Income Classification Shifting and Mispricing of Core Earnings

Transcription

1 Income Classification Shifting and Mispricing of Core Earnings Elio Alfonso Department of Accounting E.J. Ourso College of Business Louisiana State University C.S. Agnes Cheng School of Accounting and Finance The Hong Kong Polytechnic University Shanshan Pan Department of Accounting E.J. Ourso College of Business Louisiana State University September 24, 2012

2 Income Classification Shifting and Mispricing of Core Earnings Abstract: This study examines whether the market misprices core earnings (operating income before depreciation and special items) when firms use income classification shifting tactics to boost their core earnings (in short, we term these firms classification shifters ). Using a large U.S. sample across 1988 to 2010, we find that the market overprices the core earnings provided by the classification shifters. We also find evidence that the extent of accruals mispricing is stronger for the classification shifters. Our findings support recent SEC's concerns of firms' income classification shifting behavior. Given that it is relatively easier to manage earnings by using income classification shifting tactics and its significant negative impact on resource allocation, more research should be done in this area. Keywords: Earnings management, classification shifting, mispricing, and accrual anomaly. Data Availability: Data are available from public sources identified in the paper. 1

3 I. INTRODUCTION This paper intends to provide systematic evidence of negative economic consequences to shareholders when firms increase their core earnings by using classification tactics (for easy reference, we term these types of firms classification shifters ). Specifically, we investigate whether the market overvalues the core earnings of classification shifters. We are motivated by research studies' large sample evidence that firms shift their core expenses/losses to special items and discontinued operations (McVay 2006, Barua et al. 2010) and the anecdotal evidence from the SEC's investigations on shareholder losses from classification shifters. It is likely that earnings management using classification tactics is not easy to be detected 1, hence, significant inefficiency of resource allocation (e.g. mispricing of earnings) may exist. Our evidence on whether core earnings is mispriced will contribute to policy makers, investors and researchers with regard to the importance of income classification shifting. Income classification shifting is a form of earnings management where account items are intentionally misclassified to report materially false and misleading line items in income statements. Compared to accruals management and real activities management, income classification shifting is a subtle yet viable tool to manage earnings. Since managers are merely shifting items between categories, net income will remain the same. However, managers have incentives to boost core earnings because the market has become increasingly focused on core earnings that exclude non-recurring or unusual items from GAAP earnings (Bradshaw and Sloan 2002, Gu and Chen 2004). The operating expenses (or losses) that are removed from core 1, because the allocation of expenses to specific accounts can be subjective, auditors might be limited in their ability to verify the appropriate classification, and, because bottom-line income does not change, they might expend less energy on the identification or compulsory adjustments of these accounts (McVay, 2006, p505) 2

4 earnings this year could recur next year, resulting in a lower persistence of core earnings. If investors are unable to recognize the differential persistence of core earnings of firms that shift expenses (shifters) from those that do not (non-shifters), then they may overvalue the securities of shifters. The issue of income classification shifting is of high importance to the Securities and Exchange Commission (SEC) as it states, The appropriate classification of amounts within the income statement is as important as the appropriate measurement or recognition of such amounts (SEC 2000). The SEC further states that they are concerned about account classification because they have noted improper classification of line items in financial statements, especially the income statement. In the last decade, SEC has been actively pursuing companies that engage in income classification shifting. For example, on July 22, 2010, the SEC charged Dell, Inc. with using fraudulent accounting in shifting unrelated operating expenses to restructuring reserves among other charges (SEC 2010a). On April 2, 2010, the SEC charged Symbol Technologies, Inc. with misclassifying unrelated operating expenses to a restructuring charge which was recorded for its acquisition of Texlon Corporation and relocation of its manufacturing operations to new facilities (SEC 2010b). 2 In another case, on November 12, 2009, the SEC charged SafeNet, Inc. with the improper classification of ordinary operating expenses as non-recurring integration expenses (costs incurred to integrate acquired companies into current operations). The company was also charged with misclassifying a significant amount of recurring, operating expenses out of its pro forma earnings in order to meet or exceed quarterly EPS targets (SEC 2009). 3 2 See Appendix C for a detailed analysis of the restated income statements of Dell and Symbol Technologies. 3 See Appendix B for a listing of different companies that have recently been charged by the SEC for actions related to income classification shifting. 3

