VIEWS: 9 PAGES: 42 POSTED ON: 11/3/2012
Impairment Write-Offs, Discretionary Accruals, and Earnings Persistence HENRY JARVA* University of Oulu, Department of Accounting and Finance P.O. Box 4600, FIN–90014 University of Oulu, Finland. First version: 3 October 2006 This version: 16 May 2007 *Contact Address: Henry Jarva, Department of Accounting and Finance, University of Oulu, P.O. Box 4600, FIN–90014 University of Oulu, Finland. Phone: (+358) 8 553 2928; fax: (+358) 8 553 2906; e-mail: henry.jarva@oulu.fi. Acknowledgement: I appreciate comments from Juha-Pekka Kallunki and workshop participants at the University of Oulu. I also gratefully thank Weili Ge for her help with the Mishkin test using SAS. Electronic copy available at: http://ssrn.com/abstract=947676 Abstract This paper examines whether stock prices reflect the one-year-ahead earnings implications of impairment write-offs. The market appears to recognize the persistence of cash flows and nondiscretionary accruals but overestimate the persistence of impairment write-offs and discretionary accruals. We report that impairment write-offs are very common for U.S. corporations, appearing in 25.5% of all firm-year observations over the period 2002 to 2005. We document that reporting impairment write-offs is strongly linked to firm size. The results show that impairment firms have significantly lower stock return, lower asset growth, lower sales growth, higher bankruptcy risk, and a greater percentage of firms reporting losses and restructuring expenses. We show that firms with impairments have higher future stock returns than other firms, consistent with investors misunderstanding the nature of impairment write- offs. JEL classification: G10, M4 Keywords: Conservatism, earnings management, SFAS 142, SFAS 144, special items. Data availability: All data are available from public sources. Electronic copy available at: http://ssrn.com/abstract=947676 3 1. Introduction Basu (1997) defines conservatism as the extent to which current period accounting income asymmetrically incorporates economic losses, relative to economic gains. He shows that earnings are contemporaneously more sensitive to “bad news” than “good news,” where he uses firms’ stock returns to measure news. Because the recognition of gains and losses is asymmetric, this leads to systematic differences in the timeliness and persistence of earnings. Although asymmetrically timely loss recognition (conditional conservatism) is an empirically significant property of accounting earnings, its implications for future earnings, and pricing, have not been directly examined in the literature. In this paper we investigate whether stock prices rationally reflect the one-year- ahead earnings implications of impairment write-offs1, which is one of the major accrual components through which conservative accounting is facilitated. The remaining accruals are decomposed into discretionary and nondiscretionary components using a model that incorporates asymmetry in the gain and loss recognition roles of accruals. There is a large body of evidence that shows market prices do not completely impound predictable relations between current and future earnings (e.g., Bernard and Thomas 1990; Sloan 1996). Although we examine how the market prices predictable components of current earnings on future earnings, our study differs from most studies (Xie 2001; Burgstahler et al. 2002; Dechow and Ge 2006). Our primary contribution is to provide evidence whether the market rationally prices impairment write-offs with respect to their one-year-ahead earnings implications. Thus, our study sheds new light on the pricing of conditional conservatism. 1 We use the term “write-off” to refer both complete and partial downward asset revaluations. 4 The results indicate that, consistent with general view, impairment write-offs are less persistent than the remaining accrual components. The results also indicate that stock prices act as if investors do not anticipate lower persistence of impairment write- offs, leading to significant security mispricing. We show that impairment write-offs are very common and appear in 25.5% of all firm-year observations over the period 2002 to 20052. We document that reporting an impairment write-off is strongly linked to firm size. The frequency of incurring impairment write-offs in a given year is 15.8% for the smallest firms, compared with 41.9% for the largest firms. We find that impairment write-off firms have performed poorly in the year that the impairment write-off is reported. However, a positive future size-adjusted return suggests that management is taking action to turn the firm around and the market rewards these actions. The remainder of our paper is organized as follows. Section 2 discusses the background and research relating to impairment write-offs and accruals. A description of the data and sample selection is provided in Section 3. Section 4 describes the empirical results, which is followed by conclusions in Section 5. 2. Background and related research 2.1. Prior research Previous research has examined whether stock prices reflect information about future earnings contained in different components of current earnings. Sloan (1996) shows that earnings composed of accruals are less persistent than earnings composed of 2 Our sample covers the post-SFAS No. 142 and No. 144 period. 5 cash flows. He finds that investors underweight the cash flow component and overweight the accrual component’s implications for future earnings and, consequently, misprice these components of earnings. Xie (2001) shows that the overpricing of total accruals is largely due to discretionary accruals. His findings are robust after controlling for special items. Burgstahler, Jiambalvo, and Shevlin (2002) find that investors are fairly sophisticated in pricing special items but prices do not fully impound the implications of special items for future earnings. Richardson et al. (2005) show that less reliable accruals lead to lower earnings persistence and that investors do not fully anticipate lower earnings persistence, leading to significant security mispricing. They also show that the magnitude of security mispricing is directly related to the reliability of underlying accruals. Dechow and Ge (2006) show that accruals improve the persistence of earnings relative to cash flows in high accrual firms, but reduce earnings persistence in low accrual firms. They show that the low persistence of earnings in low accrual firms is primarily driven by special items. They also show that low accrual firms with large negative special items earn higher positive returns than other low accrual firms. There is a substantial body of accounting literature that examines special items and asset write-offs. In addition, controlling for the effects of special items has become a standard procedure for robustness checks. Special items refer to nonrecurring items that are unusual or infrequent, but do not meet the criteria for classification as extraordinary. Management have incentives to highlight the transitory nature of income-decreasing special items (Kinney and Trezevant 1997). Prior research indicates that asset write-offs can have an enormous impact on both accounting earnings and the book value of assets (e.g., Alciatore et al. 1998). Francis, Hanna, and Vincent (1996) find that incentives 6 play a substantial role in explaining goodwill write-offs and restructuring charges but only a small role in inventory and property, plant, and equipment (PPE) write-offs. Elliot and Hanna (1996) find that earnings response coefficients (ERCs) generally decrease in the presence of special items. Rees et al. (1996) find that abnormal accruals in the asset write-down year are a credible signal about future firm performance. Research shows that top management changes are important determinant of a write-off decision (Strong and Meyer 1987; Elliot and Shaw 1988). In conclusion, researchers seem particular interested in special items. Basu (1997) shows that the concurrent sensitivity of earnings to negative returns is two to six times as large as the concurrent sensitivity of earnings on positive returns. This finding implies that earnings more timely reflect publicly available bad news about future cash flows than good news. Basu (1997) provides evidence that asymmetric timeliness results somewhat more from accruals than from cash flows. He also finds that negative earnings changes are less persistent than positive earnings changes. Therefore, positive (negative) earnings changes have higher (lower) ERCs in a forward regression of abnormal returns on earnings. The subsequent research has extensively examined conservatism in different settings.3 There is also a large body of literature that provides evidence of earnings management using accruals (see e.g., Healy and Wahlen 1999; Dechow and Skinney 2000). Accrual accounting is subject to managerial discretion because of the flexibility allowed by the Generally Accepted Accounting Principles (GAAP). A number of papers call into question the precision and power of discretionary accruals estimates using conventional linear accruals models (e.g., Dechow et al. 1995; Fields et al. 2001; 3 See discussions of accounting conservatism in Watts (2003a; 2003b). 