Does Analyst Forecasting Behavior Explain Anomalous Stock Market Reactions to Information in Cash and Accrual Earnings Components?

Transcription

1 Does Analyst Forecasting Behavior Explain Anomalous Stock Market Reactions to Information in Cash and Accrual Earnings Components? Dana Hollie a, Phil Shane b, Qiuhong Zhao c a Louisiana State University b University of Virginia and University of Auckland c University of Missouri at Columbia Abstract This study examines whether analysts forecasting behavior can explain market inefficiencies with respect to cash and accrual earnings components. Specifically, we focus on whether analysts earnings forecasts impound persistence characteristics of the retained and distributed free cash flow components of cash, and the sales growth and efficiency change components of accruals. In a sample of relatively large firms in recent years, our evidence points toward generalized analyst underreaction across all earnings components, market underreaction to the distributed-to-debtholder component of cash, and market overreaction to the efficiency change component of accruals. Our findings are consistent with misunderstanding of the implications of leverage changes for earnings persistence. Our findings are also consistent with accounting distortions that fool investors but not financial analysts. While we find no evidence that analyst behavior explains market overreaction to accruals, we estimate that analyst behavior can explain 35% of market underreaction to the distributed-to-debtholder component of cash. JEL classification: G14, M41 Keywords: Analyst Forecast, Accruals, Cash Flows, Market Efficiency We thank the Leeds School of Business for the Hart Fellowship support in We thank I/B/E/S for providing the analysts forecast data for this study. We would also like to thank Katherine Gunny, Steve Rock and workshop participants at Colorado State University, the University of Colorado at Boulder, the 2008 Annual AAA meeting and our discussant Sudipta Basu for their helpful comments. Qiuhong Zhao thanks the Leeds School of Business for the Hart Summer Fellowship support. * addresses: danahollie@lsu.edu (D. Hollie), Phil.Shane@Virginia.edu or p.shane@auckland.ac.nz (P. Shane), and zhaoq@missouri.edu (Q. Zhao).

2 I. Introduction As information intermediaries financial analysts play an important role in conveying information to the stock market. 1 For that reason, whether analysts efficiently process accounting information remains an interesting research question. This paper investigates whether financial analysts forecasting behavior is a potential explanation for anomalous market response to information in cash and accrual earnings components. Comparatively little prior research evaluates the efficiency of financial analysts' earnings forecasts with respect to information in the cash flow and accrual components of earnings, and even less research evaluates the degree to which the behavior of analysts' earnings forecasts might explain accrual and cash flow anomalies. The purpose of this paper is to evaluate analyst forecasting behavior as a potential explanation for apparent stock market inefficiencies with respect to disaggregated cash and accrual components of accounting earnings. Our two primary research questions are: Do analysts efficiently process information in components of earnings? Does analyst forecasting behavior explain market inefficiencies with respect to information in components of cash and accrual earnings? Moreover, this study is the first to examine analysts ability to differentiate between the persistence of the retained and distributed components of cash earnings. Building on prior research (e.g., Richardson et al., 2001; Richardson et al., 2005; and Dechow et al., 2008), we examine analyst efficiency with respect to earnings disaggregated into cash and accrual components, with the accruals portion further disaggregated into asset efficiency change and sales growth components, and the cash portion further disaggregated into distributed and retained components. Given that we evaluate analysts forecasts, our sample necessarily differs from the samples used in prior studies of investor, but not analyst, response to 1 See Schipper (1991), Brown (1993) and Ramnath et al. (2008) for literature reviews describing the role of financial analysts in capital markets. 1

3 information contained in earnings components. In particular, relative to the prior returns-oriented research, our sample is constrained to more recent years and larger firms included in the I/B/E/S database. Nonetheless, we are able to replicate the key prior research findings that the market appears to overreact to the total accruals component of earnings and, in particular, to the portion of accruals due to changes in asset efficiency. 2 Contrary to Dechow et al., in our more recent sample of larger firms, we find that retained cash earnings are as persistent as distributed cash earnings, and we find that the market appears to: (i) react efficiently to the persistence of retained and distributed-to-stockholder cash earnings; and (ii) underreact to the persistence of distributed-to-debtholder cash earnings. On the other hand, Dechow et al. (2008) find that the market appears to react efficiently to the distributed portion of cash earnings and overreact to the persistence of the retained portion of cash earnings. 3 Clearly, differences in sample characteristics may play a role in differences between the evidence in our study and Dechow et al. (2008). Although beyond the scope of our study, future research might explore the influence of firm-size and analyst following on the relative persistence of retained versus distributed cash earnings, distributed-to-stockholder versus distributed-to-debtholder cash earnings, and market efficiency with respect to all three components. 4 2 Our evidence that efficiency changes drive the market overreaction to accrual earnings is in contrast to Hribar and Yehuda (2008), who provide evidence suggesting that accruals mispricing arises from misunderstanding of returns to the firm s growth opportunities, particularly during the firm s growth stage. Subsequent research might explore whether the sales growth versus efficiency change explanation for accruals mispricing depends on the stage of the firm in its life cycle. 3 However, as discussed below, our results are consistent with Dimitrov and Jain (2008) who find that the market underreacts to the association between increases in leverage and deteriorating operating performance. 4 Each earnings component can take on positive or negative values. For example, the distributed-to-debtholder component of cash earnings is positive in the case of net payments to debtholders (i.e., debt-reducing cash flows) or negative in the case of net receipts from debtholders (i.e., debt-increasing cash flows). Similarly, the change in asset efficiency and sales growth components of accrual earnings can take on positive or negative values, and both the retained and distributed-to-equity components of cash earnings can take on positive or negative values. 2