5 McVay (2006) provides large sample evidence that managers boost core earnings by shifting core expenses, such as cost of goods sold and selling, general, and administrative expenses, to non-core categories such as special items. Barua et al. (2010) further show that firms also disguise core expenses by shifting them to discontinued operations. Since the impact of core earnings is higher than the impact of non-core earnings on firm value (Lipe 1986, Elliott and Hanna 1996, Bradshaw and Sloan 2002, Abarbanell and Lehvay 2007), managers will be able to raise firm value through income classification shifting if investors cannot see through such income classification shifting techniques. To our best knowledge, no studies have provided a systematic evidence to show if the market can see through such shifting. Our paper aims to document if mispricing (overvaluing) of core earnings and accruals exists more for firms shift earnings items ( shifters ) than for firms do not shift earnings items ( non-shifters ) in the income statement. In essence, we ask: are investors fooled by core earnings that are artificially inflated by firms that engage in income classification shifting? A few studies prior to McVay (2006) have examined whether investors are misled by street earnings or pro forma earnings (the constructs, in principle, are similar to core earnings defined in our study except the inclusions/exclusions of depreciations). 4 However, there is no consensus as to whether investors accurately price street earnings and pro forma earnings. For example, Johnson and Schwartz (2005) do not find evidence of higher market multiples or stock return premiums for pro forma firms and conclude that investors are not misled by pro forma earnings. In addition, Gu and Chen (2004) find no evidence of mispricing of street earnings inclusions or exclusions. They regress future abnormal returns on quarterly earnings 4 Street earnings refers to non-gaap earnings reported by analyst-tracking services which exclude non-recurring items. Pro forma earnings refers to non-gaap earnings reported by managers which exclude certain nonrecurring items. Johnson and Schwartz (2005) find that street earnings and pro forma earnings are, on average, the same. 4

6 surprise and non-recurring items that First Call designates as included in or excluded from street earnings and find that the market s valuation appears to be complete. In contrast, Doyle et al. (2003) find that the expenses excluded from pro forma earnings are negatively related to future abnormal returns for up to three years. The authors conclude that the market does not appreciate the future cash flow implications of the excluded expenses and that the market is systematically fooled by pro forma earnings. In examining accruals anomaly, Dechow and Ge (2006) find that low accrual firms with large negative special items have more positive future abnormal returns than low accruals firms without special items. They note that their findings suggest that investors price special items as having more negative implications for future earnings than they actually do. This is in contrast to Doyle et al. (2003) who show that investors overlook the predictive power of items which are excluded from pro forma earnings. These two studies mainly look at the value implication from the non-core items and did not look at the effect of shifting. If their samples on average represent more shifting firms and the market weighs the core earnings as usual, then Doyle et al. s findings would suggest that shifters core earnings are overpriced because core earnings exclude an expense that should depress value 5. However, implication from Dechow and Ge s findings is not clear because the core earnings exclude an expense that has a positive relation to future performance. Our study attempts to provide direct evidence on the effect of shifting on mispricing of core earnings. 6 Firms that engage in income classification shifting opportunistically increase their core earnings by shifting normal operating expenses/losses to non-core accounts. In the year of the shifting, the firm will have higher core earnings and lower core expenses than without shifting. If 5 Doyle et al. (2003) find that expense excluded from pro forma earnings predicts lower future cash flows. As a result, the pro forma earnings without the exclusions in a quarter are valued more than 40% higher, suggesting that these pro forma earnings can be overpriced. 6 We also contrast accrual mispricings between shifters and non-shifters. 5

7 they do not shift again in the subsequent year, the expenses that were shifted will recur once again but in the normal operating expense section instead of in the non-recurring section. Therefore, holding everything else constant, next period s core expenses will increase and core earnings will decrease by an amount similar to the amount shifted in the prior period. This will result in lower persistence of core earnings for firms that shift. If the shifting is accomplished through accounts which are unobservable through the financial statements, investors will misperceive the persistence of core earnings and will, therefore, misprice the core earnings of shifters. Our sample spans over Our first set of analyses focuses on market mispricing of core earnings for shifters versus non-shifters. We then extend our analyses to market mispricing of accruals. Since key issue of the income classification shifting is the persistence of the over-stated core earnings after shifting will be lower than that before shifting, we use Mishkin test method (Mishkin, 1983; Sloan, 1996) to investigate if the market correctly assess the persistence of core earnings when valuing a firm for the shifters. We also use various portfolio tests to document the existence of future abnormal return when investment strategy is based on current accounting information (Sloan, 1996; Xie, 1996; Desai, et al., 2004; Cheng and Thomas, 2006). Mishkin test uses two simultaneous equations. The first equation is an earnings prediction model which generates the earnings persistence coefficient in relating current period earnings to next period earnings (we term this as the reported persistence coefficient, RPC). The second equation is a concurrent unexpected return-unexpected earnings model where the unexpected earnings is derived from the residual of the first equation and we can derive the persistence coefficients implied from the two equations (we term this as the implied persistence coefficient, 6

