Discontinued Operations: Earnings Management, Value Relevance, and the Role of ASU

Transcription

1 Discontinued Operations: Earnings Management, Value Relevance, and the Role of ASU Yuan Ji The Hong Kong Polytechnic University James Potepa The George Washington University Oded Rozenbaum The George Washington University Initial Draft: February 7, 2018 Abstract Accounting regulations require firms to separately disclose the profits and losses from portions of the company that will be or already have been eliminated, as they should have no impact on future performance. These discontinued operations are, not surprisingly, typically excluded from the definition of income used by investors, analysts, and other users of the financial statements. However, a prior study shows that this accounting treatment leads managers to shift core expenses into discontinued operations. In light of recent changes in the regulations pertaining to discontinued operations, we re-examine this relationship and discover significant changes have occurred. The new rules, which impose tighter restrictions on certain aspects of what can be considered discontinued and loosen other requirements, eliminate any significant evidence of earnings management. In addition, there is significant evidence the new rules reduce the frequency, persistence, and predictive ability of discontinued operations. Finally, the stock market starts to react to discontinued operations after this new regulation. We greatly appreciate the insights and ideas from seminar participants at Georgetown University. We would like to thank The Hong Kong Polytechnic University and The George Washington University for their financial support. All errors are our own.

2 1. Introduction Accountants attempt to classify the income statement depending on the nature and expected persistence of different types of incomes and expenses. This allows firms to distinguish between items the will recur and those that will not. Among those non-recurring items are discontinued operations, which represent the income statement impact of the components or segments of a firm that a company is attempting to sell or has sold. Unlike any other component of earnings, discontinued operations are displayed net of taxes as a line item at the very bottom of the income statement. 1 This treatment is designed to allow the users of the financial statements to easily disregard such gains and losses when predicting future cash flows, under the assumption they will not persist given that portion of the firm will be or has already been eliminated. The risk associated with allowing this treatment is that firms may shift core expenses into discontinued operations. This manipulation would be advantageous to companies because it inflates earnings before discontinued operations by making discontinued operation more negative, which most users of the financial statements typically exclude from their definition of profit. In fact, one existing paper by Barua, Lin, and Sbaraglia (2010) documents that firms regularly engage in this behavior to meet or beat earnings expectations. However, recent regulations have dramatically changed what can be considered a discontinued operation and thus the call into question whether our current understanding of discontinued operations remains accurate. New regulations have drastically impacted the reporting of discontinued operations and thus compel us to re-examine our assumptions. The Financial Accounting Standards Board introduced Accounting Standards Update (henceforth ASU ), effective fiscal 1 Extraordinary items were treated in a similar manner, but they were extremely rare and small in magnitude for at least the past decade before being completely eliminated by the Financial Accounting Standards Board in 2015 (Accountings Standards Updated ). 1

3 year 2014, to adjust the rules for recognizing discontinued operations set forth in Statement of Financial Accounting Standards 144 (henceforth SFAS 144). The most notable change is that discontinued operations must now represent a strategic shift that has (or will have) a major effect on an entity s operations and financial results (ASU ). Under the prior set of rules, a reporting segment, operating segment, a reporting unit, a subsidiary, or an asset group were all eligible to be classified as discontinued. While this change may appear to make it more difficult to categorize a component as discontinued, ASU also eliminated the requirement that the discontinued component will no longer be associated with the firm after the disposal. For a component to be classified as discontinued under the old regulations, it must have been that the operations and cash flows of the component have been (or will be) eliminated from ongoing operations of the entity as a result of the disposal transaction and the entity will not have any significant continuing involvement in the operations of the component after the disposal transaction (SFAS 144). Firms are also required to provide more detail about discontinued operations in the footnotes, which was not mandatory under SFAS 144. For a table summarizing the regulatory changes, please see Appendix C. The impact of such changes on the ability to manipulate discontinued operations as well as its other characteristics is not entirely clear. The higher threshold of a strategic shift may limit a firm s opportunity to shift core expenses into discontinued operations, both because of increased standards and the reduction in the number of opportunities for firms to claim they have discontinued operations. On the other hand, the users of financial statements may not be as open to excluding an item from their definition of profit when it may still have a financial impact on the firm after it is disposed. Moreover, loosing the constraint may allow for an increase in the frequency of discontinued operations if this was a binding constraint. The implications for the 2

4 predictive ability and market reaction to discontinued operations are similarly unclear. The increased threshold, a strategic shift, to classify a component of a firm as discontinued may increase the informativeness of and market reaction to the recognition of this item by eliminating ones that were small in magnitude or represent repeated sales under the old rules. Alternatively, the ability to classify portions of the firm that will still be economically and operationally linked after they are sold may reduce the value of separating discontinued operations out given this item will likely continue to impact earnings. Therefore, it is not clear whether the requirement that a discontinuation represents a strategic shift will outweigh the removal of the rule that a firm must sever all ties with the component that has been or will be sold. It is also unclear whether a strategic shift represents a binding constraint at all. Similarly, it is unclear whether the enhanced mandatory footnote disclosure rules will be given firms may voluntarily disclose this already. We find that ASU has significantly curtailed the use of discontinued operations to manage earnings and impacted many characteristics of this item as well. Our analysis reveals that the shifting of core expenses into discontinued operations to manage earnings has significantly declined. Specifically, we re-examine the work of Barua et al. (2010). This prior paper, the only existing study focusing on discontinued operations, finds that firms shift core expenses into discontinued operations to meet or beat common earnings benchmarks, including avoiding losses, preventing a decline in earnings, and meeting analyst s expectations. Given proforma earnings and analysts almost universally exclude discontinued operations, this method would allow managers to have core expenses excluded and thus raise their firm s earnings numbers. We confirm that there continues to be evidence of this type of classification shifting earnings management after the sample period covered in their paper, but that this disappears after the introduction of ASU

