Earnings, Cash Flows and Ex post Intrinsic Value of Equity

Transcription

1 Earnings, Cash Flows and Ex post Intrinsic Value of Equity K.R. Subramanyam Leventhal School of Accounting University of Southern California Los Angeles CA (213) Mohan Venkatachalam Graduate School of Business Stanford University Stanford, CA (650) April 2001 We appreciate financial support from University of Southern California and Graduate School of Business, Stanford University. We thank Jean Bedard, Dave Burgstahler, Karen Nelson, Per Olsen, Ken Singleton and seminar participants at the University of Wisconsin Madison for their valuable comments. We thank Yulin Long for research assistance.

2 Earnings, Cash Flows and Ex post Intrinsic Value of Equity Abstract This paper examines the relative importance of earnings and operating cash flows in equity valuation. In contrast to previous studies that use stock returns or future operating cash flows, we use ex post intrinsic value of equity as the criterion for comparison. Ex post intrinsic value of equity is determined adopting both the discounted dividend and residual income models with ex post (future) realizations of dividends/residual income over three- and five-year horizons as inputs. The advantage of using the ex post intrinsic measure is that this measure is not contaminated by the stock market s fixation on reported earnings (Sloan, 1996). Also, unlike future operating cash flows, ex post intrinsic value is a formal and comprehensive measure of the fundamental value of the firms equity. Our findings unambiguously indicate that accrual-based earnings dominates operating cash flows as a summary indicator of ex post intrinsic value. Key words: Earnings, Cash flows, Intrinsic value. Data Availability: All data used in the study are available from public sources.

3 Earnings, Cash Flows and Ex post Intrinsic Value of Equity 1. Introduction A fundamental question in accounting is the relative ability of accrual-based earnings and cash flows to provide information relevant for financial performance measurement and equity valuation. Although this issue has been extensively researched, the controversy regarding the relative superiority of cash flows versus earnings as a summary indicator of equity value and a measure of a firm s financial performance remains largely unresolved. We contribute to this important area of research by examining the relative ability of earnings and cash flows in explaining ex post intrinsic value of equity. We determine ex post intrinsic value of equity adopting both the discounted dividend and residual income models. As inputs to these models we use ex post (future) realizations of dividends/residual income over three- and five-year horizons and market value at the end of the horizon for terminal value calculations. Consistent with the Financial Standards Accounting Board s (FASB s) assertion (FASB 1978, #44), we find that accrual-based earnings dominates operating cash flows as a summary indicator of ex post intrinsic value. Prior research adopts two different approaches to investigate the relative importance of earnings and cash flows for equity valuation. Several studies use stock price as a proxy for fundamental equity value and document that earnings are superior to cash flows in explaining contemporaneous stock price or returns. For example, Dechow (1994) documents that stock returns are more highly associated with earnings than cash flows. 1 An implicit assumption in such studies is that stock prices accurately reflect the future cash flow information contained in current earnings and current operating cash flows. However, evidence in Sloan (1996) and Barth 1 Many studies also find that accruals and operating cash flows are incrementally useful beyond each other in explaining stock returns (for example, Rayburn, 1986; Bowen, Burgstahler and Daley, 1987; Ali, 1994). However, the focus of our study is examining the relative ability of aggregate earnings and operating cash flows in providing information useful for equity valuation. 1

4 and Hutton (1999) suggests that stock market participants appear to fixate on current earnings without considering the differential persistence of its accruals and cash flow components. Thus it is not clear whether the superior explanatory power of earnings (vis-à-vis cash flows) for returns arises because of the ability of earnings to better reflect value relevant information or because of the stock market fixation on bottom line earnings. A different set of studies examines the relative ability of earnings and cash flows in predicting future operating cash flows. While some of the earlier studies under this genre (Greenberg, Johnson and Ramesh, 1986; Bowen, Burgstahler and Daley, 1986) document mixed results, recent studies (for example, Barth, Cram and Nelson, 2001; Burgstahler, Jimbalvo and Pyo, 1999) use a more accurate measure of operating cash flows (direct operating cash flow measures under SFAS 95) over much larger samples and unambiguously document the superiority of operating cash flows over earnings in predicting future operating cash flows over varying horizons. 2 However, the use of a finite set of operating cash flows as a surrogate for firm value has certain limitations. Operating cash flows are not value attributes because they ignore investments in operating assets the appropriate value attribute is free cash flows (Copeland, Kohler and Murrin, 1999). Also, a finite set of future cash flows does not capture a large proportion of firm value that is represented by the terminal value (Bernard, Healy and Palepu, 2000). 3 An additional limitation of the future cash flows predictability approach (as we show 2 The primary thrust of Barth, Cram and Nelson (2001) is to show that components of earnings are informative in predicting future operating cash flows. However, they do document that aggregate operating cash flows are superior to earnings in predicting future operating cash flows over various horizons. 3 Both Barth, Cram and Nelson (2001) and Burgstahler, Jimbalvo and Pyo, 1999 examine the ability of earnings and cash flows in predicting discounted ex post operating cash flows. Additionally, Barth, Cram and Nelson also estimate terminal values by extrapolating terminal cash flows at an assumed growth rate. However, their discounted measures are not indicative of intrinsic value for two reasons. First, they discount operating cash flows and not free cash flows, hence ignoring investments. Second, their terminal value measures are mere extrapolations of the terminal cash flow. Thus, their terminal values do not add any new information beyond that already present in the finite stream of ex post cash flows. 2

5 later) is that the inferences reverse when the predictive criterion is switched to future earnings. That is, current earnings outperform current operating cash flows in predicting future earnings. Since neither future earnings nor future operating cash flows are strictly value attributes, the contradictory results make it difficult to draw unambiguous conclusions regarding the relative superiority of earnings and cash flows. We adopt a different approach. We investigate the relative importance of earnings and cash flows by examining the relative ability of the two summary measures to explain ex post intrinsic value of equity. We measure intrinsic value by using both the dividend discount model (DDM) and the residual income model (RIM). Since the purpose of our study is to evaluate the predictive ability of earnings and cash flows, we use ex post realizations of dividends/residual income as inputs to the valuation models (Penman and Sougiannis, 1998) rather than their ex ante expectations (Frances, Olsen and Oswald, 1999). Using ex ante expectations such as analyst forecasts to determine intrinsic value may not be appropriate because analysts have also been shown to fixate on current earnings (Elgers, Lo and Pfeffer, 1999). Consequently, we use three and five years ex post realizations of dividends and residual income to determine the intrinsic value of equity. To capture payoffs beyond the three- and five-year horizons, we estimate market-based terminal value measures using ex post market values at the end of the horizon. 4 We use market-based terminal value measures to abstract from arbitrary assumptions involved in estimating terminal value. We believe that the ex post intrinsic value approach that we adopt overcomes limitations inherent in the stock price/returns and the future cash flows predictability approaches adopted in prior research. First, unlike the stock price/returns approach, the ex post 4 Theoretically, the DDM and RIM approaches must produce identical intrinsic value realizations when market based terminal values are used. However, empirically the measures are not identical (although they are highly correlated) because of error in measuring both dividends and residual income. 3