8 IPC). Mishkin (1983) suggests that these two coefficients should not differ if the market is efficient. Sloan (1996) documents that these two coefficients are not significantly different. His results imply that either the market is efficient on assessing the persistence of earnings or the market is fixated on earnings. When Sloan breaks earnings into the accrual and cash flow component, he finds that the market is not efficient in assessing the persistence of accruals and cash flows, accordingly, he concludes that the market is fixated on earnings. In the context of shifters, if the market sees through the shifting, then the IPC should be lower than the RPC. Alternatively speaking, for non-shifters, IPC=RPC implies that either the market is efficient with core earnings or the market is fixated in core earnings (depending on if the market sees through persistence of accruals and cash flows); however, for shifters, IPC=RPC implies that the market is not efficient even if the mispricing tests for persistence of accruals fails. Hence, we focus first on if the market misprices core earnings followed by if mispricing of accruals also exists for shifters. In addition to Mishkin tests, we also use portfolio methods including zero-investment portfolio analysis and regression analyses (Sloan, 1996; Desai et al., 2004). Zero-investment portfolio analysis forms portfolios based on extreme deciles of core earnings/accruals to compare the difference in portfolio returns between shifters and non-shifters portfolios. Regression analysis regress future abnormal return on docile ranks of portfolios formed based on magnitude of core earnings/accruals. Identifying shifting firms is not straightforward. We identify shifters based on measures derived from previous research. First, prior studies find a positive relation between unexpected core earnings and income-decreasing special items (McVay 2006) and income-decreasing discontinued operations (Barua et al. 2010). Therefore, as our main measure of shifting, we categorize a firm as a classification shifter if the firm has positive unexpected core earnings and 7

9 negative special items or negative discontinued operations in any given fiscal year. As McVay acknowledges, the evidence of income classification shifting could also be the result of firms experiencing efficiency gains (resulting in positive unexpected core earnings) in the same year of restructuring (leading to negative special items). We account for this possibility in our sensitivity analyses where we use three alternative measures of shifting which impose additional restrictions. These restrictions include a reversal of the change in unexpected core earnings, meeting or beating prior year s core earnings by an amount equal to or greater than the amount shifted, and a combination of these two additional restrictions. We find that the core earnings of shifters persist less than that of non-shifters in forecasting both subsequent period core earnings and net income. This is consistent with our measure of shifters because when used as a temporary earnings management tool, income classification shifting will result in the recurrence of core expenses in the normal operating expense section in the subsequent period which will deflate future core earnings. Consistent with our predictions, we find that investors overprice the core earnings for shifters but not for non-shifters. We also find evidence that the market misprices accruals more for shifters. In sensitivity analyses, we find that our mispricing results are consistent with three alternative measures of income classification shifting as well as when we use a modified version of the original McVay (2006) model based off Fan et al. (2010) to estimate unexpected core earnings. We contribute to the literatures on income classification shifting, mispricing of core earnings and accrual anomaly. First, we extend the prior literature on income classification shifting by showing that investors are not able to detect the artificial inflation of core earnings. This supports the implication that managers opportunistically engage in this type of earnings manipulation because it is effective in influencing the firm s market valuation. Second, the 8

10 accrual anomaly literature shows that the market does not differentiate between the different properties of accruals and cash flows. We extend this line of literature by showing evidence that investors misprice the accruals for shifters more than that of non-shifters. Actually, when we adopt the method suggested by Desai et al.(2004), we find that mispricing of accruals is suppressed for the non-shifters but not for the shifters. Our results support the SEC s recent investigations into firms that are potentially engaging in income classification shifting. We show significant negative economic consequences of income classification shifting for a large sample of firms. There are many research studies in earnings management using discretionary accruals, our research findings call for attention in income classification shifting, a potentially very important earnings management tactics. In Section II, we provide a review of studies that are relevant to our paper and develop our hypotheses. In Section III, we outline our research design. Section IV describes our sample selection and descriptive statistics. Section V provides the results. We provide a conclusion in Section VI. II. LITERATURE REVIEW AND HYPOTHESIS DEVELOPMENT Income classification shifting is an earnings management tool used by managers to opportunistically misclassify accounting items on the income statement. Recent literature has documented that managers use income classification shifting to boost core earnings through classifying recurring expenses as special items or discontinued operations (McVay 2006, Fan et al. 2010, Barua et al. 2010). Nelson et al. (2002) found that auditors are more likely to make adjustments when earning management result in increases in current income or when the earnings management is structured to follow precise standards or unstructured with regard to 9

11 imprecise standards. 7 As income classification shifting involves placing the same income components in different categories, it does not affect the bottom-line earnings. Further, as the standards on the classification of expenses as recurring and non-recurring can be subjective, auditors are less likely to detect such misclassification and require adjustments. In a recent study, Zhang (2012) found that high quality auditors including brand name auditors, auditors with long tenure and industry specialist auditors are able to reduce accruals-based earnings management; however, their clients tend to engage in more income classification shifting. Zhang (2012) interpreted this finding as suggesting that such clients resort to income classification shifting when they are unable to use accrual-based earnings management. The same evidence is also consistent with income classification shifting being less likely to be caught even by well-trained auditors. Compared to real earnings management, income classification shifting does not entail real business transactions that can affect the actual economic performance (McVay 2006, Barua et al. 2010). Therefore, income classification shifting can be an attractive alternative for earnings management, yet it could potentially mislead investors to a greater degree. Management engages in income classification shifting with the belief that investors value core earnings higher than non-core earnings. There is ample evidence to substantiate such belief. For instance, the AICPA Special Committee on Financial Reporting are in the position that a company s core activities which are usual and recurring provide the best indicator of the future and should be reported separately from non-core activities that are transitory in nature. Practitioners including financial analysts also advocate distinguishing between recurring (core) and nonrecurring (noncore) items in firm valuations mainly because they capture different 7 Examples of structured earnings management in compliance with precise standards include leases, consolidations, the use of the equity vs. cost method where standards are precise and the transactions in these areas can be structured. Examples of unstructured earnings management dealing with imprecise standards includes reserves which tend to be unstructured and governed by standards that call for substantial judgments. See Nelson et al.(2002) for more details. 10