5 We then examine the impact of this regulatory change on the frequency of reporting discontinued operations as well as the item s persistence, value relevance, and stock market impact. We find that the frequency of reporting discontinued operation has dropped by more than a quarter after the introduction of these rules, from about 16% to 12% in a given year. We then provide evidence about significant changes in the persistence and predictive ability of this item. Discontinued operations significantly predicted future earnings prior to the rule change, but this is no longer the case. The persistence of this discontinued operations, which should be not be recurring by definition, declines significantly after the regulatory change as well. In addition, our results indicate that discontinued operations are now informative to the market. Under the prior regime, there was no market reaction to discontinued operations. The reasons for this change are unclear. It could be that investors are more likely to trust a less manipulated item, or that there is more disclosure required under the new rule. Furthermore, it is possible that the stringent rule for categorizing something as discontinued eliminates repeated, expected items and thus makes the announcement more informative. The characteristics and potential manipulation of discontinued operations have only been studied under two prior sets of regulations and rarely in sufficient detail. Barua, Lin, and Sbaraglia (2010) does take an in-depth look by showing that managers manipulate earnings by shifting core expenses into discontinued operations. That paper finds that this classification shifting method allows firms to meet or beat common earnings benchmarks. Their analysis investigates discontinued operations under two sets of regulations, but significant changes have occurred since it was published, which compel us to revisit their findings. Furthermore, the existing literature contains little evidence on the characteristics and usefulness of the information contained in discontinued operations and the evidence indicates that they are appropriately 4

6 ignored. One paper using a sample period that ended almost three decades ago shows that the discontinued operations component of earnings is an insignificant predictor of future profits (Fairfield, Sweeny, and Yohn 1996). Similarly, there is little evidence that discontinued operations impact executive compensation. Gaver and Gaver (1996) show that compensation committees generally exclude the impact of discontinued operations unless they are income increasing. However, a more recent paper re-examining this topic concludes that even profitable discontinued operations are no longer associated with CEO pay (Potepa 2018). The scant evidence using old data may not be reason enough to reevaluate our understanding of discontinued operations, but there have been significant changes in the rules for determining what qualifies as discontinued. We extent Barua et al. s (2010) finding by also providing evidence on the characteristics of discontinued operations, long ignored by prior literature, both before and after the most recent rule change. Furthermore, our study informs regulators on the effectiveness of ASU The remainder of the paper is organized as follows. The second section motivates the paper and develops our hypotheses. This is followed by our sample selection process and research design. The results of our analysis are explained in the fourth section. Finally, we conclude in the fifth section. 2. Motivation and hypotheses development Investors weigh core earnings more heavily compared to transitory earnings when valuing firms since core earnings are more persistent, and therefore more informative about the future (Bradshaw and Sloan 2002; Lipe 1986). Given the increased weight on core earnings in firm valuation, companies can manage earnings upward by shifting expenses from core earnings to line items further down the income statement. Managing earnings by shifting revenues and 5

7 expenses between income statement line items is advantageous compared to alternative earnings management methods, such as real or accrual based earnings management for at least two reasons. First, shifting revenues and expenses does not reverse like accrual-based earnings management and does not have an impact on real activity like real earnings management. Second, since net income does not change as a result of income shifting between line items on the income statement, this form of earnings management is likely to be less scrutinized by auditors and regulators. Consistent with the benefits of classification shifting, Ronen and Sadan (1975) find that firms use extraordinary items to smooth earnings before extraordinary items. In a similar vein, Barnea, Ronen, and Sadan (1976) find that firms use extraordinary items to smooth operating income. Kinney and Trezevant (1997) investigate how firms use special items to affect investor perceptions and find that while income decreasing special items are more typically disclosed as a separate line item to highlight their transitory nature, income increasing special items are usually discussed in the footnotes. However, Riedl and Srinivasan (2010) find that special items that are disclosed as separate items on the income statement are less persistent than those that are disclosed in the footnotes, which suggests that firms use these disclosure practices to inform rather than deceive investors. Finally, McVay (2006) and Fan, Barua, Cready, and Thomas (2010) find that firms shift income decreasing items from core earnings to special items. 2 The study that is most closely related to ours is Barua, Lin, and Sbaraglia (2010). Barua et al. (2010) investigates the shifting of income decreasing items into discontinued operations. Earnings management by shifting income decreasing items into discontinued operations warrants 2 While classification shifting for special items has been significantly discussed in the literature, it may be a less effective approach to classification shifting compared to discontinued operations in the sense investors, analysts, compensation committees, and other financial statement users do not automatically exclude special items from their definition of earnings. The unique below the line, after tax disclosure of discontinued operations and the consistent exclusion of this item from users of the financial statements makes it an ideal place to hide core expenses. 6