4 Additionally, we find that analysts appear to underreact to both cash and accrual components of earnings. We also find similar analyst underreaction across all three cash components of earnings (retained cash and cash distributed to both debtholders and stockholders). Furthermore, we find that analyst underreaction to the sales growth (rather than the efficiency change) component of accruals drives analysts overall underreaction to the persistence of accrual earnings. Thus, we find no evidence that analyst forecasting behavior explains apparent market overreaction to accrual earnings. 5 Our evidence of analyst underreaction to the net distributionto-debtholders component of cash earnings is consistent with our evidence of similar market underreaction. However, our empirical findings provide evidence of a point estimate that only 35% of the market underreaction can be attributed to analyst underreaction. Our evidence of analyst underreaction to accrual earnings is inconsistent with Bradshaw et al. (2001), who find that analysts overreact to the persistence of working capital accruals in a manner consistent with the market overreaction observed in Sloan (1996). However, our results are consistent with Yu (2007) who provides evidence that Bradshaw et al. s (2001) omission of the cash earnings variable explains the differences in results between the two studies. Since cash and accrual earnings are negatively correlated, a positive relation between cash earnings and subsequent forecast errors may explain the Bradshaw et al. (2001) finding that accruals are negatively related to future forecast errors. In fact, when we omit the cash variable, consistent with Bradshaw et al. (2001), we find a negative correlation between accruals and future forecast errors. However, when we include the cash variable, consistent with Yu (2007), we find that analysts forecasts underreact to the persistence of accrual earnings. 5 This interpretation is consistent with Hughes et al. (forthcoming) who find that trading strategies based on predictable analyst forecast errors are not profitable. 3

5 Our empirical results, suggesting that analysts generally underreact to earnings persistence across all components of accruals and cash flows, are consistent with the theory of analyst underreaction developed in Raedy et al. (2006). Raedy et al. (2006) develop a mathematical model predicting generalized analyst underreaction, given an asymmetric loss function that punishes analysts more severely for reversing the direction of their earnings forecasts in light of new information. This view suggests that analysts are sophisticated users of financial accounting information and, as such, understand the persistence properties of various earnings components. However, economic incentives lead analysts to underreact, and the underreaction is apparent across all earnings components. 6 On the other hand, the market includes unsophisticated investors, who apparently overreact to the persistence of some accrual earnings components and underreact to some cash earnings components, and market frictions prevent sophisticated arbitragers from fully exploiting these inefficiencies. 7 The rest of the paper is organized as follows. The next section discusses prior research, and section III describes our research design. Section IV describes our sample, section V presents our results, and section VI summarizes and concludes the paper. II. Prior Research 6 Also, see Francis and Philbrick (1993) who find evidence suggesting that economic incentives lead analysts to bias their earnings forecasts upward in the face of bad news, and see Trueman (1990) who develops a mathematical model suggesting that analysts rationally underreact to public information to disguise relatively low ability in developing timely private information. 7 Lev and Nissim (2006) and Mashruwala et al. (2006) indicate that the accrual anomaly persists and will probably endure. The explanation for why the accrual anomaly is not arbitraged away by sophisticated institutional investors is that institutions avoid extreme-accrual firms because of their attributes, such as small size, low profitability, and high risk and that sophisticated individual investors are unable to profit from trading on accruals information due to the high information and transaction costs. Also see Hirshleifer and Teoh (2003) who argue that non-diversifiable risk associated with trading against inefficient unsophisticated investor behavior limits sophisticated investor arbitrage trading that might otherwise eliminate the mispricing. 4

6 Beginning with Sloan (1996), researchers have investigated stock market efficiency with respect to information in accrual and cash flow components of earnings. While Sloan examined the market s response to operating cash flows and working capital accruals minus depreciation, the research has evolved to consider the market response to total accruals, accruals disaggregated into various components, and cash flows disaggregated into retained and distributed components. Richardson et al. (2005) find that the market overreacts to the persistence of both current (non-cash working capital) and non-current operating accruals and that trading strategies based on total operating accruals generate more profits than strategies based on current operating accruals. Richardson et al. (2001, 2006) disaggregate total operating accruals into sales growth and asset efficiency (i.e., turnover) components. Richardson et al. (2001) find that the market overreacts to the persistence of both the sales growth and asset efficiency change components of total accruals. The authors interpret this result to support both the perspective of Fairfield et al. (2003), Cooper et al. (2005) and others that the accrual anomaly emerges from diminishing returns on firms' growth opportunities and the perspective of Sloan (1996), Xie (2001) and others that the accrual anomaly emerges from accounting distortions. We extend this research by examining whether analysts forecasting behavior can explain market inefficiency with respect to total operating accruals, as well as the sales growth and asset efficiency change components of total operating accruals. Dechow et al. (2008) disaggregate the cash component of earnings and find that the market reacts efficiently to the persistence of the portion of cash earnings that the firm distributes to investors; whereas, the market overreacts to the persistence of the portion of cash earnings that the firm retains. This contradicts prior research evidence that the market underreacts to total cash earnings (e.g., Sloan 1996; Desai et al., 2004; Ahmed et al. 2006). In a sample of larger firms 5