12 economic events and should have differential persistence and predictability of future cash flows. Numerous academic studies have documented that earnings components indeed have different predictability of future earnings or cash flows (Givoly and Hayn 1992, Barth et al. 2001). Cheng and Yang (1993) found that operating income dominates net income and comprehensive income in explaining residua security returns. The operating components also receive higher weights than non-operating components in forecasting one-year-ahead profitability (Fairfield et al. 1996). In addition, the forecasting ability of earnings components is closely related to their persistence. Empirical evidence has established that investors response to core earnings is greater than to non-core earnings because core earnings have higher persistence (Lipe 1986, Kormendi and Lipe 1987, Collins and Kothari 1989, Ramakrishnan and Thomas 1998). It appears that market understands the implications of earnings persistence in firm valuation, however, Sloan (1996) finds that investors seem to fixate on the persistence of earnings, failing to recognize the differential persistence of the accrual and cash flow components of earnings. One of the research designs that Sloan (1996) adopts to test his earnings fixation hypothesis is the use of the Mishkin test. The test involves a statistical comparison of the reported persistence and implied persistence. For instance, in the Mishkin test of market expectation of earnings, the reported persistence measures the time series persistence of the reported earnings or to what degree current earnings are likely to recur in the future. The implied persistence measures the persistence of earnings that can be inferred from the market s reaction to unexpected earnings. If market is efficient (or fixated) in assessing unexpected earnings, then the implied earnings persistence should be consistent with the reported earnings persistence. Using the Mishkin test, Sloan (1996) shows that the reported persistence and the implied persistence are not statistically different for earnings. In other words, market seems to correctly 11

13 assess the persistence of earnings and do not misprice earnings. Sloan s further analyses show that the market fails to realize the lower persistence of accruals than that of cash flows. Instead, it overestimates the persistence of accruals and underestimates that of cash flows. This finding is considered evidence of mispricing of the accruals. Sloan (1996) further argues that accruals are less persistent than cash flows mostly because accruals are subject to managerial discretion and manipulation and tend to contain accounting items that do not reflect fundamental economic prospects. This conjecture is confirmed by Xie (2001) when he decomposed accruals into normal and discretionary accruals and found that the accrual anomaly is mainly caused by the mispricing of discretionary accruals, suggesting that earnings management plays a role in fooling investors into overvaluing firms with high discretionary accruals. The findings in Sloan (1996) and Xie (2001) suggest that investors seem to fixate on earnings. The stock price reflects their expectation of earnings persistence that is consistent with the reported earnings persistence. It should be noted that the earnings used in Sloan (1996) s tests is operating income after depreciation (Compustat data #178), which excludes non-recurring items such as extraordinary items, discontinued operations, special items and non-operating income. This definition of earnings is essentially that of core earnings. In other words, Sloan (1996) s findings imply that investors actually fixate on core earnings. Alternatively speaking, on average the market is able to correctly assess the persistence of core earnings and impound it in stock price Little empirical research, however, has systematically examined whether market can efficiently price the core earnings of classification shifters. According to McVay (2006), income classification shifting can have economically significant impact on security prices. For instance, firms that have been identified as opportunistic shifters can on average increase their core 12

14 earnings by half a cent per share. Firms with income-decreasing special items of at least 5% of sales can boost their core earnings by also three cents per share. Given the severe penalty by the market for firms that miss their earnings target by even one cent (Bartov et al. 2002), income classification shifting can substantially influence stock prices. Therefore, it is important to understand whether the market is able to effectively identify classification shifters and properly price them. When management artificially boost core earnings by shifting expenses from core expenses in year t, these expenses are likely to recur as core expenses in year t+1. As a result, current core earnings for classification shifters are unlikely to persist in the future, leading to lower persistence of core earnings 8. Prior research has found mixed evidence on whether the market can correctly value the core earnings and non-core earnings in street earnings or proforma earnings. For instance, Doyle et al. (2003) find that greater expenses excluded from pro forma earnings can predict lower future cash flows; however, the market fails to fully appreciate the implications of lower future cash flows on future firm value. Using both a hedge portfolio test based on the ranking of exclusions and a regression test that controls for risk and other anomalies, they find high excluded expenses are associated with significantly negative abnormal returns for up to three years. These results suggest that market can be fooled by the core earnings reported by firms. On the other hand, Gu and Chen(2004) find that the included items from street earnings are more persistent and value relevant than excluded items and the pricing difference between these two types of items cannot be used to generate abnormal returns. Johnson and 8 It should be noted that our study focuses on whether the core earnings of classification shifters are mispriced, i.e. whether the market can correctly assess the persistence of the core earnings of classification shifters and price them accordingly; however, we are muted on the relative persistence of the core earnings of classification shifters vs. nonshifters. The core earnings of shifters could be less persistent because they are inherently less persistent due to poorer firm performance or they could be more persistent because they are able to smooth their earnings through the use of income classification shifting. 13

15 Schwartz(2005) also report that the persistence of the pro forma earnings or core earnings is no different from that of the GAAP bottom line earnings. Their use of market-multiple tests and narrow-window stock returns tests leads to little evidence of mispricing of pro forma earnings. They therefore conclude that market is not misled by pro forma earnings. None of these studies on street earnings or pro forma earnings, however, addresses whether investors can correctly identify the persistence of core earnings of shifters and price them properly. When recurring expenses are buried in special items and other transitory items, market is likely to have great difficulty untangling the effect of excluded expenses that are recurring and overestimate the persistence of shifters core earnings. We therefore expect that market could misprice shifters core earnings. Hence our first hypothesis states that: H1: The core earnings of shifters are mispriced. The literature on accrual anomaly has documented that accruals are mispriced mostly because accruals are less persistent than what market expects. In the case of income classification shifting, the persistence of core earnings is expected to be lower for shifters than for non-shifters. When core earnings are boosted, both the operating accruals and operating cash flows are also artificially increased, resulting in lower persistence of operating accruals and cash flows for shifters as well. If market does not see through the lower persistence of shifters operating accruals, it may overestimate its persistence to a greater degree for shifters. Consequently, we state our second hypothesis as follows: H2: The operating accruals and cash flows of shifters are mispriced to a greater degree than those of non-shifters. The Mishkin Test 14

16 Many studies which examine the accrual anomaly, beginning with Sloan (1996), use the Mishkin (1983) test, in addition to other tests of mispricing, to examine whether investors valuation of the earnings process is significantly different from the actual underlying time-series properties. It is important to note that we do not assume that earnings, accruals, and cash flows represent the full set of information available to investors. Under the null of market efficiency, the coefficients in the forecasting equation should be equal to their corresponding coefficients in the valuation equation regardless of what other correlated omitted variables help predict earnings (Lewellen 2010). Therefore, the test of market efficiency remains valid even if there are correlated omitted variables. This argument is in contrast to Kraft et al. (2007) who argue that only if the omitted variables are rationally priced can one claim that the inferences of the Mishkin test are valid. They further show that accruals are rationally priced when one adds certain explanatory variables such as lagged sales, the level and change in sales, the level and change in capital expenditures, lagged cash flows, and lagged accruals. While Richardson et al. (2010) acknowledge that these criticisms are valid, they state that this does not invalidate the predictive ability of a given variable. The authors argue that the Mishkin test is purely a predictive regression that places some priors on the forecasting so as to allow inferences to be made about the validity of that structure. Lewellen (2010) also explains that correlated omitted variables can affect the slope on a predictive variable, but they do not affect the overall test of market efficiency. In essence, the omission of a variable can affect the magnitude of the slope on our variable of interest but cannot make returns to be predictable when they are not. Kraft et al. (2006) also claim that excluding about 1% of firm-year observations due to extreme stock returns leads to an inverted U-shaped relation between accruals and future returns which is inconsistent with the earnings fixation explanation originally proposed by Sloan (1996). However, recent 15

17 research shows that this inference is incorrect because all stock return observations other than data errors are valid observations (Teoh and Zhang 2011, Richardson et al. 2010). Specifically, Teoh and Zhang show that ex post trimming of extreme returns observations induces a spurious inverted U-shaped relation between returns and negative predictors such as accruals, net operating assets and change in net operating assets. Trimming observations other than data errors also leads to an exaggerated positive relation between returns and positive predictors such as operating cash flow and free cash flow. III. RESEARCH DESIGN Measuring Income Classification Shifting In order to categorize a firm as a classification shifter in any given year we test a firmspecific measure of shifting implied from the prior literature. McVay (2006) finds evidence of income classification shifting by showing a positive relation between unexpected core earnings and income-decreasing special items. Following McVay we estimate model (1) below crosssectionally by industry and fiscal year: CE t = β 0 + β 1 CE t-1 + β 2 ATO t + β 3 ACCRUALS t-1 + β 4 ACCRUALS t + β 5 SALES t + β 6 NEG_ SALES t + ε t (1) UE_CE t = α 0 + α 1 %SI t + ε t (2) We first estimate unexpected core earnings as the difference between the reported earnings and expected core earnings based on equation (1). CE, core earnings, is defined as operating income before depreciation divided by average total assets 9. ATO, asset turnover ratio, 9 McVay uses sales (#12) as a scaler. We scale variables by average total assets to be consistent with the scalers in our mispricing tests. Results are not sensitive to the use of scaler in our expected core earnings model. 16