6 intrinsic value methodology does not require the assumption of market efficiency. 5 Second, unlike future operating cash flows, ex post intrinsic values are a formal and comprehensive measure of the fundamental value of the firms equity. We conduct our analysis using data available post 1988 in the Compustat database for which ex post intrinsic values can be determined. We use both undeflated and deflated (using book value of equity and total assets as alternative deflators) specifications to examine the relation between earnings, operating cash flows and intrinsic value. Results from undeflated specifications unambiguously document the superiority of earnings over operating cash flows in explaining ex post intrinsic values, regardless of the model or horizon used in determining intrinsic values. However, deflated results present inference problems because the coefficient estimates for the relation between intrinsic value and both earnings and cash flows are consistently negative. The anomalous negative coefficients arise because of the large proportion of negative earnings/operating cash flows in our sample. 6 We tackle this problem in two ways. First, we estimate the empirical relation allowing for different multiples for negative and positive observations of earnings and operating cash flows. Second, we focus on only positive realizations of earnings and operating cash flows. Results from using both approaches show that current earnings dominate operating cash flows in explaining ex post intrinsic values. Our inference regarding the superior value relevance of earnings vis-à-vis cash flows is similar to that using stock price as firm value surrogates. This raises the possibility that the 5 Note that in determining the intrinsic value we use ex post market prices for calculating terminal value. While three- or five-year hence stock prices may be inefficiently priced, such mispricing is unlikely to be correlated with current earnings/cash flows (see Sloan (1996)). 6 In fact, we observe a negative relationship even when we use stock prices as our dependent variable. This is not surprising because the relation between price (or intrinsic value) and negative earnings or negative operating cash flows is theoretically undefined. Furthermore, prior research finds that the price earnings relation is different for negative and positive realizations of earnings (Burgstahler and Dichev, 1997) and that the relation between price and losses is anomalous and negative (Jan and Ou 1995; Collins, Pincus and Xie, 1998). Therefore, regressions that force the coefficients for positive and negative earnings (or operating cash flows) to be the same are mis-specified. Our deflated results are more sensitive to this problem because of the distributional characteristics of the deflated earnings and cash flows. An elaboration of this issue is deferred until later. 4

7 results obtain because of high correlation between intrinsic values and market prices. This possibility is exacerbated by our use of three- or five-year hence market values as terminal values in our intrinsic value computations. Thus we may simply be capturing the same economic phenomenon as that documented in the stock returns research. Our empirical results suggest this is unlikely our deflated intrinsic value measures are not highly correlated with market values and the superior explanatory power of earnings vis-à-vis cash flows is more pronounced for market values than for intrinsic values, suggesting some measure of functional fixation on earnings by the stock market. Nevertheless, we partition our sample on the basis of the absolute deviation of intrinsic value and market value and document that earnings is superior to cash flows even in the portfolio with the largest deviation. The ex post intrinsic value approach that we adopt is not without limitations. First, the use of ex post market values raises the possibility that the market s mispricing of accruals and cash flows may contaminate our results, although we believe this is unlikely since we use threeand five-year hence market values and the anomaly identified by Sloan (1996) does not persist beyond two years. Second, we assume a constant discount rate of 10% because of the difficulties associated with measuring the firm s true cost of capital. Finally, we adopt a levels specification, which is problematic in the presence of losses or negative cash flows. Unfortunately, the ex post intrinsic value approach is not amenable to a changes specification, which could overcome these problems. The rest of the paper is organized as follows. In the next section we provide the motivation for our analyses and relate our study to existing research. Section 3 describes the research design while section 4 presents our main findings. In Section 5 we present some concluding remarks. 5

8 2. Motivation and Related Research The accrual concept forms the cornerstone of modern accounting. Accounting standard setters maintain the superiority of accrual-based earnings over cash flows, as a summary indicator of the firm s financial status and performance. For example, the FASB s Statement of Accounting Concepts No. 1 (FASB 1978, Highlights) states that: Information about an enterprise earnings based on accrual accounting generally provides a better indication of enterprises present and continuing ability to generate favorable cash flows than information limited to the financial aspects of cash receipts and payments. Accrual based earnings is hypothesized to overcome both timing and mismatching problems inherent in cash flows, thus making it a more relevant summary indicator of a firm s value and financial performance (Dechow, 1994). Accruals, however, are considered as less reliable than cash flows because of the assumptions underlying the determination of accruals and the latitude allowed by GAAP. This often evokes strong criticism and suspicion from many economists and financial analysts on the relevance of earnings for valuation (e.g., Copeland, Kohler and Murrin, 1999). Given this debate, one of the most important questions in accounting and financial analysis is the role of accruals in generating a summary indicator of firm value or performance. As Schipper (1989) observes: one of the central questions confronted by practicing professional accountants and academic accountants [is] the influence and importance of accounting accruals in arriving at a summary measure of financial performance. 7 The relative ability of earnings and cash flows to summarize and reflect value relevant information has been extensively researched. Prior researchers have used two approaches to examine this question. One approach is to examine the relative association between stock returns 7 A related, but equally important question, is the ability of accruals to provide value relevant information incremental to that in operating cash flows. This issue has been addressed by Bowen, Burgstahler and Daley (1986, 1987), Rayburn (1986) and more recently by Barth, Cram and Nelson (2001) and Burgstahler, Jimbalvo and Pyo (1999). 6

9 and the two measures (e.g., Dechow, 1994). The evidence from this approach provides unequivocal support for the accrual-based accounting measure. That is, accrual-based earnings not only have incremental ability beyond operating cash flows in explaining stock returns but are also superior to cash flows, as a summary measure in explaining cross-sectional variation in stock returns. A maintained assumption in the stock-returns based approach is that the stock market is efficient and appropriately incorporates the economic implications of accruals and cash flow components of earnings. However, recent work by Sloan (1996) finds that stock market fails to fully incorporate the differential persistence of the accrual component of earnings relative to operating cash flows. Rather, the market fixates on bottom-line earnings, thus overpricing (underpricing) firms with higher (lower) earnings vis-à-vis operating cash flows. This raises questions about the validity of the assumption inherent in using stock returns to measure relative value-relevance. An alternative approach investigates the relative ability of the two summary measures in predicting future operating cash flows (for example, Bowen, Burgstahler and Daley, 1986; Greenberg, Johnson and Ramesh, 1986; Burgstahler, Jimbalvo and Pyo, 1999; Barth, Cram and Nelson, 2001). While the earlier results in this genre were mixed, two recent studies (Burgstahler, Jimbalvo and Pyo, 1999; Barth, Cram and Nelson, 2001) use direct measures of operating cash flows (SFAS 95) and unambiguously document that operating cash flows are superior to earnings in predicting future operating cash flows over varying horizons. Although the objective of Barth, Cram and Nelson (2001) is to examine the incremental explanatory power of accruals and its components for future operating cash flows, consistent with findings in Burgstahler, Jimbalvo and Pyo (1999), they document that operating cash flows dominate earnings as a summary measure in predicting future operating cash flows. 7

10 The choice of future cash flows as the predictive criterion for comparing current earnings and cash flows is motivated from the objective of financial accounting, which is the prediction of the magnitude and timing of prospective future cash flows (FASB, 1979, # 37). However, the primary objective of financial accounting according to FASB is the prediction of cash receipts to the users, primarily investors and creditors. 8 The prediction of cash flows to the firm assumes importance only because cash inflows to the users are related to the firms ability to generate cash (FASB, 1978, # 37). Therefore, from the perspective of equity investors, the overarching objective of financial accounting is to provide information that helps them determine the present value of future cash receipts, or fundamental values, of their investments or prospective investments which should enable them to make appropriate investment decisions. 9 The operating cash flow prediction approach has the distinct advantage of avoiding inference problems associated with using stock prices as a proxy for fundamental value. However, the use of a short horizon of future operating cash flows in understanding the relative valuation importance of earnings and cash flows has certain limitations. First, a finite set of future cash flow measures only a small portion of a firm s intrinsic value. A significant fraction of the firm value is determined by the terminal value even when the forecast horizon is as long as 8 That the prediction of cash flows to the enterprise is derived from the more fundamental objective of prediction of cash flows to the investors (or creditors) is evident from Paragraph #37 of FASB s Statement of Financial Accounting Concepts No. 1, which states that: Financial reporting should provide information to help present and potential investors and creditors and other users in assessing the amounts, timing, and uncertainty of prospective cash receipts from dividends or interest and the sale, redemption, or maturity of securities or loans. The prospect for those cash receipts are affected by an enterprise s ability to generate enough cash to meet its obligations when due and its other cash operating needs, to reinvest in operations, and to pay cash dividends and may also be affected by perceptions of investors and creditors generally about that ability, which affect market prices of the enterprise s securities. Thus financial accounting should provide information to help investors, creditors and others assess the amounts, timing and uncertainty of prospective cash inflows to the related enterprise. (FASB 1978, # 37). 9 Of course, the fundamental value approach may not be appropriate for other users such as creditors and suppliers. Hence, our inferences relate primarily to the relative importance of cash flows and earnings from an equity investor standpoint. 8