7 Kothari et al. 2005). Basu (1997) points out that accruals enable accountants to recognize bad news about future cash flow on an asymmetrically timely basis. Ball and Shivakumar (2006) investigate the role of accrual accounting in the asymmetrically timely recognition of gains and losses. Specifically, they argue that economic losses (shocks to current period cash flows plus revisions in the expectation of firms’ future cash flows) are captured by the accruals process in a more timely manner than gains. Ball and Shivakumar (2006) show that piecewise linear accruals models that incorporate asymmetry in gain and loss recognition offer a substantial specification improvement, and they explain substantially more variation in accruals than equivalent linear specifications. Their results suggest that standard linear accruals models exhibit substantial attenuation bias and are misspecified to some degree. In contrast to prior work that often focuses on the average behavior of special items, our analysis focuses on impairment write-offs.4 Our empirical tests attempt to measure the pricing of impairment write-offs in the post-SFAS No. 142 and No. 144 period. 2.2. Accounting for asset impairments The increasing frequency of large special item adjustments has attracted the attention of the business press, accounting standard-setting bodies, and regulatory authorities. The Financial Accounting Standards Board (FASB) adopted SFAS No. 142 (Goodwill and Other Intangible Assets) in June 2001. SFAS 142 requires that goodwill and intangible assets that have indefinite useful lives will not be amortized but rather 4 Burgstahler et al. (2002, 587) note that while special items are commonly viewed as transitory on average, it is not clear if individual special items have different implications for future earnings. Also Dechow and Ge (2006, 293) raise a question of whether different types of special items have different implications for earnings persistence and future returns. One reason for scarce research on individual special items is that Compustat has classified different types of special items only since 2001. 8 will be tested at least annually for impairment. Intangible assets that have finite useful lives are amortized over their useful lives. The standard provides specific guidance for testing goodwill for impairment using a two-step process. The first step is a screen for potential impairment that begins with an estimation of the fair value of a reporting unit. If the carrying amount exceeds the estimated fair value, then the second step measures the amount of impairment. The impaired amount is the difference between the carrying amount of the reporting unit’s goodwill and its implied fair value. The standard prohibits the loss recognized to exceed the carrying amount of goodwill. Also, subsequent reversal of a previously recognized impairment loss is prohibited. Under SFAS 142, the amortization expense and impairment losses for intangible assets shall be presented in income statement line items within continuing operations. Impairment losses that arise due to the initial application of the standard are to be reported as resulting from a change in accounting principle. Basu (1997, 10) argues that asset impairment recognition standards have increased the U.S.’s accounting conservatism. Watts (2003a, 217) claims that the recent adoption of FASB 142 appears inconsistent with conservative accounting and enables impairment manipulation. This is because assessing impairment requires valuation of future cash flows and those estimates are unlikely to be verifiable and contractible. The Financial Accounting Standards Board adopted SFAS No. 144 (Accounting for the Impairment or Disposal of Long-Lived Assets) in August 2001. SFAS No. 144 superseded SFAS No. 121 and replaced the amortization of goodwill with periodic assessment of goodwill impairment. The standard requires firms to a) recognize an impairment loss only if the carrying amount (book value) of a long-lived asset (asset group) is not recoverable from its undiscounted cash flows and b) measure an 9 impairment loss as the difference between the carrying amount and its fair value. SFAS 144 (par. 8) requires a long-lived asset to be tested for recoverability whenever events or changes in circumstances indicate that its carrying amount may not be recoverable. If the test is triggered then it also may be necessary to change depreciation estimates and method, or the amortization period. The recoverability of long-lived asset is tested by the entity’s own assumptions about estimates of future cash flows (SFAS 144, par. 17). Under SFAS 144, an impairment loss recognized for a long-lived asset to be held and used shall be included in income for continuing operations before income taxes. Managers’ accounting choices affect the timing and the amount of impairment recognition since subjectivity is inherent in the two-step impairment test prescribed by SFAS 142 and SFAS 144. Richardson et al. (2005, 449) point out that to estimate the amount of impairment write-offs involves great subjectivity. Riedl (2004) finds that write-offs reported after the adoption of SFAS No. 121 (predecessor of SFAS No. 144) have lower associations with economic factors, and a higher association with “big bath” reporting behavior, relative those reported prior to the standard. On the contrary, management could record impairment write-offs to smooth income in a period of higher than normal earnings to maintain a steady and predictable earnings growth. However, it is more likely that impairment write-offs lower earnings persistence. In summary, SFAS 142 and SFAS 144 give substantial latitude for management to select the timing and amount of asset impairment. 3. Data 10 The sample comprises all firms listed on the New York (NYSE), American (AMEX), and NASDAQ markets for which requisite financial and return data are available. We exclude foreign companies. Our empirical tests employ data from three sources. Financial statement data are obtained from the Worldscope database, stock return data are obtained from the Datastream, and analyst data are obtained from the I/B/E/S research files. Our sample consists of financial statement and return data for firms with required one-year-ahead, current, and lagged values during the 4-year period 2002 to 2005. The sample period is relatively short because impairment write-off data are not available in Worldscope until 2001. In addition, we exclude year 2001 because SFAS 142 and SFAS 144 are effective in fiscal years beginning after December 15, 2001 (SFAS 142, par. 48; SFAS 144, par. 49).5 The following data filters are applied to 15,864 firm-year observations during the 4- year period 2002 to 2005. Firms in the financial services, insurance, and real estate industries (SIC 6000–6999) and firms with missing SIC codes are removed (772 firm- years). Observations with less than $1 million in average total assets are deleted to control for the potential influence of observations with relatively low values of deflating variable (458 firm-years). We replace missing values of impairments with zero. Firms with positive total impairment values are deleted (48 firm-years). To minimize the effect of outliers, we delete the firm-year observations that are in the top or bottom 1 % of the distributions of the earnings, cash flows, and accruals variables (576 firm-years). Finally, we delete observations with less than 15 observations in any three-digit SIC 5 Beatty and Weber (2006) examine SFAS 142 adoption decisions, focusing on the trade-off between recording certain current goodwill impairment charges below the line and uncertain future impairment charges included in income from continuing operations. They find that both contracting and market incentives affect firms’ accounting choices relating to the trade-off between the timing and presentation of expense recognition on income statements. 