VIEWS: 7 PAGES: 54 CATEGORY: Markets / Industries POSTED ON: 3/13/2009 Public Domain
The Impact of Analysts’ Forecast Errors and Forecast Revisions on Stock Prices William Beaver,1 Bradford Cornell,2 Wayne R. Landsman,3 and Stephen R. Stubben3 April 2007 1. Graduate School of Business, Stanford University, Stanford, CA 94305. 2. California Institute of Technology, Pasadena, CA 91125 3. Kenan-Flagler Business School, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599. We thank I/B/E/S International for providing data on analysts’ earnings estimates, and the Center for Finance and Accounting Research, University of North Carolina for providing financial support. We also thank workshop participants at the 2005 Stanford Summer Camp, the University of Florida, and an anonymous reviewer for helpful comments. Corresponding author: William Beaver, william_beaver@gsb.stanford.edu. The Impact of Analysts’ Forecast Errors and Forecast Revisions on Stock Prices Abstract We present a comprehensive analysis of the contemporaneous association between security returns, quarterly earnings forecast errors, and quarter-ahead and year-ahead earnings forecast revisions in the context of a fully specified model. We find that all three variables have significant pricing effects, indicating each conveys information content. The findings hold across years, across industries, and are robust to two procedures extending the event window. Findings also show that the fourth quarter is significantly different from the other three quarters. In particular, in the fourth quarter the relative importance of the forecast error is lower, while the relative importance of the quarter-ahead forecast revision increases. We find also a marked upward shift over time in the forecast error coefficients, even in the presence of the forecast revision variables, whose coefficient also exhibit a significant but less dramatic shift. This finding is consistent with the I/B/E/S data base reflecting an improved quality of earnings forecasts, as well as an improved measure of actual earnings. 1. Introduction One of the fundamental questions in finance and accounting is the impact of earnings surprises on stock prices. The question not only is important for evaluating theories that relate reported accounting numbers to firm value, but also has widespread implications for regulation and the law. For instance, in legal disputes related to financial reporting a central issue is how much the stock price would have been affected if the company released its “correct” earnings in place of allegedly inflated earnings. Proper analysis of that issue requires an appropriately specified model of the relation between earnings innovations and stock prices. Empirical studies of this question employ analysts’ earnings forecast data as proxies for market expectations and, thereby, to measure earnings surprises. In an early paper, Cornell and Landsman (1989) demonstrate that the earnings surprise should not be identified solely with analysts’ forecast errors. They stress that a properly specified model of residual returns must simultaneously take account of both earnings forecast errors and earnings forecast revisions. They present evidence to show that if the forecast revisions are excluded, the response coefficient on the forecast error is higher because forecast revisions are in part based on forecast errors. In this paper, we present a comprehensive analysis of the relation between stock returns, analysts’ forecast errors and analysts’ forecast revisions. Despite the fact that there has been extensive new research on the relation between analysts’ forecasts and stock prices, which we review below, much of this literature has not taken account of the combined impact of forecast errors and forecast revisions. As we show, this can lead to potentially misleading results. In addition, there have been improvements in the nature, quantity, and quality of the data used to measure forecast errors and forecast revisions. Whereas Cornell and Landsman (1989) is 3 based on only three years of data and a limited number of firms, our sample comprises 20 years of data covering a much greater number of firms. Moreover, with respect to the I/B/E/S data that we use in this paper, there has been an increasing effort over time by I/B/E/S to ensure a consistency between the forecast and the realization of earnings, as well as a consistency across analysts in the earnings number being forecast. This consistency is attained by ensuring the same earnings components are included (and excluded) in the “actual” and forecasted earnings. Presumably, the effects of these efforts could alter the observed relation between security returns, forecast errors and forecast revisions. More specifically, as the I/B/E/S database becomes more successful in providing an “apples to apples” comparison, the quality of the forecast error is expected to improve because it becomes a better proxy for unexpected earnings. The resulting reduction in measurement error should affect the estimated coefficients in regression models. In addition to improving the quality of the data, I/B/E/S has extended coverage over time making the data more comprehensive. This alteration in the composition of the data may also affect the empirical estimation of the security return model. We examine whether there has been an increase or decrease over time in the information content of forecast errors and forecast revisions to assess the extent to which changes in the nature of the data have affected the observed relations. Furthermore, it has been suggested that companies have come under added pressure to “manage” earnings and that this may affect the relation between residual returns, forecast errors and forecast revisions (Matsumoto, 2002; Abarbanell and Lehavy, 2003; Burgstahler and Eames, 2003). For example, it may lead to reduced information content of the earnings forecast error over time. Proper examination of whether this occurs requires a more fully specified model that takes account of both forecast errors and forecast revisions over time. 4 Finally, it also has been suggested that managers are increasingly actively managing analysts’ expectations to avoid negative earnings surprises, which may affect the relation between residual returns, forecast errors and forecast revisions (Brown, 2001; Matsumoto, 2002). Presumably, if this activity has increased over time and adversely affected the quality of both the forecast errors as well as the forecast revisions, the change in the relation between residual returns, forecast errors and forecast revisions over time should show up over time. Analysis of these questions requires a comprehensive examination of the relation between residual stock returns in the period surrounding quarterly earnings announcements, earnings forecast errors, and revisions in quarter-ahead and subsequent year-ahead analysts’ earnings forecasts during the period from 1984 to 2003. The length of the sample period permits us to examine whether changes in the properties of the earnings forecasts result in any perceptible trends in the coefficients on the forecast error and the forecast revisions. In addition, the growth in I/B/E/S coverage also permits us to control for potential mean differences in industry effects and to examine whether the observed relation is consistent across industries. Furthermore, the availability of an I/B/E/S “actual” earnings number, which was not provided when the database first became available, permits us to compare the properties of different specifications including forecast errors based on I/B/E/S actuals versus Compustat earnings. We also examine two important specification issues: the distinct nature of fourth quarter earnings and the measurement of the residual return interval. With respect to the first issue, we consider whether the relation between residual returns, forecast errors and forecast revisions differs during the fourth quarter for a variety of reasons that we discuss later in the paper. If this is so, failure to take account of the fourth quarter effect will lead to a misspecified model and, quite likely, biased coefficients. To study this possibility, we develop specifications that permit 5 the fourth quarter slope coefficients to differ from those of the interim quarters, and that take account of the intertemporal overlap in measurement of the quarter-ahead and year-ahead forecast revisions that occurs during the fourth quarter. With respect to the residual return window, models that incorporate both forecast errors and forecast revisions face a unique data problem. The problem arises because the forecast error, by definition, is observed at the time of the earnings announcement, but the forecasts revisions are not made available until a later date. This raises two issues. The first issue is that at the time of the earnings announcement the market must use the information in the forecast error to anticipate its long-run impact, and thereby its effect on analysts’ forecast revisions, without observing the revisions. Therefore, the residual return reflects both the forecast error and the forecast revisions expected at the earnings announcement date. However, by necessity, the model includes actual forecast revisions, which likely measure the market’s expectations with error. To take account of this feature of the data, we extend the basic model in two ways. First, we extend the window over which the residual return is measured to the date at which the forecast revisions are observed. This assures us that the residual return will reflect both actual forecast errors and actual forecast revisions. A problem with this approach is that the window must be extended, on occasion, to more than two months after the earnings announcement to be sure the I/B/E/S consensus reflects forecasts made after the earnings announcement. By extending the return window, the coefficients on the forecast revisions will reflect information available subsequent to the earnings announcement. To counter this problem, the second approach turns to disaggregated data. Rather than using the I/B/E/S consensus forecasts, we employ the individual analysts’ forecasts to construct a custom consensus forecast following the 6 earnings announcement. In this way, we can shorten the window by using the subset of the individual forecasts that are available soon after the earnings announcement. The major findings are: First, in every model we estimate both the forecast error and the forecast revision coefficients are highly significant. In other words, neither the forecast error nor the forecast revisions dominate in that each provides information content not contained in the other. Second, based on twenty