11 ten years (Palepu, Healy and Bernard, 2000; Copeland, Kohler and Murrin, 2000). 10 Second, operating cash flows are not value attributes. This is because operating cash flows ignore investments in assets. The appropriate value attribute is free cash flows, which is operating cash flows less investments in operating assets (Copeland, Kohler and Murrin, 2000). Finally, as we show later in this paper, if one uses future earnings as an alternative proxy for firm value or performance (e.g., Ou and Penman, 1992), the opposite result obtains. That is, while current operating cash flows better predict future operating cash flows, current earnings are better predictors of future earnings. This conflicting result further demonstrates the difficulty in conclusively documenting the relative superiority of earnings or cash flows using either future earnings or future operating cash flows as proxy for firm value, especially since past research has used both. 11 We adopt a different approach by evaluating the ability of operating cash flows and earnings to explain ex post intrinsic value measures. That is, we consider the present value of all future realized payoffs to equity investors. We believe this approach overcomes the limitations inherent in the two approaches adopted by prior research. First, when adopting the perspective of the equity investor, our use of ex post intrinsic values as the predictive criterion derives directly from FASB s primary objective of financial accounting, which is the prediction of the magnitude, timing and uncertainty of cash receipts to the users (FASB 1978, #37). Unlike a finite set of future operating cash flows the intrinsic value measure is a more formal and comprehensive measure of cash payoffs to the investors firm value is obtained by appropriately 10 Barth, Cram and Nelson (2001) also use discounted ex post operating cash flows in some of their tests. They include a measure of terminal value by extrapolating terminal cash flows at an assumed growth rate. This discounted measure is not indicative of intrinsic value for two reasons. First, intrinsic values discount free cash flows, not operating cash flows. Operating cash flows ignore the cost of investment in operating assets. Second, their terminal value measures do not add any new information beyond that already present in the finite stream of ex post cash flows since their terminal value measures are extrapolations of the terminal cash flow. 11 Barth, Cram and Nelson (2001) and Burgstahler, Jiambalvo and Pyo (1999) are examples of papers that use future operating cash flows, while Ou and Penman (1992) and Sloan (1996) are examples of papers that use earnings. 9

12 discounting all future cash flows to investors in the infinite future. Second, ex post intrinsic value is an ex post measure unlike price, which is an ex ante measure, and hence is unlikely to be affected by stock market s perceptions, in particular, functional fixation on current earnings (Sloan, 1996). 3. Research Design and Sample Description 3.1 Intrinsic Value Measurement To determine intrinsic value we use two alternative valuation techniques: the Dividend Discount Model (DDM) and the Residual Income Model (RIM). The DDM is non-controversial and is widely accepted as a basis for representing equity value. Formally, t = ρ t+ τ τ = 1 τ P E( D ) (1) where P t is the stock price at time t, D t+τ is dividends paid at time t+τ, ρ is one plus the discount rate, and E is the market expectation operator. Note that the above expression assumes an infinite horizon. However, one can express the DDM using a finite (N periods) forecast of dividends and terminal value in the following manner: N N τ N P = ρ E( D + τ ) + ρ E( P + t τ = 1 t t N ) (2) Note that the terminal value in the DDM is the expected stock price at the end of the forecast horizon. Thus, we replace expected future dividends (price) with ex post realizations of dividends (price) to derive our measure of ex post intrinsic value: N N τ N IV _ DDM = ρ ( D + τ ) + ρ ( P + t τ = 1 t t N ) (3) where IV_DDM denotes the ex post intrinsic value using the dividend discount model and N represents the finite horizon. We compute intrinsic values using both three-year (IV_DDM 3 ) and five-year (IV_DDM 5 ) horizons. To cash dividends we add distributions through stock 10

13 repurchases to incorporate all dividends to the investors (Penman and Sougiannis, 1998). The terminal value in each case is the market value of equity at the end of the horizon. The residual income model (RIM), introduced by Preinreich (1938) and formalized later by Ohlson (1995), is an algebraic transformation of the DDM using the following accounting identity (often referred to as the clean surplus relation): BV t = BVt 1 + X t D t where BV t is book value of equity at time t, and X t is comprehensive income for period t. Substituting the above identity in the DDM (equation 1) we obtain the following residual income representation for price: τ P = BV + ρ E X τ ( ρ 1) BV τ ) (4) t t τ = 1 ( t + t + 1 where the expression X t+τ (ρ 1)BV t+τ-1 represents residual income. The RIM can also be expressed using a finite forecast horizon and terminal value in the following manner: N N τ N P t = BVt + ρ E( X t + τ ( ρ 1) BVt + τ 1) + ρ E( Pt + N BVt + τ = 1 N ) (5) The terminal value in the case of the RIM is the expected deviation of the market price from the book value of equity at the end of the forecast horizon. As in the case of the DDM, we replace expected income, book values and price by their respective ex post realizations to derive a measure of ex post intrinsic value: N N τ N IV _ RIM t = BVt + ρ ( X t + τ ( ρ 1) BVt + τ 1) + ρ ( Pt + N BVt + τ = 1 N ) (6) Because comprehensive income measures are not readily available, we use future realizations of net income. Once again, we compute intrinsic values using both three-year (IV_RIM 3 ) and fiveyear (IV_RIM 5 ) horizons. 11

14 We adopt a standard 10% discount rate for our analysis because of the lack of consensus in estimating expected rates of returns in extant literature. However, prior research has shown that results are not sensitive to the selection of the discount rate (Penman and Sougiannis, 1998). Thus, we believe that our use of a standard 10% discount rate should not materially affect our results. Unlike previous research (Penman and Sougiannis, 1998; Lee, Myers and Swaminathan, 2000), we use ex post realizations of market value for determining terminal values for our intrinsic value computations. The advantage with using ex post market values is that our measures of intrinsic value are independent of various ad hoc assumptions (such as terminal year payoffs and perpetual growth rates) that are necessary in estimating terminal values. One disadvantage with using ex post market values is that it may introduce bias in our intrinsic value measures if the three- and five-year hence stock prices inefficiently price the implications of current accruals and cash flows. That is, our approach may suffer from the same problem inherent in the stock returns based studies. However, we believe this is unlikely because the inefficient pricing of the accrual and cash flow components of earnings identified by Sloan (1996) does not persist beyond two years. 12 Since we use ex post realizations of market value at the end of the forecast horizon, our measures of intrinsic value using the DDM (equation (3)) and RIM (equation (6)) are algebraically identical. However, empirical measures of the DDM and RIM intrinsic values are different largely because of errors in measuring dividend and earnings realizations. Measurement error in dividends arises because of estimating the value and the timing of share repurchases, while the measurement error in earnings arises because reported net income does not strictly 12 We are not suggesting that the stock market efficiently prices securities three- or five-years hence. We merely maintain that the error in market pricing three- or five-years hence is uncorrelated with current earnings or cash flows. 12