18 is sales divided by average net operating assets (NOA). NOA is operating assets minus operating liabilities. ACCRUALS is the prior year s net income before extraordinary items minus cash flow operations divided by average total assets. SALES is the percent change in sales over prior year sales. NEG_ SALES is a dummy variable equal to 1 if SALES is less than 0, and 0 otherwise. The core earnings expectation model in equation (1) will produce positive unexpected core earnings if actual core earnings are greater than expected core earnings. If unexpected core earnings are positive it is possible that the firm has shifted core expenses to a non-core account in order to inflate core earnings. McVay uses equation (2) to regress unexpected core earnings (produced from the 1 st equation) on income-decreasing special items. 10 A positive α 1 provides evidence that firms, on average, shift core expenses to special items to inflate core earnings. Barua et al. (2010) find a similar positive relation between unexpected core earnings and negative discontinued operations using the McVay model. These studies provide evidence that, on average, firms shift core expenses such as cost of goods sold or selling, general and administrative expenses, to accounts that are below the line such as special items and discontinued operations. We design our main measure of income classification shifting directly based on the predicted relations between unexpected core earnings and income-decreasing special items and discontinued operations (McVay 2006, Barua et al. 2010).We categorize a firm as a classification shifter if the firm-year observation has positive unexpected core earnings and either negative special items or negative discontinued operations. All other firm-year observations are categorized as non-classification shifters. 10 Income-increasing special items are set equal to 0 and income-decreasing special items are multiplied by

19 McVay identifies one potential concern with this measure. Essentially, that the positive relation between unexpected core earnings and negative special items could also be consistent with firms experiencing efficiency gains in the year the special item is recognized by streamlining their operations or divesting unprofitable lines of business. She investigates this concern by regressing the change in unexpected core earnings in year t+1 on special items in year t. She finds a negative relation which indicates that the reversal of the change in unexpected core earnings provides further evidence of shifting in year t as opposed to alternative explanations for this relation. To address this concern as it relates to our main measure of shifting we include the reversal of the change in unexpected core earnings in two of our alternative measures of shifters in the sensitivity analysis as described below. In all of our tests we use the measure described above as our main definition of income classification shifting. We also employ three alternative measures of shifters to assess the validity of our main measure. Our 1 st alternative measure imposes the additional restriction that the firm has a negative change in unexpected core earnings in the subsequent period. Our 2 nd alternative measure imposes the additional restriction that the firm meets or beats last year s core earnings. 11 We further require that these shifters would not have met or beat last year s core earnings without income classification shifting. Our 3 rd alternative measure combines these two additional restrictions requiring the firms to meet or beat last year s core earnings and have a negative change in unexpected core earnings. These alternative measures are evaluated in the sensitivity analysis section. Tests of Mispricing 11 We use meeting or beating last year s core earnings instead of analysts forecasts because after merging our sample with I/B/E/S and including meeting or beating analysts forecasts in the categorization of classification shifters, there is insufficient variation in the categorization of shifters. 18

20 We use three different methods to identify whether stock prices act as if investors are able to see through the lower persistence of core earnings for classification shifters versus nonshifters. 12 First, we use the Mishkin test (1983) following Sloan (1996) to simultaneously estimate the reported persistence of core earnings and the implied persistence of core earnings that is reflected in stock returns. Second, we use the relation between extreme core earnings deciles and future abnormal returns to calculate the incremental return to a zero-investment portfolio of classification shifters. Third, we estimate a multivariate regression of next period abnormal returns on core earnings deciles and firm characteristics known to be associated with future returns including the book-to-market ratio, sales growth, and the operating cash flows-toassets ratio. Because shifting distorts a firm s core accruals, we repeat our mispricing tests for accruals in order to test whether investors can discern the different properties of accruals and cash flows for shifters. The Mishkin Test Our first test, the Mishkin test, is used to examine whether investors correctly price core earnings for classification shifters relative to non-shifters. 13 We begin by employing the Mishkin framework by estimating the system of equations below for core earnings following Sloan (1996) for our combined sample of shifters and non-shifters. 14 The purpose of this baseline regression is to ascertain whether Sloan s results hold in our sample prior to adding indicator variables for classification shifters in the second step. 12 McVay (2006) reports weak statistical significance (not tabulated) between future returns and income classification shifting for only firms followed by I/B/E/S. Her results lose significance when she uses the full sample of firms. Possible explanations for the difference in results between her study and our study is different time periods, different measures of income classification shifting, and differences in tests of mispricing. 13 Please refer to the Prior Literature Section (II) for a discussion of the validity of the Mishkin test. 14 Our definition of core earnings is similar to Sloan s definition of earnings. Sloan uses operating income after depreciation whereas we use operating income before depreciation, consistent with McVay (2006). 19

21 CE t+1 = γ 0 + γ 1 CE t + υ t+1 (3) AR t+1 = β(ce t+1 γ* 0 γ* 1 CE t ) + ε t+1 (4) Equation (3) is a forecasting equation that estimates the reported persistence (γ 1 ) of core earnings. Equation (4) is a valuation equation that estimates the implied persistence (γ* 1 ) that the market assigns to core earnings. We estimate all equations using the Mishkin framework in this paper jointly using an iterative generalized nonlinear least squares estimation procedure. AR is size-adjusted returns over the 12 month period beginning three months after the fiscal year-end. CE, core earnings, is operating income before depreciation divided by average total assets and is ranked evenly into deciles from 0 to 9. All ranked variables in this study are divided by 9 in order to transform the variables to range from 0 to 1. The ranked transformation(s) in this model and in our subsequent models control for any non-linearity inherent in the relation between the dependent and independent variable(s) as well as any undue influence that extreme observations can have (Cheng et al. 1992). If the implied persistence (γ* 1 ) is less than the reported persistence (γ 1 ) we assume that the market does not misprice core earnings (i.e. the market is not fixated on core earnings). This is because the reported persistence (γ 1 ) includes inflated core earnings so the true reported persistence is less than this reported figure. Likewise, if the implied persistence (γ* 1 ) is equal to or greater than the reported persistence (γ 1 ) we assume that core earnings are mispriced and that the market is fixated on the level of core earnings. We expand the Mishkin test for the benchmark core earnings model by assessing whether the market rationally prices core earnings for firms identified as classification shifters by estimating the following system of equations: CE t+1 = γ 0 + γ 1 SHIFT t + γ 2 CE t + γ 3 CE*SHIFT t + υ t+1 (5) 20

22 AR t+1 = β(ce t+1 γ* 0 γ* 1 SHIFT t γ* 2 CE t γ* 3 CE*SHIFT t ) + ε t+1 (6) Equation (5) is a forecasting equation that estimates the reported persistence (γ i ) of shifters, core earnings, and their interaction. Equation (6) is a valuation equation that estimates the implied persistence (γ* i ) that the market assigns to shifters, core earnings, and their interaction. SHIFT is equal to one if a firm-year observation has positive unexpected core earnings as estimated from equation (1) and negative special items or negative discontinued operations, and zero otherwise. All other variables are defined as mentioned above. The objective of our Mishkin test in equations (5) and (6) is to analyze whether the market misprices the core earnings of firms identified as shifters and non-shifters. If the implied persistence for non-shifters (γ* 2 ) is equal to the reported persistence for non-shifters (γ 2 ) we assume that there is no mispricing of core earnings for non-shifters. Alternatively, if the implied persistence for shifters (γ* 2 + γ* 3 ) is less than the reported persistence for shifters (γ 2 + γ 3 ) we assume that the market does not misprice core earnings for shifters (i.e. the market is not fixated on shifters core earnings). This is because, for shifters, the true reported persistence is lower than what is reported. Therefore, if the implied persistence for shifters (γ* 2 + γ* 3 ) is equal to or greater than the reported persistence (γ 2 + γ 3 ) we can assume that core earnings of shifters are overpriced and that the market is fixated on the core earnings of shifters. Next, we follow prior literature in assessing the mispricing of earnings by expanding the definition of earnings to include non-core items (Xie 2001, Cheng and Thomas 2006, Green et al. 2011). We decompose net income into core earnings and non-core earnings instead of decomposing net income before extraordinary items because the former includes all categories of non-core earnings which managers have been shown to shift. First, we assess the pricing of core 21

23 earnings and non-core earnings for the combined sample of shifters and non-shifters without indicator variables for shifters using the following Mishkin test: EARN t+1 = λ 0 + λ 1 CE t + λ 2 NCE t + υ t+1 (7) AR t+1 = β(earn t+1 λ* 0 λ* 1 CE t λ* 2 NCE t ) + ε t+1 (8) Equation (7) is a forecasting equation that estimates the reported persistence (λ i ) of core earnings and non-core earnings. Equation (8) is a valuation equation that estimates the implied persistence (λ* i ) that the market assigns to core earnings and non-core earnings. EARN is net income divided by average total assets. NCE, non-core earnings, is net income minus operating income before depreciation (core earnings) scaled by average total assets. We expand the Mishkin test by disaggregating net income into core earnings and noncore earnings and assessing whether the market misprices the core earnings of shifters and nonshifters by estimating the following system of equations: EARN t+1 = λ 0 + λ 1 SHIFT t + λ 2 CE t + λ 3 NCE t + λ 4 CE*SHIFT t + λ 5 NCE*SHIFT t + υ t+1 (9) AR t+1 = β(earn t+1 λ* 0 λ* 1 SHIFT t λ* 2 CE t λ* 3 NCE t λ* 4 CE*SHIFT t λ* 5 NCE*SHIFT t ) + ε t+1 (10) Equation (9) is a forecasting equation that estimates the reported persistence (λ i ) of core earnings, non-core earnings, and their interactions with firms identified as shifters. Equation (10) is a valuation equation that estimates the implied persistence (λ* i ) that the market assigns to core earnings, non-core earnings, and their interactions with shifters. For shifters, if the implied persistence of core earnings (λ* 2 + λ* 4 ) is less than the reported persistence of core earnings (λ 2 + λ 4 ), we assume that the market does not misprice the core earnings of shifters. This is because the reported persistence includes inflated core earnings so the true reported persistence is less than 22

24 this reported figure. Likewise, if the implied persistence (λ* 2 + λ* 4 ) is equal to or greater than the reported persistence (λ 2 + λ 4 ) we assume that the core earnings of shifters are overpriced and that the market is fixated on shifters core earnings. Next, we disaggregate earnings into accruals and cash flows to assess whether accruals and cash flows are mispriced for our combined sample of shifters and non-shifters. We employ the Mishkin framework to estimate the following system of equations prior to adding indicator variables for classification shifters: EARN t+1 = α 0 + α 1 ACC t + α 2 CFO t + υ t+1 (11) AR t+1 = β(earn t+1 α* 0 α* 1 ACC t α* 2 CFO t ) + ε t+1 (12) Equation (11) is a forecasting equation that estimates the reported persistence (α i ) of accruals and cash flows. Equation (12) is a valuation equation that estimates the implied persistence (α* i ) that the market assigns to accruals and cash flows. ACC, accruals, is calculated as net income minus cash flow from operations scaled by average total assets. CFO is calculated as operating cash flow minus extraordinary items and discontinued operations scaled by average total assets. All other variables are the same as defined above. Next, we assess whether investors rationally price the accruals and cash flows of firms categorized as shifters and non-shifters using the framework below: EARN t+1 = α 0 + α 1 SHIFT t + α 2 ACC t + α 3 CFO t + α 4 ACC t *SHIFT t + α 5 CFO t *SHIFT t + υ t+1 (13) AR t+1 = β(earn t+1 α* 0 α* 1 SHIFT t α* 2 ACC t α* 3 CFO t α* 4 ACC t *SHIFT t α* 5 CFO t *SHIFT t ) + ε t+1 (14) The originally documented accrual anomaly shows that investors overprice accruals and underprice cash flows relative to their reported persistence in forecasting future earnings. 23

25 Therefore, we expect this mispricing to hold for classification shifters and non-shifters in our sample as well. We are particularly interested in whether the market misprices the accruals and cash flows of shifters because shifters will have an inflated level of core earnings comprised of inflated core accruals and inflated core cash flows. This implies that shifters will have a lower persistence of core earnings, core accruals and core cash flows as compared to non-shifters. Therefore, we expect greater overpricing for both accruals and cash flows for shifters. In addition, shifting is very difficult to detect because firms do not normally disclose the individual line items that comprise special items or discontinued operations. Otherwise, investors would be able to see recurring operating expenses listed under non-recurring expenses and would discount the level of core earnings accordingly. Zero-Investment Portfolio Returns Our second test calculates the difference in returns to a zero-investment portfolio for classification shifters and non-shifters. Our trading strategy is to take a long position in firms in the most negative decile of core earnings in year t and a short position in firms in the most positive decile of core earnings in year t after separating the firms into shifters and non-shifters. The portfolio returns are calculated as follows: Shifter Portfolio Return t+1 = Low Decile Return t+1 High Decile Return t+1 (15) Non-Shifter Portfolio Return t+1 = Low Decile Return t+1 High Decile Return t+1 (16) Incremental Return t+1 = Shifter Portfolio Return t+1 Non-Shifter Portfolio Return t+1 (17) Equation (15) calculates the zero-investment return only for firms categorized as classification shifters. Equation (16) calculates the zero-investment return for firms not categorized as classification shifters. Equation (17) then calculates the incremental zeroinvestment return for shifters over and above non-shifters. We first run the zero-investment 24

26 return analysis for core earnings. If investors correctly price core earnings for both shifters and non-shifters, we do not expect a significant difference in the portfolio returns. On the other hand, if investors overprice core earnings more for shifters, we expect the portfolio return for shifters to be significantly greater than that of non-shifters. Next, we analyze the difference in portfolio returns for accruals. If this strategy produces positive incremental abnormal returns for extreme accrual deciles, this will support inferences from the Mishkin test that investors overprice accruals more for shifters compared to non-shifters. Multivariate Regression Our third test uses a multivariate regression to analyze the relation between accruals and future stock returns while controlling for cross-sectional differences in risk. If there is a correlation between accruals for classification shifters and cross-sectional differences in risk, then it is possible our results from the Mishkin test and the zero-investment portfolio analysis represent underlying changes in risk and not actual mispricing. To control for this possibility we include risk variables that are known predictors of returns such as book-to-market ratio, sales growth, and operating cash flows-to-price (Desai et al. 2004, Cheng and Thomas 2006). We use the following multivariate regression models: AR t+1 = δ 0 + δ 1 (-ACC t )+ δ 2 BM t + δ 3 SG t + δ 4 CFO/P t + ε t (18) AR t+1 = ρ 0 + ρ 1 SHIFT t + ρ 2 (-ACC t )+ ρ 3 (-ACC*SHIFT t ) + ρ 4 BM t + ρ 5 SG t + ρ 6 CFO/P t + ε t (19) All independent continuous variables are ranked evenly into deciles from 0 to 9. The ranked variables are divided by 9 in order to transform the variables to range from 0 to 1. BM, book-to-market ratio, is the ratio of the fiscal year-end book value of equity to market value. SG, sales growth, is the average of annual growth in sales over the most recent three years. CFO/P, is 25