years of data, we find that, even in the presence of the forecast revision variables, the coefficient of the forecast error still increases substantially over time, with a marked shift in post 1991 period. Third, in contrast, the coefficients on the two forecast revisions exhibit a similar but less dramatic shift. We present evidence suggesting that the increase in the coefficients is attributable to joint effects of the improved quality of the I/B/E/S actual earnings and analysts’ earnings forecasts over the sample period. This finding is important because it indicates that the significance of the forecast revisions in explaining the cross-sectional variation in earnings announcement residual returns is not an artifact of measurement error in the forecast error. Rather, the significance of the forecast revision coefficients is a robust finding that holds up through time despite changes in database quality and changes in the institutional features of the earnings reporting environment. Findings from separate industry regressions indicate that although there are cross-industry differences in the magnitude of coefficients on the forecast error and the forecast revisions, the basic relation holds across all industry groups. Fourth, the results further support the view that the fourth quarter is different than other quarters. The evidence is consistent with the market reacting in the fourth quarter more strongly to the change in the next quarter forecast revision and less strongly to the forecast error. This finding suggests that a revision in the quarter-ahead forecast in the fourth quarter, which is the 7 forecast revision for the first quarter of the next fiscal year, conveys more information than earlier quarters’ forecast revisions, which refer to later quarters in the same fiscal year Fifth, findings from estimations that extend the announcement event window indicate the primary results are robust, but the impact of the forecast revisions, as compared to the forecast errors, increases. This supports the notion that when the market observes the actual forecast revisions prices are adjusted to take account of the difference between the forecast revisions that are observed and the forecast revisions that were expected at the time of the earnings announcement. These increased coefficients are also consistent with the forecast revisions reflecting information available after the earnings announcement. Consistent with these arguments, the subsequent move in stock price is correlated with the observed revisions, but not necessarily with the (earlier) forecast error. To summarize, our results emphasize the importance of using a properly specified model when assessing the impact of the release of earnings information on stock prices. Models that fail to include forecast revisions, fail to take account of the changing nature of the I/B/E/S data, or fail to adjust for fourth quarter effects will produce earnings response coefficients that to not correctly characterize the relation between reported earnings and firm value. The remainder of the paper is organized as follows. In the next section, we review the key findings of the research on the relation between analysts’ forecast errors and stock returns. Section three presents the research methodology and methods for measuring the variables. Section four describes the sample data. Section five presents the results and discusses their implications. The conclusions are summarized in the final section. 2. Prior Research 8 Using I/B/E/S consensus analyst forecast data, Cornell and Landsman (1989) study the pricing effects of earnings forecast errors and earnings forecast revisions in the period surrounding quarterly earnings announcements. The key finding of their study is that both the one-quarter-ahead and one-year-ahead forecast revisions have important explanatory power in addition to the earnings surprise. An important conclusion based on their findings is that a properly specified model of residual returns in response to the release of quarterly earnings must simultaneously take account of both earnings forecast errors and earnings forecast revisions. They present evidence to show that if the forecast revisions are excluded from the basic model, the coefficient on the forecast error is higher because the error serves as a proxy for the forecast revisions and must be interpreted accordingly. In the years following the Cornell and Landsman study, surprisingly few studies have used the more completely specified model. A notable exception is Liu and Thomas (2000), which models stock returns as a function of annual forecast errors, annual forecast revisions, and an estimated annual revision in terminal value. Liu and Thomas finds that both the forecast error and forecast revisions provide incremental explanatory power. This study differs from Liu and Thomas in several respects: (1) Whereas Liu and Thomas relates annual stock returns with earnings variables, we examine the shorter-term announcement effects of the earnings variables in the spirit of an earnings announcement event study. Given the variability of stock returns, our shorter horizon tests have considerably more power. (2) Liu and Thomas examines only annual earnings; our research design measures earnings variables for annual and interim quarters. Hence, our research designs permits us to address additional issues, such as the differential behavior of the fourth quarter. (3) Liu and Thomas reports results based on pooled cross- sectional and time-series data and does not examine how the coefficients may have changed over 9 time. Further, year-by-year estimation permits the calculation of test statistics that are not affected by cross-sectional correlation in the data leading to less biased test statistics than those based on pooled estimation. (4) Liu and Thomas includes earnings variables, including revisions in long-term earnings forecasts and terminal values, that are based on the authors’ extrapolations and are not reported by I/B/E/S. Hence, the results reflect the joint effect of I/B/E/S reported variables and their extrapolations using I/B/E/S and other data. Although the number of studies that model stock returns as a function of both forecast errors and revisions is relatively small, there is a much larger literature on the properties of forecast errors and analysts’ forecasts. We briefly summarize key studies in both of this literature that provide some background to our study. A number of papers study the properties of earnings response coefficients using alternative earnings measures (Bradshaw and Sloan, 2002; Brown and Sivakumar, 2003; Lougee and Marquardt, 2004: Abarbanell and Lehavy, 2005). Of particular relevance to our study is Bradshaw and Sloan (2002), which documents that annual earnings response coefficients are higher when the forecast error is defined using I/B/E/S (i.e., “Street” earnings) rather than Compustat net income (i.e., “GAAP” earnings), and the difference in price response based on the two measures has increased over time. In particular, in the post- 1992 period there is a significant increase in the earnings response coefficient for I/B/E/S earnings forecast errors. Bradshaw and Sloan (2002) attributes these findings to analysts excluding over time an increasing number of special items from their earnings estimates, and to the increasing prevalence of special items, which predominately occur in the fourth quarter. The key distinction between our study and prior studies examining the properties of earnings response coefficients, including Bradshaw and Sloan (2002), is that we include in our regressions analysts’ forecast revisions for quarter-ahead and year-ahead earnings. Not only 10 does this permit us to study the price response to forecast revisions, but this also changes the interpretation of the coefficient on the forecast error. In particular, in a fully specified model the forecast revisions control for the future implications of the forecast error, which results in a coefficient on the forecast error that is not affected by the persistence of current earnings. This model allows us to examine if there is a shift in the earnings response coefficients in the presence of earnings forecast revisions for reasons other than a change in earnings persistence over time. Further, we are able to examine whether there has been a similar upward trend over time in the coefficients on the earnings revisions variables themselves. Neither is possible in the context of a model that contains only earnings forecast errors. One issue raised by Cornell and Landsman is whether the structural relation between the earnings variables and stock return in the fourth quarter could differ from that of the interim quarters. They raise the possibility that fourth quarter could differ because of the increased frequency of special items and because the fourth quarter result will reflect the effects of the audit process. If the fourth quarter is significantly different, and if this fact is not taken into account in the model specification, the estimated relation between stock returns and forecasts errors will not be properly measured. In the context of a model that includes only the earnings forecast error, Mendelhall and Nichols (1988) finds that the market reacts relatively less strongly to bad news in the fourth quarter because of the ability of managers to delay the reporting of bad news in earlier quarterly earnings, but which is effectively leaked to the market in earlier quarters. However, it is difficult to predict whether their results would hold in the presence of forecast revisions. Prior research examining the properties of analysts’ forecasts is substantial. Brown (1996) synthesizes a vast literature of the forecasting accuracy of analysts’ versus naïve 11 statistical models, concluding that analysts’ forecast outperform statistical models, that the forecast error has not increased over time, and that over subperiods of time analysts’ forecasts have been pessimistically, rather optimistically biased. Lys and Sohn (1990) find that even though security returns can predict a portion of the forecast revision, the analysts’ forecasts are incrementally informative. One key paper, Abarbanell and Bernard (2000), suggests that analysts’ forecasts do not fully reflect the implications of earnings forecast errors in their forecast revisions. Subsequently, Gleason and Lee (2003) document a post-revision price drift and suggest the market does not fully reflect the information content of the forecast revision. In particular, their evidence suggests that the market does not make a sufficient distinction between revisions that provide new information and those that merely move toward to