11 code combinations (846 firm-years). These procedures result in 13,164 observations for primary tests. Following earlier research, we compute accruals based on cash flow statements. Hribar and Collins (2002) show that merger and acquisition activities, foreign currency translations, and divestitures introduce significant measurement error into accruals estimated using the balance sheet approach. We calculate Accrualst by subtracting operating cash flow from earnings before extraordinary items, both taken from the cash flow statement. Pre-impairment accrualst are calculated by subtracting Impairmentst from Accrualst.6 To estimate discretionary accruals we use the piecewise linear Jones (1991) model, modified by Ball and Shivakumar (2006). We run the following regression from the pooled data separately for each three-digit SIC industry (159 industries): Pre − impairment accrualsit = α 0 + α1 ∆Salesit + α 2 PPEit + α 3Cash flowsit + α 4 DCFit + α 5Cash flowsit × DCFit + εit (1) where ∆Salesit is the change in sales for firm i in year t; PPEit is gross property, plant, and equipment; Cash flowsit is cash flow from operations and it is our proxy for gain or loss;7 DCFit is a dummy variable having the value of one if Cash flowsit is negative, zero otherwise; Cash flowsit × DCF is the interaction variable; and εit is the error term. All variables are standardized by average total assets in an attempt to reduce 6 Although we subtract impairment write-offs from total accruals, there are still many accrual components in which the timely recognition of losses is accomplished. Examples are losses on trading securities, inventory write-downs, receivables revaluation, and restructuring charges arising from attending to failed strategies or excessive head accounts (Ball and Shivakumar 2006, 213). 7 Cash flow proxy gives the highest R2 (40%) but our results are similar when we use different proxies to estimate the nondiscretionary accruals. See Ball and Shivakumar (2006) for more details. 12 heteroskedasticity. Nondiscretionary accruals, NACCit, is the fitted value from Eq. (1) and discretionary accruals, DACCit, is the residual from Eq. (1). Stock returns are measured using compounded buy-hold size-adjusted returns inclusive of dividends and other distributions. Following previous research, returns are calculated for a 12-month period beginning eight months prior to the end of the fiscal year. The size-adjusted return is calculated by deducting the value-weighted average return for firms in the same size-matched decile, where size is measured as the market value at the beginning of return cumulation period. Analyst EPS and sales forecast data are obtained from I/B/E/S. We used the median consensus forecast outstanding in the eighth month before fiscal year-end. We also obtained the number of analysts providing the forecast in the eighth month before fiscal year-end. 4. Empirical results We present results in five sections. Section 4.1 begins with descriptive statistics for our sample. Section 4.2 presents results from tests of persistence concerning the different earnings components. Section 4.3 reports the market pricing of impairment write-offs and discretionary accruals. The results of market perceptions are discussed in Section 4.4 and Section 4.5 examines the predictive ability of impairments. 4.1. Descriptive statistics 13 Panel A of Table 1 provides descriptive statistics of the financial variables used in our analysis. All variables are scaled by average total assets. The median earnings amount is 0.025, while median cash flow cash from operations is 0.069. This difference leads to negative total accruals (–0.060), mostly due to depreciation. Total impairments have a median value of 0.000. However, the mean is –0.012, and this suggests that the distributions of impairments are highly skewed. Consistent with Xie (2001) discretionary accruals are more variable than nondiscretionary accruals. For our subsample of impairment firms, the median earnings is negative (–0.006) and the median cash flow is positive (0.058). The mean value of total impairments is –0.048. The results indicate that firms have performed poorly during the fiscal year they report an impairment write-off. (Insert Table 1 about here) Panel B of Table 1 reports both Spearman and Pearson correlations. For ease of exposition, we discuss the Spearman correlations. Consistent with prior research, we document a positive correlation between earnings and total accruals (0.384). We also report a positive relation between earnings and cash flows (0.760). Consistent with Dechow (1994), who proposes that accruals offset transitory cash flow effects, our pre- impairment accruals are negatively related to cash flows (–0.203). There is a negative correlation between nondiscretionary and discretionary accruals (–0.096) which can be interpreted as an evidence of management smoothing income. Ball and Shivakumar (2006, 213) argue that the gain and loss recognition role of accruals is a source of positive correlation between accruals and current-period operating cash flow. 14 Impairments are positively correlated with cash flows (0.080) implying timely recognition of impairments. Impairments are also positively related to pre-impairment accruals (0.081) and discretionary accruals (0.078). We return to this issue in Fig. 1. Table 2 presents the relative frequency of impairments during the four-year period from 2002 to 2005.8 Impairments are fairly common, appearing in 25.5% of all firm- years. The impairment phenomenon is strongly linked to firm size. The results reveal an increasing relation between firm size and the frequency of impairments. The frequency of incurring impairment write-offs in a given year is 15.8% for the smallest firms (portfolio 1), compared with 41.9% for the largest firms (portfolio 10). In untabulated results, we observe that the probability of reporting an impairment increasing from 25.5% for all firms to 49.6% for firms reporting impairment in the previous year. This finding suggests that impairments are not as transitory as management claims them to be. The results in Table 2 indicate that large firms make more balance sheet adjustments than small firms. This is consistent with Dechow and Ge’s (2006) view that accounting rules lead to negative accruals with a balance sheet perspective (in contrast to an income statement perspective). Conservative accounting requires declining firms with large asset bases to record impairments to write assets down at fair value. (Insert Table 2 about here) Fig. 1a plots the proportion of firms reporting impairment write-offs based on ranks of accruals. Impairment write-offs are defined as impairments as a percentage of average total assets greater than 0%, 1%, 2%, or 5%. Approximately 30% of firms in 8 Ten portfolios are formed each year based on the value of firms’ total assets at the end of the previous fiscal year- end. 15 decile 1 report impairment write-offs that are greater than 5% average total assets. The results indicate that the proportion of firms reporting impairments is largest in deciles 1 and 2. (Insert Fig. 1 about here) Fig. 1b plots the proportion of firms reporting impairment write-offs based on ranks before impairment write-offs. The largest impairments are still in deciles 1 and 2. However, this ranking leads to a smoother frequency of impairments and reflects a convex pattern with larger impairment write-offs. The results show that large impairment write-offs are not just an issue for low accrual firms, but these write-offs make the total accruals exceedingly negative. In all deciles at least 8% of firms have impairment write-offs that are greater than 1% average total assets. Moreover, in all deciles except in decile 10 at least 20% of firms have reported impairment write-offs. Dechow and Ge (2006, 254) argue that large negative accruals indicate that a firm is reducing assets and downsizing, while firms with large positive accruals indicate that a firm is investing in assets, generating sales, and expanding their business. The results in Fig. 1 b imply that both “growing” and “declining” firms make large impairment write- offs if the sign of accruals is the measure of growth. However, these results support the view that declining firms are more likely to make impairment write-offs to correct their balance sheet. 