7 followed by analysts during a more recent time period, we re-examine the market's response to cash components of earnings, and extend the literature by examining whether analysts forecasting behavior can explain the market s response to the cash components of earnings as defined in Dechow et al. (2008). Contrary to Dechow et al. (2008), but consistent with Dimitrov and Jain (2008) and Penman et al. (2007), we find that the market underreacts to the persistence of the distributed-to-debtholder component of cash earnings. As described by Dimitrov and Jain (2008), if managers have private information about deteriorating operating performance, they may increase debt to provide funds to replace anticipated shortfalls. Dimitrov and Jain (2008) provide evidence that the market appears to underreact to the negative signal about future performance associated with current increases in debt (which, in our study, corresponds to a negative distributed-to-debtholder component of cash earnings). Our evidence that the market underreacts to the persistence of the distributed-to-debtholder component of current cash earnings is consistent with the evidence and discussion in Dimitrov and Jain (2008). Similarly, Penman et al. (2007) find a negative relation between the leverage component of the book-toprice ratio and subsequent stock returns. 8 In addition to reexamining the persistence and market reaction to information in cash earnings components, our study investigates the efficiency of analysts earnings forecasts with respect to information in cash earnings components, and we also investigate analyst forecasting behavior as an explanation for market inefficiencies with respect to disaggregated cash earnings. 8 Bradshaw, Richardson, and Sloan (2006) identify two competing hypotheses about the relationship between future stock returns and changes in debt: Ritter s (2003) misvaluation hypothesis and Eberhart and Siddique s (2002) wealth transfer hypothesis. Consistent with Bradshaw et al. (2006), our evidence of a positive relation between the distributed-to-debtholder component of cash earnings and both subsequent analyst forecast errors and subsequent stock return performance supports the misvaluation hypothesis. That is, our evidences suggests that analysts forecasts and stock prices are too low (high) following distributions to (from) debtholders, which means decreases (increases) in debt lead to positive (negative) future returns and forecast errors (consistent with Bradshaw et al.). 6

8 7

9 III. Research design Our investigation of analyst and investor response to cash and accrual earnings components, respectively, builds on earnings disaggregations developed by Richardson et al. (2001, 2006) and Dechow et al. (2008). Consistent with Dechow et al. (2008), we disaggregate earnings into free cash flow (cash) and accrual components as follows (firm subscripts suppressed throughout all models): RNOA t = FCF t + ΔNOA t (1) where RNOA t = operating income after depreciation, FCF t = free cash flow, and ΔNOA t = total operating accruals, all deflated by NOA t-1 (non-cash net operating assets). 9 Dechow et al. (2008) disaggregate FCF t (cash earnings) into retained and distributed components as follows: FCF t = ΔCASH t + DIST_D t + DIST_EQ t (2) where ΔCASH t = the change in cash and short-term investments (i.e., free cash flow retained), DIST_D t = free cash flow to debtholders (excluding interest), and DIST_EQ t = free cash flow to stockholders (net income minus the change in stockholders equity), all deflated by NOA t Consistent with Richardson et al. (2001, 2006), we define and disaggregate total operating accruals as follows: ACC t = SG t ΔEFF t (SG t * ΔEFF t ) (3) where: ACC t represents total operating accruals, defined as the percentage change in the firm s net operating assets; SG t = ΔSales t /Sales t-1 and represents sales growth; ΔEFF t = ΔAT t /AT t and represents the change in asset efficiency measured as (Sales t /NOA t Sales t-1 /NOA t-1 ) / (Sales t /NOA t ); and NOA t = total non-cash current and non-current operating assets less total 9 Theoretically, RNOA t in (1) should be NOI t (i.e., net operating income after tax). Following Richardson et al. (2005, 2006), we simply represent RNOA t as income after depreciation (Compustat item # 178). 10 Theoretically, distributions to preferred stockholders should be included in DIST_D. Following Dechow et al., we include distributions to/from preferred stockholders in DIST_EQ. 8

10 current and non-current operating liabilities (with cash treated as a financial asset). Richardson et al. (2006) provide an algebraic proof showing that the three terms on the RHS of (3) sum to total accruals (ACC t ). 11 The authors argue that this decomposition allows tests with the potential to distinguish between two competing explanations for the accrual anomaly: market failure to fully impound mean reversion in sales growth versus market failure to fully impound information in accruals about temporary accounting distortions. If diminishing marginal returns on investment drive the lower persistence of accruals, this should be reflected in the growth component (SG t ). In contrast, if accounting distortions or changes in real operating efficiency drive the lower persistence of accruals, then this should be reflected in the efficiency component (EFF t ). 12 Summarizing (1), (2), and (3) above, our fully disaggregated model of net income becomes: RNOA t = (ΔCASH t + DIST_D t + DIST_EQ t ) + [SG t ΔEFF t (SG t * ΔEFF t )] (4) where the first bracketed term on the RHS of (4) disaggregates free cash flow (FCF t ), and the second bracketed term disaggregates accruals (ACC t ). Following Dechow et al. (2008), we examine the persistence of the various earnings components by replacing RNOA t in model (4) with RNOA t+1 in the summary components and detailed components regression models (5) and (6) below. RNOA t+1 = θ 0 + θ 1 FCF t + θ 2 ACC t + η t+1 (5) 11 Appendix A provides detailed definitions of each variable used in the study, along with Compustat item numbers. 12 Results in Hribar and Yehuda (2008) suggest that market overreaction to the persistence of firms growth opportunities, rather than overreaction to the persistence of accounting accruals, drives the accruals anomaly, particularly for growth firms early in their life cycle. If this is the case, then we expect that, controlling for market reaction to cash earnings components, the sales growth component should drive any market overreaction to accruals. Our results do not support this prediction and, instead, suggest that the market overreaction to accruals is driven by overreaction to the persistence of the changes in asset efficiency component, rather than the sales growth component, of total accruals. 9