consensus. Another strand of analyst research has examined the contention that mangers have increasingly guided analysts’ forecasts downward so that earnings meet or beat analysts’ forecasts (Brown, 2001; Matsumoto, 2002). Presumably, to the extent that this pressure on analysts has affected their forecasts, it could also to affect the relation between residual returns, forecast errors and forecast revisions. Other research related to analysts focuses on the suggestion that companies have faced increasing pressure over time to “manage” earnings and that this may have affected the relation between residual returns, forecast errors and forecast revisions (Matsumoto, 2002; Abarbanell and Lehavy, 2003; Burgstahler and Eames, 2003). In particular, successful earnings management could affect the earnings surprise coefficient over time as earnings management increases. In addition, if analysts do not fully incorporate the effects of earnings management in their forecast revisions, this also could affect their coefficients over time as earnings management increases. The only way to explore these issues fully is in the context of a model 12 that includes both forecast errors and forecast revisions, takes account of a possible fourth quarter effect, and then examines how the coefficients change over time and across industries. These streams of research motivate our interest in examining several issues: (1) Is the rise in the coefficient on the I/B/E/S forecast still observed in the presence of the forecast error revisions? (2) Is there a similar increase in the coefficients on the forecast revisions over time? (3) In a fully specified model, is the structural relation of the model in the fourth quarter different from that of the interim quarters and has that model also changed over time? (4) Are the findings robust with respect to the alternative specifications of the announcement window? 3. Research Design 3.1 The Model Based on a valuation model that expresses equity market value as the present value of future cash flows, Cornell and Landsman derives a model where change in equity value is equal to a linear function of the cash flow forecast error and a series of revisions in expectation about future periods’ cash flows. Assuming that cash flow forecast errors (changes in future expected cash flows) are collinear with the earnings forecast errors (forecast revisions), they then derive an empirical estimation equation that appears as equation (1) below. Subsequent to Cornell and Landsman, Ohlson (1995) and Feltham and Ohlson (1995) develop a characterization of equity market value as a linear function of equity book value and the present value of future expected abnormal earnings. Moreover, Feltham and Ohlson (1996) demonstrates an equivalency between the cash flow and abnormal earnings representations. Here, we present a valuation model based on the Feltham and Ohlson abnormal earnings formulation. Empirically the stream of future expected abnormal earnings is truncated at some 13 point and a terminal value is estimated. From this price levels equation, it is straightforward to derive an expression for the unexpected security return as a function of unexpected current earnings and the change in the future expected abnormal earnings, and in the case of truncation a change in expected terminal value. In particular, the model developed by Liu and Thomas (2000) expresses unexpected security returns as: [their Equation (10)] URit = β0 + β 1UEit + β 2RAE2it +β 3RAE3it + β 4RAE4it +β 5RAE5it + β 6RTERMit + eit, where UR is the expected stock return, UE is the earnings surprise with respect to current abnormal earnings, RAE is the revised expectations about future abnormal earnings for the next four accounting periods, and RTERM is the revision in the estimated terminal value at the end of the horizon. The Liu and Thomas model is developed in context of annual returns and annual revisions in future expected earnings. In our context, which is announcement period returns for quarterly announcements, UE is represented by the forecast error on current quarterly earnings. In the most general model, there would be separate estimates for each of the revisions in future quarterly earnings for a finite period and the revision in expected terminal value. Our estimating model is a parsimonious version of the Liu and Thomas model, which, as described in detail below, reflects the structure of the I/B/E/S analysts’ forecast data, including the availability of the data, the frequency with which the forecast variables are revised and the collinearity among the forecasted variables. In particular, our estimating equation is: ARit = a0 + a1 FEit + a2 FRQit + a3 FRYit + eit (1) where AR is the unexpected security return, FE is the forecast error for current quarterly earnings, FRQ is the revision in the I/B/E/S consensus forecast for the next quarter, and FRY is the revision in I/B/E/S consensus forecast for the next fiscal year. 14 One set of potentially omitted variables is the revisions in the quarterly earnings for the remaining portion of the current fiscal year. For example, for the first quarter, FE is the first quarter forecast error, FRQ is the revision in the forecast for the second quarter, and the forecast revisions for the third and fourth quarter are omitted. There are several problems with including these additional variables. First, the number of observations for which the forecast revision is available more than one period ahead is limited. Second, the length of the remaining portion of the current fiscal year shrinks as for each of the later quarters (e.g., for the second quarter there are only two quarters remaining), and it unclear how one would incorporate the varying time horizon into estimating equation (1). Third, the revisions in forecasts for the remaining quarters are significantly correlated with one another. However, notwithstanding these difficulties, we conducted a complete specification for the first quarter for those observations where a forecast revision was reported. We found the overall explanatory power to be essentially the same as that of Equation (1). Hence, we rely on the more parsimonious form of the estimating equation. The main point to emphasize is that the coefficient on FRQ reflects the pricing multiplier that reflects the revisions for the remaining quarters as well.1 Similarly, revisions in annual earnings beyond FRY are potentially omitted variables from Equation (1). As Liu and Thomas (2000) point out, the limited availability of I/B/E/S annual forecast revisions beyond one year results in a substantial reduction in number of observations. To require additional FRY for two-years hence would reduce the sample size by 65 percent and to further require FRY terms beyond two years would reduce the sample size by over 90 percent. 1 Alternatively, we could construct our own estimates of the forecast revision for the remaining quarters based upon some extrapolation of the FRQ. This is the approach employed by Liu and Thomas (2000) to project annual forecasts beyond those reported by I/B/E/S. We have chosen not to use such an approach here because then the findings would be a joint function of I/B/E/S data and our extrapolation procedure. Moreover, in conducting preliminary calculations over our interval of revision (two months as opposed to one year in Liu and Thomas), we found the extrapolated variables to be so highly correlated with the reported variable (FRQ) that no increase in 15 Moreover, regressions including these variables does not produce any increase in explanatory power.2 Liu and Thomas (2000) constructs estimates of long-term earnings revisions based on the reported I/B/E/S short-term earnings forecasts and the I/B/E/S long-term growth rate. We examined the feasibility of using similar extrapolated variables. For our period of revision (two months versus one year), we found that the revision in long-term growth rates was zero for 68 percent of our observations. Because of this, the resulting extrapolations would produce revision variables that would either be zero or highly collinear with FRY. As a result, incorporation of these extrapolated variables would not add significantly to explanatory power and would only provide an illusion of additional variables that are in fact linear extrapolations of the FRY variable and a growth variable that is predominately zero.3 As with the possibility of including additional terms for interim quarter forecasts revisions beyond FRQ, it is more straightforward to simply include only FRY and to interpret its coefficient accordingly—namely, the coefficient also reflects the extent to which FRY is correlated with revisions in expected subsequent earnings. Further, there is no revision in terminal value calculation in Equation (1). Not only it is a purely extrapolated value in the sense that I/B/E/S does not report terminal value, but revisions in terminal value are greatly affected by revisions in the long-term growth rate, which was zero for a 68 percent of our observations. For this reason, we do not revision in terminal value in the estimating equation. Consistent with the standard approach in the literature, we measure analysts’ forecast- based variables using consensus forecasts in the I/B/E/S summary file. On the Thursday before explanatory power was provided. It is more straightforward to simply include only FRQ and to interpret its coefficient accordingly. 2 Not surprisingly, the coefficients on these variables are positive, much smaller than for FRY and closer to zero. As a result, the coefficient on FRY is slightly lower but the overall explanatory power remains the same. 3 One might dismiss 68 percent of the observations being zero as not being a sufficient reason for not using the growth rate. However, we feel the smaller propensity to update long-term forecasts is not a reflection of changing expectations and hence is a stale variable measured with considerable error. 