4.2. Tests of persistence 16 Next we investigate whether impairment write-offs and discretionary accruals are important for explaining why the accrual component of earnings is less persistent than the cash flow component. Table 3 provides the results of estimating the following regressions: Earningst + 1 = α + βEarningst + εt (2) Earningst + 1 = α + β1Earningst + β2DIt + β 3 Earningst × DIt + εt (3) Earningst + 1 = α + δ1Cash flowst + δ2Accrualst + εt (4) Earningst + 1 = α + δ1Cash flowst + δ2Pre − impairment accrualst + δ3Ιmpairmentst + εt (5) Earningst + 1 = α + δ1Cash flowst + δ2NACCt + δ3DACCt + δ4Ιmpairmentst + εt (6) where DI is a dummy variable equal to one for firms that report impairment write-offs, zero otherwise. Since we are interested in the persistence of earnings (not a return on assets) we scale earnings in period t+1 by average total assets in period t (see, Fairfield, Whisenant, and Yohn, 2003; Fairfield, 2006). The use of a contemporaneous deflator would be particularly puzzling in our setting since impairment write-offs affect both the denominator and the numerator of the dependent variable.9 The results in Table 3 indicate that the persistence parameter on earnings is equal to 0.886. This suggests that profitability follows a mean-reverting process, as documented in past research. Eq. (3) includes an interactive indicator variable equal to one for firms that report impairment write-offs. Note that after controlling for the impairments, the coefficient on earnings increases to 0.982. The interactive slope coefficient, β3 , 9 We find that our results are robust with respect to using average total assets in period t+1 as an alternative deflator for t+1 earnings. Interested readers may obtain a full set of results by contacting the author. 17 indicates that the persistence of earnings is lower for impairment firms (0.982–0.286). This finding is consistent with Basu (1997) in that conservatism affects earnings persistence. The adjusted R-square increases from 38.77% to 39.92%. (Insert Table 3 about here) Consistent with prior research, decomposing earnings into cash flows and accruals indicates that the cash flow component of earnings is more persistent than the accrual component of earnings (1.081 vs. 0.568). An untabulated F-test rejects the hypothesis that the coefficients are equal (F=483.48). Decomposing accruals into the impairment component and the pre-impairment component indicates that impairment component is less persistent than other accrual components (0.084 vs. 0.664). If impairments are purely transitory, the coefficient should move toward –1. A transitory impairment coefficient is strongly rejected (F=611.78, untabulated). Consistent with Xie (2001), further decomposing reveals that the discretionary accrual component is less persistent than the nondiscretionary component of accruals (0.581 vs. 0.928). The equality of these two coefficients is rejected (F=61.10, untabulated). In summary, these results indicate that different accrual components have incremental information content and impairment write-offs have a strong affect on earnings persistence. 4.3. Tests of the pricing of impairment write-offs and discretionary accruals Following prior research we use the Mishkin (1983) test to investigate whether investors appear to rationally price publicly available information. Our application of 18 the Mishkin (1983) test provides a statistical comparison between (a) a measure of the market’s pricing of impairment write-offs and discretionary accruals, and (b) a measure of the ability of impairment write-offs and discretionary accruals to predict one-year- ahead earnings. If the market’s valuation coefficients on accrual components significantly differ from the forecasting coefficient of these accrual components for one- year-ahead earnings, the Mishkin (1983) test would suggest that the market misprices these accrual components. Table 4 reports the results of the Mishkin (1983) tests. Panel A of Table 4 shows results of estimating Eq. (5), where we decompose earnings into cash flow from operations, pre-impairment accruals, and impairments. For cash flows, the valuation coefficient (δ*1=0.907) is not significantly smaller than the forecasting coefficient (δ1=1.089), indicating that investors correctly price the cash flow component. Investors appear to overweight the pre-impairment accruals (0.919 vs. 0.664), but this difference is not statistically significant. Investors appear to overweight the persistence of impairment write-offs (0.084 vs. 2.133). The likelihood ratio statistics strongly reject the null hypothesis of rational pricing of impairment write-offs (p<0.001). (Insert Table 4 about here) Panel B of Table 4 provides the results for regression Eq. (6). This regression decomposes pre-impairment accruals into nondiscretionary and discretionary accruals. The forecasting coefficient on nondiscretionary (discretionary) accruals is 0.927 (0.581). The valuation coefficient on nondiscretionary accruals is 0.871, while the valuation coefficient on discretionary accruals is 0.934. The results indicate that 19 investors correctly weight nondiscretionary accruals, but overweight discretionary accruals. However, the untabulated likelihood ratio statistic of 0.03 (marginal significance level = 0.859) suggest that investors do not distinguish these two components. Finally, the null hypothesis that the market rationally prices all four earnings components is rejected (p<0.001). 4.4. Market perceptions The objective of this section is to provide further insights into why the market appears to overprice impairment write-offs. Table 5 provides various measures of firm characteristics across samples formed by the magnitude of impairment write-offs relative to average total assets. We evaluate the financial conditions of firms by using Altman’s (1968) Z-score model. Altman’s Z score is computed as: Z = 1 .2 X 1 + 1 .4 X 2 + 3 .3 X 3 + 0 .6 X 4 + 1 .0 X 5 (7) where X1 is the working capital/total assets, X2 is the retained earnings/total assets, X3 is the earnings before interest and taxes/total assets, X4 is the market value of equity at previous year end/book value of total liabilities, and X5 is sales/total assets. Altman (1968, 607) renders that the Z score of 2.675 discriminates best between bankrupt and non-bankrupt firms. Panel A of Table 5 confirm that earnings, cash flows, accruals, asset growth, and sales growth declines systematically with the relative magnitude of impairment write- offs. Although discretionary accruals become more negative when the relative 20 magnitude of impairments increases, it is difficult to conclude that those firms are taking a big bath. Impairment write-off firms report a much higher percentage of losses and restructuring charges than other firms. Low Altman’s Z-scores for impairment write-off firms indicates that those firms are more likely in financial distress. These results indicate that impairment firms have performed very poorly during the fiscal year in which they report the impairment write-off. Panel B of Table 5 indicates that impairment firms have poorer market performance than other firms. This is in the spirit of Basu (1997). For example, firms that have impairment write-offs greater than 1% average total assets have size-adjusted returns of –9.4% versus 2.5% for no impairment firms. Analyst coverage is correlated with firm size and it is highest for impairment sample (70.3%). Analysts’ median EPS forecast errors indicate that no impairment firms have higher earnings than analysts expect, while impairment firms perform worse than analysts expect. Analysts may exclude impairment write-offs from their forecasts of Street earnings (Bradshaw and Sloan 2002, 46). Therefore, we also report analysts’ sales forecast errors because, unlike