11 RNOA t+1 = λ 0 + λ 1 CASH t + λ 2 DIST_EQ t + λ 3 DIST_D t + λ 4 SG t + λ 5 ΔEFF t + λ 6 (SG t * ΔEFF t ) + ξ t+1 (6) Positive θ 1 and θ 2 in model (5) indicate that the summary earnings components, FCF t and ACC t, respectively, are positively related to future earnings (i.e., they persist). Positive λ 1, λ 2, λ 3, and λ 4 and negative λ 5 in model (6) indicate that the detailed earnings components, CASH t, DIST_EQ t, DIST_D t, SG t, and ΔEFF t, respectively, are positively related to future earnings (i.e., they persist). Next, we evaluate market efficiency with respect to the various components of cash and accrual earnings. RET t+1 = α 0 + α 1 FCF t + α 2 ACC t + u t+1 (7) RET t+1 = γ 0 + γ 1 (ΔCASH t ) + γ 2 (DIST_EQ t ) + γ 3 (DIST_D t ) + γ 4 (SG t ) + γ 5 (ΔEFF t ) + γ 6 (SG t *ΔEFF t ) + e t+1 (8) where RET t+1 = firm i s raw returns, accumulated from the fifth month following the end of fiscal year t through the fourth month following the end of fiscal year t+1, minus the similarly accumulated mean return of all firms in the same size-decile as firm i (with size deciles formed as of the end of fiscal year t). 13 All empirical models exclude firm-year observations with negative values of NOA t-1 and, following Richardson et al. (2005), all financial statement ratios are winsorized at +/-1 to mitigate the influence of outliers. If, as described in Richardson et al. (2001), stock prices overreact to the persistence of sales growth and efficiency components of 13 For sensitivity analysis, we accumulate returns starting from the beginning of the fourth month following the end of fiscal. The results are qualitatively the same. 10

12 total accruals, then we expect α 2 < 0 in (7) above, and we expect γ 4 < 0 and γ 5 > 0 in (8) above. 14 If, as described in Dechow et al. (2008), stock prices overreact to the persistence of the retained component of cash earnings, then we expect γ 1 < 0 in (8) above. 15 To evaluate any inefficiency in analysts earnings forecasts following the release of financial statement information containing cash and accrual earnings components, we estimate summary model (9) and components model (10) below. FE t+1 = β 0 + β 1 FCF t + β 2 ACC t + ε t+1 (9) FE t+1 = δ 0 + δ 1 (ΔCASH t ) + δ 2 (DIST_EQ t ) + δ 3 (DIST_D t ) + δ 4 (SG t ) + δ 5 (ΔEFF t ) + δ 6 (SG t *ΔEFF t ) + ω t+1 (10) where FE t+1 = (A t+1 F t+1 ) / P t ; A t+1 = actual earnings for year t+1 per I/B/E/S; F t+1 is the most recent individual analyst forecast of year t+1 earnings, released prior to the beginning of the return accumulation period; and P t is the adjusted price taken from the same I/B/E/S monthly report containing F t If analysts issue efficient forecasts following firms release of their financial statements containing cash and accrual components of earnings, then the information in those financial statements should not predict analysts forecast errors (FE t+1 ), and the coefficients on the information variables in (9) and (10) should equal zero. The intercept term (β 0 or δ 0 ) 14 Controlling for sales growth, increases in asset efficiency (ΔEFF t ) correspond to decreases in NOA t and decreases in accruals [see model (3) above]. Thus, if investors overreact to the persistence of accruals components of earnings, then increases in ΔEFF t (and decreases in accruals) should correspond to lower stock prices, higher subsequent returns (as correction occurs), and γ 5 > 0 in model (8). 15 Rather than using the framework developed by Mishkin (1983), we use a conventional OLS model to test the rational expectations hypotheses. Kraft et al. (2007) show that OLS and the Mishkin test generate identical inferences in accounting settings when samples are large. As described in Kraft et al. (2007), to obtain estimates of the difference between actual and implied market estimates of the persistence of earnings components, the residual terms from models (5) and (6) are added as independent variables in models (7) and (8). Then the difference between actual and implied market persistence estimates are derived by dividing each component variable coefficient by the residual variable coefficient in models (7) and (8). 16 When more than one individual analyst forecast occurs on the day we use the median of those forecasts. 11