16 the third Friday of each month, I/B/E/S calculates the consensus forecast as the mean or median of all outstanding estimates for a particular fiscal period. Forecasts are available for a variety of fiscal periods, including the current quarter, the next quarter, the current fiscal year, and the next fiscal year. Additional horizons are available, but analysts’ forecasts for these periods are less frequent. The ideal measurement of the response of security prices to earnings announcements and earnings forecast revisions would use a consensus forecast made just prior to an earnings announcement, and another made just after. However, consensus forecasts are compiled only monthly. Preannouncement forecasts, then, are the most recent consensus forecasts compiled before the earnings announcement date. Postannouncement forecasts are compiled the second month after the earnings announcement. Consensus forecasts for the first month after the earnings announcement are not used because they may contain individual forecasts issued both before and after the earnings announcement. As shown in the hypothetical example in figure 1, preannouncement forecasts are gathered on the last forecast date before the earnings announcement, March 19. In general, the time between the preannouncement forecast date and the earnings announcement will vary up to one month. Since the April forecast period might contain forecasts made both before and after the earnings announcement, postannouncement forecasts are instead gathered on May 21. Abnormal stock returns are calculated from the close of the preannouncement date, March 19, until one trading day after the earnings announcement, March 24, and abnormal stock returns over the extended window regressions described in section 3.3 are calculated until the end of the postannouncement period, May 21. 17 Our initial tests are based on the Cornell and Landsman regression given by equation (1) above, where i, t are indices referring to a sample firm and an announcement quarter. ARit = the abnormal stock return for firm i associated with quarterly earnings announcement t. ARit is measured from the close of the day of the announcement of the most recent I/B/E/S consensus forecast prior to the earnings announcement date (which we refer to as the last day of the preannouncement forecast period) through the trading day following the earnings announcement (see figure 1). The abnormal return is computed by subtracting the compounded daily mean return for the corresponding size decile, rdec, from the compounded daily firm return, r, over the period described above. That is, ARit = Π s (1 + rs ) − Π s (1 + rsdec ). FEit = the forecast error for firm i and quarterly earnings announcement t. FEit, which is measured over the same time interval as ARit, is given by (EPSit – E(EPSit |θ0))/Pit, where EPSit is the realized quarterly earnings per share taken from I/B/E/S, E(EPSit |θ0) is the median preannouncement I/B/E/S consensus forecast of EPSit, and Pit is the security price of firm i on the last day of the preannouncement forecast period (θ0 refers to the set of information available on the preannouncement forecast date).4 FRQit = the forecast revision for firm i for quarter t+1, made subsequent to the earnings announcement for quarter t. FRQit is given by (E(EPSi,t+1 |θ2) – E(EPSi,t+1 |θ0))/Pit, where E(EPSi,t+1 |θ0) is the preannouncement forecast of EPS for quarter t+1, and E(EPSi,t+1 |θ2) is the postannouncement forecast of EPS for quarter t+1. θ2 refers to the set of 18 information available at the postannouncement date. As discussed above, this is the second, not the first I/B/E/S consensus forecast available after the earnings announcement. FRYit = the forecast revision for firm i for the next fiscal year. FRYit is given by (E(EPSYi,t+k |θ2) – E(EPSYi,t+k |θ0))/ Pit, where E(EPSYi,t+k |θ0) is the preannouncement forecast of EPS for the fiscal year which ends in quarter t+k, and E(EPSYi,t+k |θ2) is the postannouncement forecast of EPS for the fiscal year ending in quarter t+k.5 Note that the number of quarters ahead for the subsequent fiscal year depends on the quarter of observation. For example, if the current quarter is the first quarter of the year, the subsequent fiscal year begins with k =4 and ends with k=7 quarters ahead, but if the current quarter is the third quarter of the year, the subsequent fiscal year is begins with k=2 and ends with k=5 quarters ahead. 3.2 Incorporation of by Year and by Industry Fixed Effects We estimate equation (1) several ways. These include (a) a pooled estimation with year and industry fixed effects, where year is determined by the quarter end date and industry is based on industry groupings used in Barth, Beaver, Hand, and Landsman (2005) (see table 1); (b) year- by-year estimations with industry fixed-effects; and (c) for each industry, year-by-year estimations. The fixed effects are included to capture sources of time dependence or cross sectional dependence of a particular form (i.e., a constant for a given year and a constant for a 4 All variables used to compute FE, FRQ, and FRY are adjusted for stock splits and stock dividends. 5 AR and the two forecast revisions, FRQ and FRY, are affected by the information revealed in the earnings release, θ1. However, AR does not reflect information in the postannoucement period, θ2. 19 given industry across years). We assess statistical significance of coefficients in the year-by-year estimations using Fama-MacBeth (1973) t-statistics and Z1 and Z2 statistics.6 3.3 Measurement of “Actual” Earnings per Share Cornell and Landsman estimate equation (1) measuring earnings forecast errors using a Compustat measure of actual earnings per share, earnings per share before extraordinary items, which we hereafter refer to as the Compustat “actual”. Because I/B/E/S forecasts and I/B/E/S actual earnings are measured more similarly, i.e., exclude similar items, the I/B/E/S constructed forecast error is expected to be a better measure of earnings surprise. We assess whether this is the case directly by estimating equation (1) using FE_COMP in place of FE, where: FE_COMPit = the forecast error calculated as FEit, except EPSit is earnings before extraordinary items taken from Compustat, divided by shares outstanding taken from I/B/E/S. Even though the forecast errors measured using consistent I/B/E/S actuals likely have more explanatory power, it is still possible that the market derives additional insight from the information conveyed by the Compustat actuals. This may occur, for instance, if the Compustat actuals provide information about GAAP related variables, such as special items, that the market considers relevant, at least in some circumstances, but which are not included in the earnings measure reported by I/B/E/S. To examine this possibility, we estimate equation (2) which adds the term ADJ, the difference between the Compustat and I/B/E/S actuals: ARit = a0 + a1 FEit + a2 FRQit + a3 FRYit + a4 ADJit + eit (2) 6 The Fama-MacBeth t-statistic = β /( stddev( β ) / ( N − 1) ) , where N is the number of years. Z1 equals 1 / N ∑ N=1 t j j / k j /( k j − 2) , where tj is the t-statistic for year j, kj is the degrees of freedom, and N is the number of years. Z2, which equals t /( stddev(t ) / ( N − 1) , corrects for potential upward bias in Z1 arising from lack of independence of parameters across industries. See Barth (1994). 20 If the Compustat actuals provide additional information to the market, the ADJ coefficient, a4, should be significantly different from zero. However, for the reasons indicated, we expect a4 to be less than a1. 3.4 Impact of the Fourth Quarter In their original paper, Cornell and Landsman conjecture that estimating the basic model across all four quarters was potentially misleading because the fourth quarter could be different than the other three quarters. They argue, for instance, that analysts might wait until year end to revise year-ahead forecasts and that the market might place more weight on annual forecast errors because annual financial results are audited. Although they produce some preliminary results to support those conjectures, it is based on a sample of only three years and uses Compustat actuals. There are reasons other than those suggested by Cornell and Landsman for believing that the fourth quarter may be unique. First, as one moves from the first to the fourth quarter, the forecasting horizon implicit for FRY becomes shorter. It is reasonable to expect that as the forecasting horizon becomes shorter the perceived precision and hence response coefficient would increase. This horizon is shortest at the time of the announcement of fourth quarter results, which is actually sometime within the first quarter of the next year. Second, the information environment, as well as the nature of quarterly earnings, may differ in the fourth quarter. For example, fourth quarter earnings contain more adjustments and special charges than the prior quarters, in part because of auditing of the annual financial statements. It is possible that these items are leaked to the market in earlier quarters (Mendelhall and Nichols, 1988), which could result in a lower response coefficient for the fourth quarter forecast error relative to the other quarters. Also, more information, in the form of a full set of financial statements, more 21 elaborate management discussion and analysis, and potentially more information gathering by analysts may also occur. As a result, the fourth quarter is more than simply another “interim” report. It is, in fact, the final quarter in the firm’s annual financial statements. Similarly, the quarterly forecast revision, FRQ, is more than simply a forecast for a later quarter in the same fiscal year. It is, in fact, a forecast of the first quarter of the next fiscal year. Third, in addition to these substantive reasons, there are econometric reasons for separating the fourth quarter. For the first three quarters, there is no temporal overlap between FRQ and FRY. However, in the fourth quarter, FRQ is a component of FRY. Hence, the interpretation of the coefficients differs for the fourth quarter. To take account of these possibilities, we consider estimating a version of equation (1) that permits the intercept and FE, FRQ and FRY coefficients to differ for fourth quarter earnings announcements: ARit = a0 + a1 FEit + a2 FRQit + a3 FRYit + a4 Dit + a5 DFEit + a6 DFRQit + a7 DFRYit + eit where D it is an indicator variable that equals one (zero) if the announcement is made in the fourth (interim) quarter, and DFE, DFRQ, and DFRY are interactions between D and the corresponding three variables.7 In this model, the full impact of the forecast error and the quarter-ahead and year-ahead forecast revisions in the fourth quarter is α 1 + α 5 , α 2 + α 3 + α 6 + α 7 , and α 3 + α 7 , respectively. The reason the full impact of the quarter-ahead revision is more complicated is that an increase in FRQ mechanically increases FRY.8 To 7 In principle, coefficients for the explanatory variables could change each quarter as the forecast horizon approaches the end of the fiscal year. Findings from untabulated estimations of equation (1) indicate that coefficient point estimates for FE, FRQ, and FRY for the first three quarters are similar in magnitude, while those for the fourth quarter substantially differ from those of the three interim quarters. 8 Figure 3 illustrates the measurement of FE, FRQ, and FRY by quarter and the temporal overlap of FRQ and FRY that occurs in the fourth quarter. 