earnings, sales are not affected by impairment write-offs. Results also indicate that sales forecasts are too optimistic for impairment firms. The poor performance of impairment firms appears to be a surprise for analysts. (Insert Table 5 about here) Table 5 Panel C reports the stock price performance and investor recognition in period t+1. The size-adjusted return of impairment firms is 7.8% whereas for no impairment firms the return is 1.3%. A similar finding holds in other impairment 21 subsamples. Note that impairment write-offs are accounting entries and do not require any action to be taken by management to improve performance. However, a positive size-adjusted return suggests that management is taking action to turn the firm around and the market rewards these actions. Investors appear to believe that poor performance of impairment firms will continue because realized earnings outperform analysts’ forecasts. Increase in analyst coverage (change in number of EPS/sales forecasts) is largest for no impairment firms and lowest for the 5% subsample. This result suggests that an analyst response is correlated with firm performance. Finally, Panel D compares selected values that are statistically different between firms that report impairment write-offs and firms that have no impairments. All performance measures confirm that impairment firms have performed significantly worse than firms without impairments. Results on the market show that analysts’ median sales forecast error in year t and EPS forecast error in year t+1 is significantly different between the samples. Change in the number of EPS forecasts in year t + 1 is significantly smaller for impairment firms. For impairment firms the mean size-adjusted return is not statistically smaller in period t, but it is significantly higher in period t+1. Table 5 suggests that firms that recognize impairments differ from other firms. Table 6 investigates whether impairments predict future returns after controlling for other potential determinants of future returns. We include the level of total accruals as a control variable because Sloan (1996) shows that there is a negative relation between accruals and future returns. The results indicate that impairments continue to explain future returns after controlling for other variables. Size (measured as the natural logarithm of the fiscal year-end market value of equity) is negative and significant suggesting that smaller 22 firms earn higher future returns. There is a negative relation between total accruals and future size-adjusted returns. Book-to-market ratio is significantly negative, and this suggests that value stocks perform slightly worse in the future. There is a negative relation between current size-adjusted returns and future size-adjusted returns. The analyst EPS forecast indicator variable is positive suggesting that firms followed by analyst earn higher future returns. These results demonstrate that predictive ability of impairments is not subsumed by these other variables. In summary, impairment firms have performed poorly, have higher bankruptcy risk, and earn negative size-adjusted stock returns in the year that an impairment write-off is reported. Analysts give pessimistic earnings forecasts for impairment firms suggesting that they believe that poor performance will continue. Investors seem to believe that impairment firms are more risky or misunderstand the nature of impairments and, therefore, impairment firms earn higher returns than other firms. 4.5. Predictive ability of impairments Barth, Cram, and Nelson (2001) show that different accrual components reflect different information relating to future cash flows. If impairments incorporate into current earnings information about changes in expected future cash flows, they should improve the ability of earnings (and its cash flow and accruals components) to predict future cash flows. By definition an impairment loss must be recognized if the carrying amount of a long-lived asset is not recoverable from its undiscounted cash flows. We next investigate whether cash flows and accruals contain incremental information for impairment write-off firms about the next period’s cash flows. We adopt a similar 23 regression than Ball and Shivakumar (2006, 236) to test the predictive ability of impairments. In particular, if management recognizes an impairment loss in a timely manner, we expect that this specification conveys information about future cash flows. Table 7 provides the following regression: Cash flowst + j = α + δ1Cash flowst − 1 + δ2Accrualst − 1 + δ 3Cash flowst + δ 4 Accrualst + δ 5 DIt + δ 6Cash flowst × DIt + δ 7 Accrualst × DIt + εt + j (8) j = 1 to 3. Cash flowst–1 and Accrualst–1 control for expected cash flows at the beginning of year t. DIt is a dummy variable equal to one when the magnitude of impairments is greater than 1% average total assets, zero otherwise. Variables are standardized by average total assets in period t. If impairments incorporate information about expected future cash flow changes, the specification in Eq. (8) will better predict future cash flow. The incremental coefficients, δ6 and δ7, during impairment years will be negative, because recognition of impairment will incorporate a multiperiod cash flow effect, not simply current-year effects. We offer no predictions for δ5 in Eq. (8). (Insert Table 7 about here) The results in Panel A of Table 7 indicate that the incremental coefficient δ6 on current year cash flows during impairment years is negative and significant for each of the three future-year cash flow. The accruals in impairment-recognition years incorporate incremental cash flow effects for year t+1 and t+2. The negative coefficient 24 on the interaction terms indicates that firms with impairment suffer a multiperiod reduction in cash flows. This is consistent with impairments incorporating information about expected future cash flow changes. Elliot and Shaw (1988) point out that, “In general, economic events precede accounting recognition; an event occurs and then it is disclosed. For material write-offs, this sequence implies that assets suffer an impairment of value, management realizes that impairment, and then an accounting entry is created to record the impairment.” Hayn and Hughes (2006) find that goodwill write-offs lag behind the economic impairment of goodwill by an average of three to four years. To examine conjecture that firms delay disclosing asset impairment issue we re-estimate Eq. (8). We give dummy variable value of one when the magnitude of impairments is greater than 1% average total assets for period t+1 but zero for period t. Our untabulated results do not support the view that firms delay recognition of impairment write-offs. However, this result should be interpreted with caution and we leave this issue for future research. 5. Conclusions Our research relates closely to two streams of accounting research. The first stream of research is based on Sloan (1996). Sloan (1996) shows that stock prices do not fully reflect information about future earnings in accruals and cash flows. He finds that investors appear to overweight (underweight) total accruals (cash flows) when forming future earnings expectations. The second stream of research is based on Basu (1997). He shows that earnings report publicly available bad news about future cash flows more 25 timely than good news. Basu (1997, 16) argues that accruals incorporating write-offs are more likely to reflect conservatism than others. This paper examines whether stock prices reflect the one-year-ahead earnings implications of impairment write-offs, which is one of the major accrual components through which conservative accounting is facilitated. Recognition of impairment write- offs indicates that the carrying amount of the asset have exceeded its recoverable amount. The remaining accruals we decompose into nondiscretionary and discretionary accruals using the piecewise linear Jones (1991) model, modified by Ball and Shivakumar (2006). We show that impairments and discretionary accruals are less persistent than other accruals, which, in turn, are less persistent than cash flow components or earnings. The market appears to recognize the persistence of cash flows but overestimate the persistence of impairment write-offs. Moreover, the market is fairly sophisticated in pricing the one-year-ahead earnings implications of the nondiscretionary accruals but overweight discretionary accruals. Although investors appear to misprice impairment write-offs, we do not claim that conservative accounting is the reason for this. We report that impairment write-offs are very common for U.S. corporations, appearing in 25.5% of all firm-year observations over the period 2002 to 2005. We document that reporting an impairment write-off is strongly linked to firm size. The frequency on incurring impairment write-offs in a given year is 15.8% for the smallest firms, compared with 41.9% for the largest firms. Firms that report impairment write- offs have performed very poorly. The results show that impairment firms have significantly lower stock returns, lower asset growth, lower sales growth, higher bankruptcy risk, and a greater percentage of firms reporting losses and restructuring 26 expenses. The average change in analyst coverage is lowest for firms that report large impairment write-offs. This result suggests that analysts believe that poor performance of impairment firms will continue. We document that firms with impairments have higher future stock returns than other firms. This is consistent with the conjecture that investors misunderstand the nature of impairments and the actions that the management is taking to turn the firm around. Finally, empirical results reveal that cash flows and accruals in impairment write-off firms contain incremental information about future cash flows. Our conclusions are subject to a number if limitations. First, our findings should be interpreted with caution because our sample period is relatively short. Second, we focus on the overall impairment write-offs and do not examine its components. Our study addresses the average behavior of impairment write-offs. For example, while both goodwill write-offs and PPE write-offs indicate that the carrying amount of the asset have exceeded its recoverable amount, their implications for future performance may be different. This raises questions for future research. 27 References Alciatore, M., Dee, C. C., Easton, P., & Spear, N. (1998). Asset write-downs: a decade of research. Journal of Accounting Literature 17, 1–39. Altman, E. I. (1968). Financial ratios, discriminant analysis, and the prediction of corporate bankruptcy. The Journal of Finance 23(4), 589–609. Ball, R. & Shivakumar, L. (2006). The role of accruals in asymmetrically timely gain and loss recognition. Journal of Accounting Research 44(2), 207–242. Barth, M. E., Cram, D. P. & Nelson, K. K. (2001). Accruals and the prediction of future cash flows. The Accounting Review 76(1), 27–58. Basu, S. (1997). The conservatism principle and the asymmetric timeliness of earnings. Journal of Accounting and Economics 24, 3–37. Beatty, A. & Weber, J. (2006). Accounting discretion in fair value estimates: an examination of SFAS 142 goodwill impairments. Journal of Accounting Research 44(2), 257–288. Bernard, V. L. & Thomas, J. K. (1990). Evidence that stock prices do not fully reflect the implications of current earnings for future earnings. Journal of Accounting and Economics 13, 305–340. Bradshaw, M. T. & Sloan, R. G. (2002). GAAP versus the Street: an empirical assessment of two alternative definitions of earnings. Journal of Accounting Research 40(1), 41–66. Burgstahler, D., Jiambalvo, J., & Shevlin, T. (2002). Do stock prices fully reflect the implications of special items for future earnings? Journal of Accounting Research 40(3), 585–612. 28 Dechow, P. M. (1994). Accounting earnings and cash flows as measures of firm performance: the role of accounting earnings. Journal of Accounting and Economics 18, 2–42. Dechow, P. M. & Ge, W. (2006). The persistence of earnings and cash flows and the role of special items: implications for the accrual anomaly. Review of Accounting Studies 11(2/3), 297–303. Dechow, P. M. & Skinner, D. J. (2000). Earnings management: reconciling the views of accounting academics, practitioners, and regulators. Accounting Horizons 14(2), 235–250. Dechow, P. M., Sloan, R. G. and Sweeney, A. P. (1995). Detecting earnings management. The Accounting Review 70(2), 193–225. Elliot, J. A. & Hanna, J. D. (1996). Repeated accounting write-offs and the information content of earnings. Journal of Accounting Research 34(supplement), 135–155. Elliot, J. A. & Shaw, W. H. (1988). Write-offs as accounting procedures to manage perceptions. Journal of Accounting Research 26(supplement), 91–119. Fairfield, P. M. (2006). Discussion of “The persistence of earnings and cash flows and the role of special items: implications for the accrual anomaly.” Review of Accounting Studies 11(2/3), 297–303. Fairfield, P. M., Whisenant, J. S., & Yohn, T. L. (2003). The differential persistence of accruals and cash flows for future operating income versus future profitability. Review of Accounting Studies 8, 221–243. Fields, T. D., Lys, T. Z. & Vincent, L. (2001). Empirical research on accounting choice. Journal of Accounting and Economics 31, 255–307. 29 Financial Accounting Standards Board (2002). Statement of financial accounting concepts No. 142: goodwill and other intangible assets. Norwalk, CT: FASB. Financial Accounting Standards Board (2002). Statement of financial accounting concepts No. 144: accounting for the impairment or disposal of long-lived assets. Norwalk, CT: FASB. Francis, J., Hanna, J. D., & Vincent, L. (1996). Causes and effects of discretionary asset write-offs. Journal of Accounting Research 34(supplement), 117–134. Hayn, C. & Hughes, P. (2006). Leading indicators of goodwill impairment. Journal of Accounting, Auditing and Finance (forthcoming). Healy, P. M. & Wahlen, J. M. (1999). A review of the earnings management literature and its implications for standard setting. Accounting Horizons 13(4), 365–383. Hribar, P. & Collins, D. W. (2002). Errors in estimating accruals: implications for empirical research 40(1), 105–134. Journal of Accounting Research 40(1), 105–134. Jones, J. (1991). Earnings management during import relief investigations. Journal of Accounting Research 29, 193–228. Kinney, M. & Trezevant, R. (1997). The use of special items to manage earnings and perceptions. Journal of Financial Statement Analysis 3(1), 45–53. Kothari, S. P., Leone, A. J., & Wasley, C. E. (2005). Performance matched discretionary accrual measures. Journal of Accounting and Economics 39, 163–197. Mishkin, F. (1983). A Rational Expectations Approach to Macroeconomics: Testing Policy Effectiveness and Efficient-Market Models. Chicago, IL: University of Chicago Press. 30 Rees, L., Gill, S., & Gore, R. (1996). An investigation of asset write-downs and concurrent abnormal accruals. Journal of Accounting Research 34(supplement), 157–169. Richardson, S. A., Sloan, R. G., Soliman, M. T., & Tuna, I. (2005). Accrual reliability, earnings persistence and stock prices. Journal of Accounting and Economics 39, 437–485. Riedl, E. J. (2004). An examination of long-lived asset impairments. The Accounting Review 79(3), 823–852. Sloan, R. G. (1996). Do stock prices fully reflect information in accruals and cash flows about future earnings? The Accounting Review 71(3), 289–315. Strong, J. S. & Meyer, J. R. (1987). Asset write-downs: managerial incentives and security returns. Journal of Finance 42(3), 643 – 661. Watts, R. L. (2003a). Conservatism in accounting part I: explanations and implications. Accounting Horizons 17(3), 207–221. Watts, R. L. (2003b). Conservatism in accounting part II: evidence and research opportunities. Accounting Horizons 17(3), 287–301. White, H. (1980). A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica 48(4), 817–838. Xie, H. (2001). The pricing of abnormal accruals. The Accounting Review 76(3), 357– 373. 