13 captures analysts optimism/pessimism. For example, a significantly negative coefficient for the intercept term would indicate systematically optimistic analysts forecasts. To assess whether analyst forecasting behavior potentially explains market inefficiencies with respect to information in cash and accrual components of earnings, we estimate summary model (11) and components model (12). RET t+1 = ρ 0 + ρ 1 FCF t + ρ 2 ACC t + ρ 3 FE t+1 + ρ 4 FREV t+2 + e t+1 (11) RET t+1 = φ 0 + φ 1 (ΔCASH t ) + φ 2 (DIST_EQ t ) + φ 3 (DIST_D t ) + φ 4 (SG t ) + φ 5 (ΔEFF t ) + φ 6 (SG t )*(ΔEFF t ) + φ 7 FE t+1 + φ 8 FREV t+2 + μ t+1 (12) where FREV t+2 = ( t+1 F t+2 t F t+2 ) / P t ; t+1 F t+2 is the first individual analyst forecast of t+2 earnings issued after the end of the return accumulation period; and t F t+2 is the most recent forecast of t+2 earnings issued prior to the beginning of the return accumulation period. Following Richardson et al. (2001, 2005), we use models (7) and (8) above to assess market efficiency with respect to earnings components in our sample and during our time period. If analyst forecasting behavior explains market inefficiency with respect to the information in cash and accrual components of earnings, then we expect the coefficients on the cash and accrual earnings information variables in models (9) and (10) to have the same signs and similar significance levels as the coefficients on the same information variables in models (7) and (8), and we expect the coefficients on the same information variables in models (11) and (12) to equal zero. As described in Shane and Brous (2001), if analyst forecasting behavior explains market inefficiency with respect to information about future earnings, then adding the forecast error and forecast revision variables to the returns regression should make the coefficients on the information variables go to zero. 12

14 IV. Sample Since we rely on I/B/E/S to measure the forecast error (FE t+1 ) and forecast revision (FREV t+2 ) variables in models (9) through (12), our sample represents larger firms and a more recent time period relative to the sample and time periods in Dechow et al. (2008) and Richardson et al. (2001, 2005). Our time period spans the years ; whereas, the Richardson et al. (2001) time period spans , the Richardson et al. (2005) time period spans , and the Dechow et al. (2008) time period spans To estimate the variables in our models, we obtain: financial statement data from Compustat; returns data from CRSP; and earnings forecasts, actual earnings, and stock prices from I/B/E/S. Table 1 describes our sample selection procedure. Our initial sample finds 186,928 firm-year observations on Compustat s Annual Industrial, Research, and Full Coverage files between the years, 1988 and Following Richardson et al. (2005) and Richardson et al. (2006), we exclude firms in the financial services industry (32,923 firm-years with SIC codes in the range ). We omit 56,001 observations without Compustat data needed to compute our cash flow and accrual earnings component variables or with negative net operating assets. We lose 29,649 observations missing the CRSP data needed to compute our returns variable. We lose another 51,661 firm-year observations without I/B/E/S data needed to compute our forecast error variable (FE), and we lose 7,929 observations without I/B/E/S data needed to compute our forecast revision variable (FREV), resulting in a final sample of 8,755 firm-years spanning the time-frame. {Insert Table 1 about here} 13

15 V. Results Descriptive statistics Table 2 presents descriptive statistics, including univariate correlations between variables used in the study. The sample firms exhibit relatively high sales growth (15.8% on average), slightly optimistic analyst forecasts (FE < 0), and horizon-dependent optimism (FREV < 0). Most firm-years are associated with distributions of cash to stockholders (median DIST_EQ = 0.109), increases in debt (median DIST_D = ), and positive free cash flow (median FCF = 0.1). The latter result is inconsistent with Dechow et al. (2006) whose sample has a negative median free cash flow (median FCF = ). Negative free cash flow is not necessarily an indication of a financially distressed firm. Many young firms put a lot of their cash into investments, which diminishes their free cash flow. However, a firm should have a good reason for having negative free cash flows and the firm should be earning a sufficiently high rate of return on its investments. While free cash flow does not receive as much scrutiny as earnings, it is considered by some experts to be a better indicator of a firm s financial health. Dechow et al. (2008) find that accruals are less persistent than cash flows, on average, only in their subsample of negative FCF firm-years. While our sample is characterized by firms with relatively more persistent accrual earnings, as described in table 3 below, we still find that cash earnings (FCF) are more persistent than accrual earnings (ACC) in our sample. {Insert Table 2 about here} Panel B of Table 2 provides pair-wise correlations among the variables used in this study. On a univariate basis, it appears that changes in asset efficiency, rather than sales growth, explain the overall negative correlation between accruals and the subsequent year s returns. Analysts also appear to overestimate the persistence of the changes in asset efficiency earnings component, as 14

16 we observe a positive correlation between changes in asset efficiency and subsequent forecast errors. Thus, on a univariate basis we see some indication that analyst forecasting behavior could explain market overreaction to the persistence of the change in asset turnover component of accrual earnings. The univariate correlations also provide some evidence of market and analyst underreaction to the persistence of the distributed portion of free cash flow. Table 3 describes returns to trading strategies with positions in top and bottom deciles of each detailed earnings component: FCF t, ACC t, SG t, ΔEFF t, ΔCASH t, DIST_EQ t and DIST_D t. Table 3 also describes differences in analyst forecast errors in the top versus bottom deciles of the same variables. Panel A shows a 10.4% hedge portfolio return based on a trading strategy taking a long (short) position in stocks in the highest (lowest) decile of cash earnings; however this hedge return is not statistically significant, as the two-tailed p-value associated with the difference in mean returns in the highest versus lowest decile is Panel A shows that analysts forecast errors are significantly larger in the highest decile relative to the lowest decile of cash earnings. Thus, on a univariate basis, it appears that analysts forecasts underreact to the persistence of cash earnings. {Insert Table 3 about here} Panel B shows a 13.2% hedge portfolio return based on a trading strategy taking a long (short) position in stocks in the lowest (highest) decile of net operating accruals. Panels C and D suggest that this 13.2% hedge portfolio return is driven by market overreaction to the efficiency change component (ΔEFF) of net operating accruals. We find a statistically significant 12.4% hedge portfolio return based on a trading strategy taking a long (short) position in stocks in the highest (lowest) decile of ΔEFF. Recall that the highest decile contains stocks where asset turnover increased due to either an increase in sales (holding net operating assets constant) or a 15