8 a specific investigation for multiple reasons. First, discontinued operations are presented in the last line of the income statement. Therefore, shifting operating expenses into discontinued operations will increase operating income, core earnings, and income from continued operations. Second, there are limited disclosure requirements for discontinued operations, which provide fertile ground to masking the shifting of operating expenses into discontinued operations. Third, ASU drastically changes the rules of measurement and recognition of discontinued operations. The change warrants a reinvestigation of whether firms manage earnings by shifting expenses to discretionary expenses because it changes firms opportunity to shift expenses into discontinued operations and serves as an exogenous shock that can improve the identification of the research design. The Financial Accounting Standards Board introduced ASU , effective fiscal year 2014, to adjust the rules for recognizing discontinued operations set forth in SFAS 144. The most notable change is that discontinued operations must now represent a strategic shift that has (or will have) a major effect on an entity s operations and financial results (ASU ). Under the prior set of rules, a reporting segment, operating segment, reporting unit, subsidiary, or asset group were all eligible to be classified as discontinued. In addition, ASU eliminated the requirement that the discontinued component will no longer be associated with the firm after the disposal. For a component to be classified as discontinued under the old regulations, it must have been that the operations and cash flows of the component have been (or will be) eliminated from ongoing operations of the entity as a result of the disposal transaction and the entity will not have any significant continuing involvement in the operations of the component after the disposal transaction (SFAS 144). 7

9 It is unclear whether ASU will limit firms ability to shift expenses from core earnings to discontinued operations. On the one hand, the new rule sets a more stringent threshold, which requires discontinued operations to represent a material strategic shift. This change can make it harder for firms to shift core expenses to discontinued operations. On the other hand, ASU eliminated the requirement that the discontinued component will no longer be associated with the firm after the disposal. This change can potentially allow for more costs to be recorded as discontinued operations. Furthermore, it is unclear whether the new requirement to limit discontinued operations to material strategic shifts is binding. It is possible that firms will be able to define expenses as a strategic shift and shift those expenses to discontinued operations. Under the prior set of rules, a reporting segment, operating segment, a reporting unit, a subsidiary, or an asset group were all eligible to be classified as discontinued operations. Under ASU , it is possible that firms may claim that these are all significant strategic shifts. For example, Lee (2014) finds that more stringent restrictions on restructuring charges imposed by SFAS 146 lowered, but did not eliminate, expense shifting into restructuring charges. Based on the above arguments, our first hypothesis is stated in the null form: Hypothesis 1: The association between unexplained core earnings and negative discontinued operations does not change following ASU The persistence of discontinued operations may also change following ASU Prior to ASU , firms could include a component of an entity in its discontinued operations. This definition allows firms to define as discontinued operations a component that has independent cash flows and may allow more persistence in recognizing discontinued operations. However, since the firm can still have remaining interests in the discontinued operations, it is possible that the persistence of discontinued operations will increase, or even reverse if more transactions can be classified as discontinued operations. Lastly, as we note 8

10 earlier, it is possible that the new rules requiring a material strategic shift are not binding, and therefore the persistence of discontinued operations will remain unchanged following ASU This leads to our second hypothesis, also stated in the null form: Hypothesis 2: The persistence of discontinued operations does not change following ASU Earnings management is a tool to mislead investors by portraying a false picture, which investors then use to project the future. Therefore, a key feature of earnings management is that its performance does not persist. Accrual-based earnings management, by definition, must reverse (Jones 1991; Cohen and Zarowin 2010). Real-activity earnings management also cannot persist. For example, if firms artificially lower R&D expense to meet an earnings benchmark, R&D expense should pick up in the future, otherwise the decrease in R&D spending may be a legitimate response to adverse economic changes (Ji and Rozenbaum 2017). Therefore, following McVay (2006) and Barua et al. (2010), we posit that to the extent that firms shift core expenses to discontinued operations, we should observe a negative association between negative discontinued operations and future pre-discontinued operations income. If ASU limits firms ability to shift core expenses to discontinued operations, we expect to find the negative association between negative discontinued operations and future pre-discontinued operations income is attenuated, leading to our third non-directional hypothesis stated in the null form: Hypothesis 3: The negative association between negative discontinued operations and future pre-discontinued operations income does not change following ASU Lastly, we examine how the market reacts to the news of discontinued operations at earnings announcement. If investors believe that the discontinued operations are expected to persist in the future, the market reaction should be stronger compared to expectations of a transitory event. In addition, the change in the market response to discontinued operations at 9

11 earnings announcements is ambiguous. If ASU restricts discontinued operations to significant strategic shifts, then the frequency of discontinued operations is likely to decrease and not persist into the future. Therefore, the market reaction to discontinued operations may be lowered. However, ASU also allows firms to recognize discontinued operations even if they will continue to maintain an interest in the continued operations. In that case, discontinued operations may contain information about the future and the market reaction will be stronger after ASU Therefore, our last hypothesis stated in the null form is: Hypothesis 4: The market reaction to discontinued operations at earnings announcement remains unchanged following ASU Sample selection and research design 3.1 Sample selection Our interest in examining this regulatory change leads us to examine the full Compustat universe of firms around the implementation of this rule. We begin with all the firms in the Compustat database between 2012 and We exclude all observations during 2014, the year during which the regulations changed. This ensures that our current period lagged variables are measured during the same set of regulations. This leads to an initial sample of 23,054 firm-year observations. We require firms to have sufficient data to run unexpected earnings regression as shown later in equations (1), which reduces our sample to 15,442 firm year observations. We then combine this with returns data from CRSP and analyst data from IBES. Certain tests requiring information about analyst expectations and lagged variables data will reduce the size of the sample. All variables are winsorized at the 1 st and 99 th percentiles. 10