22 account directly for the temporal overlap between FRQ and FRY in the fourth quarter, we estimate the following model: ARit = a0 + a1 FEit + a2 FRQit + a3 FRY*it + a4 Dit + a5 DFEit + a6 DFRQit + a7 DFRY*it + eit (3) where FRY* equals FRY for announcement quarters 1, 2, and 3, and FRY − FRQ for announcement quarter 4. In this model, the full impact of the quarter-ahead forecast revision for the fourth quarter is α 2 + α 6 , while that for the year-ahead forecast revision for the fourth quarter is α 3 + α 7 . 3.5 Extending the Event Window Ideally the forecast error and forecast revisions should be measured over the same time period, so that the market reacts to all three simultaneously. Because of the reporting lag and analyst aggregation issues, this ideal is not met. The aggregation problem arises because analysts do not release their forecasts simultaneously. A measure of a consensus forecast requires individual forecasts to be aggregated over time, and over time subsequent events that are unrelated to the earnings announcement may influence forecast revisions. Reporting lag refers to the time between an analyst forecast and its inclusion in the database. This becomes a problem when a forecast should be included in the current consensus but is not added to the database until after it is calculated. Cornell and Landsman address these problems by extending the measurement window for the forecast revisions. As a result, whereas FE is measured over the same time period as AR, the measurement periods of FRQ and FRY extend several weeks beyond the earnings announcement event window. This feature of the data raises the issue that at the time of the earnings announcement the market must use the information in FE to anticipate its long-run impact, and thereby its effect on analysts’ forecasts without observing the forecasts. 23 Therefore, AR reflects both the forecast error and the expected forecast revisions. Because the estimating equations include actual forecast revisions, FRQ and FRY, even if the market’s expectation of the forecast revisions is unbiased, FRQ and FRY measure these expectations with error, thereby biasing their coefficients towards zero.9 We address the non-simultaneous variable measurement issue by modifying the basic model in two ways. First, we extend the window over which the abnormal return is measured to the date at which the forecast revisions are observed. This assures us that the residual return will reflect both actual forecast errors and actual forecast revisions. Other things equal, with forecast revisions better aligned in time with the residual return, we would expect their coefficients to increase. However, the problem with this approach is that the window must be extended, on occasion, up to two months after the earnings announcement to make it unlikely that the postannoucement consensus I/B/E/S forecast is sensitive to preannouncement forecasts made by individual analysts. The example illustrated in figure 1 shows the event window runs from March 23 until May 21. The longer event window has the effect of causing both the stock return and the forecast revisions to reflect information that is unrelated to the earnings announcement. While this may increase the slope coefficients on the forecast revisions, the longer event window results in a regression that moves the research question away from discerning the informational effects of the earnings announcement. Therefore, we develop a second approach that utilizes disaggregated data. Rather than using the I/B/E/S consensus forecasts, we employ the individual analysts’ forecasts to construct a consensus forecast following the earnings announcement. The I/B/E/S detail file contains 9 Because FRQ and FRY are measured beyond the abnormal return event window, this also raises the possibility that forecast revisions are (at least partially) responding to the abnormal return, thereby creating a potential endogeneity issue. Cornell and Landsman (1989, p. 686) recognize this, but argue there is no economic reason to believe that the 24 individual analysts’ forecasts that can be combined to create custom consensus forecasts at any date and for any time interval. This permits us to shorten the postannoucement event window considerably relative to that associated with the consensus forecasts. The shorter event window mitigates the effects of the aggregation problem by shrinking the forecast periods and aligning them more closely to the earnings announcement. The timeline in figure 2 illustrates how the detail data are used to construct the forecast revisions and to compute the abnormal return over the extended event window. Each constructed forecast revision is computed as the median of analysts’ forecasts available 19 trading days or less after the earnings announcement less the median of analysts’ earnings forecasts made from 20 trading days through 1 trading day prior to the earnings announcement. We use median forecasts rather than mean forecasts to lessen the effect of “stale” forecasts, i.e., those which may be out of date.10 The return window is computed from the day of the earnings announcement through 19 trading days after the earnings announcement. The forecast error is simply I/B/E/S actual earnings less the median earnings forecast made in the twenty trading days prior to the earnings announcement. As shown in the hypothetical example in figure 2, preannouncement forecasts are gathered over the twenty trading day period ending the trading day before the earnings announcement, March 22. Postannouncement forecasts are gathered over the twenty trading day period beginning on the earnings announcement date, or March 23 to April 20. Abnormal stock returns over the extended window are calculated from the close of the preannouncement date, March 22, until the end of the postannouncement period, April 20, a twenty-day window. information contained in analysts’ recommendations can be costlessly discerned by observing the change in price when earnings are released. 10 Note that I/B/E/S consensus forecasts likely suffer from effects of stale forecasts, as I/B/E/S includes all available forecasts to construct their consensus measure. For this reason, we use the consensus median rather than mean. 25 As we discuss below, one cost of using detail forecasts is that there is a significant reduction in sample size because we require new forecasts to be issued both before and after the announcement. Thus, the benefit of shortening the event window may be offset to some extent by the loss of precision associated with a smaller estimation sample. 4. Sample Data Firm-quarters included in the sample meet four criteria: 1. Median monthly earnings forecasts, actual earnings, and earnings announcement dates are available from the I/B/E/S summary forecast data file for quarters ending between 1984 and 2003. 2. Quarterly earnings are available on the Compustat file for the same period. Consistent with prior research, earnings is measured as income before extraordinary items and discontinued operations. 3. Daily security price and return data are available from the CRSP file for each earnings announcement “event” interval (defined above). 4. To mitigate the effects of outliers, for abnormal return, forecast error and forecast revisions, we treat as missing observations that are in the extreme top and bottom one percentile (Kothari and Zimmerman, 1995; Collins, Maydew and Weiss, 1997; Fama and French, 1998; Barth, Beaver, Hand, and Landsman, 1999). Table 1, panels A and B, report descriptive statistics and correlations among the variables used in the study; panel C reports annual descriptive statistics for the forecast errors using I/B/E/S actuals and Compustat actuals, and is discussed in detail in section 5 below. Panel A reveals that mean abnormal return is positive and, consistent with prior research (Abarbanell and Lehavy, 2003), mean forecast errors are negative. In addition, means for both forecast revisions 26 are negative. Panel B reveals that the forecast errors and forecast revisions are correlated with abnormal return and with each other. Panel C shows that sample observations are increasing throughout the sample period until 1998, which is consistent with I/B/E/S coverage expanding over time. Untabulated statistics reveal that sample observations are drawn from a wide variety of industries, with Financials (14.6%) and Computers (14.2%) comprising the largest percentage. 5. Results 5.1 Pooled Estimation To get an overview at the outset, we begin with a pooled model that covers the entire data set. Table 2, panel A presents the results for the pooled regression of equation (1), estimated over 1984 to 2003 with year and industry fixed effects. Consistent with the original findings of Cornell and Landsman, all three coefficients are highly significant.11 In part because of the immense sample, over 150,000 observations, all of the t-statistics exceed twenty. The finding implies that each variable conveys information to the market not contained in the other. In particular, the significance of the forecast error coefficient above the theoretical value of 1 implies that the market perceives that the forecast revisions do not fully capture the implications of the forecast error for future earnings (cash flows) performance. 12 We expect the forecast revision variables to be significant, because prices are viewed as a function of future earnings. 11 The term significant indicates statistical significance at the 0.05 level or less using a two-sided test. Some of our tests clearly have directional predictions, e.g., we predict the pooled forecast error coefficient to be positive. However, we adopt a two-sided convention because our tests involving changes in coefficient magnitudes over time are associated with two-sided predictions. 12 Note that the theoretical value 1 is based on the assumption that a forecast error would be priced on a dollar-for- dollar basis after controlling for its implications for future earnings by inclusion of the forecast revisions. Finding the forecast error coefficient exceeds 1 is also consistent with the estimating equation not including all forecast revisions. To assess this possibility, we also estimated specifications including forecast revisions of two-quarter- ahead and two-year-ahead earnings forecasts. Untabulated findings from these regressions indicate that forecast error coefficients falls closer to 1 when both additional forecast revision variables are included. However, significance levels tend to be smaller, largely because data availability constraints associated with the additional regressors result in higher regression standard errors. In addition, the coefficients on the two-quarter-ahead and two- year-ahead earnings forecast revisions, while often significant, are of much smaller magnitude than the one-quarter- ahead and one-year-ahead revisions. 