17 decrease in accruals (holding sales growth constant). Since, we see only a 0.2% return to a trading strategy based on sales growth (panel C), we attribute the 12.4% return to the trading strategy based on changes in asset efficiency to a denominator effect (i.e., changes in accruals, holding sales growth constant). On the other hand, financial analysts earnings forecasts do not differ significantly between the top and bottom deciles of total accruals (ACC) or accrual components (SG, ΔEFF). In our sample of relatively large profitable and positive free cash flow firms, contrary to Dechow et al. (2008), we find no evidence of returns to a trading strategy based on the retained cash component of free cash flow (panel E). We do, however, find some evidence (panel G) of returns to a trading strategy taking long (short) positions in stocks with high (low) free cash flow to debtholders (7.6%, p-value = 0.018). This suggests that the market efficiently impounds (underreacts to) information about the persistence of the retained (distributed-to-debtholder) component of cash earnings. On the other hand, our univariate tests suggest that financial analysts earnings forecasts underreact to the persistence of the distributed-to-stockholder component of cash earnings (panel F). Returns analysis Before examining market efficiency with respect to earnings components, we report the persistence characteristics of earnings components for our sample firms and time period. Panel A of Table 4 shows that the coefficients relating cash and accrual earnings to next year RNOA are and 0.420, respectively, and both coefficients are highly significant. Similarly, as shown in table 4 panel B, each detailed cash and accrual earnings component is strongly related to next year s RNOA, and all components have positive persistence. 16

18 {Insert Table 4 about here} Table 5 describes the behavior of stock returns following publication of information in accrual and cash components of earnings. In panel A, like Richardson et al. (2001, 2005), we find a significantly negative relation between total accruals and following year returns, suggesting market overreaction to the persistence of accrual earnings. Like Richardson et al. (2001, 2005), we do not find a significant relation between cash earnings and subsequent returns. Compared to Richardson et al. (2001, 2005), the coefficient on the total accruals variable is smaller in our sample. We attribute this difference to mitigation of the negative relation between accruals and subsequent returns during the time period. 17 {Insert Table 5 about here} Panel B of Table 5 replicates the Richardson et al. (2001) analysis of the relation of cash flow and disaggregated current fiscal year accrual earnings components with following year returns. In both our replication and the Richardson et al. (2001) study, the cash flow component of earnings is not significantly related to future returns, and the change in asset efficiency component of accruals is positively related to future returns. In our sample, the coefficient on the sales growth component of accruals although negative is not significantly different from zero; whereas, Richardson et al. find a significantly negative relation between sales growth and following year returns. The coefficient on the change in asset efficiency component of accruals is only about half the size of the coefficient in the Richardson et al. (2001) study. 18 Nonetheless, using our smaller sample of larger firms in a more recent time period, we are able to replicate the 17 Our results do not change when we remove our forecast error and forecast revision data constraint. However, when we maintain that constraint but eliminate the years , the estimated coefficient on the accruals variable is similar in magnitude than the coefficient in the Richardson et al. study. 18 The analysis of the relation of future returns with current accruals disaggregated into sales growth and change in asset efficiency components appears only in the Richardson et al. (2001) working paper and not in any subsequent published papers. 17

19 Richardson et al. (2001) finding that the market apparently overreacts to the persistence of the change in the efficiency component of total accruals. 19 Panels C and D of Table 5 extend the analysis to include the relation between disaggregated cash flow and accrual information and next year s returns. Panel C shows that, in our sample, we replicate the result in Dechow et al. (2008) that the market overreacts to the persistence of the accrual component of current year earnings and, consistent with Richardson et al. (2001), panel D shows the overreaction to ΔEFF t drives the overreaction to ACC t. Panel D of Table 5 reports the results of estimating our full returns model (8), which regresses following year abnormal returns (RET t+1 ) on all three current year cash components of earnings (ΔCASH, DIST_EQ, and DIST_D) and all three current year accrual components of earnings (SG, ΔEFF and SG*ΔEFF). Consistent with our results reported above, estimates of coefficients of the full model (8) suggest that in our sample the market underreacts to the persistence of the net distributed-to-debtholder cash earnings component (DIST_D); whereas Dechow et al. (2008) find that the market overreacts to the persistence of the retained cash earnings component (ΔCASH). Analyst forecast errors To assess the efficiency of analysts earnings forecasts following the release of financial statements containing information about the cash and accrual components of the prior year s earnings, we estimate models (9) and (10). Table 6 reports estimates of our regressions of analysts forecast errors on prior year cash and accrual components of earnings. Given the results reported above, if analyst forecasting behavior is consistent with the market response to total and 19 Recall that, holding sales growth constant, an increase in asset turnover (efficiency) corresponds to a decrease in accruals, so a positive coefficient on the efficiency change variable indicates that the market has overreacted to the persistence of this component of accruals. 18