12 3.2 Research design Testing Hypothesis 1 Our first hypothesis tests the relation between unexpected core earnings and negative discontinued operation following ASU Generally, our research design follows closely the one implemented by Barua et al. (2010). This approach builds on McVay (2006) to measure unexpected core earnings and unexpected change in core earnings. We use the following models to estimate unexpected core earnings and unexpected change in core earnings: CE t = α 0 + α 1 CE t-1 + α 2 ATO t + α 3 Accruals t-1 + α 4 Accruals t + α 5 Sales t + α 6 Neg_ Sales t + ε t, (1) CE t = β 0 + β 1 CE t-1 + β 2 CE t-1 + β 3 ATO t + β 4 Accruals t-1 + β 5 Accruals t + β 6 Sales t + β 7 Neg_ Sales t + µ t, (2) where CE is core earnings, measured as sales (REVT) minus cost of goods sold (COGS) and selling, general and administrative expenses (XSGA), scaled by sales (REVT). ATO is asset turnover ratio, defined as sales (REVT) scaled by the average of net operating assets at year t and year t-1 (NOA t + NOA t-1 )/2. Accruals is operating accruals, defined as net income before extraordinary (NI-XI) minus cash from operations (OANCF), scaled by sales (REVT). Sales t is percentage change in sales, computed as sales at year t minus sales at year t-1, deflated by sales at year t. Neg_ Sales t is equal to 1 if Sales t is less than zero, zero otherwise. We estimate Eq. (1) and (2) by industry and year to generate the expected value of core earnings and the expected change in core earnings for firm i in year t. For years when there is a discontinued operations, we replace the variables with ones from Compustat s restated file, if available. This ensures the results are not driven mechanically by one period excluding those disconitnued operation when prior periods include that part of the firm. The unexpected core earnings and unexpected change in core earnings are the difference between the actual value and 11

13 the predicted value of Eq. (1) and (2), respectively. Detailed variable definitions are provided in Appendix A. Barua et al. (2010) document that classification shifting takes place when firms have negative discontinued operations. Building on this result, we further explore the association between unexpected core earnings and negative discontinued operations after ASU (Hypothesis 2). Following Barua et al. (2010), we use the following ordinary least square (OLS) regressions to test H1, which examines whether classification shifting using negative discontinued operation (%DO_NEG) is used to meet or beat earnings benchmarks after the rule change. We further modify their model by including an indicator POST to measure the post rule effect and the variable of interest is the three-way interaction between %DO_NEG, MEET&BEAT and POST. UE_CE t = µ 0 + µ 1 %DO_NEG t + µ 2 MEET&BEAT t + µ 3 %DO_NEG * MEET&BEAT t + µ 4 %DO_NEG * MEET&BEAT*POST t + µ 5 POST t + µ 6 SIZE t + µ 7 BM t + µ 8 ACCRUALS t + µ 9 OCF t + µ 10 ROA t + ε t, (3a) UE_ CE t+1 = π 0 + π 1 %DO_NEG t + π 2 MEET&BEAT t + π 3 %DO_NEG * MEET&BEAT t + π 4 %DO_NEG * MEET&BEAT*POST t + π 5 POST t + π 6 SIZE t + π 7 %DO_NEG t+1 + π 8 BM t + π 9 ACCRUALS t + π 10 OCF t + π 11 ROA t + ƞ t+1, (3b) where UE_CE t is unexpected core earnings in year t. %DO_NEG t is the percentage of discontinued operations when discontinued operations are negative, and zero otherwise. UE_ CE t+1 is the unexpected change in core earnings in year t+1. We define MEET&BEAT t in three different ways. In the first specification, MEET&BEAT t is equal to one if firm i's earnings per share is greater than or equal to zero, and 0 otherwise. In the second specification, MEET&BEAT t is equal to 1 if the change in firm i's earnings per share is greater than or equal to zero, and 0 otherwise. In our last specification, MEET&BEAT t is equal to 1 if firm i has positive earnings surprise, defined as actual earnings per share minus the latest median consensus analyst 12

14 forecast, and zero otherwise (Matsumoto, 2002). POST t equals to one after the year 2014, and 0 otherwise. To control for income decreasing discontinued operations at year t+1, we include %DO_NEG t+1 in Eq. (3b). We also control for firm-level characteristics that influence unexpected core earnings, including book-to-market ratio (BM t ), firm size (SIZE t ), accruals(accruals t ) operating cash flow (OCF t ) and return on assets (ROA t ). Lastly, we consider industry fixed effects to control for unobserved industry variations in the panel data, and report standard errors clustered by firm Testing Hypothesis 2 The persistence of discontinued operations may also change after the rule. Our Hypothesis 2 examines the influence of ASU on the persistence of discontinued operation accounts. We employ the following ordinary least square (OLS) regression: %DO t+1 = Ɵ 0 + Ɵ 1 %DO + Ɵ 2 %DO * POST + Ɵ 3 POST + υ t+1, (4) where the dependent variable %DO t+1 is discontinued operation of firm i at year t+1. The variable of interest is Ɵ 2. A negative (positive) Ɵ 2 indicates that the persistence decreases (increases) post ASU All the other variables are as previously defined. If firms manage discontinued operations to shift core expenses to discontinued operation then we would expect a different effect of the new rule on positive and negative discontinued operations. Specifically, we posit that the rule affects the persistence of positive and negative discontinued operations differently. Therefore we separately examine the persistence of negative (positive) discontinued operations using Eq. (4) Testing Hypothesis 3 We further explore the predictive ability of discontinued operations before and after the rule change. A key feature of earnings management is that its performance does not persist. 13