27 Hence, price changes are expected to be related to the market’s revision in expected future earnings, and analysts’ forecast revisions are expected to serve as proxies for those expectational changes. Note that the two forecast revisions, for the next quarter and for the next year, also compete with each other for incremental explanatory power. However, neither dominates and each is significant. This implies that analysts are able to distinguish the implications of information arriving at the time of quarterly earnings announcements in terms of the short-run and longer-run implications for future earnings. There are several important implications of our findings for valuation research and its practical applications. For the large body of research that examines the effects of information on valuation, our finding that both forecast errors and forecast revisions are significant determinants of security price responses to earnings announcements implies that a fully specified model should incorporate both forecast errors and forecast revisions. In particular, forecast errors and forecast revisions reflect different information and must be included as separate variables when analyzing the security price effects of earnings announcements. This distinction is also of practical interest to analysts, whose role is to understand how information is reflected in prices and to evaluate the implications of that information for the value of a security. It is also important in security litigation contexts, where allegations relate to the separate security price effects of forecast errors and forecast revisions. Moreover, the finding that the relative importance of the forecast error and forecast revision differs in the fourth quarter also has important research and practical implications. From a research perspective, the finding indicates the importance of permitting the pricing effects of forecast errors and revisions to differ for interim and annual earnings announcements. 28 A practical implication in a securities litigation context is estimates of damages typically aggregate all quarters. Our results demonstrate that when assessing the price effects of an alleged change in either forecast errors or forecast revisions, it is potentially misleading to assume a homogenous price response and aggregate across quarters. 5.2 Year-by-year Estimations The issues discussed in the introduction, however, suggest that the pooled regression hides potential changes in the relation between forecast errors, forecast revisions, and residual returns. To examine this possibility, table 2, panel A, also presents results from annual estimations of the model. There is an increase in the number of observations per year, reflecting the increased coverage by the I/B/E/S database with decrease in the 2000-2003 period presumably reflecting the reduction in the number of firms forecasted due to the economic downturn.13 The results show that the relation is robust. Except for the FRQ coefficient in 1984, a year in which there are only 1,013 observations, all of the coefficients are positive and significant. To compare the annual results with those from the pooled regression, we use a Fama- MacBeth (1973) procedure. In particular, table 2, panel A reports the mean coefficient across all the years, and Fama-MacBeth t-statistics and Z1 and Z2 statistics to assess statistical significance of the coefficients over time. Because the Fama-MacBeth t-statistic does not use cross-sectional data within a given year to calculate the standard error used in its calculation, and the Z2 statistic corrects for the effect of cross-sectional dependence in the data, each is a less biased test statistic than Z1 in the presence of cross-sectional dependence in the data. The test statistics are not 13 McNichols and O’Brien (1999) investigate in detail the reasons why analyst coverage might be dropped. Basically, unfavorable information increases the likelihood of an analyst dropping a stock rather than continuing to report, which would have required a downward revision in forecasts. This self censoring could potentially affect the distribution of forecast errors and forecast revisions, although it is less clear how it would affect the coefficients. 29 likely to be affected by time-series dependence in the data because the event-window abnormal returns are non-overlapping in time and expected to be serially uncorrelated. In addition, we further expect both the forecast errors and the forecast revisions, which are separated by a year, are also serially uncorrelated.14 The comparison reveals that the Fama-MacBeth t-statistic and Z2 statistics are noticeably lower than those in the pooled regression, which is consistent with positive cross sectional dependence in the data even after extracting fixed effects. However, as in the pooled regression, the magnitude of the forecast error coefficient is larger than that of either of the forecast revisions. Table 2, panel B, presents regression summary statistics from a specification that includes only the forecast error as a regressor. Comparison of the FE coefficients in panel B to those in panel A indicate that on average the FE increases approximately 60 percent when the forecast revisions are excluded from the estimating equation. The increase is not surprising given the positive correlation between forecast errors and forecast revisions. When the forecast revisions are excluded, the forecast errors pick up some of their impact on stock prices. These findings underscore the importance of including forecast revisions when explaining earnings announcement period stock returns. Failure to do so results in an incorrectly specified model when assessing the relation between forecast errors and residual returns. Returning to panel A, which presents findings from the specification including the forecast revisions and the forecast error, we also find evidence that the coefficients exhibit an interesting pattern. Similar to Bradshaw and Sloan (2002) and Abarbanell and Lehavy (2005), 14 Adjacent-quarter forecast errors and forecast revisions would be expected to have some slight serial correlation (see Abarbanell and Bernard, 2000, among others). However, each year’s variables are separated from the next year’s variables by an average of four quarters. 30 there appears to be an abrupt shift in the FE coefficients beginning in 1991. For example, from 1984 through 1990, none of the coefficients is above one, while from 1991 through 2003 none of the coefficients is below one. Needless to say, this is statistically significant using a simple binomial test. Prior studies observe this may reflect an improvement in the quality of actual earnings as reported by I/B/E/S in the sense of increasing the consistency between what is being forecasted and what is included in actual. A second possibility is an improvement in the quality of the earnings forecast in the sense of producing more consistency across analysts that comprise the consensus. Either could induce this shift in coefficients. Although this shift in similar to that found in prior research, there is an important difference in that the shift is still observed in the presence of the forecast revisions variables. Hence, the observed shift is not explained solely by a temporal change in the persistence of earnings. To the best of our knowledge, no prior research has examined whether a similar shift exists in forecasts revisions. To the extent that the shift in the FE coefficient is attributable to improvements in the quality of the forecasts, we might expect to see that improvement reflected in the coefficients on the forecast revisions as well. Both FRQ and FRY coefficients are higher in the post 1991 period, although the shift is not as dramatic. For FRQ (FRY), 5 (5) coefficients are below one while 2 (2) are above one from 1984-1990. For 1991-2003, 3 (4) coefficients are below one while 10 (9) are above one. Using a binominal test, both coefficients are significantly greater in the post-1991 period at less than the 0.05 level. Hence the improvements in the nature of the database or the underlying quality of analysts’ forecasts appears to be a partial explanation for the increase in the FE coefficients as well as the increase in the FRQ and FRY coefficients.15 15 Other possible reasons for the upward trend in the forecast error coefficient include enhanced earnings management and increased management’s guiding of analysts’ forecasts (Matsumoto, 2002; Abarbanell and Lehavy, 2003; Burgstahler and Eames, 2003), and increasing exclusion of transitory items over time by I/B/E/S (Bradshaw and Sloan, 2002). 31 One way to assess the importance of improvement in consistency of measures of forecast and actual earnings is to use another measure of actual earnings. The obvious alternative is to use Compustat data to measure actual earnings. Table 3, panel A reports the results from estimation of equation (1) which employs forecast errors measured using Compustat actuals, FE_COMP, instead of FE. The most striking feature of the results is that the upward trend in the forecast error coefficient disappears entirely. This is consistent with the hypothesis that increase in the coefficients observed when using the I/B/E/S actuals is attributable to the success of the effort by I/B/E/S to match the forecasts and the actuals on a more consistent basis. Second, the coefficients on the forecast revisions are somewhat larger when FE_COMP is used as the forecast error. This implies that the revisions are picking up some of the variance left unexplained by use of an incorrect measure of the forecast error. To further investigate the impact of measurement consistency of the components of the forecast error, we examine the forecast errors directly. The findings are presented in table 1, panel C. The results strongly support the hypothesis. Whereas the forecasts errors computed from Compustat data, FE_COMP, show no evidence of a downward trend—if anything they appear to increase— there is a pronounced downward trend in the magnitude of the I/B/E/S consensus forecast errors, FE, particularly in the first ten years of the sample. This matches the period over which the FE coefficients increase in Table 3, panel A. The evidence strongly suggests that the trend in the FE coefficient is likely attributable to the success of the efforts by I/B/E/S to more accurately align the actuals that are reported with the measure that analysts forecast. This underscores the importance of defining actual and forecast earnings in precisely the same fashion. 