20 disaggregated cash and accrual earnings, then we expect a negative coefficient on total accruals in model (9), a positive coefficient on the efficiency change component of accrual earnings in model (10), and a positive coefficient on the distributed-to-debtholder component of cash earnings in model (10). {Insert Table 6 about here} Panel A shows the results of estimating summary model (9), which regresses the error in forecasts following the release of prior year financial statements on prior year cash (FCF) and accrual (ACC) summary earnings components. The significantly positive coefficients on both variables suggest that analysts earnings forecasts underreact to the persistence of both cash and accrual earnings; whereas, we found in Table 5 that the market appears to overreact to the persistence of accrual earnings. Thus, it does not appear that analyst forecasting behavior can explain the apparent market overreaction to the persistence of accruals. Furthermore, panel B indicates that the underreaction to accrual earnings extends to both the sales growth and asset efficiency change components of total accruals, although the coefficient on the efficiency change component is only marginally significant. Nonetheless, analyst behavior does not appear to be responsible for investor overreaction to the persistence of the efficiency change component of accrual earnings. Panel C reports estimates of the coefficients in our full model (10). As shown in panel C, analyst underreaction to cash earnings extends to all three components: ΔCASH, DIST_EQ, and DIST_D. The positive coefficient on DIST_D offers a possible explanation for the market s underreaction to the distributed-to-debtholders component of cash earnings documented in panel D of Table 5. Overall, we find that analyst behavior can only partially explain apparent market inefficiencies in processing information about the persistence of accrual and cash earnings. 19

21 Analyst forecasting behavior as an explanation for market inefficiency To summarize the results above, we find some evidence of market overreaction to total accruals (panel A of Table 5), market overreaction to the asset efficiency change component of total accruals (panel D of Table 5), and market underreaction to the distributed-to-debtholders component of cash earnings (panel D of Table 5); and we find evidence of general analyst underreaction to accrual and cash components of earnings, with the possible exception of the efficiency change component of accruals (panel C of Table 6). Table 7 adds forecast error and forecast revision variables to the returns summary and component models estimated in Table 5. Like the returns variable, the forecast error (FE) and forecast revision (FREV) variables are computed with reference to forecasts issued following firms publication of financial statements containing the earnings component information variables: ACC, SG, ΔEFF, FCF, ΔCASH, DIST_EQ, and DIST_D. If analyst forecasting behavior explains the market overreaction to ACC t and ΔEFF t and the market underreaction to DIST_D t, then we expect the coefficients on those variables to approach zero when we add FE t+1 and FREV t+1 to the returns model. 20 {Insert Table 7 about here} As described in panel A of Table 7, without FE t+1 and FREV t+2 in model (7), the coefficient relating ACC t to RET t+1 equals and is significantly less than zero (p-value < 0.05). Since estimating models (9) and (10) in Table 6 produces no evidence that analysts overreact to any earnings components, it is not surprising that adding FE t+1 and FREV t+1 to model (7) does not cause the coefficient on ACC t to move much towards zero. In fact, panel A of Table 7 shows that 20 See Shane and Brous (2001) for a detailed explanation of this approach to gauging the degree to which analyst forecasting behavior explains market inefficiencies. 20

22 the coefficient on ACC t in model (11) is , still significantly negative (p-value < 0.05). Similarly, the coefficient relating ΔEFF t to RET t+1 equals 0.11 without FE t+1 and FREV t+1 in model (8) (indicating market overreaction to the persistence of changes in asset efficiency, p- value < 0.01), and adding FE t+1 and FREV t+2 to model (8) does not cause the coefficient on ΔEFF t to move much towards zero. In fact, the coefficient on ΔEFF t in model (12) is 0.104, still significant at less than the 0.01 level. On the other hand, since the market underreacts to the persistence of DIST_D t (panel D of Table 6), and analysts similarly underreact (panel C of Table 6), we expect the coefficient on DIST_D t to move towards zero when we add FE t+1 and FREV t+2 to model (8). In fact, the coefficient relating DIST_D t to RET t+1 is without FE t+1 and FREV t+2 in the model (p-value < 0.05), and it decreases by 35% to and is no longer statistically significant when we add FE t+1 and FREV t+1 to the model. Thus, we estimate that analyst forecasting behavior potentially explains 35% of the market s underreaction to the distributed-to-debtholder component of cash earnings. VI. Conclusion Richardson et al. (2001, 2005) provide evidence suggesting that stock prices overreact to the persistence of total operating accruals, and Richardson et al. (2001) provide evidence that the apparent stock price overreaction to the persistence of total operating accruals extends to two important accrual earnings components: sales growth and change in asset efficiency. Dechow et al. (2008) provide evidence that stock prices overreact to the persistence of cash earnings, and Dechow et al. find that overreaction to the persistence of the retained cash component of earnings drives the market s overreaction to total cash earnings. Since both Dechow et al. (2008) 21