15 Therefore, negative discontinued operations should be negatively associated with future prediscontinued operations income if companies shift core expenses to discontinued operations. In other words, discontinued operations were able to predict future earnings prior to the rule change. If ASU restrains firms ability to shift core expenses to discontinued operations, the predictability of negative discontinued operations is expected to decrease significantly (Hypothesis 3). We use the following model specification to test our conjectures: PRE_DO_INCOME t+1 = ɸ 0 + ɸ 1 PRE_DO_INCOME t + ɸ 2 %DO t + ɸ 3 %DO * POST t + ɸ 4 POST t + Ʊ t+1, (5) where the dependent variable PRE_DO_INCOME is net income before discontinued operations (NI-DO), scaled by sales (REVT) of firm i at year t+1. PRE_DO_INCOME is the same variable at year t. All the other variables are as previously defined. Similar to testing Hypothesis 2, we separately examine negative and positive discontinued operations Testing Hypothesis 4 Our final hypothesis tests how the market reacts to discontinued operations at earnings announcements before and after ASU Ex-ante, it is not clear the change in the market response to discontinued operations at earnings announcements. If negative discontinued operations are expected to persist into the future, then the market reaction is expected to be negative. However, even if discontinued operations are transitory, the market reaction should be negative, but at a lower magnitude, because of the costs associated with discontinuing operations. On the other hand, if the company is discontinuing non-profitable operations, then the market reaction may be positive if investors believe that future profitability will increase. To summarize a stronger negative market reaction is consistent with expense shifting from core earnings to discontinued operations. We employ the following model: 14

16 CAR i,t = γ 0 + γ 1 SUR i,t + γ 2 DO i,t + γ 3 DO_NEG i,t + γ 4 POST i,t + γ 5 DO_NEG *POST +γ 6 DO*NEG_DO i,t + γ 7 DO* DO_NEG *POST i,t + γ 8 SIZE i,t + γ 9 BM i,t + χ i,t (6) where CAR is the cumulative market-adjusted abnormal returns in the 3-days surrounding firm i's earnings announcement for year t s results. SUR is the earnings surprise. DO is the amount of discontinued operations deflated by price. DO_NEG is an indicator variable equal to 1 if discontinued operations are negative, and zero otherwise. POST is an indicator variable that takes the value of 1 post ASU , and zero otherwise. Lastly, SIZE and BM are firm size and book-to-market, respectively. Detailed variable definitions are presented in Appendix A. 4. Results 4.1 Descriptive statistics Table 1 Panel A presents the descriptive statistics for our sample. The mean (median) core earnings, CE, is (0.105). While the mean CE Barua et al. (2010) observe is marginally positive, the median CE is almost identical, with a value of The minor differences could be a result of the different sample period. The mean and median values of discontinued operations (%DO) are both These results suggest that (1) discontinued operations are uncommon and (2) there are both positive and negative discontinued operations. The means of %DO_NEG and %DO_POS are and , respectively (please keep in mind %DO is multiplied by -1). These results are similar to the findings observed by Barua et al. (2010). The mean of MEET&BEAT1, an indicator variable that captures firms with positive earnings, is 60.9%, which implies that about 40 percent of our firm-year observations incur a loss. The mean of MEET&BEAT2, an indicator variable that identifies firms with earnings growth is 52.0%. MEET&BEAT3 identifies firms that meet or beat the analysts consensus. We find that 63.7 percent of our observations achieved that benchmark. Our control variables, SIZE, BM, ACCRUALS, OCF, and ROA have comparable values to values found by Barua et al. (2010). 15

17 We present the descriptive statistics for firm-years with reported discontinued operations in Panel B of Table 1. The mean (median) core earnings, CE, are (0.103), significantly higher than the mean for the entire sample. These results provide initial evidence consistent with expense shifting from core earnings to discontinued operations. With respect to the control variables, SIZE, BM, ACCRUALS, and OCF, the means for the subsample of firms reporting discontinued operations are comparable to the means for the entire sample. 4.2 Testing Hypothesis 1: The association between unexpected core earnings and negative discontinued operations after ASU We present the results for testing whether expense shifting from core earnings to discontinued operations decreased following ASU in Tables 2 and 3. Table 2 reports the levels specification and Table 3 displays the changes specification. Since our focus is testing whether core expenses are shifted into discontinued operations, we present 2 columns: column 1 includes all discontinued operations and column 2 examines negative discontinued operations. For brevity, we only narrate the results in column 2. The coefficient estimate on %DO_NEG is (t-statistic = 7.72), which suggests that prior to ASU firms did shift core expenses into discontinued operations. The coefficients estimate on %DO_NEG_POST is (tstatistic = -2.26), which suggests that the practice to shift core expenses into discontinued operations declined, but was not eliminated, following ASU Table 3 presents a future change in core earnings specification. The underlying premise of this model is that recording discontinued operations is not a recurring activity. Therefore, if a firm shifts core expenses into discontinued operations in one period, it is unlikely it will be able to do so in the next fiscal period and current discontinued operations should predict a decrease in next period s core earnings. For brevity, we narrate the results in column 2, where discontinued operations are only defined for negative values. The coefficient estimate on %DO_NEG is

18 (t-statistic = -2.14), which is consistent with core expense shifting to discontinued operations prior to ASU The coefficient estimate on %DO_NEG_POST is (t-statistic = 1.77), which suggests that core expense shifting to discontinued operations stopped following ASU In summary, the analyses in Tables 2 and 3 are a replication of the most basic findings of Barua et al. (2010). We are able to show that the initial findings of that paper are present in our sample, but that there is a significant decline in the level of classification shifting for most specifications. However, it is insufficient to show that there is just a change in the level of classification shifting. Therefore, we examine in the next two tables whether the rule impacts classification shifting when firms meet or beat analyst expectations (i.e., when they have incentives to engage in classification shifting). Table 4 provides the results of testing the association between unexpected core earnings and the use of discontinued operations to meet or beat analyst expectations before and after ASU Note that as Barua et al. (2010) document that firms reporting income-decreasing (or negative ) discontinued operations use discontinued operations to increase unexpected core earnings, we mainly focus on income-decreasing discontinued operations (%DO_NEG) in H1. We employ three proxies to measure earnings benchmarks, which are: positive earnings per share at year t (Column 1), positive change in earnings per share at year t (Column 2), and earnings surprise, defined as the difference between actual earnings per share and the median of the latest consensus analyst estimate (Column 3). Column 1 replicates the OLS regression in Barua et al. (2010), with the addition of an indicator to capture the post ASU period, as well as interaction term. Results are in general consistent with those in Table 4 of Barua et al. (2010). Specifically, %DO_NEG is positive and significant (coefficient estimate = and t-statistic = 7.74), suggesting that firms 17

19 with more income-decreasing discontinued operations are also more likely to have positive earnings per share. The variable of interest is the interaction between %DO_NEG, MEET&BEAT and POST, which is negative and significant in Column 1 (coefficient estimate = and t- statistic = -3.19). However, there is positive but insignificant coefficient on the interaction of %DO_NEG and MEET&BEAT, which does not allow us to claim there is significant evidence of classification shifting to avoid losses in the pre-asu period. Control variables are also in line with the literature. The coefficient estimate on OCF is positive and significant (coefficient estimate = and t-statistic = 15.68), suggesting that firms with higher operating cash flow also have higher unexpected core earnings. Furthermore, return on assets (ROA) are positively associated with unexpected core earnings, suggesting that firms with good performance usually have more unexpected core earnings. When using the a positive change in EPS as the performance benchmark, presented in Column 2, we find no conclusive evidence of classification shifting to meet prior year based expectations. Specifically, the coefficient estimate on %DO_NEG*MEET&BEAT is (t-statistic = 0.59), and is statistically insignificant. Column 3 reports the results using analysts consensus estimate as the benchmark. We find that the variable of interest, the interaction between %DO_NEG, MEET&BEAT and POST is negative and significant (coefficient estimate = ; t-statistic = -5.42), suggesting that after ASU , companies reduce the usage of discontinued operations to meet or beat analysts expectations. The interaction term between %DO_NEG and MEET&BEAT and POST is significantly positive (coefficient estimate = 0.113; t-statistic = 4.24), which is consistent with expense shifting to discontinued operation to meet or beat analysts consensus estimate prior to ASU Controls variables have the predicted signs as those in Column 1with the exception of BM, which is insignificant. Overall, results in Table 4 indicate that in contrast to the 18

20 findings in Barua et al. (2010) in the pre-rule period, ASU significantly curtails the usage of discontinued operations to manage core earnings to meet or beat certain earnings benchmarks. Next, we re-examine their changes specification, which those authors argue is the more rigorous test. Table 5 reports the results examining the relation between the unexpected change in core earnings (UE_ CE) and the negative discontinued operations (%DO_NEG) pre- and post- ASU Similar to Table 4, we use three earnings benchmarks. The benchmark in column 1 is avoiding a loss. The coefficient estimate on our first variable of interest, %DO_NEG*MEET&BEAT, is negatively significant (coefficient = ; t-statistic = ). These results imply that higher values of negative discontinued operations are predictors of negative changes to next period s core earnings prior to ASU The coefficient estimate on %DO_NEG*MEET&BEAT*POST is (t-statistic = 2.04), which suggests that the improvement in core earnings are persistent in future earnings following ASU , and that expense shifting to discontinued operations stopped following ASU Column 2 reports the results using the a positive change in EPS as the earnings benchmark. Similar to what Barua et al. (2010) document, we find that the interaction term between %DO_NEG, MEET&BEAT and POST is negative and significant, but post ASU , the 3-way interaction (%DO_NEG*MEET&BEAT*POST) becomes positively significant (coefficient estimate = 0.710; t-statistic = 2.48), consistent with the results in Column 1 that negative discontinued operations no longer reverse in the post-rule period. The results are equally strong in Column 3, which examines the classification shifting of discontinued operations to meet analyst expectations. We find that significant use of classification shifting to meet or beat analyst forecasts (coefficient estimate = ; t-statistic = -2.90), but that this 19

21 behavior is eliminated after the introduction of ASU (coefficient estimate = 0.814; t- statistic = 4.23) In summary, Table 4 and Table 5 provide consistent evidence that the shifting of core expenses into discontinued operations to manage earnings in order to meet or beat common earnings benchmarks (including avoiding losses, preventing a decline in earnings and meeting analysts earnings expectations) has been eliminated after ASU Testing Hypothesis 2:The persistence of discontinued operations following ASU Table 6 presents the results of testing the persistence of discontinued operations before and after ASU Column 1 reports the overall discontinued operations, column 2 and 3 report the results from negative and positive discontinued operations, respectively. The variable of interest is the interaction term of %DO and POST, which is negative and significant in Column 1 (coefficient estimate = ; t-statistic = -1.70). This suggests that the persistence of discontinued operations declines after the rule change. We do not find a significance change in the persistence of positive discontinued operation, as presented in Column 2 (t-statistic = -1.07). In addition, the coefficient estimate on %DO is statistically insignificant (coefficient estimate = 0.101; t-statistic = 1.62). These results are consistent with our expectations, since firms have no incentive to shift revenues and gains to discontinued operations, and even more so on a persistent basis. In Column 3, we find persistence in negative discontinued operations prior to ASU (coefficient estimate = 0.106; t-statistic = 2.54). However, the persistence in negative discontinued operations is eliminated following the rule change (coefficient estimate = ; t-statistic = -1.74). These results suggest that the overall decline in the persistence of discontinued operations, as shown in Column 1, is primarily driven by the income-decreasing component. Overall, results in Table 5 suggest that the persistence of 20

22 the discontinued operations, which should not be recurring by its definition, significantly declines after the regulatory change for fiscal year Testing Hypothesis 3: The association between earnings and prior discontinued operations following ASU Results of H3 are presented in table 7, in which we test the predictive ability of discontinued operations before and after the regulatory change. Similar to Table 6, Column 1 reports the overall results, while Columns 2 and 3 present the negative and positive discontinued operations, respectively. In Column 1, both %DO and the interactive term between %DO and POST are statistically insignificant. Specifically, the coefficient estimate on %DO is (t-statistic = -1.01) and the coefficient estimate on %DO*POST is (t-statistic = 0.86). The lack of statistically significant results may be driven by the pooling of negative and positive discontinued operations and their expected opposite predictive ability of operating income. Our focus is on negative discontinued operations, since those are the ones that are likely manipulated for classification shifting. When focus on negative discontinued operations in Column 3, we find that the main effect is negative and significant (coefficient estimate = ; t-statistic = -3.23), consistent with the argument that negative discontinued operations are negatively associated with future pre-discontinued net income prior to the rule change. However, this is no longer the case postrule, i.e., we find the interaction term of %DO and POST positive and significant (coefficient estimate = 5.201; t-statistic = 2.57). 4.5 Testing Hypothesis 4: The change in market reaction to negative discontinued operations following ASU We present the results for testing the change of the market reaction to negative discontinued operations following ASU in Table 8. The coefficient estimate on the 21

23 earnings surprise, SUR, is positive and significant (coefficient = 0.010; t-statistic = 6.19), as expected. The variable of interest is DO*DO_NEG *POST. We find that the coefficient estimate on that variable is negatively significant, with a value of (t-statistic = -2.29). Given that the discontinued operations that are applied to this coefficient estimate are negative, the results imply that higher negative discontinued operations are associated with a more positive market reaction following the regulation change. These results suggest the market views negative discontinued operations following the rule change as a signal for the firm discontinuing low profitability operations. If the investors believed that firms continue to shift core expenses to discontinued operations following the rule change, we would have expected to observe a positive coefficient estimate following ASU Lastly, the coefficient estimate on SIZE is positive and significant (estimate = 0.002; t-statistic = 3.69), consistent with prior studies. To summarize, we find that the market reaction to negative discontinued operations is positive subsequent to the regulation change, which is consistent with investors believing that the negative discontinued operations are transitory and represent the discontinuing of low-profitability operations which will lead to a future increase in profitability. 5. Conclusion Until discovering evidence of classification shifting expenses into discontinued operations, this income statement item was typically dismissed given that any associated profits and losses should not impact the firm going forward. Assuming firms had followed the rules correctly, many of the users of the financial statements including investors and analysts would appropriately discount this item. Except for the findings of Barua et al. (10), there was little evidence that the users of financial statements should be concerned with this item. However, the ability of managers to shift core expenses into the typically discounted discontinued operations 22

24 would be concerning. After showing that such activities are used to meet or beat earnings benchmarks, the implications were not further investigated. The recent introduction of ASU compels us not only to re-examine the findings of their paper under the new regime but further understand what implications this rule change has on the properties of discontinued operations. This new standard did not have clear ex-ante implications given it simultaneously increased the rigidity of certain aspects of the rules on discontinued operations while easing others. Our findings indicate the rule change had a dramatic impact on the reporting of discontinued operations. It significantly curtailed the frequency of the item, reducing the likelihood of reporting one by more than a quarter. At the same time, we find that evidence of classification shifting to meet or beat earnings benchmarks is significantly reduced after the introduction of ASU There is also evidence that the persistence and predictive ability of discontinued operations has significant declined. This is consistent with our other findings that indicate discontinued operations are more genuinely a product of the portions of the firm that have or will be eliminated and thus have less of an impact on firm performance going forward. Finally, we find that there exists a market reaction to the reporting of discontinued operations for the first time. Prior studies as well as findings from the pre-asu time period indicate that the market disregards this item (Fairfield et al. 1996). However, the new rule leads the market to positively react to news of negative discontinued operations. Possibly the higher threshold of a strategic shift now makes the announcement genuinely useful to investors in valuing firms. The current draft of this paper narrowly examines the impact of a regulatory change, ASU , on the use of discontinued operations to manage earnings as well as the 23