15 adhere to clean surplus, although deviations from clean surplus are not large on average (Dhaliwal, Subramanyam and Trezevant, 1999). Because neither measure of intrinsic value is measured perfectly, we perform our analysis using both measures Sample and Descriptive Statistics We compute intrinsic values for a sample of all available firms in the 1999 Compustat database for the period We restrict the analysis to this time period for two reasons. First, operating cash flow measures using balance sheet information are known to suffer from significant measurement error (Barth, Cram and Nelson, 2001; Burgstahler, Jiambalvo and Pyo, 1999; Collins and Hribar, 2000). Therefore, we confine our analysis to firm-years starting from 1988 when SFAS 95, that requires explicit operating cash flow information, became effective. Second, at least three subsequent years of data are required to compute intrinsic values this limits our sample to years 1996 and earlier. For the analysis pertaining to intrinsic values that require a five-year forecast horizon our sample is further restricted to the period We delete all firm-years with negative book value of equity and negative intrinsic values. We also delete firm-years where any of the key variables are in the extreme 1% of their respective distributions. Our final sample consists of firm year observations of 8038 distinct firms. 14 Table 1, Panel A provides descriptive information regarding the dependent and independent variables. For our empirical analysis we consider both undeflated and deflated (separately by book value and total assets) specifications. Hence, we report descriptive statistics in both undeflated and deflated forms. Earnings (EARN) is defined as income before extraordinary items. Operating cash flow (OCF) is defined as net cash flow from operating 13 Empirically both the measures are highly correlated, indicating that these differences are not economically significant. However, we report analysis using both measures for completeness. 14 Note that our sample sizes vary because of different horizons and different deflators in estimating intrinsic value. While for the total assets deflation we have observations, for the book value of equity deflation we have slightly lower (37561) observations. For samples using intrinsic value measures based on a five-year horizon the sample sizes are even smaller. 13

16 activities obtained from the cash flow statement adjusted for extraordinary items and discontinued operations. 15 Table 1 reports that the mean and median EARN are lower than that of OCF in both deflated and undeflated forms mainly because of depreciation and amortization. Also, both OCF and EARN having a significant proportion (about 25%) of negative observations. The means of market value (MKTV) are smaller than that of the intrinsic value measures using a three-year horizon (IV_DDM 3 and IV_RIM 3 ), while the medians of MKTV are higher than those of the intrinsic values. The means of the undeflated five-year horizon intrinsic values (IV_DDM 5 and IV_RIM 5 ) are somewhat higher than that of either the three-year horizon intrinsic values or market values, although the means of the deflated intrinsic values and market values are comparable. It is comforting that the intrinsic value measures using both DDM and RIM approaches are comparable despite potential measurement error problems discussed earlier. In particular, the DDM and RIM intrinsic value measures for both three-year and five-year horizon and across deflated and undeflated specifications are highly correlated, with the Pearson correlation averaging over 0.95 (see Panel B of Table 1). This further indicates that the measurement error in computing the intrinsic value measures is economically insignificant. 16 Panel B also reveals that the various intrinsic value measures are highly correlated with market values for the undeflated specification (about 0.90 on average). However, the correlation is much lower when the variables are deflated (about 0.50 on average). 15 We adjust cash flows from operating activities (Compustat # 308) by the extraordinary component of operating cash flows. The extraordinary component of operating cash flows is obtained by subtracting the accrual component of extraordinary and discontinued items (Compustat # 124) from the extraordinary and discontinued items in net income (Compustat # 48). 16 Alternatively, this could mean that the measurement errors under both computations are highly correlated. While possible, we believe this is unlikely because the sources of the measurement error under the two methods are very different the measurement error in IV_DDM arises from inaccurate valuation of stock repurchases while that in IV_RIM arises from violation of clean surplus accounting. 14

17 4. Results 4.1 Preliminary Analyses Initially, we present evidence on the relative explanatory power of earnings and operating cash flows for contemporaneous market values, future operating cash flows and future earnings. Largely, this analysis is similar to that in prior research such as Dechow (1994), Barth, Cram and Nelson (2000) and Burgstahler, Jiambalvo and Pyo (2000). Nevertheless, we present results from this analysis for the following reasons. First, we replicate prior research for completeness and to verify whether consistent results are obtained in our sample. Second, it serves to highlight the ambiguities in inferences when we use future operating cash flows and future earnings as predictive criterion to determine which of the two is a superior summary measure. We show that while current operating cash flows better predict future operating cash flows, current earnings are superior predictors of future earnings. Finally, the contemporaneous market value analysis allows us to identify and describe certain peculiarities in our data that create inference problems in our deflated models. This issue is important because similar data problems arise in our analysis using intrinsic values Explanatory power of earnings and operating cash flows for market value of equity Panel A of Table 2 reports explanatory power of pooled cross-sectional undeflated, equity book-value deflated and total assets deflated regressions of contemporaneous market value on earnings and operating cash flows. In the undeflated regressions, the explanatory power of EARN (R 2 =73.68%) is significantly higher than that of OCF (R 2 = 63.70%) the Vuongstatistic for the difference is 8.85 and significant at the 1% level. However, in both deflated models, notice that the coefficients on earnings and cash flows are contrary to expectations for example, the coefficient on earnings is 1.56 (-1.72) and that of cash flows is 1.39 (-2.27) under book value (total assets) deflation model. This makes the explanatory power from these models non-interpretable. 15

18 The anomalous negative coefficients for both CFO and EARN in deflated specifications underscores a basic problem with our full sample regressions. As discussed earlier, our sample contains a significant proportion (about 25%) of negative realizations of EARN and OCF (see Table 1). Prior research reports that the earnings-price relation is different for positive and negative earnings (Burgstahler and Dichev, 1997). Moreover, market value is anomalously negatively related to earnings for losses (Jan and Ou, 1995; Collins, Pincus and Xie, 1998). Thus, the higher proportion of negative observations potentially leads to anomalous negative coefficients and low explanatory power. In other words, our full sample results are biased because we force the coefficients on negative and positive realizations of OCF and EARN to be the same. The bias depends on the relative variances of the distribution of positive and negative realizations. 17 This misspecification, however, is not severe in the undeflated regression. This is because the variance of positive earnings observations relative to negative earnings observations is higher in the undeflated specification than in the deflated specifications Let us consider the equation of the form Y= a 0 + a 1 X + ε [i], where Y is market value of equity and X is earnings (or cash flows). Estimating equation [i] assumes identical coefficients for both positive and negative realization of X. However, prior research suggests that the coefficient for negative realizations of X is negative, while it is positive for positive realizations of X. To examine the implications of restricting the same coefficient on both positive and negative realizations of X, we can rewrite equation [i] as a structural (stacked) equation of the form (Y 1 Y 2 )' = (X 1 X 2 ) (b 1 b 2 )' + (ε 1 ε 2 )' [ii], where X 1 represents positive observations of X and X 2 represents negative observations of X. For simplicity, let us assume that this equation [ii] is estimated in deviation from means form (to avoid intercept effects). Then, b 1 = (X' 1 X 1 ) -1 X' 1 Y 1 and b 2 = (X' 2 X 2 ) -1 X' 2 Y 2 (see Johnston, 1984 pages 209,210). However, in fitting the restricted model where b 1 = b 2 = a 1, the coefficient, a 1, is (X' 1 X 1 + X' 2 X 2 ) -1 (X' 1 Y 1 + X' 2 Y 2 ). Note that in difference form X 1 Xs are scalars and denote variances. Through algebraic manipulation we can represent a 1 as a linear combination of b 1 and b 2 : a 1 = b 1 * (X' 1 X 1 )/(X 1 1 X 1 + X 1 2 X 2 ) + b 2 * (X' 2 X 2 )/(X' 1 X 1 + X' 2 X 2 ), i.e., a 1 = b 1 * (σ 2 X1) /(σ 2 X1 + σ 2 X2) + b 2 * (σ 2 X1) /(σ 2 X1 + σ 2 X2). Thus, the pooled coefficient will be a weighted average of the coefficients on negative and positive quadrants where the weighting is proportional to their respective variances. 18 The variance for undeflated positive earnings (cash flows) is (53929) whilst that for negative earnings (cash flows) is 320 (83). Thus, the ratio of variance of negative observations relative to total variance is 2.4%, 0.2% for earnings and cash flows respectively. This implies that the weight on negative observations is relatively small. However, for variables deflated by book value (total assets), the variance of positive earnings observations is (0.002) while that for negative observations is (0.062). This suggests that the weight on negative earnings observations is 98% (97%), which is very high relative to the ratios from the undeflated models. Similar variance ratios are obtained for operating cash flows. This explains why the bias arising from the presence of negative observations is more pronounced for our deflated models. 16

19 We address this problem in two ways. First, we allow for different slope coefficients for negative and positive realizations of EARN and OCF. The results of this analysis are reported in Panel A of Table 2. For completeness, we present results both for deflated and undeflated specifications. There is only marginal improvement in the explanatory power of EARN and OCF in the undeflated specification (76.01% versus 73.68% for EARN and 63.83% versus 63.70% for OCF). However, for deflated specifications notice that the R 2 s improve considerably. While the explanatory power of both OCF and EARN increases substantially, the improvement is more pronounced for EARN. Specifically, for the book value deflated models the R 2 for EARN (OCF) increases from 4.93% (3.49%) to 15.28% (10.26%), while for the total assets deflated models the R 2 increases from 3.76% (5.48%) to 22.26% (14.44%). Now, under both deflators EARN significantly outperforms OCF in explaining variation in market value of equity. Also, in all models, the slopes are positive for positive realizations of the explanatory variables while negative for negative realizations, consistent with prior research (Jan and Ou, 1995; Collins, Pincus and Xie, 1998). 19 While allowing for different slopes for negative and positive values of EARN and OCF takes into account the differential price-earnings (or price-cash flow) relation for negative and positive values of the explanatory variables, it is still not correctly specified. This is because the slopes for negative OCF and EARN are negative, which is contrary to expectation. An alternative approach that can lead to correctly specified regressions is limiting the sample to positive earnings and operating cash flows. However, this approach leads to a loss of generality 19 Collins, Pincus and Xie, 1998 report that the anomalous negative coefficient for negative earnings becomes positive with the inclusion of book value in the regression. Our book-value-deflated results are similar to a regression that includes book value as an additional variable in the regression, because the intercept proxies for book value. We still report negative coefficients for negative earnings. This apparent inconsistency arises because Collins, Pincus and Xie use lagged book value in their regressions their results are consistent with a negative coefficient for losses if contemporaneous book value is used in the regression. Notwithstanding, we use lagged book value as an alternative deflator and find no change in our inferences. Thus the differential results could also arise because we use a deflated specification compared to an undeflated specification by Collins, Pincus and Xie. 17

20 of our results these results are applicable only to profits and positive operating cash flows. Results from regressions of market value on only positive realizations of OCF and EARN are also reported in Panel A of Table 2. Once again, there is no appreciable change in the results of the undeflated models the R 2 s for EARN and OCF are 77.17% and 63.00% respectively and the difference is significant at the 1% level. Also, for deflated specifications we find that the explanatory power for EARN is significantly higher than that for OCF the R 2 s for EARN are 28.41% (36.06%) and OCF is 5.08%(9.92%) for the book value (total assets) deflators. The difference in explanatory power between EARN and OCF is statistically significant at the 1% level across both deflators. Overall, our results relating to contemporaneous market value are broadly consistent with prior research that shows that earnings are superior to operating cash flows in explaining contemporaneous stock returns (Dechow, 1994) Explanatory power of current earnings and operating cash flows for future operating cash flows and earnings We next investigate the relative predictive ability of earnings and operating cash flows for future operating cash flows and future earnings. We use future operating cash flows and earnings as alternative predictive criteria because neither future operating cash flows nor future earnings are strictly value attributes but both variables have been used in the literature to evaluate value relevance. For example, Barth, Cram and Nelson (2001) and Burgstahler, Jimbalvo and Pyo (1999), among others, use future operating cash flows as the predictive criterion. Alternatively, Ou and Penman (1992), Sloan (1996), Aboody, Barth and Kaznik (1999) and Dhaliwal, Subramanyam and Trezevant (1999), among others, use future earnings as surrogates for firm value/performance. Since there is no ex ante basis for deciding which of the 20 Our analysis considers a levels framework with market value, rather a changes framework with returns, as the dependent variable because our intrinsic value approach precludes a changes specification. However, Kothari and Zimmerman (1996) show that price-based and return-based frameworks are equivalent. 18

21 two is inherently superior we examine both measures. Results from examining the predictive ability of the two summary measures for future operating cash flows and future earnings are presented in Panel B and Panel C of Table 2, respectively. As with prior work we consider one- two- and three-periods ahead earnings and operating cash flows as the dependent variables. Once again, we report results from both undeflated and deflated (book-value and total assets) specifications. 21 From Panel B of Table 2 it is evident that operating cash flows better predict future operating cash flows than current earnings, consistent with prior research. This result obtains across one- two- and three-period ahead operating cash flows. However, when the prediction criterion is future earnings rather than future operating cash flows, the results are reversed. Specifically, for all horizons and specifications, with one exception (one-period ahead earnings deflated by total assets) the explanatory power of EARN is significantly greater than that of OCF at the 0.01 level (see Panel C of Table 2). In summary, the preliminary results regarding the relative value relevance of earnings and operating cash flows are inconclusive. Of particular importance, note that the results reverse when the prediction criterion is switched from future operating cash flows to future earnings. This highlights the difficulties with measures like future operating cash flows or future earnings over short horizons as a value surrogate. 4.2 Explanatory Power of Operating Cash Flows and Earnings for Ex post Intrinsic Value Given the inconclusive nature of prior results, we use the ex post intrinsic value approach that has advantages over either the stock price/returns or the future operating cash flow/earnings 21 Unlike the price (or intrinsic value) specification, the anomalous negative coefficient does not arise when we use future operating cash flows/earnings are the dependent variable. This is probably because, unlike market (or intrinsic) values, future earnings or cash flows can have negative realizations. 19

22 approaches. This approach measures the present value of all future payoffs to the investor as the criteria for comparing the relative superiority of earnings and operating cash flows. We present results using intrinsic value estimates obtained from two alternative valuation approaches the dividend discount model (DDM) and the residual income model (RIM) and over two different forecast horizons three and five years. Panel A of Table 3 reports the relative explanatory powers of regressions of alternative intrinsic value measures on operating cash flows (OCF) and earnings (EARN). As before, we present results for both undeflated and deflated (by book value and total assets) specifications. In the undeflated regressions, the R 2 s for EARN are significantly higher than those for OCF for both intrinsic value measures and for both forecast horizons. For example, the R 2 s for the EARN (OCF) are 64.48% (55.82%), 65.96% (56.80%), 61.04% (52.68%) and 62.67% (53.45%) for the models with IV_DDM 3, IV_RIM 3, IV_DDM 5 and IV_RIM 5 respectively as the dependent variables. Across all intrinsic value measures EARN significantly (at the 1% level) outperforms OCF in an R 2 sense. The results from both deflated specifications are mixed, with extremely low explanatory power (less than 1%). However, as with the market value based results reported in Table 2 these results are non-interpretable because most of the OCF and EARN have negative signs, which is opposite to that predicted by valuation theory. As explained earlier, these anomalous results obtain because the relation between value and EARN (OCF) is negative for negative realizations of EARN (OCF) and the inordinate weights placed on negative observations due to the high variance of negative observations relative to that for positive observations (see footnote 17). Accordingly, we repeat our analysis after allowing for different coefficients for positive and negative realizations of the respective independent variables and after deleting all negative observations of EARN and OCF. 20

23 Panel B of Table 3 reports the results of regressions where negative and positive observations of the independent variable are allowed different coefficients. The undeflated results are similar to that in Panel A, with marginal improvement in explanatory power of both EARN and OCF the explanatory power of EARN is higher than that of OCF for all intrinsic value measures, and all differences are significant at the 0.01 level. Unlike Panel A where we have mixed results, for both the book value deflated and the total assets deflated specifications, EARN consistently generates higher R 2 s than OCF. Specifically, for the book value deflated models the R 2 s for the EARN (OCF) are 9.03% (4.42%), 8.64% (3.56%), 9.40% (4.55%) and 8.98% (3.68%) for the models with IV_DDM 3, IV_RIM 3, IV_DDM 5 and IV_RIM 5 respectively as the dependent variables and the differences in R 2 are significant at the 1% level. Qualitatively similar results obtain for the total assets deflation. The explanatory power of EARN (OCF) is 14.17% (7.60%), 13.63% (6.29%), 13.35% (6.81%) and 13.51% (6.23%) for the models with IV_DDM 3, IV_RIM 3, IV_DDM 5 and IV_RIM 5 respectively as the dependent variables. In Panel C, we report explanatory powers of regressions with OCF and EARN respectively as the explanatory variables for a sub-sample where observations with negative values for either OCF or EARN have been deleted. Once again the undeflated results are similar to that in Panels A and B. For the book value deflation, EARN s explanatory power is significantly higher than that of OCF in all models. In particular, the R 2 s for EARN (OCF) are 13.68% (4.22%), 14.39% (4.44%), 12.31% (3.69%) and 12.93% (3.81%) for the models with IV_DDM 3, IV_RIM 3, IV_DDM 5 and IV_RIM 5 respectively as the dependent variables and all differences are significant at the 0.01 level. Qualitatively similar results obtain for the models with total assets deflation. Specifically, the explanatory power for models with EARN (OCF) as the independent variables are 21.22% (8.40%), 22.01% (9.06%), 18.07% (7.80%) and 18.81% (8.27%) for the models with IV_DDM 3, IV_RIM 3, IV_DDM 5 and IV_RIM 5 respectively as the dependent variables. 21

24 Overall, our results are consistent with the relative superior ability of earnings vis-à-vis operating cash flows in explaining ex post intrinsic value. Given the consistent results obtained across various specifications we believe that our results provide unambiguous evidence on the relative superiority of accrual based earnings as a summary measure that reflects the fundamental value of the firm s equity. 4.3 Additional Analysis We argue that our ex post intrinsic value approach abstracts from inference problems inherent in the stock price approach, viz., stock market fixation on bottom line earnings. However, our intrinsic value measures are highly correlated with market values in their undeflated form, although the correlation is lower after deflation (see Panel B of Table 1). Therefore, it is reasonable to question whether our results provide any additional insights beyond that provided by stock returns research. More specifically, are our inferences similar to those of the stock returns studies because we investigate a different economic phenomenon that provides similar inferences, or are we documenting the same economic phenomenon measured differently? The above question can be answered to some extent without additional analysis. We recognize that the stock market does fixate on current earnings (Sloan, 1996) and our design specifically ensures that our intrinsic value measures are not biased by this problem. 22 Therefore, theoretically our measures of intrinsic value capture a construct that is sufficiently different from stock price. However, this does not imply that intrinsic values empirically capture a construct that is different from stock price. A comparison of our stock price (Table 2, Panel A) and intrinsic values (Table 3) regressions, however, indicates that stock prices do reflect some 22 We use three/five year hence market values for determining terminal value, and it is true that the stock market three or five years hence will fixate on the earnings of that year. However, as we argue earlier, this fixation merely adds measurement error to our intrinsic value rather than bias, because the mispricing of accrual and cash flows three or five year hence is not correlated to current earnings or cash flows which are our variables of interest. 22

25 measure of functional fixation that is not present in the intrinsic values. This inference can be drawn by noticing that magnitude of the superior explanatory power of earnings over cash flows is substantially larger for the market value models than the intrinsic value models (these results are particularly pronounced for the deflated specifications). Nevertheless, we conduct additional analysis to substantiate that we are documenting an economic phenomenon that is distinctly different from that documented by the stock price studies. Specifically, we partition our sample into three sub-samples (HIGH, MID and LOW) ranking them based on the extent of valuation errors. As in Penman and Sougiannis (1998), we define valuation error as the absolute value of the difference between market value and ex post intrinsic value scaled by market value. Table 4, Panel A reports correlation between market values and intrinsic values for the third of our sample with high absolute errors (HIGH) and that with the lowest valuation errors (LOW). 23 While the correlation for our undeflated measures are only marginally lower for the HIGH group compared to the LOW group (averages of 0.84 for the HIGH and 0.98 for the LOW groups), probably because of spurious correlation induced by scale, the correlation is dramatically different for our deflated measures. Specifically, the correlation ranges from 0.94 to 0.98 for the LOW group compared to a range of 0.23 to 0.30 for the HIGH group. This clearly suggests that for observations in the HIGH valuation error group, our (deflated) intrinsic value measures are sufficiently different from market value. We next present results of regression of alternative intrinsic value measures on earnings and cash flows separately for firms with HIGH and LOW valuation errors. For brevity we restrict our presentation to intrinsic values based on the three year horizon and only for specifications that allow separate coefficients for positive and negative observations (Panel B) and for specifications that only considers firm observations with positive cash flows and earnings 23 For brevity, we report results pertaining only to the three-year horizon intrinsic value measures. However, all our results are qualitatively similar for the five-year horizon. 23

26 (Panel C). 24 We find that earnings dominate cash flows for both the LOW and the HIGH error sub-samples and for all specifications. The results for the HIGH error group are particularly interesting because they show that our full sample results are not attributable to the high correlation between market values and our measures of intrinsic value. Rather these results imply that earnings are superior to cash flows in predicting ex post intrinsic values, which are a distinctly different economic construct from market values. 5. Concluding Remarks The objective of our study is to examine the relative importance of earnings and operating cash flows in equity valuation. We document that earnings are superior to cash flows in terms of explanatory power for ex post intrinsic values. This evidence supports the FASB s assertion that accrual based earnings is superior to cash flows in providing information about future cash flows to the investors (FASB 1978, #44). The salient feature of our study is the use of ex post intrinsic value of equity to examine this question. We believe that the ex post intrinsic value approach that we adopt contributes to this debate by providing unambiguous evidence about the superiority of earnings over operating cash flows as a summary indicator of fundamental equity value. Unlike stock returns, our measures of intrinsic value are not contaminated by the stock market s fixation on reported earnings and the consequent mispricing of the accruals and cash flow components of earnings (Sloan, 1996). Also, the intrinsic value measure that we use captures the present value of all future cash receipts to the investors and hence provides a more formal and comprehensive measure of fundamental equity value than measures such as future operating cash flows (or future earnings) used in prior research. In this sense, the intrinsic value approach is a more direct 24 Inferences are unchanged if we use the intrinsic values based on the five year horizon. 24

27 test of the FASB s assertion that accrual earnings are superior to cash flows in predicting cash flows to the investor. Nevertheless, our approach is subject to the following caveats. First, our intrinsic value measure may be criticized because our measure is not completely independent of stock prices. That is, we use ex post market value to compute terminal values. However, we believe that this approach is better than using ad hoc assumptions that may be otherwise required to determine terminal values. Also, our results are unlikely to be influenced by the anomaly identified by Sloan (1996) because the anomaly does not persist beyond two years. Second, we adopt a constant discount rate of 10% for all firms because prior research has shown that the results are not sensitive to selection of discount rates (Penman and Sougiannis, 1998) and because there is presently no consensus regarding estimating expected rates of return. This undeniably introduces measurement error into the intrinsic value estimations. Despite the measurement error, ex post intrinsic values are probably a more appropriate proxy for fundamental equity values than one or more future years operating cash flows. Third, our levels approach suffers from model misspecification problems when we consider negative realizations of earnings and cash flows. Adopting a changes specification could resolve this problem. Unfortunately, our research design precludes such an attempt. This limitation, to some extent, reduces the generalizability of our results. 25

28 References Ali,A.,1994. The incremental information content of earnings,working capital from operations, and cash flows. Journal of Accounting Research 32: Barth, M.E., Cram, D., and K. Nelson Accruals and the prediction of future cash flows. The Accounting Review 76: Beaver,W.H.,and R.E.Dukes Interperiod tax allocation,earnings expectations,and the behavior of security prices. The Accounting Review 47: Bowen, R.M., Burgstahler, D., and L.A. Daley Evidence on the relationships between earnings and various measures of cash flow. The Accounting Review 61: Bowen, R.M., Burgstahler, D., and L.A. Daley The incremental information content of accruals versus cash flows. The Accounting Review: 62: Burgstahler and Dichev Earnings, adaptation and equity value. The Accounting Review 72: Burgstahler, D., J. Jiambalvo, and Y. Pyo The informativeness of cash flows for future cash flows. Working paper, University of Washington. Copeland, T., T. Kohler and J. Murrin Valuation: Measuring and Managing the Value of Companies, 3 rd edition, John Wliey & Sons, New York. Dechow, P. M Accounting earnings and cash flows as measures of firm performance: the role of accounting accruals. Journal of Accounting and Economics 18: Elgers, P.T., Lo, M.H., and R. Pfeffer Analysts incorporation of the differential persistence of cash and accrual earnings components in forecasting annual earnings. Working paper, University of Massacusetts. Francis J., P. Olsen and D.R. Oswald Comparing the accuracy and explainability of dividend, free cash flow, and abnormal earnings equity value estimates. Journal of Accounting Research, Spring: Greenberg, Johnson and Ramesh, Earnings versus cash flow as a predictor of future cash flow measures. Journal of Accounting, Auditing & Finance, Jan, C.L., and J. Ou The role of negative earnings in the valuation of equity stocks. Working paper, New York University and Santa Clara University. Johnston, J Econometric Methods, Third Edition. McGraw-Hill Inc. 26

29 Lee, C.M.C., J. Myers, and B. Swaminathan What is the intrinsic value of the Dow? Journal of Finance 53, Ohlson. J Earnings, book values, and dividends in equity valuation. Contemporary Accounting Research (Spring): Palepu, K., P. Healy and V. Bernard Business Analysis and Valuation: Using Financial Statements, Text and Cases, Second Edition. Southwestern. Penman S. and T. Sougiannis A comparison of dividend, cash flow, and earnings approaches to equity valuation. Contemporary Accounting Research, Rayburn, J The association of operating cash flow and accruals with security returns. Journal of Accounting Research 24: Schipper, K Commentary on earnings management. Accounting Horizons 3: Sloan, R Do stock prices fully reflect information in accruals and cash flows about future earnings? The Accounting Review 71: Vuong, Q.H Likelihood ratio tests for model selection and non-nested hypotheses. Econometrica 57:

30 Table 1 Descriptive Statistics and Correlation Matrix Panel A: Descriptive Statistics Variable N Deflator Mean Median Std Dev Skewnes s Maximum Minimu m % Positive OCF Undeflated ($ m) % Book Value of Equity % Total Assets % EARN Undeflated ($ m) % Book Value of Equity % Total Assets % MKTV Undeflated ($ m) Book Value of Equity Total Assets IV_DDM Undeflated ($ m) Book Value of Equity Total Assets IV_RIM Undeflated ($ m) Book Value of Equity Total Assets % IV_DDM Undeflated ($ m) Book Value of Equity Total Assets IV_RIM Undeflated ($ m) Book Value of Equity Total Assets

31 Table 1 continued Panel B: Correlation statistics Undeflated Deflated by Book Value Deflated by Total Assets Corr (MKTV, IV_DDM 3 ) Corr (MKTV, IV_RIM 3 ) Corr (IV_DDM 3, IV_RIM 3 ) Corr (MKTV, IV_DDM 5 ) Corr (MKTV, IV_RIM 5 ) Corr (IV_DDM 5, IV_RIM 5 ) Notes: 1. OCF represents operating cash flows, EARN represents income before extraordinary items, MKTV represents market value of equity, IV_DDM 3, IV_DDM 5 (IV_RIM 3, IV_RIM 5 ) represents intrinsic value using dividend discount model (residual income model) for three- and fiveyear horizons respectively. 2. All correlation coefficient estimates are statistically significant at the 1% (two-tailed). 29

32 Table 2 Relative Explanatory Power of Earnings and Cash Flows for Contemporaneous Stock Price and Future Earnings/Cash Flows Undeflated Deflated by Book Value Deflated by Total Assets Dependent Variable Explanatory Power Vuong Explanatory Power Vuong Explanatory Power Vuong OCF t EARN t Statistic OCF t EARN t Statistic OCF t EARN t Statistic Panel A: Market Value of Equity All Firm-Years 63.70% 73.68% 8.85* 3.49% 4.93% 3.57* 5.48% 3.76% 7.42* [N=38590, 37561, 38512] (6.28) (13.55) (-1.39) (-1.56) (-2.27) (-1.72) All Firm Years Separate Coefficients for 63.83% 76.01% 11.27* 10.26% 15.28% 7.23* 14.44% 22.26% 13.08* Negative and Positive Quadrants (-4.85) (-10.69) (-3.50) (-2.91) (-5.20) (-4.25) [N=38590, 37561, 38512] (6.31) (14.00) (1.65) (7.82) (3.68) (11.81) Firm years with positive EARN and OCF 63.00% 77.17% 13.00* 5.08% 28.41% 18.43* 9.92% 36.06% 29.50* [N=24518, 24339, 24452] (6.37) (13.93) (2.18) (11.28) (5.64) (15.24) Panel B: Future Operating Cash Flows All Firm-Years One Year Ahead OCF 70.66% 55.99% 5.44* 19.61% 10.44% 6.61* 29.87% 22.36% 7.67* [N=38590, 37561, 38512] (0.96) (1.72) (0.56) (0.39) (0.69) (0.55) All Firm-Years Two Years Ahead OCF 61.31% 49.09% 3.24* 11.96% 4.36% 5.33* 22.38% 17.24% 7.26* [N=38590, 37561, 38512] (0.97) (1.75) (0.61) (0.33) (0.72) (0.58) All Firm-Years Three Years Ahead OCF 55.00% 43.12% 4.42* 7.17% 3.05% 3.92* 10.49% 8.22% 2.50** [N=38590, 37561, 38512] (1.05) (1.89) (0.68) (0.40) (0.79) (0.64) Panel C: Future Earnings All Firm-Years One Year Ahead EARN 52.87% 67.18% 9.34* 14.88% 24.97% 4.71* 24.07% 27.71% 1.60 [N=38590, 37561, 38512] (0.40) (0.92) (0.45) (0.55) (0.67) (0.66) All Firm-Years Two Years Ahead EARN 45.78% 55.84% 7.67* 8.26% 11.86% 3.37* 16.02% 18.40% 4.39* [N=38590, 37561, 38512] (0.42) (0.92) (0.45) (0.51) (0.72) (0.71) All Firm-Years Three Years Ahead EARN 39.84% 44.59% 3.39* 5.63% 7.19% 2.05** 9.90% 11.47% 2.47** [N=38590, 37561, 38512] (0.44) (0.93) (0.55) (0.59) (0.83) (0.82) Notes: 1. OCF represents operating cash flows, EARN represents income before extraordinary items, MKV represents market value of equity. *,** represent significant Vuong statistics at the 1% and 5% level respectively. 2. Coefficient estimates are presented in parenthesis with significance at 1% level (two-tailed) represented in bold. For specifications where separate coefficients for positive and negative quadrants are allowed the coefficients for negative and positive coefficients are presented parenthetically in that order. 30

33 Table 3 Relative Explanatory Power of Earnings and Cash Flows for Alternative Measures of Ex post Intrinsic Value Dependent Variable Undeflated Deflated by Book Value Deflated by Total Assets Explanatory Power Vuong Explanatory Power Vuong Explanatory Power Vuong OCF t EARN t Statistic OCF t EARN t Statistic OCF t EARN t Statistic Panel A: All Firms Years IV DDM % 64.48% 8.26* 0.09% 0.91% 4.98* 0.57% 0.53% 0.55 [N=38590, 37561, 38512] (6.86) (14.75) (-0.28) (-0.86) (-0.88) (-0.77) IV RIM % 65.96% 8.79* 0.07% 0.09% % 0.00% 0.12 [N=38590, 37561, 38512] (6.91) (14.95) (0.23) (-0.24) (-0.03) (0.01) IV DDM % 61.04% 6.20* 0.00% 0.66% 3.07* 0.01% 0.04% 0.98 [N=23516, 23089, 23502] (6.88) (15.04) (-0.08) (-0.87) (-0.11) (-0.28) IV RIM % 62.67% 6.75* 0.27% 0.00% 2.25** 0.53% 0.40% 1.39 [N=23516, 23089, 23502] (6.91) (15.19) (0.51) (-0.05) (0.90) (0.76) Panel B: All Firm Years With Separate Coefficients for Positive and Negative Quadrants IV DDM % 66.63% 10.56* 4.42% 9.03% 11.18* 7.60% 14.17% 15.36* [N=38590, 37561, 38512] (-4.87) (-12.35) (-2.43) (-2.38) (-3.98) (-3.37) (6.87) (15.25) (2.82) (9.73) (5.41) (13.08) IV RIM % 68.07% 11.01* 3.56% 8.64% 13.27* 6.29% 13.63% 17.96* [N=38590, 37561, 38512] (-4.56) (-11.93) (-1.54) (-1.68) (-2.73) (-2.39) (6.94) (15.44) (2.78) (9.73) (5.43) (12.82) IV DDM % 62.80% 7.60* 4.55% 9.40% 9.07* 6.81% 13.35% 12.53* [N=23516, 23089, 23502] (-7.13) (-13.56) (-2.87) (-2.81) (-4.11) (-3.82) (6.91) (15.47) (3.07) (10.47) (5.95) (13.33) IV RIM % 64.39% 8.07* 3.68% 8.98% 10.82* 6.23% 13.51% 14.40* [N=23516, 23089, 23502] (-7.02) (-12.94) (-1.70) (-1.85) (-2.44) (-2.45) (6.94) (15.63) (3.00) (10.47) (5.97) (13.09) 31

34 Table 3 continued Dependent Variable Panel C: Only Firm Years with Positive EARN and OCF Undeflated Deflated by Book Value Deflated by Total Assets Explanatory Power Vuong Explanatory Power Vuong Explanatory Power Vuong OCF t EARN t Statistic OCF t EARN t Statistic OCF t EARN t Statistic IV DDM % 67.62% 11.95* 4.22% 13.68% 16.46* 8.40% 21.22% 20.40* [N=24518, 24339, 24452] (6.80) (15.15) (3.38) (13.24) (7.21) (16.22) IV RIM % 68.88% 12.26* 4.44% 14.39% 17.15* 9.06% 22.01% 20.52* [N=24518, 24339, 24452] (6.88) (15.35) (3.33) (13.05) (7.21) (15.92) IV DDM % 63.30% 8.02* 3.69% 12.31% 12.62* 7.80% 18.07% 14.80* [N=15969, 15858, 15879] (6.86) (15.35) (3.38) (13.53) (7.45) (16.01) IV RIM % 64.79% 8.45* 3.81% 12.93% 13.20* 8.27% 18.81% 15.11* [N=15969, 15858, 15879] (6.89) (15.50) (3.30) (13.30) (7.39) (15.69) Notes: 1. OCF represents operating cash flows, EARN represents income before extraordinary items, IV_DDM 3, IV_DDM 5 (IV_RIM 3, IV_RIM 5 ) represents intrinsic value using dividend discount model (residual income model) for three- and five-year horizons respectively. *,** represent significant Vuong statistics at the 1% and 5% level respectively. 2. Coefficient estimates are presented in parenthesis with significance at 1% level (two-tailed) represented in bold. For specifications where separate coefficients for positive and negative quadrants are allowed the coefficients for negative and positive coefficients are presented parenthetically in that order. 32

35 Table 4 Relative Explanatory Power of Earnings and Cash Flows for Alternative Measures of Ex post Intrinsic Value Conditional on Magnitude of Valuation Errors Panel A: Pearson Correlation Between Alternative Intrinsic Value Measures and Market Value Undeflated Deflated by Book Value Deflated by Total Assets All Firm Years Corr (MKTV, IV_DDM 3 ) Corr (MKTV, IV_RIM 3 ) LOW ERROR Corr (MKTV, IV_DDM 3 ) Corr (MKTV, IV_RIM 3 ) HIGH ERROR Corr (MKTV, IV_DDM 3 ) Corr (MKTV, IV_RIM 3 )

36 Table 4 (continued) Dependent Variable Undeflated Deflated by Book Value Deflated by Total Assets Explanatory Power Vuong Explanatory Power Vuong Explanatory Power Vuong OCF t EARN t Statistic OCF t EARN t Statistic OCF t EARN t Statistic Panel B: All Firm Years With Separate Coefficients for Positive and Negative Quadrants IV DDM 3 LOW ERROR 64.98% 75.06% 7.01* 12.50% 22.29% 5.89* 15.74% 29.41% 10.35* [N=12863,12520,12837] (-6.44) (-15.45) (-3.50) (-3.15) (-5.24) (-4.42) (6.36) (13.81) (2.63) (9.01) (4.67) (12.27) IV DDM 3 HIGH ERROR 43.67% 59.79% 5.40* 2.61% 6.39% 7.41* 5.28% 10.89% 8.90* [N=12863,12520,12837] (-5.11) (-11.43) (-1.34) (-1.72) (-2.93) (-2.57) (9.06) (23.06) (3.68) (12.69) (7.40) (17.07) IV RIM 3 LOW ERROR 65.87% 75.18% 6.74* 10.32% 21.33% 7.14* 15.06% 29.43% 11.31* [N=12863,12520,12837] (-6.61) (-16.45) (-3.01) (-2.85) (-4.84) (-4.00) (6.38) (13.93) (2.57) (9.29) (4.90) (12.30) IV RIM 3 HIGH ERROR 47.03% 63.12% 5.36* 2.84% 7.19% 7.86* 5.29% 11.57% 9.56* [N=12863,12520,12837] (-5.38) (-10.10) (-0.42) (-0.94) (-1.35) (-1.36) (9.58) (23.57) (3.62) (12.81) (7.36) (11.52) Panel C: Only Firm Years with Positive EARN and OCF IV_DDM 3 LOW ERROR 64.22% 76.33% 7.58* 8.57% 31.27% 11.88* 13.50% 40.52% 16.85* [N=6390,6085,6113] (6.44) (13.90) (2.84) (11.54) (6.11) (14.52) IV_DDM 3 HIGH ERROR 43.71% 62.95% 6.84* 3.26% 10.56% 8.95* 7.12% 17.06% 10.38* [N=6390,6085,6113] (8.46) (21.96) (3.94) (15.93) (8.64) (19.47) IV_RIM 3 LOW ERROR 64.42% 76.46% 7.93* 8.56% 31.88% 12.61* 14.61% 41.27% 17.88* [N=6390,6085,6113] (6.33) (13.96) (2.86) (11.76) (6.37) (14.66) IV_RIM 3 HIGH ERROR 47.24% 64.68% 6.78* 3.13% 10.48% 9.22* 7.43% 16.80% 9.92* [N=6390,6085,6113] (8.81) (21.85) (3.70) (15.10) (8.49) (18.48) 34

37 Table 4 continued Notes: 1. OCF represents operating cash flows, EARN represents income before extraordinary items, IV_DDM 3 (IV_RIM 3 ) represents intrinsic value using dividend discount model (residual income model) for three-year horizon. *,** represent significant Vuong statistics at the 1% and 5% level respectively. 2. Coefficient estimates are presented in parenthesis with significance at 1% level (two-tailed) represented in bold. For specifications where separate coefficients for positive and negative quadrants are allowed the coefficients for negative and positive coefficients are presented parenthetically in that order. 3. HIGH (LOW) ERROR represents observations in the top (bottom) third of firm observations ranked on the basis of valuation error. Valuation error is defined as the absolute value of the difference between stock price and expost intrinsic value measure scaled by stock price. 35