31 Table 1 Summary statistics for 13,164 firm-year observations for the period 2002 to 2005. Panel A: Descriptive statistics Full Sample (N=13,164) Total Impairments ≠ 0 Sample (N=3,360) Variable Mean Std dev 25% Median 75% Mean Std dev 25% Median 75% Earnings –0.064 0.271 –0.089 0.025 0.072 –0.117 0.292 –0.173 –0.006 0.049 Pre–impairment earnings –0.052 0.257 –0.071 0.028 0.074 –0.069 0.248 –0.106 0.014 0.060 Total accruals –0.087 0.151 –0.115 –0.060 –0.020 –0.133 0.182 –0.159 –0.083 –0.039 Pre–impairment accruals –0.075 0.136 –0.106 –0.055 –0.017 –0.086 0.139 –0.118 –0.061 –0.025 Cash flows 0.023 0.201 –0.016 0.069 0.131 0.017 0.186 –0.025 0.058 0.119 Total impairments –0.012 0.059 –0.000 0.000 0.000 –0.048 0.109 –0.038 –0.009 –0.002 NACC –0.075 0.069 –0.102 –0.069 –0.041 –0.077 0.067 –0.101 –0.070 –0.043 DACC 0.000 0.118 –0.025 0.008 0.042 –0.009 0.122 –0.033 0.002 0.034 32 Table 1 continued Panel B Pearson (above diagonal) and Spearman (below diagonal) correlations Pre- Total impairment Total Earnings Cash flows accruals accruals NACC DACC impairments Earnings 0.835 0.681 0.612 0.461 0.440 0.329 Cash flows 0.760 0.166 0.133 0.264 0.000 0.118 Total accruals 0.384 –0.160 0.921 0.475 0.789 0.433 Pre-impairment accruals 0.325 –0.203 0.959 0.504 0.864 0.047 NACC 0.156 –0.128 0.505 0.518 0.000 0.051 DACC 0.328 –0.074 0.686 0.717 –0.096 0.024 Total impairments 0.209 0.080 0.225 0.081 0.025 0.078 All correlations are significant at <0.01 level except for the Pearson correlation between cash flows and NACC, and the Pearson correlation between NACC and DACC. The full sample consists of 13,164 firm-year observations from 2002 to 2005. The subsample where total impairments ≠ 0 consist of 3,360 observations. The variables are defined as follows: Earnings is earnings before extraordinary items. Pre- impairment earnings is earnings minus total impairments. Cash flows is cash flow from operations. Total accruals is earnings before extraordinary items minus cash flow from operations. Pre-impairment accruals is total accruals minus total impairments. Total impairments equals the sum of impairment of fixed assets, impairment of goodwill, impairment of other intangibles, and impairment of PPE. NACC is nondiscretionary accruals and it is the fitted value from the Eq. (1). DACC is discretionary accruals and it is the residual from the Eq. (1). All above variables are scaled by average total assets 33 Table 2 Frequency of impairments by portfolios ordered by firm size, where firm size is measured as total assets at the end of previous fiscal year-end Number of Number of % of impairment Portfolio firm-years impairment years years All portfolios 13,164 3,360 25.5 1 (smallest firms) 1,314 208 15.8 2 1,317 212 16.1 3 1,316 246 18.7 4 1,317 288 21.9 5 1,316 305 23.2 6 1,318 330 25.0 7 1,317 375 28.5 8 1,316 409 31.1 9 1,317 436 33.1 10 (largest firms) 1,316 551 41.9 34 (a) 45 % 40 % Frequency of Impairments 35 % 30 % 25 % 20 % 15 % 10 % 5% 0% 1 2 3 4 5 6 7 8 9 10 Rank of Total Accruals Frequency of Impairments greater than 5% Frequency of Impairments greater than 2% Frequency of Impairments greater than 1% Frequency of Impairments greater than 0% (b) 35 % 30 % Frequency of Impairments 25 % 20 % 15 % 10 % 5% 0% 1 2 3 4 5 6 7 8 9 10 Rank of Pre-impairment Accruals Frequency of Impairments greater than 5% Frequency of Impairments greater than 2% Frequency of Impairments greater than 1% Frequency of Impairments greater than 0% Fig. 1 Frequency of impairments across accrual deciles (a) Based on ranks of total accruals. (b) Based on pre-impairment accruals. The sample consist 13,164 firm-years from 2002 to 2005. Firm-year observations are ranked annually and assigned in ascending order to decile portfolios based on total accruals and pre-impairment accruals. Decile 1 consists of firms with the most negative accruals. Decile 10 consists of firms with the most positive accruals 35 Table 3 Results from ordinary least squares regressions of future earnings on cash flows, accruals, and impairments for 13,164 firm-year observations for the period 2002 to 2005 One-year-ahead earnings Variable (2) (3) (4) (5) (6) Intercept 0.009 0.001 –0.023 –0.022 –0.002 (5.15) (0.40) (–6.05) (–5.61) (–0.28) Earnings 0.886 0.982 (30.76) (26.58) DI 0.022 (5.81) Earnings × DI –0.286 (–4.98) Cash flow 1.081 1.089 1.066 (44.17) (44.44) (42.05) Total accruals 0.568 (10.42) Pre-impairment accruals 0.664 (10.03) Impairments 0.084 0.082 (0.90) (0.87) NACC 0.928 (10.63) DACC 0.581 (6.76) Adjusted R2(%) 38.77 39.92 40.94 41.57 41.84 Figures in parenthesis denote t-statistics based on the heteroskedasticity-consistent covariance matrix (White, 1980). A t-statistic of 2.58 implies a significance level of 0.01 using a two-tailed test. A t-statistic of 1.96 implies a significance level of 0.05 using a two-tailed test. Earnings is earnings before extraordinary items. Cash flows is cash flow from operations. Total accruals is earnings before extraordinary items minus cash flow from operations. Impairments equals the sum of the impairment of fixed assets, impairment of goodwill, impairment of other intangibles, and impairment of PPE. Pre-impairment earnings is earnings minus impairments. Pre-impairment accruals is total accruals minus impairments. NACC is nondiscretionary accruals and it is the fitted value from the Eq. (1). DACC is discretionary accruals and it is the residual from the Eq.(1). All variables are scaled by average total assets 36 Table 4 Nonlinear generalized least squares estimation (the Mishkin test) of the market pricing of cash flows, nondiscretionary accruals, discretionary accruals, and impairments with their respect to their implications for one-year-ahead earnings Panel A: Earningst+1 = α + δ1Cash flowst + δ2Pre-impairment accrualst + δ3Impairmentst + εt+1 Size-adjusted returnt+1 = β(Earningst+1 – α – δ*1Cash flowst – δ*2Pre-impairment accrualst – δ*3Impairmentst) + εt+1 Forecasting coefficients Valuation coefficients Parameter Coefficient estimate (t-statistic) Parameter Coefficient estimate (t-statistic) δ1 1.089 (84.00) δ*1 0.907 (8.83) δ2 0.664 (34.88) δ*2 0.919 (6.09) δ3 0.084 (1.92) δ*3 2.133 (5.73) β 0.521 (14.58) Test of market efficiency Null hypothesis Likelihood ratio statistic Marginal significance level δ1 = δ*1 3.15 0.076 δ2 = δ*2 2.85 0.091 δ3 = δ*3 34.75 <0.001 All δ* = δ* 11.11 0.004 All δj = δ*j 39.02 <0.001 j = 1 to 3 37 Table 4 continued Panel B: Earningst+1 = α + δ1Cash flowst + δ2NACCt + δ3DACCt + δ4Impairmentst + εt+1 Size-adjusted returnt+1 = β(Earningst+1 – α – δ*1Cash flowst – δ*2NACCt – δ*3DACCt – δ*4Impairmentst) + εt+1 Forecasting coefficients Valuation coefficients Parameter Coefficient estimate (t-statistic) Parameter Coefficient estimate (t-statistic) δ1 1.065 (80.20) δ*1 0.911 (8.65) δ2 0.927 (23.97) δ*2 0.871 (2.85) δ3 0.581 (26.63) δ*3 0.934 (5.38) δ4 0.082 (1.88) δ*4 2.135 (5.72) β 0.521 (14.54) Test of market efficiency Null hypothesis Likelihood ratio statistic Marginal significance level δ1 = δ*1 2.14 0.144 δ2 = δ*2 0.03 0.856 δ3 = δ*3 4.18 0.041 δ4 = δ*4 34.83 <0.001 All δ* = δ* 11.13 0.011 All δj = δ*j 40.37 <0.001 j = 1 to 3 Earnings is earnings before extraordinary items. Cash flows is cash flow from operations. Total accruals is earnings before extraordinary items minus cash flow from operations. Impairments equals the sum of impairment of fixed assets, impairment of goodwill, impairment of other intangibles, and impairment of PPE. Pre-impairment accruals is total accruals minus impairments. Pre-impairment accruals is total accruals minus impairments. NACC is nondiscretionary accruals and it is the fitted value from Eq. (1). DACC is discretionary accruals and it is the residual from Eq. (1). All above variables are scaled by average total assets Annual returns are calculated from the start of the fourth month subsequent to the fiscal year-end. The size-adjusted return is calculated by deducting the value-weighted average return for firms in the same size-matched decile, where size is measured as the market value at the beginning of return cumulation period. 