32 Even though the forecast errors measured using consistent I/B/E/S actuals have more explanatory power, it is still possible that the market derives additional insight from the information conveyed by the Compustat actuals. Equation (2), which includes an additional term, ADJ, to capture the difference between the Compustat actual and the I/B/E/S actual tests this proposition directly. The findings reported in table 3, panel B reveal that ADJ does increase the explanatory power of the regression, but not a great deal. Its coefficient, a4, is positive in every year but one, and is significantly so in over half of the years. In addition, the Z2 statistic is highly significant. Nonetheless, the average coefficient is only 0.16, which is several orders of magnitude less than the other coefficients. This low coefficient is consistent with prior research by Elliott and Hanna (1996), among others, which shows that special items and other transitory items are priced at much less than a dollar for dollar basis. In this regard, the coefficient on FE_COMP reported previously can be thought of as being biased toward zero because it consists of two components, one of which has a coefficient of 0.16. 5.3 Cross-industry Results When estimating earnings response coefficients, researchers often implicitly assume that firms in different industries can be treated identically. That is, forecast errors and forecast revisions have the same impact on stock prices independent of the underlying business in which the company operates. However, this may not be the case. For example, the pressure to guide or manage earnings may be greater in one industry because of competitive factors are regulatory concerns. To test the assumption that the coefficient are equal across industries, we return to the fundamental model given by equation (1) and use the full time series to estimate a series of regressions across industries. 33 Untabulated results for the industry regressions reveal that the results observed for the pooled sample holds generally in every industry. In addition, most of the coefficients for the individual industries are quite close to the cross industries means. There are, however, some exceptions. Insurance and real estate are found to be low coefficients while those for the retail restaurant industry are high. We leave to future research an effort to determine whether these differences are stationary and, if so, what accounts for them. 5.4 The Impact of the Fourth Quarter Table 4 reports the pooled and year-by-year estimation results for equation (3), which takes into account the temporal overlap between FRQ and FRY in the fourth quarter. Both panels reveal similar insights. In particular, with the overlap eliminated, the coefficient on DFRQ is positive and significant, whereas that on DFRY* is insignificant. By definition, the coefficient on DFE is unaffected. Focusing on the year-by-year results, the mean incremental coefficient for FE is −0.50, which implies that that the total earnings response coefficient for fourth quarter forecast errors is 0.78 (1.28−0.50), which is now less than that for FRQ (1.84 = 1.10 + 0.74) or FRY* (1.14 = 1.00 + 0.14.). The lower coefficient implies that the market responds less to FE in the fourth quarter. This is consistent with more information arriving in the fourth quarter, such as more elaborate management discussion and analysis and more comprehensive year end reviews by analysts, that is reflected in the forecast revisions but not in the forecast errors. It is also consistent with the fourth quarter I/B/E/S actual containing more transitory factors that the first three quarters. For example, if some year-end adjustments are implicitly imbedded in revenue or expenses and not explicitly shown as a special charge, it would be difficult for I/B/E/S to extract these effects when forming an I/B/E/S actual. 34 These findings contrast with that of Cornell and Landsman, which finds that the forecast error has significant explanatory power only in the fourth quarter, the quarter-ahead forecast revision has no explanatory power in the fourth quarter, but the year-ahead forecast revision coefficient is significantly larger in the fourth quarter than in the interim quarters. Nonetheless, the results confirm the conjecture that the fourth quarter is different than other quarters. Consequently, properly specified models of the reaction of stock prices to forecast errors and forecast revisions must not only include the revisions to be properly specified, they must also take account of the unique nature of the fourth quarter. 5.5 Event Window Extensions As discussed in section 3.3, our research design uses forecast revisions that occur after the earnings announcement window ends. This results in FRQ and FRY likely measuring the market’s expectations of analysts’ forecasts with error, thereby biasing their coefficients towards zero. This section presents findings from two approaches designed to mitigate the effects of this problem by extending the event window so that the abnormal return and forecast revisions are measured over the same time period. Table 5, panel A, presents results from estimations based on consensus I/B/E/S forecasts that extend the return window to the end of revision period. Table 5, panel B, presents results from estimations based on the consensus forecast we construct from I/B/E/S detail data; the forecast error, quarter-ahead and year-ahead forecast revisions constructed from the detail data are denoted FE_DET, FRQ_DET, and FRY_DET. For these regressions, each constructed forecast revision is computed as the median of analysts’ forecasts available 19 trading days or less after the earnings announcement less the median of analysts’ earnings forecasts made from 20 trading days through 1 trading day prior to the earnings 35 announcement. The return window is computed from the day of the earnings announcement through 19 trading days after the earnings announcement. The results for the pooled regression reported in table 5, panel A, are similar to those reported in table 2, in that all three regressors, FE, FRQ, and FRY have significantly positive coefficients. That is, each informational variable has pricing effects incremental to the others. However, relative to the findings reported in table 2, panel A, and consistent with the conjecture that the expanded return window better captures the price effects of FRQ and FRY, the coefficients on these variables increase substantially. For example, the coefficient on FRY essentially doubles from 1.02 to 2.01. Furthermore, in contrast to the basic estimation results reported in table 2, in table 5, panel A, both the FRQ and FRY coefficients are larger than the FE coefficient. The results of the year-by-year estimations yield similar insights from the pooled results. Relative to the year-by-year results from the basic estimations reported in Table 2, panel A, there is general increase in the FRQ and FRY coefficients, with overall means increasing from 1.08 and 0.99 to 1.76 and 1.89. In addition, the FRQ and FRY coefficients are generally larger than the FE coefficient, where the overall mean FE coefficient is 1.31. The results indicate the presence of a slight increase in the coefficient of FE over time, although it is far less noticeable than when the shorter event window is examined. Hence, the basic conclusions are robust to this extension with the added insight that, as expected, sensitivity coefficients on FRQ and FRY are higher when the event window is extended.16 16 We also examined an alternative specification to alleviate the disparity in the timing of the forecast revision variables and announcement return. Specifically, we include as an additional explanatory variable the abnormal stock return beginning the day after the announcement period return through the date of the postannouncement earnings forecast. This postannouncement return will reflect information arriving after the earnings announcement but available to analysts (and the market) up to the time when the postannouncement earnings forecast is provided. Under these assumptions, this return will be correlated with the measurement error in the forecast revision variables, and therefore its inclusion will potentially reduce the revision variable coefficients (Brown, Griffin, Hagerman, and 36 The findings for the pooled sample and year-by-year estimations reported in table 5, panel B, show that coefficients on all three regressors are significant. Hence, the basic findings are robust to this second extension as well. The magnitude of the coefficients can be compared with those reported in panel A of tables 2 and 5. However, a word of caution is required. Whereas the findings presented in panel A of tables 2 and 5 were based on essentially the same firm-quarter observations, the number of observations used to estimate the table 5, panel B, regression results is considerably smaller. The substantial reduction in sample size occurs because many of the individual analysts do not provide a quarter-ahead and year-ahead forecast in the twenty trading days before each quarterly earnings announcement. Hence, FRQ_DET and FRY_DET are not available for many firm-quarters. The reduction in sample size reduces the efficiency of the coefficient estimates. Notwithstanding these caveats the basic findings are robust to this second extension as well. The most noticeable difference between the coefficients in reported in table 8 and those reported in the earlier tables is the substantial decrease in the FRY_DET coefficient, which suggests that the potential limitations of this constructed consensus outweigh the potential benefits. The coefficients on FE_DET and FRQ_DET are similar to those reported in table 8. 6. Concluding Remarks We offer the first comprehensive analysis, both over time and across industries, of security returns and analyst forecast errors that also takes account of forecast revisions and a possible fourth quarter effect. We find that all analysts forecast errors, quarter ahead forecast revisions and year ahead forecast revisions all have significant pricing effects, indicating each Zmijewski, 1987; Collins, Kothari, Shanken, and Sloan, 1994). When we conduct such an estimation, the coefficient on the added return variable is not significantly different from zero and the revision coefficients are unaffected by the inclusion of this variable. Hence, our basic finding is robust to this specification as well. This is consistent with the measurement error being uncorrelated with the subsequent security returns. 