Earnings Management Via Intraperiod Tax Allocations: The Case of Discontinued Operations

Transcription

1 Earnings Management Via Intraperiod Tax Allocations: The Case of Discontinued Operations Steven E. Kaplan David G. Kenchington Brian S. Wenzel Arizona State University August 20, 2015 Abstract We examine whether firms reporting discontinued operations engage in biased intraperiod tax allocations by moving tax expense out of continuing operations and into discontinued operations (or, conversely, moving tax benefits out of discontinued operations and into continuing operations). We contend that managers use discontinued operations as an opportunity to upwardly manage the reported amount of after-tax earnings from continuing operations. Consistent with our expectations, we find that compared to years immediately preceding or following the period of discontinued operations, the effective tax rate from continuing operations is significantly lower during the years that firms report incomeincreasing discontinued operations. As further support of our earnings management argument, we find that firms with income-increasing discontinued operations are more likely to engage in biased intraperiod tax allocations to meet or just beat the median analyst forecast and are less likely to engage in this biased behavior when they are subject to higher quality audits. This paper is the first to document that classification shifting occurs for firms reporting income-increasing nonrecurring items, whereas prior research has only found evidence that firms shift expenses into income-decreasing nonrecurring items (McVay 2006 and Barua et al. 2010).

2 Introduction Classification shifting, one of three methods of earnings management 1, involves the deliberate misclassification of items within the income statement (McVay 2006, p. 501.) While classification shifting does not change the current period s net income, it does lead to off-setting changes in income statement components, such as between income from continuing operations and discontinued operations. While research examining classification shifting is growing (McVay 2006; Barua et al. 2010; Fan et al. 2010; Haw et al. 2011; Fan and Liu 2015), whether managers use intraperiod tax allocations to boost earnings from continuing operations has not been examined. An intraperiod tax allocation involves apportioning a corporation s total income tax expense for a period to the various components of its net income (Wahlen et al. 2013, p. 5-19). Therefore, when reporting discontinued operations, managers must allocate total tax expense between income from continuing operations and discontinued operations. In this study, we provide evidence on whether managers intraperiod tax allocations are biased when they report discontinued operations. Managers generally have incentives to boost earnings from continuing operations, in part, because investors tend to value persistent new earnings more highly than less persistent new earnings (Lipe 1986; Kormendi and Lipe 1987; Collins and Kothari 1989) As earnings from continuing operations are more persistent than earnings from discontinued operations (Curtis et al. 2014), managers may have incentives to develop biased intraperiod tax allocations in order to report higher income from continuing operations. In addition, intraperiod tax allocations present an opportunity to engage in earnings management. In general, determining tax expense involves complex judgments 1 The other two methods of earnings management involve accruals and real activities. 2

3 and estimates that are subject to manager discretion (Dhaliwal et al. 2004, Cazier et al. 2015). More specifically, the intraperiod tax allocation is one of the more complex areas within ASC 740 (EY 2014 p. 242). Given this opportunity, management may bias downward the amount of tax allocated to continuing operations, which will lower the firm s effective tax rate (ETR) on continuing operations. Thus, we focus on ETRs to capture tax classification shifting. For firms reporting discontinued operations, we examine firms ETR on continuing operations from before, during, and after the discontinued operations period to capture tax classification shifting. 2 As discussed further below, we expect that ETRs for firm years with an income-increasing discontinued operation (i.e. discontinued operations reporting positive income) will be lower compared to firm years preceding and following the reporting of discontinued operations. Further, for firm years with an income-increasing discontinued operation, we expect the association to be stronger for firms that meet or just beat analysts consensus forecast and weaker for firms that have higher quality audits. In contrast, for firm-years with an income-decreasing discontinued operation, we do not expect firm year ETRs to change. Our results show that the ETR for firms recognizing income-increasing discontinued operations is significantly lower during the discontinued operations period than in the period immediately preceding and immediately following the period of discontinued operations. Specifically, we find that a one standard deviation increase in the size of income-increasing discontinued operations is associated with an ETR that is 69 basis points lower during the period of discontinued operations when compared to the 2 For the remainder of the paper, when we discuss ETRs we are referring to the ETR on continuing operations unless specifically noting otherwise. 3

4 period immediately preceding the discontinued operations. Further, we find that the reduction in the ETR reverses when comparing the period of discontinued operations to the period immediately following it, indicating that once the opportunity for tax shifting is finished the decrease in the ETR disappears. We believe our evidence shows that managers of firms recognizing income-increasing discontinued operations opportunistically increase after-tax income from continuing operations by shifting tax expense from continuing operations to discontinued operations. 3 Further, we are unable to find evidence that managers engage in biased intraperiod tax allocations when reporting income-decreasing discontinued operations (i.e. discontinued operations reporting negative income). We believe that these results are likely due to a combination of two factors: (1) shifting tax expense into income-decreasing discontinued operations increases the likelihood that the firm will report a positive tax expense on a loss component of earnings, which is likely to draw additional attention and scrutiny from the external auditors; and (2) given that Barua et al. (2010) find support for classification shifting of core expenses out of income from continuing operations and into income-decreasing discontinued operations, it may be that managers for these firms are reluctant to also use tax accounts to further manage earnings. It is important to note that our insignificant results for income-decreasing discontinued operations help to rule out the possibility that our significant results for income-increasing discontinued operations are driven by some kind of mechanical relationship. 4 3 For ease of exposition, we primarily discuss moving tax expense out of continuing operations and into discontinued operations. However, it is also possible that firms move tax benefits (for example, credits) out of discontinued operations and into continuing operations. 4 For example, it is possible that designating a portion of the firm s assets as discontinued could make future taxable income higher for the continuing operations and thus lead to a release of a valuation allowance. This could lead to a one-time reduction in a firm s ETR, consistent with the pattern we are documenting. 4

5 To provide further evidence that our results are due to deliberate and biased intraperiod tax allocations, we examine a setting where the pressure to manage earnings is high. Specifically, we look at whether or not the firm meets or just beats the median analyst forecast during the period of discontinued operations. Prior literature documents that firms receive market rewards for meeting or just beating the analysts forecast (Barth et al. 1999; Brown and Caylor 2005), hence pressure to meet these forecasts is high. Also, research shows that the tax accounts are frequently managed to meet or just beat the analyst consensus forecast (Dhaliwal et al. 2004; Graham et al. 2012; Cazier et al. 2015). Therefore, we expect tax expense shifting to be stronger for firms reporting incomeincreasing discontinued operations that meet or just beat the analyst consensus forecast. We find evidence consistent with this prediction. We also examine whether earnings management via intraperiod tax allocations is lower when a firm has a higher quality audit. We predict that an audit will be of higher quality if a Big N auditor performs the audit (DeFond et al. 2014).We find earnings management via intraperiod tax allocations is lower when a firm has a higher quality audit, consistent with our prediction. Finally, we examine whether the adoption of SFAS No. 144 (FAS 144, now codified under ASC topic 360) increased classification shifting through biased intraperiod tax allocations. Prior to the adoption of FAS 144, firms were only able to report discontinued operations if they were disposing of a business segment (e.g. a major line of business or geographical area of operations). With the implementation of FAS 144, the FASB reduced the threshold for recognition of an asset disposal to qualify as discontinued However, if this were the case we would expect the pattern on average to be stronger for firms that have income-decreasing discontinued operations, which we do not find. 5

6 operations. This could provide more opportunities to classify operations as discontinued. However, prior research has found that the size of discontinued operations decreased after FAS 144 (Barua et al. 2010), which may reduce the opportunity for earnings management. Therefore, it is not clear whether managers use of biased intraperiod tax allocations will change as a result of FAS 144. We do not find evidence of a change in the post FAS 144 regime. Our findings contribute to academic literature in several ways. Our study is the first to examine whether and when managers are likely to use intraperiod tax allocations as a mechanism for classification shifting. In addition, we provide further evidence of how managers engage in classification shifting when reporting discontinued operations. Prior research has only found evidence that classification shifting occurs for firms reporting income-decreasing nonrecurring items (McVay 2006; Fan et al. 2010; Barua et al. 2010). 5 In contrast, by focusing on the potential for biased intraperiod tax allocations, we provide evidence that classification shifting also occurs for firms reporting income-increasing nonrecurring items (i.e. discontinued operations). 6 Thus, classification shifting is likely to occur for both income-increasing and income-decreasing discontinued operations, but there are key differences in how classification shifting occurs. Finally, our research contributes by responding to the call made by Graham et al. (2012) to provide empirical evidence regarding the placement of the tax expense on the income statement on management s 5 Specifically, McVay (2006) and Fan et al. (2010) find that managers engage in classification shifting by moving expenses from core earnings to income-decreasing special items. Barua et al. (2010) reports that managers shift core expenses from continuing operations to income-decreasing discontinued operations. 6 We do not examine special items in our paper because special items are reported as part of income from continuing operations and thus no opportunity for tax expense shifting exists. In contrast, extraordinary items are reported below income from continuing operations and require the use of intraperiod period tax allocations. We did not include extraordinary items in our analysis because they are rare. For instance, even the losses incurred from the September 11, 2001 terrorist attacks did not meet the standard of extraordinary item recognition (FASB 2001). 6

7 choice to manage the tax expenses (p. 424). Our evidence suggests that managers use intraperiod tax allocations to boost income from continuing operations. The rest of the paper is organized as follows. The next section discusses the hypotheses. Section three describes the methodology used in the study, and section four presents the empirical results. The last section offers a conclusion. Hypothesis Development Classification Shifting One approach to managing earnings involves classification shifting. Under this approach, managers intentionally misclassify revenues and/or expenses within the income statement, but do not change the firm s net income (McVay 2006). Relative to accrual or real activities based approaches to earnings management, classification shifting has two attractive features. First, classification shifting has no effect on earnings outside of the current period. That is, when using classification shifting, managers avoid implicit costs of settling up, due to the impact of current actions on future or past earnings. In contrast, managing earnings upwards in the current period with accrual-based or real activities tends to lower future or past earnings (McVay 2006). Second, classification shifting is nearly costless to implement. 7 In contrast, managing real activities generally involves the misuse of economic resources, which could damage the firm s future. Previous research on classification shifting as an earnings management tool is growing, but still fairly limited. An early study was conducted by Barnea et al. (1976). This study provides evidence that managers use extraordinary items to smooth operating 7 In our setting, the most obvious cost to classification shifting is its effect on the reported earnings of the discontinued operation (i.e., earnings are lower due to increased tax expenses or decreased tax benefits). 7

8 income. However, as Barua et al. (2010) note, the recognition of an event as extraordinary under GAAP is highly restrictive, thus diminishing the availability of extraordinary items as a means to engage in classification shifting. More recently, McVay (2006) investigates the extent to which managers engage in classification shifting by moving expenses from core earnings (i.e. revenues less costs of goods sold and selling, general, and administrative expenses) to special items. As expected, results are consistent with managers opportunistically shifting expenses away from core earnings and to income-decreasing special items. She provides additional evidence that managers are more likely to engage in classification shifting to meet analyst forecasts. McVay (2006) notes the use of contemporaneous accruals in her expectations model as a weakness that potentially introduces bias to her results. Fan et al. (2010) correct for this bias by using lagged and contemporaneous returns, and provide further evidence demonstrating classification shifting between core earnings and special items. Fan and Liu (2015) extend McVay (2006) by examining when firms misclassify cost of goods sold verses selling, general, and administrative expenses. Using an international setting, Haw et al. (2011) find that classification shifting is pervasive in East Asian countries, especially among family controlled firms. However, they also find that strong investor protection laws and quality external auditors reduce misclassification behavior. Barua et al. (2010) extend research on classification shifting by focusing on a different economic event, discontinued operations. Specifically, Barua et al. (2010) examines classification shifting involving core earnings and discontinued operations. Using core expectation models similar to McVay (2006), Barua et al. (2010) reports that managers shift core expenses from continuing operations to income-decreasing 8

9 discontinued operations. Finally, Lail et al. (2014) find that firms mask their true performance by misallocating expenses between core operating segments and the corporate/other segment. Calculation of Income Tax Expense for Continuing and Discontinued Operations When firms report discontinued operations, the total tax expense must be allocated between continuing and discontinued operations using the with-and-without approach (ASC to 14). This approach generally requires the firm to take the following steps: (1) compute the total tax expense on all items of income and loss for the period ( with all items of income); (2) compute the tax expense on income from continuing operations only ( without the other income or loss items); and (3) subtract the tax expense on income from continuing operations (step 2) from the total tax expense (step 1) and allocate the residual tax expense to the other income item(s). 8 The amount of tax expense allocated to continuing operations is reported as a separate line item on the face of the income statement. In contrast, for income statement items reported after income from continuing operations (i.e., extraordinary items, or in our setting, discontinued operations), the face of the income statement shows a line item reporting an amount that is net of tax. The associated tax expense on these items is generally reported as part of a footnote or in the text of the income statement. Importantly, intraperiod tax allocations do not alter the firm s total income tax expense, but simply the amount of tax expense allocated to continuing versus discontinued operation. Thus, if managers inappropriately over allocate 8 The intraperiod allocation becomes more nuanced when there are multiple other non-continuing income items with different tax attributes (e.g. gains and losses) which would cause the separate computation of tax expenses to not equal the total tax expense computed under step (1). While not a focus of this paper, the allocation rules require the firm to compute the tax expense (or more likely tax benefit) on the total loss items first, ratably allocate this tax benefit amongst the loss items, and then allocate the residual tax expense ratably to the gain items. 9

10 tax expense to discontinued operations, then tax expense to continuing operations must be inappropriately under allocated, which, in turn, increases income from continuing operations. The process of performing intraperiod tax allocations is difficult and complex (EY 2014 and PWC 2015), which consequently, we contend provides managers an opportunity to manage earnings. Classification shifting using tax expense The methodologies used in McVay (2006), Fan et al. (2010), Barua et al. (2010), Haw et al. (2011), and Fan and Liu (2015) use pre-tax income from continuing operations. 9 However, tax expense is often relatively large, and its calculation is complex, often involving judgment, suggesting that managers would be interested in using tax expense as a mechanism for earnings management. Another reason tax expense shifting may be used to manage earnings is analyst earnings forecasts are generally some derivation of after-tax income from continuing operations (Bradshaw and Sloan 2002; So 2013). Because investors utilize analyst earnings forecasts, taxes should play an important role in firms earnings management decisions. Previous research shows that managers use the discretion and complexity in reporting tax expense as an earnings management tool (Dhaliwal et al. 2004; Krull 2004; Schrand and Wong 2003). However, the method of earnings management in these studies is not classification shifting. 9 Ronen and Sadan (1975) and Barnea et al. (1976) both analyze adjusted versions of income from continuing operations, after accounting for taxes. The data they use is from 1951 to 1970, and is limited by the inclusion of only four industries: Paper, Chemicals, Rubber and Airlines. In addition, they don t specifically analyze the effect of the tax accounts. McVay (2006), Fan et al. (2010), and Barua et al. (2010) focus on core earnings of the firm, defined as sales less cost of goods sold and selling, general, and administrative expenses. Thus, these three studies specifically exclude taxes on income from continuing operations. 10

11 Because of the structure of the income statement, there are relatively few opportunities in which intraperiod tax allocations could be used for classification shifting. Fundamentally, few economic events trigger intraperiod tax allocation. For example, research by McVay (2006) and Fan et al. (2010) examines classification shifting involving special items. Because core expenses and special items are included in income from continuing operations, both are also included in the determination of tax expense for continuing operations, and thus it is not possible to use biased tax allocations to engage in classification shifting. In contrast, Barua et al. (2010) examines classification shifting using discontinued operations. Because discontinued operations are reported separately from income from continuing operations, there is an opportunity to use biased tax allocations to engage in classification shifting when reporting discontinued operations. We are aware of only one prior study that investigates earnings management through classification shifting involving tax expense. Robinson (2010) investigates FASB guidance from Emerging Issues Task Force No (EITF 94-1) (FASB 1995). In short, EITF 94-1 requires a firm to account for its low-income housing investments under the equity method, with the amortization of this investment classified as an operating expense. However, if the firm purchases a tax benefit guarantee, then the firm can account for this investment under the effective yield method, where the amortization of this investment is treated as tax expense. Thus, managers have the discretion to purchase a tax benefit guarantee, and by doing so, to reclassify expenses from pre-tax income from continuing operations to tax expense. Her results indicate that firms are willing to purchase the guarantee to increase their pre-tax income from continuing operations. Our current study differs from Robinson (2010) in two key ways: (1) The classification shift in Robinson 11

12 does not involve intraperiod tax allocations, and (2) the classification shift in Robinson is costly to implement, whereas no explicit costs are involved to implement the classification shift in our study. 10 Recall, when reporting discontinued operations, tax expense must be divided between income from continuing operations and discounted operations, using a process called intraperiod tax allocation. We argue that the with-and-without rules for intraperiod tax allocations offer managers an opportunity to engage in classification shifting, as a means to manage earnings. For example, a firm could allocate some of the R&D tax credit or Section 199 domestic manufacturing deduction generated by the discontinued operations to the continuing operations. 11 Such shifting would be, in spirit, similar to the core expense shifting documented by McVay (2006) and Barua et al. (2010). Further, relative to other methods, using intraperiod tax allocations to engage in classification shifting is almost costless and does not impact past or future accounting periods. In addition, managers generally have incentives to boost earnings from continuing operations because these continuing earnings are more persistent than earnings from discontinued operations (Curtis et al. 2014) and because investors tend to value persistent new earnings more highly than less persistent new earnings (Lipe 1986; Kormendi and Lipe 1987; Collins and Kothari 1989). On the other hand, income from discontinued operations is not considered persistent, which is why the income statement separately reports this event after income from continuing operations on the income statement. 10 As Robinson (2010) notes, the average fee accounts for 15.8% of the total investment. Additionally, she notes that, from a pure cost-benefit analysis, the low-risk nature of such investments does not warrant insurance fees of this magnitude. Because classification shifting in Robinson s setting is costly, the earnings management she documents also involves the use of real resources. 11 It is important to note that in order to change the ETR, tax shifting would have to involve permanent booktax differences. 12

13 While Barua et al. (2010) find that classification shifting only occurs with incomedecreasing discontinued operations, we examine both income-increasing and incomedecreasing discontinued operations. While we predict firms with income-increasing discontinued operations will shift tax expense to boost earnings from continuing operations, we do not expect firms will engage in biased intraperiod tax allocations to engage in earnings management for firms reporting income-decreasing discontinued operations. We base our views for income-decreasing discontinued operations on two factors: (1) shifting tax expense into income-decreasing discontinued operations increases the likelihood that the firm will report a positive tax expense on a loss component of earnings, which would likely draw additional scrutiny from the external auditors 12 ; and (2) given that Barua at al. find support for classification shifting of core expenses out of income from continuing operations and into income-decreasing discontinued operations, it may be that managers for these firms are reluctant to also use tax accounts to further manage earnings. Therefore, we might not find an on-average relation between taxes and income-decreasing discontinued operations. Our formal hypotheses are thus: H1a: Relative to the years preceding and following discontinued operations, we expect firms ETR on income from continuing operations will be lower for income-increasing discontinued operations firm years. H1b: For income-decreasing firm years, we do not expect firms ETR on income from continuing operations to differ from the years preceding and following. 12 While shifting tax expense into income-decreasing discontinued operations may draw additional scrutiny from auditors, shifting tax benefits out of income-decreasing discontinued operations into income from continuing operations may not. Therefore, it is possible we could see the ETR on continuing operations decline when firms are reporting income-decreasing discontinued operations. However, for the two reasons listed in the text we do not expect biased tax allocations when firms report income-decreasing discontinued operations. 13

14 To provide further evidence that our results are due to deliberate and biased intraperiod tax allocations, we examine a setting where the pressure to manage earnings is high. Specifically, CEOs report that the consensus analyst earnings forecast is a primary benchmarks that managers feel pressure to achieve (Graham et al. 2005). Similarly, Brown and Caylor (2005) provide evidence that investors reward (and penalize) firms for meeting (missing) analysts earnings forecasts. Research also finds that managers specifically utilize tax accounts to report earnings that meet or beat analyst earnings forecasts (Dhaliwal et al. 2004; Krull 2004; Frank and Rego 2006; Cazier et al. 2015). Barua et al. (2010) find evidence of classification shifting of core operating expenses only for firms that meet or beat analyst forecasts. Consequently, we propose the following hypothesis: H2: The effect of income-increasing discontinued operations on lowering firms ETR on income from continuing operations will be stronger when the firm meets or just beats analysts forecasts. We also examine whether earnings management via intraperiod tax allocations is lower when a firm has a higher quality audit. We predict that an audit will be of higher quality if a Big N auditor performs the audit. We expect audit quality to be higher when a firm hires a Big N auditor because in theory Big N auditors have stronger incentives to provide higher audit effort. These incentives arise due to reputational capital concerns (DeAngelo, 1981), higher litigation risk, and greater regulatory scrutiny (DeFond et al. 2014). Big N auditors are also likely to employ more experts and to develop superior audit methodologies, leading to higher proficiency (Dopuch and Simunic, 1982). Additionally, Big N auditors should be more independent as their diverse customer base reduces their financial dependence on a particular client. In general, the empirical audit literature supports the prediction that Big N auditors provide higher quality audits (Lennox and 14

15 Pittman 2010; Chan and Wu 2011; Becker, DeFond, Jiambalvo, and Subramanyam 1998; Francis, Maydew, and Sparks 1999; Kim, Chung, and Firth 2003; Ball, Jayaraman, Shivakumar 2012). For these reasons we hypothesize: H3: The effect of income-increasing discontinued operations on lowering firms ETR on income from continuing operations will be weaker when the firm has a Big N auditor. Financial reporting rules for discontinued operations changed with fiscal years beginning after December 15, Prior to the change in accounting rules, the reporting of discontinued operations was based on APB Opinion No. 30. Under this opinion, firms were allowed to classify an asset disposal as discontinued operations only if the disposition was a business segment (e.g. major lines or business, customer class) (APB 1973). With the implementation of SFAS No. 144 (FAS 144) for fiscal years beginning after December 15, 2001, the FASB modified the threshold for recognizing an asset disposal as discontinued operations to disposals of a component of an entity. Under FAS 144, components of an entity include such disposals as an asset group, an operating segment, a reportable segment, a reporting unit, or a subsidiary. Barua et al. (2010) contends that the change to FAS 144 will generally increase managers ability to classify an asset disposal as a discontinued operation as well as their ability to engage in classification shifting when reporting discontinued operations. However, their results indicate that, post-fas 144, the magnitude of classification shifting of core expenses into income-decreasing discontinued operations, while still statistically significant, decreased. They attribute this unexpected decrease to a decline in the average size of discontinued operations post-fas 144 (arguing that smaller discontinued operations provide less opportunity to shift expenses into). Overall, it is unclear whether 15

16 FAS 144 will change managers ability to engage in biased intraperiod tax allocations. Consequently, we propose the following research question. RQ1: Will the effect of income-increasing discontinued operations on lowering firms ETR on income from continuing operations for discontinued operations firm years differ after the implementation of SFAS No. 144? Methodology We test our hypotheses by comparing the ETR on continuing operations during periods of discontinued operations to ETRs during periods immediately preceding and immediately following periods of discontinued operations. Only firms that report discontinued operations at some point in time are included in our study. These firms are allowed to have one, two, or three consecutive years of discontinued operations (i.e. the string of discontinued operations), as shorter strings likely relate to an individual discontinued operation. We require that every recognized discontinued operation within the string must be of the same character to retain consistent incentives and opportunities across the string (i.e. all string observations must be either income-increasing or incomedecreasing). Further, firms with a string of discontinued operations must have an equal number of observable periods preceding (i.e the pre period) and following (i.e. the post period) a period of discontinued operations, to provide consistent on average ETRs in both periods. We allow these pre and post periods to range form one to three consecutive years (i.e. the pre and post strings). Each firm is allowed to have multiple strings of discontinued 16

17 operations, so long as each unique string has a pre and post string that does not overlap with a different string of discontinued operations. 13 We employ the following base model, or some derivation of this model, to test our hypotheses:!"#!" =!! +!!!"_!"#!" +!!!"_!"#!" +!(!"#$%"&') +! (1) The dependent variable is firm i's GAAP ETR in year t. The variable of interest is DO_POS. Following Barua et al. (2010), DO_POS is the amount of income-increasing discontinued operations, scaled by sales, that firm i recognizes in year t. A negative coefficient on!! suggests that GAAP ETRs are lower during periods of income-increasing discontinued operations, as compared to periods immediately preceding and following a string of discontinued operations. DO_NEG captures the effect on income-decreasing discontinued operations on a firm s ETR. We expect!! will be insignificant. The vector CONTROLS contains common control variables from prior tax research, including firm size (SIZE), profitability (ROA), debt (LEV), gross property, plant and equipment (PPE), gross intangibles (INTAN), an indicator for research and development expenditures (RD_IND), advertising expense (AD), an indicator for net operating loss carryforwards (NOL_IND), foreign income (FOR), and special items (SPI). We include period and two-digit SIC industry fixed-effects in all specifications. We also cluster standard errors by firm. Appendix A provides detail regarding how each variable is calculated. 13 Thus, firms can have more than one unique discontinued operation string in our sample. Each discontinued operation string must have usable preceding period and post period strings. It is therefore possible that a post period year observation for one discontinued operation string is also a preceding period year observation for a separate discontinued operation string, or vice versa. In such instances, that unique firm-year observation is only included once in our sample. 17

18 Data We begin with all Compustat firms from 1992 through We restrict the sample to begin in 1992 as that was the first year Statement of Financial Accounting Standards 109 Accounting for Income Taxes (SFAS No. 109) was implemented. Of the 217,058 firm-year observations with non-missing assets collected during this time period, 22,232 report discontinued operations, roughly 10.2% of the total observations (a comparable percentage to Barua et al. 2010). Thus, discontinued operations aren t an entirely uncommon event. Of these 22,232 firm-year observations, 10,330 are unique firmperiods of discontinued operations (that is, firm-years with consecutive years of recognizing discontinued operations, henceforth denoted as a discontinued operation string), suggesting that, on average, a firm reporting discontinued operations will do so over 2.15 consecutive years (untabulated). There are 7,194 unique firms within this full Compustat sample; thus, it does not appear uncommon that firms report separate strings of discontinued operations within a lifetime. In addition to our SFAS No. 109 limitation, we restrict our sample in several ways. Firms must be incorporated in the United States (FIC = US). Following McVay (2006) and Barua et al. (2010), firm-year observations must have total sales (SALE) of at least 1 million, since several variables are scaled by sales. Similarly, we require firm-year observations to have total assets (AT) of at least 1 million. In addition, we exclude firms in utility industries (SICs ), because Compustat does not collect advertising expenditures for these firms. Consistent with other tax research, and to ensure the ETR can be meaningfully interpreted, we require firm-year observations to have positive, pre-tax income from continuing operations (PI). We also require tax expense (TXT) to be non- 18

19 negative and less than pre-tax income from continuing operations (i.e. the firm must have an ETR of less than 100%). Continuous variables are winsorized at 1% and 99%. Table 1 depicts time-series trends of discontinued operations. Panel A includes all Compustat observations from 1992 through 2013 that meet our control variable restrictions. Panel B includes our final sample with all variable and discontinued operation string restrictions. We separate each panel into income-increasing and income-decreasing discontinued operations observations. In both panels, the mean scaled magnitude of discontinued operations (discontinued operations (DO) / total sales (SALE)) is statistically smaller post-sfas No. 144, consistent with SFAS No. 144 lowering the threshold for designation of a discontinued operation (p-values <0.01). Additionally, it appears that over time firms are recognizing more income-increasing discontinued operations (as compared to income-decreasing discontinued operations). Panel A of Table 2 provides descriptive statistics for all firm-year observations in our sample set. The average and median ETRs are 33.68% and 35.64%, respectively, both relatively close to the top federal statutory corporate tax rate of 35%. Panel B of Table 2 presents descriptive statistics for only those firm-years that recognize discontinued operations, split between income-increasing and income-decreasing discontinued operations. Firm-years with income-increasing discontinued operations have, on average, lower ETRs (at 32.04%) than firm-years with income-decreasing discontinued operations (at 35.26%), which are statistically different at the 1% level (t-stat of 5.42). Further analysis suggests that firm-years with income-increasing discontinued operations tend to be larger, have more intangible assets, spend more on R&D, have higher foreign income, and have more negative special items. 19

20 Table 3 includes the Pearson correlation matrix for the variables included in Equation 1. DO_POS is negatively related to ETR, consistent with our intuition in H1. However, DO_NEG is positively related to ETR. This result indicates splitting our main variable of interest into income-increasing or income-decreasing discontinued operations is important. Empirical Results H1: Do ETRs change during periods of discontinued operations? Our first hypothesis (H1) predicts a negative (null) relation between the ETR and income-increasing (income-decreasing) discontinued operations. We employ Equation 1 to test our H1 predictions, the results of which are presented in Table 4. Panel A of Table 4 presents the results when including firm-year observations from before, during, and after the discontinued operations period in the same regression. Column 1 includes only those firm-year observations that are connected to a string of income-increasing discontinued operations (DO_POS). Results are consistent with the ETR being lower during periods of income-increasing discontinued operations. Column 2 includes only those firm-year observations that are connected to a string of income-decreasing discontinued operations (DO_NEG). We do not find evidence that the ETR is lower during periods of incomedecreasing discontinued operations. Column 3 includes all firm-year observations (those that are connected with either a string of income-increasing discontinued operations or income-decreasing discontinued operations), with results consistent with Columns 1 and 2. Specifically, after controlling for factors known to impact the ETR, the results in Column 3 suggest the ETR on continuing operations is significantly lower during periods when the 20

21 firm is reporting discontinued operations, when compared to the periods preceding and following the discontinued operation. 14 If the results in Panel A are due to classification shifting of tax expense between income from continuing operations and discontinued operations, we would expect to the see the ETR change between the pre period and the period of discontinued operations, and subsequently see the ETR revert back between the period of discontinued operations and the post period. To analyze this, we split our sample set into two groups and rerun Equation 1 with our split sample sets. The results are presented in Panel B of Table 4. Column 1 of Panel B depicts the results from our sample set that includes only periods of discontinued operations and the pre periods, while Column 2 depicts the results from our sample set that includes only periods of discontinued operations and the post periods. A negative coefficient on DO_POS in Column 1 of Panel B would suggest that the ETR is lower during periods of discontinued operations as compared to the pre periods, while a negative coefficient on either DO_POS in Column 2 would suggest that the ETR is lower during periods of discontinued operations as compared to the post periods. Consistent coefficients in both columns would provide evidence that the ETR shifts between pre period to the discontinued operations periods, and reverts back between the discontinued operations period and post period. The results in Panel B of Table 4 are consistent with our H1 predictions. Specifically, the coefficient of on DO_POS in Column 1 suggests that a one standard deviation increase in income-increasing discontinued operations results in a reduction to the ETR of 69, while the coefficient of on DO_POS in Column 2 14 We compute this by taking the coefficient on DO_POS ( ) and multiplying it by the standard deviation in DO_POS from Panel A of Table 2 (0.0188). 21

22 suggests that a one standard deviation increase in income-increasing discontinued operations results in a reversion to the ETR of 63 basis points. 15 Together, these results suggest that the ETR decreases during periods of income-increasing discontinued operations (when the opportunity to engage in classification shifting is present) and then subsequently increases after periods of income-increasing discontinued operations (once the opportunity to engage in classification shifting is removed). As predicted, we do not find result for DO_NEG, suggesting that the ETR is not statistically different during periods of income-decreasing discontinued operations as compared to the pre and post periods. H2: The influence of the incentive to meet analyst earnings forecasts. Our second hypothesis (H2) predicts that the relationship between taxes on income from continuing operations and discontinued operations is stronger when the firm faces an incentive to manage after-tax income from continuing operations, namely when the firm faces pressure to meet analyst forecasts. To test this prediction, we augment Equation 1 as follows:!"#!" =!! +!!!"_!"#!" +!!!"_!"#!" +!!!"#!" +!!!"_!"#!" (2)!"#!" +!!!"_!"#!"!"#!" +!(!"#$%"&') +! The variable MJB is an indicator equal to 1 when the firm meets or just beats (within $0.01) the median analyst forecast, zero otherwise. We then interact MJB with DO_POS and DO_NEG. Given our results from Equation 1, that ETRs are lower during periods of 15 The basis point computation is computed similar to that in Panel A. We use the standard deviation of each separate dataset in our calculations. Specifically, the standard deviation in the untabulated dataset including only discontinued operations and the pre periods is , while the standard deviation in the untabulated dataset including only discontinued operations and the post periods is also

23 income-increasing discontinued operations, a negative and significant coefficient on DO_POS*MJB in Equation 2 would suggest that ETRs are managed even more during periods of income-increasing discontinued operations when firms meet or just beat analyst forecasts. The results of Equation 2 are presented in Table 5. We restrict our sample to periods of discontinued operations that only span one year, to isolate those single instances where firms have a specific opportunity (i.e. discontinued operations) and specific incentive (i.e. analyst forecast) to manage earnings, relative to the surrounding years where the opportunity does not exist. We further allow only one pre and one post period for each firm-period string, although in unreported robustness tests we obtain quantitatively similar results to those discussed below when we relax this requirement. We obtain analyst forecasts from I/B/E/S. These restrictions reduce our sample size for Equation 2 to 1,424 firm-year observations. Consistent with our predictions, and with the H1 results, the coefficient on DO_POS*MJB is negative and statistically significant. Per our hypothesis, this suggests that the presence of an incentive to meet an earnings threshold (i.e. an analyst forecast) coupled with a means to meet that target (i.e. separately stated discontinued operations) results in a shift of tax expense away from income from continuing operations and into income-increasing discontinued operations. The results for DO_NEG*MJB are not significant at any conventional level. This result is unsurprising, since we argue that the presence of income-decreasing discontinued operations don t provide the same opportunity to shift taxes as income-increasing discontinued operations do. In summary, we find support for H2 when firms recognize income-increasing discontinued operations. 23

24 H3: The influence of external monitoring. Our third hypothesis predicts the relation between taxes on income from continuing operations and discontinued operations is reduced when the firm receives a higher quality audit. We argue an audit will be higher quality if the firm uses a Big N external auditor (H3). To investigate these hypotheses, we augment Equation 1 as follows:!"#!" = α0#+#β1do_posit#+#β2do_negit#+#β3bignit#+# (3) β4#do_posit*bignit+β5do_negit*bignit#+#γ(controls)#+#ε# When testing H3, BIGN is defined as an indicator equal to 1 when firm i uses a BIG N external auditor in time t, zero otherwise. Given our results from Equation 1, that ETRs are lower during periods of income-increasing discontinued operations, a positive coefficient on DO_POS*BIGN would suggest that monitoring from the external auditor attenuates the negative main result finding. The results of Equation 3 are provided in Table 6. Consistent with our prediction, the coefficient on DO_POS * BIGN is positive and statistically significant. Specifically, the positive coefficient on DO_POS * BIGN suggests that the use of a Big N auditor attenuates the negative relation between taxes on income from continuing operations and income-increasing discontinued operations. Similar to our prior results, we find no statistically significant interactions when firms recognized income-decreasing discontinued operations (DO_NEG). 24

25 RQ1: The influence of the implementation of SFAS No Where H2 looks to increased incentives to engage in classification shifting of tax expense, and H3 looks at the attenuation of classification shifting due to external monitoring, our research question looks to increased opportunity. Specifically, we look to whether that the implementation of SFAS No. 144 increased the magnitude of classification shifting of tax expense, due to more firms separately recognizing discontinued operations. To investigate this, we augment Equation 1 as follows:!"#!" =!! +!!!"_!"#!" +!!!"_!"#!" +!!!"#144!" + (4)!!!"_!"#!"!"#144!" +!!!!_!"#!"!"#144!" +!(!"#$%"&') +! The variable FAS144 is an indicator equal to one if the fiscal year begins after December 15, 2001 (the implementation date of SFAS No. 144), zero otherwise. We then interact FAS144 with DO_POS and DO_NEG. Given our results for H1, a negative and significant coefficient on DO_POS*FAS144 would suggest that SFAS No. 144 had an incremental effect on the opportunity for firms to engage in classification shifting of tax expense via income-increasing discontinued operations. Table 7 presents the results of Equation 4. We return to our main sample of 4,847 firm-year observations for this regression. While our main variable of interest, DO_POS, remains negative and statistically significant, the interaction of DO_POS * POST144 is statistically insignificant. Therefore, we find no evidence that FAS144 either increased or decreased the relation between discontinued operations and ETRs. 25

26 Conclusion This study investigates whether firms use discontinued operations to increase aftertax income from continuing operations by shifting tax expense out of income from continuing operations and into discontinued operations. Analyzing the ETR in the period preceding the reporting of discontinued operations, in the period of discontinued operations, and in the period following the reporting of discontinued operations, we provide evidence that, on average, the ETR is lower during the period of reporting incomeincreasing discontinued operations, but no such statistically significant relation exists with income-decreasing discontinued operations. We interpret this as evidence that, on average, firms with income-increasing discontinued operations shift tax expense out of income from continuing operations. This result complements the results of Barua et al. (2010), who show that firms engage in core expense classification shifting by using income-decreasing discontinued operations. We also investigate whether incentives to manage after-tax income from continuing operations magnify this relation. We present evidence that the decline in ETR during the period of income-increasing and discontinued operations is greater for those firms that met or just beat the median analyst forecast. We interpret this as evidence that, for those firms with a greater incentive to manage earnings, engage in larger classification shifting of tax expense out of income from continuing operations. We also provide evidence that classification shifting through the tax expense declines when firms have higher quality audits. Finally, we do not find that tax expense classification shifting changed after implementation of SFAS No These results highlight an under-investigated method of 26