25 implications on certain characteristics of this item. Future drafts will detail the impact on the analyst treatment of this item as well as how discontinued operations are discussed during conference calls. Preliminary results for these tests indicate that analysts exclude discontinued operations from their definition of earnings, both before and after the rule change. This supports the findings in Barua et al. (2010) and the earlier time period in our paper in that the ability to meet or beat analyst forecasts is dependent on discontinued operations being excluded from their definition of earnings. We also find evidence from conference calls that managers and those asking questions adjust their discussion of discontinued operations. For the pre-asu period, managers were less likely to mention discontinued operations and callers were more likely to ask about it when there was evidence of classification shifting. This relationship does not persist after the introduction of the new rule. 24

26 References Barnea, A., J. Ronen, and S. Sadan Classificatory smoothing of income with extraordinary items. The Accounting Review 51(1), Barua, A., S. Lin, and A. Sbaraglia Earnings Management Using Discontinued Operations. The Accounting Review 85(5), Bradshaw, M. T., and R. G. Sloan GAAP versus the street: An empirical assessment of two alternative definitions of earnings. Journal of Accounting Research 40(1), Cohen, D. A., and P. Zarowin Accrual-based and real earnings management activities around seasoned equity offerings. Journal of accounting and Economics 50(1), Fairfield, P., R. Sweeney, and T. Yohn Accounting Classification and the Predictive Content of Earnings. The Accounting Review 71(3), Fan, Y., Barua, A., Cready, W. M., and W. B. Thomas Managing earnings using classification shifting: Evidence from quarterly special items. The Accounting Review 85(4), Financial Accounting Standards Board Accounting for the impairment or disposal of longlived assets. Statement of Financial Accounting Standards No Norwalk, CT: FASB. Financial Accounting Standards Board Accounting for costs associated with exit or disposal activities. Statement of Financial Accounting Standards No Norwalk, CT: FASB. Financial Accounting Standards Board Accounting Standards Update Norwalk, CT: FASB. Financial Accounting Standards Board Accounting Standards Update Norwalk, CT: FASB. Gaver, J., and K. Gaver The Relation between Nonrecurring Accounting Transactions And CEO Cash Compensation. The Accounting Review 73(2), Ji, Y., and O. Rozenbaum Do Investors Unravel Real Earnings Management to Meet or Narrowly Beat Analysts Expectations? Evidence from Conference Calls. Working paper Jones, J. J Earnings Management During Import Relief Investigations. Journal of Accounting Research 29(2), Kinney, M. and R. Trezevant The use of special items to manage earnings and perceptions. Journal of Financial Statement Analysis 3, Lee, Y. G An examination of restructuring charges surrounding the implementation of SFAS 146. Review of Accounting Studies 19(2), Lipe, R. C The information contained in the components of earnings. Journal of Accounting Research 24, McVay, S. E Earnings management using classification shifting: An examination of core earnings and special items. The Accounting Review 81(3), Potepa, J The Treatment of Special Items in Determining CEO Cash Compensation, Working Paper Riedl, E. J., and S. Srinivasan Signaling firm performance through financial statement presentation: An analysis using special items. Contemporary Accounting Research 27(1), Ronen, J., and S. Sadan Classificatory smoothing: Alternative income models. Journal of Accounting Research 13(1)

27 Appendix A: Variable Definitions Variable CE CE ATO SALES Neg_ SALES UE_CE UE_ CE %DO_NEG MEET&BEAT1 Definition Revenue (REVT) less Cost of Goods Sold (COGS) and SG&A (XSGA) scaled by Revenue (REVT). Current period CE less lagged CE. Revenue (REVT) divided by the average of current and lagged net operating assets Current period revenue (REVT) less lagged revenue scaled by lagged revenue. SALES if SALES<0 and 0 otherwise. Unexpected core earnings, defined as the difference between actual and expected core earnings. The expected core earnings are estimated from the Eq. (1). Unexpected change in core earnings, defined as the difference between actual change in core earnings and expected change in core earnings. The expected change in core earnings are estimated from Eq. (2). Discontinued operations (DO) deflated by sales (REVT) and multiplied by -1, if reported discontinued operations are negative. (Compustat). An indicator equals to one if earnings per share (IB) is positive, and 0 otherwise (Compustat). MEET&BEAT2 An indicator equals to one if change in earnings per share is positive, and 0 otherwise (Compustat). MEET&BEAT3 POST SIZE BM ACCRUALS OCF ROA PRE_DO_INCOME CAR SUR DO DO_NEG An indicator equals to one if actual earnings per share is greater than latest consensus median estimate (I/B/E/S). An indicator equals to one if fyear >2014, and 0 otherwise (Compustat). Natural logarithm of total assets (AT) (Compustat). Ratio of book value (AT-LT) to market value (MKVALT) (Compustat). Net income before extraordinary items (NI-XI) minus cash from operations (OANCF), scaled by sales (REVT) (Compustat). Operating cash flows (OANCF), scaled by lagged total assets (AT) (Compustat). Income before extraordinary items (NI-XI), divided by average total assets (Compustat). Income before discontinued operations (NI -DO), deflated by sales (REVT) (Compustat). The cumulative market-adjusted abnormal returns in the 3-day around earnings announcement day (CRSP). Actual earnings per share minus the most recent analysts consensus estimate, scaled by the median of most recent analysts consensus estimate (I/B/E/S). The value of discontinued operations, deflated by year-end stock price (Compustat and CRSP). An indicator equals to one if discontinued operations are negative, zero otherwise. 26

28 Appendix B: Examples of Income Statement with Discontinued Operations 27