23 and Richardson et al. (2001) suggest that the market does not efficiently impound information about the persistence of earnings components, an important question is whether financial analysts fail to distinguish the persistence of various earnings components and, thereby, publish earnings forecasts that create market inefficiency. We find little evidence to support the view that financial analysts are responsible for market inefficiency with respect to the persistence characteristics of the various earnings components evaluated by Dechow et al. (2008) and Richardson et al. (2001). Instead, consistent with economic theories of analyst underreaction (e.g., Raedy et al. 2006, Trueman 1990), we find a generalized analyst underreaction to earnings components. Our sample of relatively large firms followed by analysts in a recent time period produces results that differ somewhat from the evidence of market inefficiencies in Dechow et al. (2008) and Richardson et al. (2001). In particular, we do not find the overreaction to sales growth documented by Richardson et al. (2001), and we find that the market underreacts to the persistence associated with cash earnings distributed to debtholders, whereas, Dechow et al. (2008) find that the market overreacts to cash earnings retained for reinvestment. Our evidence is consistent with market underreaction to the persistence of earnings components corresponding to cash flows that increase or decrease leverage. Our evidence is also consistent with accounting distortions causing temporary accounting changes in asset efficiency (turnover) that fool investors but not financial analysts. 22

24 Appendix A Variable definitions (Firm i subscripts suppressed; Item #s refer to Compustat) RET t+1 = the annual buy and hold size-adjusted return. The size-adjusted return is calculated by subtracting the value-weighted average return for all firms in the same size-matched decile, where size is measured as market capitalization at the beginning of the return accumulation period. The return accumulation period begins four months after the end of fiscal year t. FE t+1 = the signed forecast error, calculated as (A t+1 F t+1 ) / P t. A t+1 is actual earnings for fiscal year t+1, per I/B/E/S. F t+1 is the most recent I/B/E/S individual forecast prior to the beginning of the return accumulation period. When more than one forecast occurs on that day, F t+1 is the median of all forecasts occurring on that day. P t is the adjusted price taken from the same I/B/E/S report month containing F t+1. FREV t+2 = ( t+1 F t+2 t F t+2 ) / P t ; where t+1 F t+2 = the first individual analyst forecast of t+2 earnings issued after the end of the return accumulation period; and t F t+2 = the most recent forecast of t+2 earnings issued prior to the beginning of the return accumulation period ACC t = (NOA t NOA t-1 ) / NOA t-1 = the percentage change in Non-cash Net Operating Assets, defined as total operating accruals for fiscal year t. NOA t = Total operating assets Total operating liabilities for fiscal year t, where cash and investments are defined as financial assets [Item #6 Item #1 Item #32 (Item #181 Item #34 Item #9)]. SG t = sales growth, calculated as (Sales t / Sales t-1 ) 1, where Sales is Item #12. EFF t = the deflated change in asset efficiency, calculated as (AT t AT t-1 ) / AT t, where AT t = Sales t / NOA t. RNOA t = return on net operating assets for fiscal year t, calculated as operating income after depreciation (Item #178) deflated by NOA t-1. FCF t = free cash flow deflated by NOA t-1, calculated as RNOA t ACC t. DIST_EQ t = net equity distributions, calculated as ( Item #6 Item #181 - Item #32 Item #178) deflated by NOA t-1. DIST_D t = net debt distributions, calculated as ( Item #9 + Item #34) deflated by NOA t-1. 23

25 DIST t = net distributions to capital providers, calculated as DIST_EQ t + DIST_D t. CASH t = change in cash balance, calculated as Item #1 deflated by NOA t-1. 24

26 References Ahmed, A., Nainar, S. M., Zhou, J., Do analysts' earnings forecasts fully reflect the information in accruals? Canadian Journal of Administrative Sciences 22, Ahmed, A., Nainar, S.M., Zhang, F., Further evidence on analyst and investor misweighting of prior period cash flows and accruals. The International Journal of Accounting 41, Barth, M., Hutton, A., Analyst earnings forecast revisions and the pricing of accruals. Review of Accounting Studies 9, Bradshaw, M., Richardson, S., Sloan, R., Do analysts and auditors use information in accruals? Journal of Accounting Research 39, Bradshaw, M., Richardson, S., Sloan, R., The relation between corporate financing activities, analysts forecasts and stock returns. Journal of Accounting and Economics 42, Brown, L., Earnings forecasting research: its implications for capital markets research. International Journal of Forecasting 9, Chan, L., Karceski, J., Lakonishok, J., The level and persistence of growth rate. Journal of Finance 58, Cooper, M., Gulen, H., Schill, M., What best explains the cross-section of stock returns? Exploring the asset growth effect. Working paper, Purdue University. Dechow, P., Richardson, S., Sloan, R., The persistence and pricing of the cash component of Earnings. Journal of Accounting Research 46, Dimitrov, V., Jain, P The value-relevance of changes in financial leverage beyond growth in assets and GAAP earnings. Journal of Accounting, Auditing and Finance 23, Drake, M., Myers, L., Analysts accrual-related over-optimism: do experience and broker Affiliation play a role? Working paper, Texas A&M University. Desai, H., Rajgopal, S., Venkatachalam, V., Value-glamour and accrual mispricing: one anomaly or two? The Accounting Review 79: Eberhart, A., Siddique, A., The long-term performance of corporate bonds (and stocks) following seasoned equity offerings. The Review of Financial Studies 15, Elgers, P., Lo, M., Pfeiffer, R., Analysts vs. Investors weightings of accruals in forecasting annual earnings. Journal of Accounting and Public Policy 22: