUB coefficients, SUB-t and SUB-r. While some of the information in these categories is redundant to information disclosed elsewhere in the footnotes about equity method accounting, many items are BTDs from flow-through investee. Thus, while these flow-through BTDs might be informative if details about them were provided, the descriptions are typically inadequate to provide any information, much like that of OTHER-t and OTHER-r. Thus, we do not expect that investors use these disclosures for price formation purposes. 4. Empirical Results 4.1. Descriptive Statistics This section presents our findings from estimating equation (1). We begin with descriptive statistics for the regression variables in Table 5. The first column indicates the number of non-zero observations for that regression variable. The second (third) column indicates the mean (median) value in millions of dollars for the non-zero observations. The last column presents the actual variable used in the regression equation, expressed in percentages $7.19) = $53.08; this implicitly assumes that we expected the state effect to be a decrease in tax of $7.19 and instead Hewlett Packard reported an increase in tax of $45.89 so our “surprise” is a $53.08 increase in tax. The actual STATERATE regression variable for 2007 is equal to 0.05% of beginning MVE. 26 (i.e., the average in the last column includes observations with a value of zero). All regression variables are scaled by the beginning value of the firm’s market value of equity. For example, 2,171 of the 3,890 firm-year observations have a non-zero value for ΔBENEFITS. The mean (median) for those 2,171 non-zero observations is $6.6 ($0.5) million. The mean value for ΔBENEFITS in the regression equation is 0.005%. The dependent variable is R, 16-month market-adjusted returns ending four-months after the fiscal year-end. R averages 9.0% during the investigation period. The mean annual change in pretax book income (ΔPTBI) is 1.7% and in total cash flows (ΔCF) is 0.5% of beginning MVE. Not surprisingly, the annual changes in BTDs are not a large percentage of MVE. The sum of all of the temporary differences (ΔTemporary) are generally negative (mean ΔTemporary = -0.402%), consistent with deferred tax expenses exceeding deferred tax income, on average, as expected. The sum of the accrual quality temporary differences (ΔAQ BTD) also are negative, on average (mean ΔAQ BTD = -0.392%); however, the sum of the temporary differences related most closely to tax liabilities are usually positive (mean ΔTAX BTD-t = 0.032%). Total permanent differences are generally positive, albeit small (mean ΔPermanent = 0.011%), indicative of permanent differences usually increasing the effective tax rate. Those permanent differences that potentially communicate information about the firm’s tax liabilities also are positive (mean ΔTAX BTD-r is 0.027%). Looking at specific BTD categories, the largest means in absolute value are all temporary differences: ΔREG-t (0.477%), ΔEXP (-0.420%), ΔINVENT (0.268%), and ΔINTANG (-0.262%). The largest permanent difference in absolute value is M&Ar at 0.122%. 27 4.2. Primary Findings Table 6 presents summary statistics from estimating equation (1). Column A shows that positive returns are associated with increases in pretax book income and increases in total cash flows, as expected. Column B adds the change in total temporary differences and total permanent differences as explanatory variables. Column C splits the temporary and permanent differences into accrual quality temporary BTDs (AQ BTD) to test the first hypothesis of the paper, TAX temporary BTDs (TAX BTD-t) and TAX permanent BTDs (TAX BTD-r) to test the second hypothesis, and non-informative BTDs (NO INFO BTD) to test the third hypothesis. Finally, Column D presents coefficient estimates for each of the 25 BTD categories that may contain information. As expected, the sign on ΔTemporary in Column B is positive and significant at the 0.05 level using a two-tailed test. This is consistent with the market favorably viewing increases in deferred tax assets and decrease in deferred tax liabilities. As mentioned above, the temporary differences include some that we expect are informative about the quality of the firm’s financial reports and others that potentially shed light on the firm’s tax obligations. Turning first to the accrual quality BTDs, we find that the coefficient on ΔAQ BTD in Column C is positive and significant at the 0.01 level, as predicted. This is consistent with Hypothesis 1, i.e., investors assessing the accrual quality of the firm by examining the BTDs arising from accounts where managers enjoy discretion in the recognition of income and expenses. Column D shows the results for each of the nine AQ BTDs. We find that three (ΔEXP, ΔINTANG, and ΔREV) are significantly different from zero at the 0.01 level. Two (ΔEXP and ΔREV) are positive, as predicted, indicating that stock returns are increasing as these deferred tax assets (liabilities) increase (decrease). We interpret these findings as evidence that investors 28 infer information about the accrual quality of firms through changes in the BTDs arising from general business expenses and revenue recognition. At least some items in ΔEXP (accrued or deferred expenses, including frequent flyer plans, direct marketing, policy acquisition, and contingent rent) and ΔREV (such as deferred revenue, customer deposits, contract accounting, and deferred gain on sales of assets) likely are followed by investors as indicators of accrual quality. These results are the strongest evidence in this study that the BTDs convey information to investors that is incremental to accounting earnings and cash flows. In contrast, the coefficient on ΔINTANG has the wrong sign, contrary to our predictions. Four other AQ BTDs also have a negative coefficient, though none of them is significantly different from zero. We have no explanation for these results. The fact that five of the nine accounts that we thought were potential sources of incremental information about accrual quality have the wrong sign causes us considerable pause. Overall, we interpret the results in Columns C and D as providing only weak support for Hypothesis 1. The BTD categories where we find the predicted results, EXP and REV, do appear to contain accounts, such as miscellaneous reserves, that are easily managed and were the conjecture of prior studies, e.g., Hanlon, (2005), that did not have disaggregated BTD data. However, few other temporary accounts, if any, appear to be priced by the market, and we cannot explain the one that is (ΔINTANG). Turning next to our tests of Hypothesis 2, we first consider those temporary differences that speak to the firm’s tax liabilities. We find that the coefficient on ΔTAX BTD-t (which aggregates the five temporary differences that concern tax liabilities) is not significantly different from zero. However, the coefficients on ΔM&A-t and ΔSTATE are positive, as predicted. These results imply that stock returns are negatively associated with changes in deferred tax positions 29 that suggest higher than expected taxes from merger and acquisitions or state income taxes, but not with the other tax-related temporary differences. Thus, we again conclude that, at best, investors are only using select items from the deferred tax positions to infer information about the firm’s tax liabilities. We do not have an explanation for why they would find ΔM&A-t and ΔSTATE more informative than the others. We now turn to the permanent differences related to tax payments, which are expected to be negative, indicative of investors interpreting as good news those items that communicate an unexpected reduction in tax payments. Contrary to expectations, the coefficients on total permanent differences (ΔPermanent) in Column B and ΔTAX BTD-r in Column C are positive and significant at conventional levels. We have no explanation for these results because they imply that investors view higher effective tax rates favorably. However, when we examine the seven permanent differences individually in Column D, we find that four have negative coefficients and none is significantly different from zero. One possibility for our failure to detect anything for some of the individual permanent differences is a lack of power arising from few non-zero values (e.g., there are only 91 non-zero observations for ΔCOLI and only 92 for ΔDMD). Taken together, the coefficients on the changes in the five TAX temporary BTDs and the seven ETR permanent BTDs provide little evidence that investors are using the BTDs to assess the current and future taxes facing the firm, inconsistent with our second hypothesis. With only two of the 12 individual tax BTDs significant in the predicted direction, we conclude Hypothesis 2 is not supported and that there is little evidence that the BTDs arising from accounts that center on the firm’s taxes are important sources of incremental information to investors. 30 Finally, we turn to Hypothesis 3 regarding the NO INFO BTDs. Recall that four BTD categories include so little description that we do not anticipate any association between changes in these BTDs and stock returns. Consistent with that expectation, we find that the coefficient on total ΔNO INFO BTD in Column C is not significantly different from zero. However, one of the four individual coefficients (ΔSUB-t) is positive and significant in Column D. We have no explanation for this result. We cannot ascertain what the market is pricing with this account. Finding evidence that the market responds to a largely nondescript account is particularly troubling, given the difficulty in finding significance in the other accounts. Until resolved, this result raises concerns about whether some of the significant results above are spurious. All in all, we find spotty support, at best, for the notion that BTDs are important sources of information that is incremental to pretax book income and total cash flows. Based on the detailed disclosures from the statements of deferred tax positions and rate reconciliations, the most compelling case that we can make is that a few accounts appear to have incremental information content, namely those involving revenue recognition and expensing (where managers can shift book income across periods) and two tax-related temporary differences. However, taking the empirical results as a whole, we can only make a weak case in support of the market’s pricing BTDs. The new disaggregated BTD data used in this study do not yield clear, robust findings, suggesting that the association between returns and BTDs is more nuanced than previously thought. Finding significance in the wrong direction is particularly troubling. Prior studies using aggregate BTDs were unable to separate the BTDs that should result in a positive association from those that should result in a negative association. As a result, findings of either sign could 31 be claimed to support rational market pricing, even though the research design was incapable of ruling out the possibility that the direction was wrong. Naturally the same concerns that face all studies with anything less than overwhelming statistical significance apply here. We may have a poorly specified model, inadequate power, and measurement error in the variables that creates the false impression that few relations are significant in the predicted direction. That said, the model expands one already established in the literature; the power is sufficient for some variables to load; and the higher quality of the data should result in less measurement error than prior studies in this field have faced. 4.3. Supplemental Analyses with Non-BTDs and Redundant Accounts As discussed above, we limit our analysis to those differences between book and taxable income that potentially provide incremental information. In this section we include non-BTDs (e.g., state and foreign tax rate differentials) and redundant BTDs (accounts where the tax footnote restates information already in the financials) as explanatory variables and re-estimate equation (1). The reason for this supplemental test is that the measure of BTDs used in most studies (pretax book income less grossed-up current tax expense, sometimes adjusted for changes in the net operating loss carryforwards) includes both non-BTDs and redundant BTDs. It is possible that prior studies have reported that BTDs are incrementally informative, when, in fact, the accounts that the market prices are either not BTDs or redundant ones. In the latter case, significance would occur because the model omits the information found elsewhere in the financial reports. Table 7 presents summary statistics from estimating (1) after including the nine nonBTDs and the seven redundant BTDs and reproduces Table 6, Columns B and C for comparison. 32 First, we recompute ΔTemporary and ΔPermanent, to include the appropriate redundant BTDs, and then add a new explanatory variable for the non-BTDs (NON-BTD). Column B′ shows that the coefficient on the new ΔTemporary is positive but not significant. We find that the coefficient on the new ΔPermanent is positive and significant, similar to the result obtained before adding the non-BTD items. We also find that the coefficient on ΔNON-BTD is negative and nearly significant at the 0.06 level. Column C′ shows the results when we break apart ΔTemporary and ΔPermanent into ΔAQ BTD, ΔTAX BTD-t, ΔTAX BTD-r, and ΔNO INFO BTD, as before, and put the redundant BTDs into a separate variable. We find that the coefficient on ΔREDUNDANT BTD is not significantly different from zero. The coefficient on ΔAQ BTD has become insignificant, though barely at the 0.06 level. The coefficient on ΔTAX BTD-r remains positive and highly significant. The coefficient on ΔNON-BTD is negative, but now significant at the 0.02 level. This is consistent with the market favorably viewing lower taxes. The coefficient on ΔNO INFO BTD remains insignificant. In untabulated analysis, we replace the summary measures with the 41 individual BTD categories. All of the significant coefficients in Table 6, Panel D remain significant with ΔBENEFITS joining them with a positive coefficient. None of the other insignificant coefficients in Table 6 becomes significant when the non-BTD and redundant BTDs are added to the regression. Not surprisingly, the only significant non-BTDs are the rate differentials, foreign (ΔFORr) and state (ΔSTATERATE). What is troubling and puzzling is that the coefficient on ΔSTATERATE is positive, consistent with the market viewing a boost to the effective tax rate arising from higher state taxes as good news. A possible explanation (although we find it 33 unlikely) is that high state taxes are more informative about accrual quality (if a firm pays high state taxes, then its book earnings are more likely to be of high quality) than about actual state tax obligations. The coefficient on ΔFOR-r is negative, as expected. ΔADA is the only redundant coefficient that is significantly different from zero (p-value of 0.01). Its positive coefficient indicates that the market views increases to the allowance for doubtful accounts as good news. This is consistent with the ADA BTD being an account that investors view as indicative of accrual quality. However, since all of the information in the ADA BTD is reported elsewhere in the financial reports, we doubt that this result would hold if we included that other information about the allowance for doubtful accounts in the regression model and then tested for an incremental loading on the ADA BTD variable. 19 The evidence from this supplemental analysis is consistent with the market pricing nonBTDs but not redundant BTDs. Furthermore, the presence of non-BTDs appears to shrink the magnitude of the coefficients on temporary BTDs and particularly those that we have identified as potentially informative about accrual quality. In other words, the coefficients on aggregated temporary BTDs and AQ BTDs appear significantly greater than zero when non-BTDs are excluded from the analysis. When non-BTDs are included, they and the permanent differences appear less incrementally informative. Our findings may shed light on a longstanding puzzle between the findings in two prominent papers in this field, Hanlon (2005) and Lev and Nissim (2004). Studying temporary differences alone, Hanlon (2005) reports that temporary differences are useful for assessing accrual quality (as measured by persistence). Evaluating temporary, permanent and non-BTDs together, Lev and Nissim (2004) report that temporary differences are not significant for 19 We could undertake such a test, but it is not clear that proving this point justifies the costly hand-collection of detailed disclosures about the allowance for doubtful accounts that would be required. 34 purposes of assessing earnings quality (as measured by earnings growth), but the residual permanent and non-BTD differences are significant. The findings in this study potentially reconcile those seemingly inconsistent results because the temporary differences in our study appear incrementally informative until non-BTDs are introduced, then they lost their significance, although marginally. At a minimum, this potential reconciliation of Hanlon (2005) and Lev and Nissim (2004) warrants further investigation. 4.4. Extensions There is no reason to expect that investors scrutinize the BTDs of all firms the same, as implicitly assumed in the tests above. Thus, we undertake three (untabulated) extensions in an attempt to identify firms whose BTDs might be analyzed more closely for accrual quality or taxes. In our first extension, we identify a set of firms whose accrual quality would appear in some question. Following Blaylock, et al. (2010), we identify suspected earnings managers as firms whose discretionary accruals (in absolute value) place them in the top quartile of their industry-year, where industry-years are computed using all firms in the Compustat database. 20 We then interact an indicator for these observations (which comprise 7% of our firm-years) with ΔPTBI, ΔCF and ΔAQ BTD and include the three interactions and the indicator as explanatory variables. Finding a positive coefficient on ΔAQ BTD will be interpreted as evidence that investors assess accrual quality more through the AQ BTDs for suspected earnings managers than they assess it through the AQ BTDs with other firms. We find that the interaction with ΔAQ BTD is positive (0.88) and significant at the 0.001 level, while the coefficient on the standalone 20 Another possible indicator of accrual quality is whether firms have restated their earnings. Unfortunately, those data are not available in computer-readable form before 2000. Thus, we would forgo a substantial portion of our sample if we used that indicator. 35 ΔAQ BTD, representing the firms not suspected (as much) of manipulation, is no longer significant. These results are consistent with the market paying close attention to the AQ BTDs of companies whose accrual quality is questionable and ignoring the AQ BTDs for other companies. In our second extension, we sort the firm-years between the 2,544 with good news (earnings increased that year) and the 1,342 with bad news (earnings decreased that year) and repeat the tests. For the good news observations, we find that the coefficient on ΔAQ BTD is positive (0.419) and highly significant. For the bad news observations, we find that the coefficient on ΔAQ BTD turns negative (-0.263) and significant at the 0.05 level. This is contrary to our expectations. However, it is consistent with investors reacting unfavorably toward a big bath. That is, investors may interpret a decline in accounting earnings from the previous year, combined with an increase (decrease) in deferred tax assets (liabilities), which indicates that taxable income did not decline as much as book income, as evidence of a big bath and drive prices downward. 21 In our third extension, we anticipate that investors will place more weight on the changes in the tax-related BTDs of unusually gifted tax avoiders. Since such firms have demonstrated an ability to sustain low effective tax rates, their value depends critically on maintaining low taxes. As a source of incremental information, changes in the tax-related BTDs might be a red flag to investors that taxes will be rising in the future. To test this proposition, we identify firms as particularly successful at tax avoidance if its five-year cash taxes paid as a percentage of pretax book income (cash ETR) is in the lowest quintile of its industry-year, where industry-years are computed using all firms in the Compustat database. As with the first extension, we include an 21 We also sort firm-years between the 3,476 with net profits and the 414 with net losses. We find that the coefficient on ΔAQ BTD is not significant for either subset. 36 indicator variable for the 8% of the sample that are particularly adept tax avoiders and interact it with the other explanatory variables, excepting ΔAQ BTD and ΔNO INFO BTD since they should be unaffected by tax considerations. We find that the coefficients on the interactions involving ΔTAX BTD-t and ΔTAX BTD-r are not significantly different from zero. Therefore, this extension provides no evidence that investors in companies with a history of low cash ETRs scrutinize the tax-related BTDs any differently than they do those in other companies. This is not particularly surprising, given the primary test failed to detect much evidence that investors use the BTDs to assess a firm’s tax obligations. 4.5. Econometric Considerations We finish with a few econometric points. First, as mentioned above, unlike Hanlon, et al. (2005), we include total cash flows as a control variable. Although its coefficient is always highly significant, as expected, excluding the variable from the model has no qualitative impact on the coefficients of primary interest. The same is true for year and industry indicator variables. Second, Hanlon, et al. (2005) conduct regressions for each year and then analyze the coefficients from those regressions. Unlike Hanlon, et al. (2005) who enjoy a larger sample and fewer explanatory variables, we concluded that we did not have enough observations to conduct similar yearly tests for our primary analysis, given the large number of explanatory variables in our study. Thus, the significance of our coefficients may be overstated because of crosssectional correlation in the pooled cross-sectional regressions. However, as a sensitivity check, we conduct annual regressions. As expected, we find no significant explanatory variables, except pretax book income and total cash flows. However, when we examine the annual coefficients, we find an interesting pattern. The coefficients on 37 ΔTemporary and ΔAQ BTD move inversely with those for pretax book income and total cash flows (correlations of the annual coefficients range from -61% to -72%). In other words, it appears that the ΔTemporary and ΔAQ BTD variables have their greatest explanatory power when the associations between returns and pretax book income and returns and total cash flows are weakest. The economics behind these relations are unclear. One possibility is that investors focus most intently on these AQ BTDs for information about accrual quality when their confidence in pretax income and cash flows is diminished. More study is warranted. Finally, we employ no clustering or firm fixed effects in our model. That said, we are not overly concerned about understated standard errors because we are taking first differences throughout the analysis, which should eliminate some of the econometric problems that would otherwise impede our analysis. 5. Conclusion To our knowledge, this is the first study that uses details about book-tax differences from the statements of deferred tax positions and rate reconciliations to test for the incremental information content of specific BTDs. Prior BTD studies have suffered from measurement error because the lack of computer-readable data have forced them to estimate BTDs in a manner that have prevented clean, precise tests of the impact of BTDs on financial reporting and tax avoidance. This study uses thousands of hand-collected items from the tax footnote to determine which BTDs matter to investors. We find a little evidence that investors use at least a few BTDs to assess the accrual quality of companies. In particular, investors appear to focus on items concerning general business expense and revenue recognition to assess firms’ accrual quality. Likewise, we find a 38 little evidence that investors use the BTDs to assess the current or future tax liabilities of firms, primarily around mergers and acquisitions and state taxes. We find a number of puzzling results. For example, the regression results would suggest that investors view higher effective tax rates arising from larger state tax outlays favorably. We do not believe that this is the case; however, at this point, we cannot explain these results. More study is needed. This paper provides the first examination of new data. They are invaluable for disaggregating book-tax differences and attempting to understand the specific accounts that comprise a firm’s book-tax differences. However, at this point, the results suggest that we may not understand book-tax differences as well as previously thought. Until we can better understand the specific accounts that are driving previous BTD inferences, we are hesitant to place too much reliance on prior claims that the market prices BTDs. In fact, the primary take away from this study is that earlier claims about strong associations between BTDs and returns may not hold when examined using disaggregated information about specific BTDs. 39 References Altshuler, R., A. J. Auerbach, M. Cooper, and M. Knittel, 2008. “Understanding U.S. Corporate Tax Losses,” NBER working paper 14405. Badertscher, B. A., J. D. Phillips, M. Pincus, and S. O. Rego, 2009. “Earnings Management Strategies and the Trade-Off between Tax Benefits and Detection Risk: To Conform or Not to Conform?” The Accounting Review 84(1): 63-97. Barth, M. E., W. H. Beaver, and W. R. Landsman, 1998. “Relative Valuation Roles of Equity Book Value and Net Income as a Function of Financial Health,” Journal of Accounting and Economics 25(1): 1-34. Blaylock, B., T. J. Shevlin, and R. J. Wilson, 2010. “Tax Avoidance, Large Positive Book-Tax Differences and Earnings Persistence,” University of Washington working paper. Chen, S., X. Chen, Q. Cheng and T. Shevlin, 2010. “Are Family Firms More Tax Aggressive than Non-Family Firms?” Journal of Financial Economics 95(1): 41-61. Dechow, P. M., W. Ge, and C. M. Schrand, 2009. “Understanding Earnings Quality: A Review of the Proxies, Their Determinants and Their Consequences”, University of Pennsylvania working paper. Desai, M. A., 2003. “The Divergence between Book and Tax Income,” in Tax Policy and the Economy 17: 169-206 (ed. J. Poterba). Desai, M. A. and D. Dharmapala, 2006. “Corporate Tax Avoidance and High-Powered Incentives,” Journal of Financial Economics 79(1): 145-179. Desai, M. A. and D. Dharmapala, 2009. “Corporate Tax Avoidance and Firm Value” in The Review of Economics and Statistics 91(3): 537-546. Dyreng, S., M. Hanlon, and E. Maydew, 2008. “Long run corporate tax avoidance,” The Accounting Review 83(1): 61-82. Frank, M. M., L. J. Lynch, and S. O. Rego, 2009. “Tax Reporting Aggressiveness and Its Relation to Aggressive Financial Reporting,” The Accounting Review 84(2): 467-496. Graham, J. R., J. S. Raedy, and D. A. Shackelford, 2010. “Research in Accounting for Income Taxes,” University of North Carolina working paper. 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. 40 Hanlon, M., S. K. LaPlante, and T. Shevlin, 2005. “Evidence for the Possible Information Loss of Conforming Book Income and Taxable Income,” Journal of Law and Economics 48(2): 407-442. Hanlon, M. and S. Heitzman, 2010. “Tax Research: Real Effects, Reporting Effects, and Governance,” MIT Sloan School of Management working paper. Hanlon, M. and J. Slemrod, 2009. “What does Tax Aggressiveness Signal? Evidence from Stock Price Reactions to New Announcements about Tax Shelter Involvement,” Journal of Public Economics 93(1-2): 126-141. Keating, A. S., and J. L. Zimmerman, 2000. “Depreciation-policy Changes: Tax, Earnings Management, and Investment Opportunity Incentives,” Journal of Accounting and Economics 28: 359-389. Lev, B. and D. Nissim, 2004. “Taxable Income, Future Earnings, and Equity Values,” The Accounting Review 79(4): 1039-1074. Lisowsky, P., 2009. “’Seeking Shelter’: Empirically Modeling Tax Shelters Using Financial Statement Information,” University of Illinois College of Business working paper. Mills, L., 1998. “Book-tax differences and Internal Revenue Service Adjustments,” Journal of Accounting Research 36(2): 343-356. Mills, L. and K. Newberry, 2001. “The Influence of Tax and Nontax Costs on Book-Tax Reporting Differences: Public and Private Firms,” Journal of the American Taxation Association 23(1): 1-19. Phillips, J., M. Pincus, and S. O. Rego, (2003), “Earnings Management: New Evidence Based on Deferred Tax Expense,” The Accounting Review 78(2): 491-521. Poterba, J., N. Rao and J. Seidman, 2010. “Deferred Tax Positions and Incentives for Corporate Behavior around Corporate Tax Changes,” MIT working paper. Seidman, J. K., 2010. “Interpreting the Book-Tax Gap as Earnings Management or Tax Sheltering,” University of Texas working paper. Wilson, R., 2009. “An Examination of Corporate Tax Shelter Participants,” The Accounting Review 84(3): 969-999. 41 Appendix A Description of Variables Temporary Items that may contain information Variable Descriptive Name Accruals Quality? BENEFITS Employee Benefits Y EXP Expenses Y EXTRACT Extractive Industries Y FOR-t Foreign Differences INTANG Related to Owned Intangible Property Y INVENT Inventory Y LEASE Y M&A-t Related to Leased Property Merger, Acquisition, Divestitures, and Restructuring NOL NOL and other Carryforwards OTHER-t Vague Disclosures PPE REG-t Related to Owned Tangible Property Regulated Entities Y REV STATE Revenue State Differences Y SUB-t Subsidiary-related Infrequent disclosures or disclosures that cross categories UNUSUAL-t Related to Tax? Y Y Y Y Y Y Explanation of line item disclosures Accrued or deferred compensation, including vacation pay, stock compensation, health care costs, etc. Accrued or deferred expenses, including frequent flyer plans, direct marketing, policy acquisition, contingent rent, etc. Accrued environmental obligation, Cost of removal, Decommisioning, etc. Foreign affiliates, Foreign investments, Translation adjustments, DISC, Anticipated repatriation, etc. Amortization, Goodwill impairments, Acquired intangibles including IPRD, Mortgage Servicing, etc. Capitalized inventory, Inventory valuation methods, Intercompany profit in inventory, Excess or obsolete inventory, etc. Lease transactions, Lease reserves, Gains on sale and leasebacks, Capital leases, Operating leases, etc. Deferred gain on sale of business, Merger related costs, Deferred acquisition costs, Discontinued operations, etc. Net Operating Loss Carryforwards, including foreign and state, charitable contribution carryforwards, etc. Other, Miscellaneous, Accrued charges, Items not currently deductible, etc. Accelerated depreciation, Asset basis difference, Capitalized interest, Property valuation, Asset Impairments, etc. Deferred fuel costs, Regulatory assets, Rate deferrals, etc. Deferred revenue, Customer discounts, Contract accounting, Deferred gain on sale of assets, etc. Deferred state and local income and franchise taxes Basis difference in unconsolidated subs, Equity in earnings of affiliates, Investment in JV, Partnership related, etc. Combined disclosures such as "Accounts receivable and inventory", infrequent disclosures Redundant citation Appendix A Description of Variables Permanent Items that may contain information Accruals Quality? Variable Descriptive Name Related to Tax? Redundant citation Y Y Explanation of line item disclosures Corporate owned life insurance, Changes in cash surrender value, etc. Dividends received deduction, Dividends on ESOP stock, etc. Domestic production deduction, Manufacturing exemption Nondeductible acquisition costs, Merger related costs, Divestitures, Exit costs, Gain on disposal of business, Restructuring charges, etc. Other items, Permanent differences, etc. Regulatory disallowances, Allowance for funds used during construction, Full normalization, Mirror construction work in progress, etc. Equity earnings, Gain on sale of subsidiary stock, Impairment of investments in equity method subs, etc. Tax exempt income, Nontaxable income, Tax preferred investments, etc. Infrequent disclosures COLI DIVS DMD Corporate Owned Life Insurance Non-taxable Dividends Domestic Manufacturing Deduction Y Y Y M&A-r OTHER-r Merger, Acquisition, Divestitures, and Restructuring Vague Disclosures Y REG-r Regulated Entities Y SUB-r Subsidiary-related TEI UNUSUAL-r Tax Exempt Income Infrequent disclosures Permanent? Explanation of line item disclosures Redundant citation ASC 310-10-50-6 through 10. SEC Reg S-X, Rule 12-09 Redundant Items Variable Descriptive Name ADA Allowance for Doubtful Accounts AMORT-GW Amortization of Goodwill Temporary? Y Y Accounts receivable, Allowance for doubtful accounts, etc. Amortization of goodwill, Impairment of nondeductible goodwill, Tax effect of purchase accounting, etc. ASC 350-20-50-1 and 2 Acquired in process R&D, Amortization of Intangibles, Nondeductible amortization, etc. ASC 350-30-50-1 through 3 required on income statement Effect of minority holdings ASC 320-10-50-2, ASC 820-1050, ASC 825-10-50-10 Basis differential of investments, Derivatives, FAS 115 through 19 adjustment, Investment valuation, etc. Y Y Pensions and other post-employment or post-retirement benefits ASC 715-30-50-1 Warranty reserves and accruals ASC 460-10-50-8 Y AMORT-OTH Other Amortization Y MININT Minority Interest Y MTM Mark-to-Market Accounting Pensions and Other Post-Employement Benefits Warranty OPEB WARRANTY 43 Appendix A Description of Variables Non-Book/Tax Differences Included in Statement of Deferred Tax Included in Rate Reconciliation? Positions? Variable Descriptive Name ADJTR Adjustments to deferred tax balances AUDIT CREDITS-t Audit Adjustments Tax Credits CREDITS-r Tax Credits Y ETI Extraterritorial Income Y FOR-r Foreign Differences Y REPAT STATERATE VA Effect of Repatriations State Rate Differential Valuation Allowance Y Y Y Y Y Y 44 Explanation of line item disclosures Adjustment of prior year provision, Effects of enacted rate and law changes, Adjustment to valuation allowance, etc. Adjustment to prior year tax liability, Audit settlements, Change in reserve for tax contingencies, FIN 48, etc. Credit carryforwards, including foreign and state Credits including disclosures related to Investment Tax Credits (ITC) and amortization of ITC, excludes foreign tax credits which are included in FOR-r Benefit from export incentives, Extraterritorial income exclusion, Foreign sales corporation, etc. Foreign rate differential, Foreign exchange, Foreign tax credits, Withholding taxes, etc. Dividend repatriation, Earnings repatriation, APB 23, AJCA repatriation tax, Foreign dividends, etc. Effect of state and local taxes Valuation allowances Redundant citation Table 1 Sample Means ($M of tax) for Temporary Differences Collected from the Schedule of Deferred Tax Positions Positive values denote increases (decreases) in deferred tax assets (deferred tax liabilities). Negative values denote decreases (increases) in deferred tax assets (deferred tax liabilities). Variable ADA BENEFITS CREDITS-t ENVIRON EXP FOR-t INTANG INVENT LEASE M&A-t MTM NOL OPEB OTHER-t PPE REG-t REV STATE SUB-t UNUSUAL-t VA WARRANTY TOTAL (n=3890) Mean 3.0 18.8 12.4 (1.4) (2.3) (1.1) (20.4) 0.2 (2.3) (3.9) (2.6) 20.0 (0.2) 15.1 (47.2) (0.7) 1.4 (0.0) (9.1) 1.7 (31.7) 0.2 (50.1) (n=317) (n=349) (n=342) (n=351) (n=317) (n=297) (n=287) (n=275) (n=278) (n=275) (n=268) (n=266) (n=261) 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 % ne 0 1995 1996 26 1.8 1.3 1.7 7.0 (1.9) (3.2) 15.3 (1.7) (4.8) 2.3 (7.9) (6.2) 37.6 55 (1.1) 13.3 19.8 8.7 (4.0) 5.6 20.6 58.1 (57.6) 15.2 30.2 208.2 (56.4) 53 (5.3) (8.3) 1.3 12.4 8.6 2.5 19.3 51.4 33.6 32.0 14.7 10.3 0.6 9 2.2 (0.9) (1.3) (0.9) (1.0) (13.4) 5.8 (2.3) (1.1) (0.0) 0.4 (27.6) 22.1 15 2.6 2.7 3.3 (1.3) (8.5) 4.5 (7.0) (6.2) (12.3) (7.0) (7.8) 5.7 (2.1) 20 (0.5) (0.8) 2.7 3.8 4.1 3.7 0.5 (0.4) (24.3) (13.7) 19.3 (8.9) (3.1) 36 0.5 (6.5) (3.6) (3.2) (45.5) (25.6) (7.8) 55.8 (60.7) (20.4) (58.6) (87.9) (16.5) 34 0.7 (0.8) 0.8 1.5 (0.6) 1.5 0.5 (2.4) 0.6 (4.7) 4.2 (5.2) 6.4 16 (3.2) (0.5) (9.4) (7.6) (5.3) (16.2) (2.2) (10.1) 4.8 1.2 (4.3) 14.4 14.3 21 (2.9) (8.5) (7.1) 2.1 (11.0) (10.5) 18.1 (0.2) (8.2) 6.0 0.3 (24.1) (3.8) 29 (37.9) 2.8 (20.4) (14.9) (17.2) 23.1 36.2 14.4 (41.1) (14.6) 47.7 (7.2) 7.4 43 (3.2) 5.1 (0.2) 11.6 17.8 24.9 37.1 27.8 25.8 50.0 24.1 27.0 26.7 40 0.6 (17.9) (9.9) (3.1) (20.5) (13.8) (9.4) 65.9 (19.3) (39.4) 13.0 129.5 (64.6) 99 8.9 (4.4) 8.6 20.3 (3.8) 7.8 26.5 12.2 14.2 2.9 25.4 10.1 80.1 90 (16.6) (34.0) (28.2) (25.6) (55.9) (88.1) (57.9) (42.2) (62.3) (78.4) (21.7) (102.4) (14.9) 10 1.6 7.0 (0.6) 1.0 (2.6) (2.0) (4.2) (7.1) (5.8) (7.3) (8.3) 4.2 13.0 31 (1.3) (2.4) 2.9 (2.9) 1.4 8.2 3.2 9.1 8.2 2.0 (13.8) 1.1 4.1 11 (1.6) (1.1) (1.5) 0.2 1.6 (0.5) 0.5 2.0 (1.9) 0.1 (0.2) (0.5) 3.5 14 (2.3) 0.5 (2.3) (4.3) (25.2) (36.3) 6.4 27.0 (19.3) 0.6 (2.1) (62.0) (3.6) 41 (0.6) (3.0) 3.1 9.2 6.7 (6.2) (23.3) 2.3 5.3 22.3 41.8 (26.9) (8.4) 57 4.4 6.0 (1.7) (1.5) (15.5) (8.6) (25.5) (89.3) (39.8) (48.3) (41.0) (47.5) (147.4) 3 0.3 (0.4) (0.1) 0.5 0.3 0.0 14.8 7.0 1.8 3.7 (27.2) 0.7 0.3 (53.1) (50.8) (42.0) 13.0 (178.2) (142.8) 67.5 171.1 (264.2) (95.5) 28.1 4.9 (104.6) All temporary differences are calculated as the one-year change in the temporary difference category of deferred tax positions; figures remain in dollars of tax. ADA: bad debt expense (includes allowance for finance receivables if filer is a bank). BENEFITS: bonus accrual expense; accrued compensation expense; accrued vacation expense; accrued healthcare expense; employee stock plan expenses; workers compensation expense; other employment benefits; items that contain combinations of employment benefits, other post-employment benefits, and pensions. CREDITS-t: credit carryforwards. ENVIRON: asset retirement/abandonment obligation expenses; accrued environmental expenses; decommissioning reserves/removal costs; depletion expenses; emission allowances; mineral reserves; exploration costs. EXP: advertising/marketing/promotion expenses; member acquisition costs; policy acquisition costs; loan origination costs; travel reward program expenses; other expenses. FORt: currency translation adjustments; taxes on unremitted profits; deductions related to foreign subsidiaries; deferred foreign income taxes; foreign inventory adjustments. INTANG: acquired intangible assets (including goodwill); start-up & organization costs; capitalized research and development; software development costs; mortgage 45 servicing rights; franchise costs; license costs; research and development write-offs. INVENT: capitalized inventory under UNICAP; intercompany profit in inventory, inventory valuation adjustments. LEASE: capital lease adjustments by the lessee; gains from sale-leaseback transactions; lease-related expenditures. M&A-t: acquisition/merger-related costs; gains/losses on the sale of subsidiaries or business units, restructuring and reorganizational costs; discontinued operations. MTM: unrealized gains/losses on available-for-sale securities; unrealized gains/losses on non-available-for-sale securities; items that contain both available-for-sale and not available-for-sale securities; it includes derivative and hedging activities. NOL: net operating loss carryforwards; capital loss carryforwards; charitable contribution carryforwards. OPEB: pension expenses; other post-employment benefit expenses. OTHER-t: line items too vague to categorize (e.g., “other adjustments”); includes items which are partially but not fully categorizable (e.g., “accounts receivable and inventory”). PPE: difference between book and tax depreciation or property; allowance for funds used during construction; asset impairments/write-downs; asset valuation reserves; assets held for sale; items that contain both depreciation and intangible amortization. REG-t: adjustments unique to regulated entities; unrecovered gas and fuel costs; unamortized nuclear costs; statutory contingency reserves; rate refunds and rate case costs; power purchase agreements. REV: accrued dividends and royalties, advance payments; contract accounting differences; intercompany receivables; customer deposits, customer discounts; unearned revenue; other deferred income/revenue; completed contract method of accounting; present value of future profits; deferred lease revenue. STATE: state and local franchise and income taxes; valuation allowances on state deferred taxes. SUB-t: subsidiary, partnership, and joint venture investments; equitymethod subsidiary investments; earnings from partnerships and joint ventures. UNUSUAL-t: all line items not included in another category. They are material enough to warrant separate disclosure for that firm but occur too infrequently in the sample to warrant a separate category. VA: change in the valuation allowance. WARRANTY: warranty reserves and accruals. 46 Table 2 Sample Means ($M of tax) for Temporary Differences Collected from the Schedule of Deferred Tax Positions Positive values denote increases (decreases) in deferred tax assets (deferred tax liabilities). Negative values denotE decreases (increases) in deferred tax assets (deferred tax liabilities) (n=465) variable ADA BENEFITS CREDITS-t ENVIRON EXP FOR-t INTANG INVENT LEASE M&A-t MTM NOL OPEB OTHER-t PPE REG-t REV STATE SUB-t UNUSUAL-t VA WARRANTY TOTAL (n=58) (n=212) (n=174) (n=84) (n=161) (n=198) (n=623) (n=365) (n=309) (n=417) (n=653) Mining Food Textile Chemical Pharma Extraction Durables Computers Transport Utilities Retail Banks & Insurance (n=121) Services 0.0 3.9 0.9 0.0 0.0 (0.0) (0.1) 32.1 (0.1) 0.3 0.3 1.1 0.0 13.5 (0.9) 0.0 (0.2) (0.0) 1.0 8.8 (25.7) 1.8 (0.0) 14.5 0.2 (0.0) (0.3) (4.4) (28.0) (0.7) 0.4 (3.5) (0.3) 10.5 1.4 (0.0) (6.3) 0.0 0.4 0.2 (9.5) 10.9 (9.1) 0.0 0.1 5.2 8.0 (16.2) (1.1) (1.2) (5.7) (0.0) 0.2 (0.4) 0.1 5.9 1.5 1.4 (5.0) 0.0 (1.6) 0.7 (0.1) 2.4 (10.6) 0.0 (0.0) 7.5 50.5 (0.0) 1.9 1.7 (150.4) 2.4 0.6 (6.6) 2.6 2.3 (4.3) 33.2 (10.2) 0.0 1.9 0.0 (2.4) 5.3 (15.0) 0.0 1.5 48.6 39.2 (0.5) 3.7 (16.8) (46.8) 13.4 0.3 (7.7) 1.8 11.6 (5.0) 45.9 (8.9) 0.0 7.7 3.9 (5.3) 18.7 (22.2) 0.0 0.1 21.3 37.2 (14.9) 0.0 (5.6) (10.1) 1.3 (2.2) (1.4) 2.5 29.9 16.1 17.5 (314.2) (5.0) (0.6) 1.1 (19.3) (0.8) (38.4) (0.0) (0.3) 9.1 3.5 0.0 1.1 7.6 (7.8) (0.9) 1.0 (4.2) (0.9) 40.6 8.5 14.7 3.9 0.0 2.4 (0.3) 0.1 (1.8) (104.5) 0.2 0.7 20.2 43.9 0.0 (0.0) (19.9) (4.6) 2.0 (2.0) (7.7) (1.6) 12.1 (10.0) 14.9 3.5 0.0 21.4 0.4 (0.3) 1.3 (45.7) 1.7 (3.6) 50.8 (2.2) 1.4 3.0 1.8 (36.9) 0.0 (8.9) (3.3) 12.0 62.1 (14.7) 19.6 (186.8) (0.2) (3.2) (1.2) (94.2) 3.3 (42.3) 0.0 (0.1) 2.3 3.3 (0.0) (0.0) (0.7) (1.1) (1.3) (6.0) (0.3) 1.6 9.9 3.5 9.8 (70.3) (4.4) 0.2 (2.1) (0.6) 0.2 (6.7) 0.0 0.3 6.0 3.9 (0.0) 1.9 (0.0) (6.1) (5.3) 0.0 (0.1) 0.6 4.4 0.1 5.9 (21.0) 0.0 (2.1) 0.6 (0.1) 2.7 (6.8) 0.0 25.0 46.9 6.3 0.0 (26.7) 6.0 (49.0) (0.0) (0.9) (10.5) (9.1) 19.5 (1.3) 17.4 (21.7) 0.1 (3.7) (0.2) 1.4 7.6 (2.4) 0.0 0.8 6.4 1.9 0.0 (0.0) 0.0 (0.3) 0.0 (0.5) (5.0) (3.1) 6.3 0.0 3.5 (33.6) 0.0 3.1 0.1 0.6 (11.2) (7.7) 0.0 36.7 (23.7) (16.4) (79.0) 83.1 (285.4) (27.9) 30.5 (243.4) (62.7) (15.3) 4.6 (38.6) All temporary differences are calculated as the one-year change in the temporary difference category of deferred tax positions; figures remain in dollars of tax. ADA: bad debt expense (includes allowance for finance receivables if filer is a bank). BENEFITS: bonus accrual expense; accrued compensation expense; accrued vacation expense; accrued healthcare expense; employee stock plan expenses; workers compensation expense; other employment benefits; items that contain combinations of employment benefits, other post-employment benefits, and pensions. CREDITS-t: credit carryforwards. ENVIRON: asset retirement/abandonment obligation expenses; accrued environmental expenses; decommissioning reserves/removal costs; depletion expenses; emission allowances; mineral reserves; exploration costs. EXP: advertising/marketing/promotion expenses; member acquisition costs; policy acquisition costs; loan origination costs; travel reward program expenses; other expenses. FORt: currency translation adjustments; taxes on unremitted profits; deductions related to foreign subsidiaries; deferred foreign income taxes; foreign inventory adjustments. INTANG: acquired intangible assets (including goodwill); start-up & organization costs; capitalized research and development; software development costs; mortgage 47 servicing rights; franchise costs; license costs; research and development write-offs. INVENT: capitalized inventory under UNICAP; intercompany profit in inventory, inventory valuation adjustments. LEASE: capital lease adjustments by the lessee; gains from sale-leaseback transactions; lease-related expenditures. M&A-t: acquisition/merger-related costs; gains/losses on the sale of subsidiaries or business units, restructuring and reorganizational costs; discontinued operations. MTM: unrealized gains/losses on available-for-sale securities; unrealized gains/losses on non-available-for-sale securities; items that contain both available-for-sale and not available-for-sale securities; it includes derivative and hedging activities. NOL: net operating loss carryforwards; capital loss carryforwards; charitable contribution carryforwards. OPEB: pension expenses; other post-employment benefit expenses. OTHER-t: line items too vague to categorize (e.g., “other adjustments”); includes items which are partially but not fully categorizable (e.g., “accounts receivable and inventory”). PPE: difference between book and tax depreciation or property; allowance for funds used during construction; asset impairments/write-downs; asset valuation reserves; assets held for sale; items that contain both depreciation and intangible amortization. REG-t: adjustments unique to regulated entities; unrecovered gas and fuel costs; unamortized nuclear costs; statutory contingency reserves; rate refunds and rate case costs; power purchase agreements. REV: accrued dividends and royalties, advance payments; contract accounting differences; intercompany receivables; customer deposits, customer discounts; unearned revenue; other deferred income/revenue; completed contract method of accounting; present value of future profits; deferred lease revenue. STATE: state and local franchise and income taxes; valuation allowances on state deferred taxes. SUB-t: subsidiary, partnership, and joint venture investments; equitymethod subsidiary investments; earnings from partnerships and joint ventures. UNUSUAL-t: all line items not included in another category. They are material enough to warrant separate disclosure for that firm but occur too infrequently in the sample to warrant a separate category. VA: change in the valuation allowance. WARRANTY: warranty reserves and accruals. 48 Table 3 Sample Means ($M of tax) for Items from the Rate Reconciliation Positive (negative) values denote reconciling items that increase (decrease) the effective tax rate. Variable ADJTR AMORT-GW AMORT-OTH AUDIT COLI CREDITS-r DIVS DMD ETI FOR-r M&A-r MININT OTHER-r REG-r REPAT STATERATE SUB-r TEI UNUSUAL-r TOTAL (n=3890) (n=3890) Mean $M Mean %PI 19.5 -2.4% 13.8 2.7% 6.9 0.5% (9.0) -0.5% (0.1) 0.0% (12.1) -0.8% (2.6) -0.2% (0.6) 0.0% (3.3) -0.3% (25.8) -1.9% 0.6 0.2% (0.2) 0.0% (3.9) -0.3% 0.7 0.1% 3.7 0.0% 25.0 2.3% (3.8) -0.2% (10.0) -1.1% 2.3 -0.2% 1.0 -2.0% % ne 0 26 17 9 18 2 32 11 2 9 50 7 2 92 7 8 84 6 12 16 (n=317) (n=349) (n=342) (n=351) (n=317) (n=297) (n=287) (n=275) (n=278) (n=275) (n=268) (n=266) (n=261) 2001 2002 2003 2004 2005 2006 2007 1995 1996 1997 1998 1999 2000 (4.6) (1.9) (0.2) (2.9) (0.3) 2.6 14.9 51.4 15.4 34.2 19.6 8.4 148.4 2.9 3.2 8.0 7.0 7.3 9.1 12.1 74.0 10.4 3.5 2.5 2.8 47.0 4.3 5.3 7.1 15.0 8.5 8.0 8.7 5.7 9.2 3.3 2.6 4.8 4.4 (1.2) (2.7) (2.5) (3.6) (1.2) (2.6) 2.7 (5.7) (9.8) (19.6) (39.0) (31.7) (9.9) (0.2) (0.2) (0.2) (0.0) 0.1 0.9 0.0 (0.2) (0.2) (0.1) (0.4) (0.5) (0.9) (5.7) (5.8) (6.8) (6.6) (8.4) (9.9) (14.0) (13.2) (14.5) (16.0) (21.6) (19.2) (22.9) (1.1) (1.5) (1.4) (1.3) (1.5) (1.8) (1.7) (3.6) (3.3) (3.8) (4.6) (4.7) (5.1) 0.0 (0.0) (0.1) (0.2) (0.3) (0.4) (1.5) (2.5) (3.1) (1.3) (1.6) (1.8) (2.1) (2.7) (3.1) (3.4) (4.0) (3.0) (8.4) (8.0) (3.8) (1.5) 2.2 4.2 2.0 (8.1) (15.4) (17.6) (18.0) (30.4) (45.4) (58.5) (62.3) (56.2) (62.2) (1.1) (1.0) (5.6) 0.6 3.3 9.8 9.8 (5.1) (2.4) (1.3) (0.4) (0.4) 1.8 (0.4) 0.2 (0.3) (1.0) (0.5) (0.3) (0.4) (0.3) 0.4 0.6 0.0 (0.0) (0.0) (0.6) (3.0) (4.1) (2.9) (1.0) (2.1) (0.6) (0.9) (7.3) (6.8) (9.6) (7.1) (7.4) 0.5 1.1 1.5 1.5 1.4 1.9 0.7 0.0 0.2 0.1 0.0 (0.3) (0.8) 2.5 0.6 (0.2) 0.6 0.7 0.1 1.0 0.7 (2.1) 10.3 25.3 6.6 6.4 17.3 17.8 17.8 17.9 27.0 28.1 19.6 20.3 25.3 21.3 38.5 44.5 38.1 (3.0) (4.2) (2.6) (2.5) (1.8) 6.3 1.0 0.4 (8.9) (10.6) (13.7) (6.2) (6.1) (4.2) (4.1) (5.2) (5.6) (7.5) (8.5) (9.1) (9.1) (9.5) (11.1) (17.5) (21.7) (24.3) 0.4 0.8 0.3 1.1 2.7 0.8 0.0 (1.6) 4.0 26.0 6.0 (3.6) (5.5) 6.7 7.2 5.8 7.2 10.8 21.7 23.1 78.2 (41.9) (37.0) (84.1) (90.8) 96.4 All reconciling items are presented in dollars of tax, calculated as the effect on the effective tax rate times income before taxes. ADJRET: provision to tax return adjustments, adjustment of deferred taxes due to changes in rates or laws, change in the valuation allowance. AMORT-GW: amortization and impairments of goodwill; effect of nondeductible goodwill; negative goodwill amortization; purchase accounting adjustments. AMORT-OTH: amortization of intangible assets; nondeductible amortization; includes acquired purchased in-process research and development. AUDIT: audit resolutions and settlements; adjustments of tax reserves or prior year adjustments; FIN 48 and other tax contingencies. COLI: company-owned life insurance. CREDITS-r: all credits including state, foreign, investment, and alternative minimum tax; credit carryforwards; expiration of credits. DIVS: non-taxable dividends, including dividends-received deduction. DMD: domestic production activities deduction. ETI: benefits of FSC and ETI. FOR-r: effects of foreign rates higher or lower than U.S. statutory rate; currency translation adjustments; foreign tax holidays; foreign withholding tax. MININT: minority interest. M&A-r: merger costs, Gains/losses on the sale or disposal of subsidiaries, restructuring charges. UNUSUAL-r: all line items not included in another category. They are material enough to warrant separate disclosure for that firm but occur too infrequently in the sample to warrant a separate category. OTHER-r: line items too vague to categorize (e.g., “nondeductible expenses”); includes items which are partially but not fully categorizable (e.g., nondeductible goodwill and other). 49 REG-r: items unique to regulated entities. REPAT: effects of repatriating foreign income. STATERATE: effects of U.S. state income tax rates. SUB-r: equity earnings in affiliates; gain on subsidiary IPO; consolidation differences; write-downs of equity investments, permanent items flowed-through from subsidiaries. TEI: non-taxable interest and dividends. 50 Table 4 Sample Means ($M of tax) for Items from the Rate Reconciliation Positive (negative) values denote reconciling items that increase (decrease) the effective tax rate. (n=465) variable ADJTR AMORT-GW AMORT-OTH AUDIT COLI CREDITS-r DIVS DMD ETI FOR-r M&A-r MININT OTHER-r REG-r REPAT STATERATE SUB TEI UNUSUAL-r TOTAL (n=58) (n=212) (n=174) (n=84) (n=161) (n=198) (n=623) (n=365) (n=309) (n=417) (n=653) Mining Food Textile Chemical Pharma Extraction Durables Computers Transport Utilities Retail Banks & Insurance (n=121) Services 21.7 3.3 0.0 (1.3) 0.0 (0.1) 0.0 (1.1) (0.7) 0.5 0.0 0.0 2.6 0.0 (0.1) 12.7 0.0 0.0 (0.9) (2.8) 14.4 4.7 (28.5) 0.0 (1.6) 0.0 (0.1) (2.1) (58.9) 1.2 0.0 (8.8) 0.0 12.6 37.3 (3.6) (0.3) 1.4 1.5 8.4 1.0 (4.8) (0.1) (7.7) (0.3) (0.6) (2.4) (10.4) 1.1 (2.6) (9.7) 0.0 4.5 10.9 0.0 0.0 4.4 (3.2) (0.0) 17.6 (7.9) (0.0) (8.5) (1.0) 0.0 (2.5) (57.3) 0.9 0.0 (5.6) 0.0 (3.6) 6.9 (13.3) 0.0 (2.5) 6.2 7.1 47.4 (8.7) 0.0 (21.2) (7.9) (0.2) 0.0 (375.9) 5.9 (2.8) 7.2 0.0 32.9 29.2 (2.9) 0.0 5.7 (11.1) 2.7 0.5 (17.0) (0.0) (20.8) (1.4) (2.0) (2.4) 365.2 (2.6) 0.0 (25.2) 0.0 6.5 31.7 (62.1) 0.0 (0.7) 94.5 7.6 2.6 (11.9) (0.0) (11.3) (2.6) (0.6) (11.0) (22.2) (0.7) 0.1 (2.9) 0.0 1.3 6.5 (1.7) (0.2) (4.3) 43.1 8.4 13.4 (14.8) 0.0 (12.2) (0.1) (0.9) (9.7) (87.8) 8.5 0.0 2.7 0.0 6.6 17.7 (0.3) (4.4) (2.8) 2.0 100.8 26.1 (6.7) (0.0) (7.5) (1.1) 0.0 (1.6) (3.2) (0.8) 0.1 (3.5) (0.0) 1.4 57.6 7.3 (0.1) 32.4 2.7 3.4 1.1 (4.7) (0.1) (29.4) (0.9) (2.0) 0.0 (1.7) (1.4) 0.2 (4.0) 6.3 1.0 24.5 0.0 (0.1) 0.7 0.2 3.5 0.7 (1.7) (0.1) (1.1) (0.4) (0.0) (0.1) (8.8) 0.9 0.1 (0.9) 0.0 0.4 25.5 (0.5) (0.3) (0.0) 0.1 3.1 0.9 (7.5) (0.7) (19.1) (8.2) 0.0 0.3 (31.1) (0.3) (0.1) (6.3) 0.0 0.3 36.4 (1.4) (71.8) (3.2) 3.4 13.4 3.0 (2.4) (0.1) (7.2) 0.1 0.0 0.0 (3.9) 0.4 0.3 (0.2) 0.0 0.7 16.5 0.7 (2.7) 10.8 36.8 (35.1) (6.7) (80.2) (277.8) 261.3 43.0 (32.6) 203.1 (4.5) 17.4 (108.5) 32.8 All reconciling items are presented in dollars of tax, calculated as the effect on the effective tax rate times income before taxes. ADJRET: provision to tax return adjustments, adjustment of deferred taxes due to changes in rates or laws, change in the valuation allowance. AMORT-GW: amortization and impairments of goodwill; effect of nondeductible goodwill; negative goodwill amortization; purchase accounting adjustments. AMORT-OTH: amortization of intangible assets; nondeductible amortization; includes acquired purchased in-process research and development. AUDIT: audit resolutions and settlements; adjustments of tax reserves or prior year adjustments; FIN 48 and other tax contingencies. COLI: company-owned life insurance. CREDITS-r: all credits including state, foreign, investment, and alternative minimum tax; credit carryforwards; expiration of credits. DIVS: non-taxable dividends, including dividends-received deduction. DMD: domestic production activities deduction. ETI: benefits of FSC and ETI. FOR-r: effects of foreign rates higher or lower than U.S. statutory rate; currency translation adjustments; foreign tax holidays; foreign withholding tax. MININT: minority interest. M&A-r: merger costs, Gains/losses on the sale or disposal of subsidiaries, restructuring charges. UNUSUAL-r: all line items not included in another category. They are material enough to warrant separate disclosure for that firm but occur too infrequently in the sample to warrant a separate category. OTHER-r: line items too vague to categorize (e.g., “nondeductible expenses”); includes items which are partially but not fully categorizable (e.g., nondeductible goodwill and other). REG-r: items unique to regulated entities. REPAT: effects of repatriating foreign income. STATERATE: effects of U.S. state income tax rates. SUB-r: equity earnings in 51 affiliates; gain on subsidiary IPO; consolidation differences; write-downs of equity investments, permanent items flowed-through from subsidiaries. TEI: non-taxable interest and dividends. 52 Table 5 Mean of non-zero observations Fortune 250 firms, 1994-2007 n 3890 Mean (%) 9.038 Δ PTBI Δ CF 3890 3890 Mean ($M) Median ($M) 135.125 63.066 22.919 10.050 Δ Temporary Δ AQ BTD Δ BENEFITS Δ ENVIRON Δ EXP Δ INTANG Δ INVENT Δ LEASE Δ PPE Δ REV Δ UNUSUAL-t 3890 3853 2171 360 603 1458 1368 672 3516 1279 1713 17.149 9.812 (6.627) 15.543 (1.060) 2.761 0.889 8.844 6.809 0.084 7.008 0.317 (0.136) 0.500 0.457 (0.337) (0.523) 0.081 0.100 (0.095) 0.247 0.338 (0.402) (0.392) (0.005) (0.057) (0.420) (0.262) (0.268) 0.040 (0.191) (0.015) 0.101 Δ TAX BTD-t Δ FOR-t Δ M&A-t Δ NOL Δ REG-t Δ STATE-t 2960 832 914 1790 406 455 8.414 1.446 13.394 1.328 19.678 2.399 1.254 (0.594) 2.993 1.962 (0.117) 0.131 0.032 0.000 0.104 (0.090) 0.477 (0.002) Δ Permanent Δ TAX BTD-r Δ COLI Δ DIVS Δ DMD Δ M&A-r Δ REG-r Δ TEI Δ UNUSUAL-r 3757 1828 91 435 92 444 269 508 732 (2.651) (2.574) (1.048) (2.340) (8.933) 4.994 (1.581) (5.602) (2.345) (0.100) (0.376) (0.494) (0.366) (7.850) 0.713 0.556 (0.749) (0.080) 0.011 0.027 (0.011) (0.001) (0.052) 0.122 (0.014) (0.016) 0.096 Δ NO INFO BTD Δ OTHER-t Δ SUB-t Δ OTHER-r Δ SUB-r 3885 3857 586 3671 293 (0.323) 2.164 (7.422) (1.343) (1.106) 0.251 0.346 0.062 (0.034) (0.652) (0.059) (0.055) 0.078 (0.015) (0.030) Variable R Mean (% beg MVE) 0.017 0.005 The dependent variable is Rit , which denotes the market-adjusted return for firm i and is defined as the compound (with dividend) return less the compound return on the value-weighted market portfolio. Returns are calculated over a 16-month period, starting at the beginning of fiscal year t and ending four months after the end of the fiscal year t. ∆PTBI it is the change in pretax book income less minority interest. ∆CFit is the change in total cash flows. Δ Temporary is the sum of the changes in the temporary difference categories: AQ BTD, TAX BTD-t, OTHER-t, and SUB-t. AQ BTD is the sum of the changes in the accrual quality BTD categories: BENEFITS, ENVIRON, EXP, INTANG, INVENT, LEASE, PPE, REV and UNUSUAL-t. TAX BTD-t is the sum of the changes in the tax BTD categories: NOL, FOR-t, M&A-t, REG-t, and STATE. Δ Permanent is the sum of the changes in the permanent difference categories: TAX BTD-r, OTHER-r, and SUB-r. TAX BTD-r is the sum of the changes in the TAX BTD-r categories: COLI, DIVS, DMD, M&A-r, UNUSUAL-r, REG-r, and TEI. NO INFO BTD is the sum of the changes in the categories in which the information is too difficult to decipher to be informative: OTHER-t, SUB-t, OTHER-r, and SUB-r; these BTDs are included in Δ Temporary and Δ Permanent as appropriate. The variables for each category formed from accounts in the statement of deferred tax positions are computed as the difference between the values in year t and year t-1 less the difference between the values in year t-1 and year t-2. These variables are: BENEFITS: bonus accrual expense; accrued compensation expense; accrued vacation expense; accrued healthcare expense; employee stock plan expenses; workers compensation expense; other employment benefits; items that contain combinations of employment benefits, other post-employment benefits, and pensions. ENVIRON: asset retirement/abandonment obligation expenses; accrued environmental expenses; decommissioning reserves/removal costs; depletion expenses; emission allowances; mineral reserves; exploration costs. EXP: advertising/marketing/promotion expenses; member acquisition costs; policy acquisition costs; loan origination costs; travel reward program expenses; other expenses. FOR-t: currency translation adjustments; taxes on unremitted profits; deductions related to foreign subsidiaries; deferred foreign income taxes; foreign inventory adjustments. INTANG: acquired intangible assets (including goodwill); start-up & organization costs; research and development; software development costs; mortgage servicing rights; franchise costs; license costs; research and development write-offs. INVENT: capitalized inventory under UNICAP; intercompany profit in inventory, inventory valuation adjustments. LEASE: capital lease adjustments by the lessee; gains from sale-leaseback transactions; lease-related expenditures. M&A-t: acquisition/merger-related costs; gains/losses on the sale of subsidiaries or business units, restructuring and reorganizational costs; discontinued operations. NOL: net operating loss carryforwards; capital loss carryforwards; charitable contribution carryforwards. OTHER-t: line items too vague to categorize (e.g., “other adjustments”); includes items which are partially but not fully categorizable (e.g., “accounts receivable and inventory”). PPE: difference between book and tax depreciation or property; allowance for funds used during construction; asset impairments/write-downs; asset valuation reserves; assets held for sale; items that contain both depreciation and intangible amortization. REG-t: adjustments unique to regulated entities; unrecovered gas and fuel costs; unamortized nuclear costs; statutory contingency reserves; rate refunds and rate case costs; power purchase agreements. REV: accrued dividends and royalties, advance payments; contract accounting differences; intercompany receivables; customer deposits, customer discounts; unearned revenue; other deferred income/revenue; completed contract method of accounting; present value of future profits; deferred lease revenue. STATE: state and local franchise and income taxes; valuation allowances on state deferred taxes. SUB-t: subsidiary, partnership, and joint venture investments; equity-method subsidiary investments; earnings from partnerships and joint ventures. UNUSUAL-t: all line items not included in another category. They are material enough to warrant separate disclosure for that firm but occur too infrequently in the sample to warrant a separate category. The variables for each category formed from items in the rate reconciliation are computed as the dollar value in the rate reconciliation in year t less the dollar value in the rate reconciliation in year t-1. These categories are: COLI: company-owned life insurance. DIVS: non-taxable dividends including dividends-received deduction. DMD: domestic production activities deduction. M&A-r: merger costs, Gains/losses on the sale or disposal of subsidiaries, restructuring charges. OTHER-r: line items too vague to categorize (e.g., “nondeductible expenses”); includes items which are partially but not fully categorizable (e.g., nondeductible goodwill and other). REG-r: items unique to regulated entities. SUB-r: equity earnings in affiliates; gain on subsidiary IPO; consolidation differences; writedowns of equity investments, permanent items flowed-through from subsidiaries. TEI: non-taxable interest and dividends. UNUSUAL-r: all line items not included in another category. They are material enough to warrant separate disclosure for that firm but occur too infrequently in the sample to warrant a separate category. 54 Table 6 Estimated Regression Coefficients from Regressing 16-month returns on pretax book income, total cash flows and book-tax differences Fortune 250 firms, 1994-2007 pred intercept Δ PTBI Δ CF (+) (+) Δ Temporary Δ AQ BTD Δ BENEFITS Δ ENVIRON Δ EXP Δ INTANG Δ INVENT Δ LEASE Δ PPE Δ REV Δ UNUSUAL-t (+) (+) (+) (+) (+) (+) (+) (+) (+) (+) (+) Δ TAX BTD-t Δ FOR-t Δ M&A-t Δ NOL Δ REG-t Δ STATE-t (+) (+) (+) (+) (+) (+) Δ Permanent Δ TAX BTD-r Δ COLI Δ DIVs Δ DMD Δ M&A-r Δ REG-r Δ TEI Δ UNUSUAL-r (-) (-) (+/-) (-) (-) (-) (-) (-) (-) A 0.139 0.370 ** 0.757 ** B 0.141 0.430 ** 0.758 ** 0.176 C 0.140 0.421 ** 0.746 ** D 0.120 0.575 ** 0.812 ** * 0.245 ** 0.466 (0.102) 0.994 ** (0.763) ** (0.682) (1.344) (0.138) 2.356 ** 0.449 (0.001) 0.107 1.516 ** (0.089) (0.027) 4.809 * 0.912 * Δ NO INFO BTD Δ OTHER-t Δ SUB-t Δ OTHER-r Δ SUB-r ** (*) significant at the 0.01 (0.05) level, using a two-tailed test 1.416 ** 5.303 6.236 (19.733) 1.040 (1.516) (23.753) (1.478) 0.004 (0.027) 1.429 ** 1.518 (2.294) The dependent variable is Rit , which denotes the market-adjusted return for firm i and is defined as the compound (with dividend) return less the compound return on the value-weighted market portfolio. Returns are calculated over a 16-month period, starting at the beginning of fiscal year t and ending four months after the end of the fiscal year t. ∆PTBI it is the change in pretax book income less minority interest. ∆CFit is the change in total cash flows. Δ Temporary is the sum of the changes in the temporary difference categories: AQ BTD, TAX BTD-t, OTHER-t, and SUB-t. AQ BTD is the sum of the changes in the accrual quality BTD categories: BENEFITS, ENVIRON, EXP, INTANG, INVENT, LEASE, PPE, REV and UNUSUAL-t. TAX BTD-t is the sum of the changes in the tax BTD categories: NOL, FOR-t, M&A-t, REG-t, and STATE. Δ Permanent is the sum of the changes in the permanent difference categories: TAX BTD-r, OTHER-r, and SUB-r. TAX BTD-r is the sum of the changes in the TAX BTD-r categories: COLI, DIVS, DMD, M&A-r, UNUSUAL-r, REG-r, and TEI. NO INFO BTD is the sum of the changes in the categories in which the information is too difficult to decipher to be informative: OTHER-t, SUB-t, OTHER-r, and SUB-r; these BTDs are included in Δ Temporary and Δ Permanent as appropriate. The variables for each category formed from accounts in the statement of deferred tax positions are computed as the difference between the values in year t and year t-1 less the difference between the values in year t-1 and year t-2. These variables are: BENEFITS: bonus accrual expense; accrued compensation expense; accrued vacation expense; accrued healthcare expense; employee stock plan expenses; workers compensation expense; other employment benefits; items that contain combinations of employment benefits, other post-employment benefits, and pensions. ENVIRON: asset retirement/abandonment obligation expenses; accrued environmental expenses; decommissioning reserves/removal costs; depletion expenses; emission allowances; mineral reserves; exploration costs. EXP: advertising/marketing/promotion expenses; member acquisition costs; policy acquisition costs; loan origination costs; travel reward program expenses; other expenses. FOR-t: currency translation adjustments; taxes on unremitted profits; deductions related to foreign subsidiaries; deferred foreign income taxes; foreign inventory adjustments. INTANG: acquired intangible assets (including goodwill); start-up & organization costs; research and development; software development costs; mortgage servicing rights; franchise costs; license costs; research and development write-offs. INVENT: capitalized inventory under UNICAP; intercompany profit in inventory, inventory valuation adjustments. LEASE: capital lease adjustments by the lessee; gains from sale-leaseback transactions; lease-related expenditures. M&A-t: acquisition/merger-related costs; gains/losses on the sale of subsidiaries or business units, restructuring and reorganizational costs; discontinued operations. NOL: net operating loss carryforwards; capital loss carryforwards; charitable contribution carryforwards. OTHER-t: line items too vague to categorize (e.g., “other adjustments”); includes items which are partially but not fully categorizable (e.g., “accounts receivable and inventory”). PPE: difference between book and tax depreciation or property; allowance for funds used during construction; asset impairments/write-downs; asset valuation reserves; assets held for sale; items that contain both depreciation and intangible amortization. REG-t: adjustments unique to regulated entities; unrecovered gas and fuel costs; unamortized nuclear costs; statutory contingency reserves; rate refunds and rate case costs; power purchase agreements. REV: accrued dividends and royalties, advance payments; contract accounting differences; intercompany receivables; customer deposits, customer discounts; unearned revenue; other deferred income/revenue; completed contract method of accounting; present value of future profits; deferred lease revenue. STATE: state and local franchise and income taxes; valuation allowances on state deferred taxes. SUB-t: subsidiary, partnership, and joint venture investments; equity-method subsidiary investments; earnings from partnerships and joint ventures. UNUSUAL-t: all line items not included in another category. They are material enough to warrant separate disclosure for that firm but occur too infrequently in the sample to warrant a separate category. The variables for each category formed from items in the rate reconciliation are computed as the dollar value in the rate reconciliation in year t less the dollar value in the rate reconciliation in year t-1. These categories are: COLI: company-owned life insurance. DIVS: non-taxable dividends including dividends-received deduction. DMD: domestic production activities deduction. M&A-r: merger costs, Gains/losses on the sale or disposal of subsidiaries, restructuring charges. OTHER-r: line items too vague to categorize (e.g., “nondeductible expenses”); includes items which are partially but not fully categorizable (e.g., nondeductible goodwill and other). REG-r: items unique to regulated entities. SUB-r: equity earnings in affiliates; gain on subsidiary IPO; consolidation differences; writedowns of equity investments, permanent items flowed-through from subsidiaries. TEI: non-taxable interest and dividends. UNUSUAL-r: all line items not included in another category. They are material enough to warrant separate disclosure for that firm but occur too infrequently in the sample to warrant a separate category. 56 Table 7 Estimated Regression Coefficients from Regressing 16-month returns on pretax book income, total cash flows and book-tax differences Fortune 250 firms, 1994-2007 intercept Δ PTBI Δ CF Table 6 pred B 0.141 (+) 0.430 ** (+) 0.758 ** Δ Temporary Δ AQ BTD Δ TAX BTD-t (+) (+) (+) 0.176 * Δ Permanent Δ TAX BTD-r (-) (-) 0.912 * Δ NON-BTD Table 6 C 0.140 0.421 ** 0.746 ** B' 0.144 0.461 ** 0.780 ** P>z 0.04 0.04 0.140 0.740 * (0.153) Δ NO INFO BTD Δ REDUNDANT BTD P>z 0.12 C' 0.132 0.483 ** 0.765 ** P>z 0.245 ** (0.001) 0.01 0.99 1.416 ** 0.00 P>z 0.188 (0.010) 0.06 0.94 0.03 0.06 0.004 0.98 1.810 ** 0.00 (0.210) * 0.02 (0.067) 0.059 0.73 0.76 ** (*) significant at the 0.01 (0.05) level, using a two-tailed test The dependent variable is Rit , which denotes the market-adjusted return for firm i and is defined as the compound (with dividend) return less the compound return on the value-weighted market portfolio. Returns are calculated over a 16-month period, starting at the beginning of fiscal year t and ending four months after the end of the fiscal year t. ∆PTBI it is the change in pretax book income less minority interest. ∆CFit is the change in total cash flows. The variables for each category formed from accounts in the statement of deferred tax positions are computed as the difference between the values in year t and year t-1 less the difference between the values in year t-1 and year t-2. Δ Temporary is the sum of the changes in the temporary difference categories: AQ BTD, TAX BTD-t, OTHER-t, and SUB-t. AQ BTD is the sum of the changes in the accrual quality BTD categories: BENEFITS, ENVIRON, EXP, INTANG, INVENT, LEASE, PPE, REV and UNUSUAL-t. TAX BTD-t is the sum of the changes in the tax BTD categories: NOL, FOR-t, M&A-t, REG-t, and STATE. The variables for each category formed from items in the rate reconciliation are computed as the dollar value in the rate reconciliation in year t less the dollar value in the rate reconciliation in year t-1. Δ Permanent is the sum of the changes in the permanent difference categories: TAX BTD-r, OTHER-r, and SUB-r. TAX BTD-r is the sum of the changes in the TAX BTD-r categories: COLI, DIVS, DMD, M&A-r, UNUSUAL-r, REG-r, and TEI. NON-BTD is the sum of changes in the categories that are related to a difference in expected tax and actual tax rather than a difference in book income and taxable income: ADJTR, AUDIT, CREDITS-t, CREDITS-r, ETI, FOR-r, REPAT, STATERATE, and VA-t. NO INFO BTD is the sum of the changes in the categories in which the information is too difficult to decipher to be informative: OTHER-t, SUB-t, OTHER-r, and SUB-r. REDUNDANT BTD is the sum of changes in the categories that are also disclosed elsewhere in the financial statements: ADA, AMORT-GW, AMORT-OTH, MININT, MTM, OPEB, and WARRANTY. Items in NO INFO BTD and REDUNDANT BTD are included in Δ Temporary and Δ Permanent as appropriate. 58