38 Table 5 Descriptive statistics for samples of firms formed by the magnitude of impairment write-offs relative to average total assets. Panel A: Values of select characteristics Full Sample IM = 0 IM ≠ 0 IM > 1% IM > 2% IM > 5% (N=13,164) (N=9,804) (N=3,360) (N=1,602) (N=1,165) (N=728) Earnings (mean) –0.064 –0.046 –0.117 –0.246 –0.304 –0.400 Cash flows (mean) 0.023 0.025 0.017 –0.039 –0.060 –0.085 Total accruals (mean) –0.087 –0.072 –0.133 –0.207 –0.244 –0.316 NACC (mean) –0.075 –0.075 –0.077 –0.087 –0.091 –0.099 DACC (mean) 0.000 0.003 –0.009 –0.023 –0.025 –0.032 Asset growth (median) 0.049 0.066 –0.003 –0.089 –0.137 –0.206 Sales growth (median) 0.084 0.097 0.049 0.006 –0.008 –0.023 Percentage of loss firms (%) 38.7 34.1 52.0 77.0 85.3 93.5 Percent having restructuring expense (%) 18.4 14.8 28.9 30.5 31.4 31.9 Altman’s Z-score (median) 3.00 3.28 2.29 1.36 0.88 0.22 Total assetst–1 (million $) (median) 200.1 153.0 408.9 170.0 131.6 119.4 Book to market (median) 0.438 0.425 0.470 0.481 0.479 0.492 Panel B: Stock price performance and investor recognition in period t Full Sample IM = 0 IM ≠ 0 IM > 1% IM > 2% IM > 5% (N=13,164) (N=9,804) (N=3,360) (N=1,602) (N=1,165) (N=728) Return (mean) 0.311 0.332 0.249 0.232 0.240 0.256 Size-adjusted return (mean) 0.018 0.025 –0.003 –0.094 –0.091 –0.086 Percent having EPS forecast (%) 62.6 59.9 70.3 57.1 54.2 51.8 Analyst EPS forecast error (median) 0.005 0.008 –0.005 –0.037 –0.042 –0.036 No. of EPS forecasts (median) 5 5 7 5 5 5 Percent having sales forecast (%) 57.2 54.6 64.8 52.3 49.4 48.1 Analyst sales forecast error (median) 0.004 0.010 –0.008 –0.041 –0.050 –0.072 No. of sales forecasts (median) 4 3 4 4 4 4 39 Table 5 continued Panel C: Stock price performance and investor recognition in period t+1 Full Sample IM = 0 IM ≠ 0 IM > 1% IM > 2% IM > 5% (N=13,164) (N=9,804) (N=3,360) (N=1,602) (N=1,165) (N=728) Return (mean) 0.387 0.367 0.445 0.609 0.670 0.714 Size-adjusted return (mean) 0.029 0.013 0.078 0.118 0.135 0.142 Percent having EPS forecast (%) 64.5 62.7 69.6 54.2 50.6 48.1 Analyst EPS forecast error (median) 0.013 0.010 0.025 0.035 0.008 0.037 Change in no. of EPS forecasts (%) 19.6 22.6 12.2 3.6 1.0 –6.7 Percent having sales forecast (%) 61.2 59.5 66.0 51.4 47.9 44.6 Analyst sales forecast error (median) 0.010 0.011 0.007 0.001 –0.003 –0.008 Change in no. of sales forecasts (%) 43.6 45.7 38.5 26.3 23.5 16.5 Panel D: Descriptive statistics for firms without impairment write-offs and firms with impairment write-offs IM = 0 IM ≠ 0 Independent variable (N=9,804) (N=3,360) Difference (p-value) DACC (mean)a 0.003 –0.009 0.012 (<0.001) b Asset growth (median) 0.066 –0.003 0.069 (<0.001) Sales growth (median)b 0.097 0.049 0.048 (<0.001) c Percentage of loss firms (%) 34.1 52.0 –17.9 (<0.001) Percent having restructuring expense (%)c 14.8 28.9 –14.1 (<0.001) Altman’s Z-score (median)b 3.28 2.29 0.99 (<0.001) Analyst EPS forecast errort (median)b 0.008 –0.005 0.013 (0.146) Analyst sales forecast errort (median)b 0.010 –0.008 0.018 (<0.001) b Analyst EPS forecast errort+1 (median) 0.010 0.025 –0.015 (<0.001) Analyst sales forecast errort+1 (median)b 0.011 0.007 0.004 (0.406) a Change in no. of EPS forecasts (%) 22.6 12.2 10.4 (<0.001) Change in no. of sales forecasts (%)a 45.7 38.5 7.2 (0.007) a Size-adjusted returnt (mean) 0.024 –0.003 0.027 (0.465) Size-adjusted returnt+1 (mean)a 0.013 0.078 –0.065 (0.008) 40 a Reported numbers are means and p-value is based on t-test. b Reported numbers are medians and p-value is based on Wilcoxon signed-rank test. c Reported numbers are means and p-value is based on chi-square test. The full sample covers 13,164 firm-year observations for the period 2002 to 2005. Subsamples are formed based on the relative magnitude of impairments to average total assets. Earnings is earnings before extraordinary items. Cash flows is cash flow from operations. Total accruals is earnings before extraordinary items minus cash flow from operations. NACC is non-discretionary accruals and it is the fitted value from Eq. (1). DACC is discretionary accruals and it is the residual from Eq. (1). Asset growth is calculated as (total assetst – total assetst–1)/total assetst–1. Sales growth is calculated as (salest – salest–1)/salest–1. Percent of loss firms is the percentage of firm-years that have negative earnings before extraordinary items. Percent having restructuring expense is the percentage of firm-years that have reported restructuring expenses. Altman’s Z-score is calculated as Z = 1.2 (working capital/total assets) + 1.4 (retained earnings/total assets) + 3.3 (earnings before interest and taxes/total assets) + 0.6 (market value of equity at the end of previous year/total liabilities) + 1.0 (sales/total assets). Total assetst–1 is the total assets at year t–1. Book to market is the ratio of the book value of equity to the market value of equity at the end of fiscal year t. Percent having EPS/sales forecast refers to the percentage of firm-year observations where there is an analyst EPS/sales forecasting during the fourth month after the fiscal year-end of year t in IBES. Analyst EPS forecast error equals (Realized EPSt – Analyst EPS forecastt)/Realized EPSt–1 obtained from IBES. Analyst sales forecast error equals (Realized salest – Analyst sales forecastt)/Realized salest–1 obtained from IBES. No. of EPS/sales forecasts is the number of analysts that issue EPS/sales forecast during the eighth month before the fiscal year-end of year t. Change in no. of EPS/sales forecast is the percentage change in the number of analysts that issue EPS/sales forecast, which is calculated as (No. of EPS/sales forecasts at year t+1 – No. of EPS/sales forecasts at year t)/No. of EPS/sales forecasts at year t. Annual returns are calculated from the start of the eighth month before the fiscal year-end. The size-adjusted return is calculated by deducting the value-weighted average return for firms in the same size-matched decile, where size is measured as the market value at the beginning of the return cumulation period. 41 Table 6 Future size-adjusted returns on impairment write-offs and other predictors of returns Variable One-year-ahead size-adjusted return Intercept 0.545 (4.83) Impairment write-offs indicator 0.063 (2.54) Size –0.031 (–4.81) Total accruals –0.149 (–2.03) Size-adjusted return from t –1 to t –0.029 (–3.75) EPS forecast indicator 0.088 (2.88) ∆EPS forecasts × EPS forecast indicator –0.022 (–1.12) Book to market –0.003 (–3.03) No. of obs. 13,081 Adj. R2 (%) 0.44 The t-statistics are in the parentheses. Impairment write-off indicator is an indicator variable, taking the value of 1 when the firm reports impairment write-offs. Size is the natural logarithm of the market value of equity measured at fiscal year-end. Total accruals is earnings before extraordinary items minus cash flow from operations. EPS forecast indicator is an indicator variable, taking the value of 1 when there is an analyst EPS forecast during the fourth month after the fiscal year-end of year t in IBES. ∆EPS forecasts is the percentage change in the number of analysts that issue EPS forecast, which is calculated as (No. of EPS forecasts at year t+1 – No. of EPS forecasts at year t)/No. of EPS forecasts at year t. Book to market is the ratio of the book value of equity to the market value at the end of fiscal year t. Annual returns are calculated from the start of the fourth month subsequent to the fiscal year-end. The size-adjusted return is calculated by deducting the value-weighted average return for firms in the same size-matched decile, where size is measured as the market value at the beginning of the return cumulation period. 42 Table 7 The ability of impairments to predict future cash flows for the period 2002 to 2005. Panel A Cash flowst+1 Cash flowst+2 Cash flowst+3 Variable I II I II I II Intercept 0.031 0.031 0.048 0.050 0.061 0.061 (13.66) (12.37) (12.43) (11.58) (8.01) (7.23) Cash flowt–1 0.005 0.003 0.005 0.004 –0.002 –0.004 (0.21) (0.13) (0.10) (0.07) (–0.02) (–0.05) Accrualst–1 –0.011 –0.009 –0.016 –0.014 –0.017 –0.017 (–1.38) (–1.24) (–1.04) (–0.88) (–1.13) (–1.05) Cash flowt 0.896 0.933 0.966 1.019 1.077 1.135 (35.51) (36.84) (17.66) (18.84) (10.69) (10.89) Accrualst 0.206 0.268 0.306 0.410 0.275 0.367 (8.55) (8.04) (7.03) (7.05) (4.55) (4.62) DIt 0.009 0.004 0.012 (1.20) (0.32) (0.58) Cash flowt × DIt –0.228 –0.303 –0.278 (–5.80) (–4.22) (–2.34) Accrualst × DIt –0.107 –0.194 –0.137 (–2.30) (–2.55) (–1.21) No. of obs. 13,146 13,146 9,530 9,530 6,064 6,064 2 Adj. R (%) 54.58 55.27 37.67 38.51 23.66 24.01 Figures in parenthesis denote t-statistics based on the heteroskedasticity-consistent covariance matrix (White, 1980). A t-statistic of 2.58 implies a significance level of 0.01 using a two-tailed test. A t-statistic of 1.96 implies a significance level of 0.05 using a two-tailed test. Cash flows is cash flow from operations. Accruals is earnings before extraordinary items minus cash flow from operations. DI is a dummy variable (N=1,602), taking the value of 1 when the magnitude of impairment is greater than or equal to the 1% average total assets. Variables are standardized by average total assets in period t.