37 conveys incremental information content. This finding is remarkably robust, holding across years, industries, and two procedures extending the event window. Further, findings from the expanded event window tests reveal that although forecasts revision coefficients increase, they do not do so at the expense of the forecast error coefficients. Over the narrower event window, the forecast error has the highest coefficient, while over the longer window the forecast revisions have larger coefficients than the forecast error. In addition, we document that the fourth quarter is significantly different from the other three quarters. In particular, the relative importance of the forecast error is lower (although still highly significant), while the relative importance of the quarter-ahead forecast revision increases. We attribute this difference not only to the nature of fourth quarter earnings but also to the enhanced information available near year-end that conveys information about the next fiscal year. Finally, we document an increase in the forecast error coefficient over time even in the presence of the forecast revision variables and a similar but less dramatic shift in the forecast revision coefficients. Consistent with prior research, the evidence supports the view that the I/B/E/S measure of actual earnings is superior for calculating forecast errors because it is more comparable to the earnings being forecasted by analysts. In this respect, the quality of that number has improved over time, underscoring the importance of defining the forecast and the actual earnings measure in precisely the same fashion. Moreover, the shift in the forecast revision coefficients is consistent with improvements in the analysts forecast being a factor in explaining the upward shift in both the forecast error and forecast revision coefficients. Finding the forecast revisions continue to play a significant role in explaining earnings announcement period returns after controlling for improvement in the measurement of the 38 forecast error over time suggests that the significance of the forecast revision coefficients is a robust finding that holds up through time despite changes in database quality and changes in the institutional features of the earnings reporting environment. Thus, the findings from this study underscore the importance of including forecast revisions in addition to forecast errors, and allowing for a fourth quarter effect, when examining how stock returns are affected by earnings announcements. A model that fails to take account of these factors is likely to produce biased results and give a misleading impression of the impact of earnings surprises on stock prices. 39 References Abarbanell, J. and V. Bernard (2000). “Is the US Stock Market Myopic?” Journal of Accounting Research 38, 221-242. Abarbanell, J., and R. Lehavy (2003). “Can Stock Recommendations Predict Earnings Management and Analysts’ Earnings Forecast Errors?” Journal of Accounting Research 41, 1-31. Abarbanell, J., and R. Lehavy (2005). “Letting the ‘Tail Wag the Dog’: The Debate over GAAP versus Street Earnings Revisited.” Contemporary Accounting Research (forthcoming). Barth, M.E. (1994). “Fair Value Accounting: Evidence from Investment Securities and the Market Valuation of Banks.” The Accounting Review 69, 1-25. Barth, M.E., W.H. Beaver, J.M. Hand, and W.R. Landsman (1999). “Accruals, Cash Flows, and Equity Values.” Review of Accounting Studies 3: 205-229. Barth, M.E., W.H. Beaver, J.M. Hand, and W.R. Landsman (2005). “Accruals, Accounting– Based Valuation Models, and the Prediction of Equity Values.” Journal of Accounting, Auditing & Finance 20, 311-345. Bradshaw, M.T., and R.G. Sloan (2002). “GAAP versus The Street: An Empirical Assessment of Two Alternative Definitions of Earnings.” Journal of Accounting Research 40, 41-66. Brown, L. (1996) “Analyst Forecasting Errors and Their Implications for Security Analysis: An Alternative Perspective” Financial Analysts Journal 52, 40-47. Brown, L. (2001). “A Temporal Analysis of Earnings Surprises: Profits versus Losses.” Journal of Accounting Research 39, 221-241. 40 Brown, L., P. Griffin, R. Hagerman, and M. Zmijewski (1987). “An Evaluation of Alternative Proxies for the Market’s Assessment of Unexpected Earnings.” Journal of Accounting and Economics 9, 159-193. Brown, L. and K. Sivakumar (2003). “Comparing the Value Relevance of Two Operating Income Measures.” Review of Accounting Studies, 561-572. Burgstahler, D., and M. Eames (2003). “Earnings Management to Avoid Losses and Earnings Decreases: Are Analysts Fooled?” Contemporary Accounting Research 20, 253-272. Collins, D., S.P. Kothari, J. Shanken, and R. Sloan (1994), “Lack of Timeliness and Noise as Explanations for the Low Contemporaneous Returns-Earnings Association.” Journal of Accounting and Economics 18, 289-324. Collins, D.W., Maydew, E.L., and I.S. Weiss (1997). “Changes in the Value–Relevance of Earnings & Equity Book Values Over The Past Forty Years.” Journal of Accounting and Economics 24, 39-67. Cornell, B., and W. Landsman (1989). “Security Price Response to Quarterly Earnings Announcements and Analysts’ Forecast Revisions.” The Accounting Review 64, 680-692. Elliott, J., and D. Hanna (1996). “Repeated Accounting Write-Offs and the Information Content of Earnings.” Journal of Accounting Research 34, 135-155. Fama, E.F., and K.R. French (1998). “Taxes, Financing Decisions, and Firm Value.” Journal of Finance 53, 819-843. Fama, E.F., and J.D. MacBeth (1973). “Risk, Return, and Equilibrium: Empirical tests.” Journal of Political Economy 81, 607-636. Feltham, J., and J. Ohlson (1995). “Valuation and Clean Surplus Accounting for Operating and Financial Activities,” Contemporary Accounting Research 11, 689-731. 41 Feltham, J., and J. Ohlson (1996). “Uncertainty Resolution and the Theory of Depreciation Measurement,” Journal of Accounting Research 34, 209-234. Gleason, C. and C. Lee (2003). “Analyst Forecast Revision and Market Price Discovery.” The Accounting Review 78, 192-225. Kothari, S.P., and J. Zimmerman. (1995). “Price and Return Models.” Journal of Accounting and Economics 20, 155-192. Liu, J. and J. Thomas (2000). “Stock Returns and Accounting Earnings.” Journal of Accounting Research 38, 71-101. Lougee, B. and C. Marquardt (2004). “Earnings Informativeness and Strategic Disclosure: An Empirical Examination of ‘Pro Forma’ Net Income.” The Accounting Review 79, 769-795. Lys, T. and S. Sohn (1990). “The Association between Revisions of Financial Analysts’ Earnings Forecasts and Security-Price Changes.” Journal of Accounting and Economics 13, 341-363. Matsumoto, D. (2002). “Management’s Incentives to Avoid Negative Earnings Surprises.” The Accounting Review 77, 483-514. McNichols, M., and P. O’Brien (1999). “Self Selection and Analyst Coverage.” Journal of Accounting Research (Supplement) 37, 167-199. Mendelhall, R. and W. Nichols (1988). “Bad News and Differential Market Reactions to Announcements of Earlier-Quarters Versus Fourth-Quarter Earnings.” Journal of Accounting Research 26, 63-86. Ohlson, J. (1995). “Earnings, Book Values and Dividends in Security Valuation,” Contemporary Accounting Research 11, 661-687. 42 FIGURE 1 Earnings Announcements and the Timing of I/B/E/S Summary Forecasts: An Example Using a Hypothetical Announcement on March 23 MAR 23: Earnings Announcement Date MAR 19: End of APR 16: MAY 21: End of Preannouncement End of April Postannouncement Forecast Period Forecast Period Forecast Period Return Cumulation Period Extended Return Cumulation Period FIGURE 2 Earnings Announcements and the Timing of I/B/E/S Custom Summary Forecasts: An Example Using a Hypothetical Announcement on March 23 MAR 23: Earnings Announcement Date FEB 22 – MAR 22: MAR 23 – APR 20: Pre-announcement Post-announcement Forecast Period Forecast Period Return Cumulation Period Extended Return Cumulation Period 43 FIGURE 3 Measurement of Primary Variables by Quarter Y1Q1 Y1Q2 Y1Q3 Y1Q4 Y2Q1 Y2Q2 Y2Q3 Y2Q4 First FE FRQ FRY Quarter Second FE FRQ FRY Quarter Third FE FRQ FRY Quarter Fourth FE FRQ Quarter FRY 44 TABLE 1 – Summary Statistics (n = 156, 993) Panel A: Descriptive Statistics Percentiles Mean Std. Dev. 25 50 75 AR 0.004 0.096 –0.045 0.001 0.049 FE –0.001 0.008 –0.001 0.000 0.001 FE_COMP –0.002 0.020 –0.002 0.000 0.003 FRQ –0.001 0.005 –0.002 0.000 0.000 FRY –0.003 0.012 –0.003 0.000 0.001 Panel B: Correlations (Pearson above diagonal, Spearman below) AR FE FE_COMP FRQ FRY AR 0.159 0.099 0.165 0.194 FE 0.238 0.381 0.327 0.314 FE_COMP 0.184 0.573 0.206 0.210 FRQ 0.201 0.324 0.259 0.551 FRY 0.249 0.384 0.306 0.507 45 TABLE 1 – Continued Panel C: Forecast Errors Over Time FE FE_COMP Year N Mean Q1 Median Q3 Mean Q1 Median Q3 1984 1,013 –0.286 –0.471 –0.042 0.286 –0.071 –0.307 0.024 0.420 1985 2,944 –0.267 –0.410 –0.062 0.153 –0.124 –0.344 –0.005 0.316 1986 3,659 –0.245 –0.346 –0.043 0.133 –0.186 –0.293 0.024 0.320 1987 3,254 –0.243 –0.297 –0.018 0.131 –0.103 –0.260 0.039 0.316 1988 3,555 –0.175 –0.270 0.000 0.201 –0.097 –0.233 0.052 0.353 1989 4,721 –0.245 –0.345 –0.034 0.143 –0.226 –0.333 0.015 0.300 1990 5,036 –0.222 –0.328 –0.031 0.132 –0.243 –0.363 0.012 0.283 1991 5,774 –0.156 –0.238 0.000 0.126 –0.227 –0.268 0.028 0.288 1992 6,664 –0.104 –0.167 0.000 0.135 –0.163 –0.181 0.051 0.319 1993 8,075 –0.079 –0.145 0.000 0.136 –0.146 –0.178 0.060 0.313 1994 10,021 –0.072 –0.125 0.000 0.144 –0.104 –0.145 0.080 0.347 1995 10,866 –0.075 –0.106 0.014 0.144 –0.197 –0.192 0.067 0.326 1996 12,265 –0.073 –0.075 0.021 0.126 –0.226 –0.176 0.063 0.282 1997 13,691 –0.039 –0.049 0.027 0.126 –0.183 –0.138 0.071 0.308 1998 13,632 –0.069 –0.052 0.018 0.114 –0.316 –0.246 0.049 0.264 1999 12,784 –0.050 –0.036 0.029 0.144 –0.229 –0.203 0.059 0.298 2000 10,650 –0.033 –0.023 0.035 0.161 –0.386 –0.342 0.038 0.282 2001 9,820 –0.033 –0.051 0.024 0.142 –0.670 –0.607 –0.005 0.197 2002 9,877 0.029 0.000 0.041 0.180 –0.308 –0.279 0.049 0.288 2003 8,692 0.031 –0.021 0.043 0.188 –0.111 –0.202 0.061 0.316 AR = stock return measured from the close of the earnings forecast date to the close of the weekday following the earnings announcement, less the mean return for the firm’s corresponding size decile over the same period; FE = realized I/B/E/S quarterly earnings per share less the median preannouncement I/B/E/S consensus forecast; FE_COMP = same as FE, except realized earnings is taken from Compustat—income before extraordinary items divided by the number of shares outstanding taken from I/B/E/S; FRQ = earnings forecast revision for the subsequent fiscal quarter, calculated as the I/B/E/S median forecast after the earnings announcement less the I/B/E/S median forecast before the announcement; FRY = earnings forecast revision for the subsequent fiscal year, calculated as the I/B/E/S median forecast after the earnings announcement less the I/B/E/S median forecast before the announcement. Forecast variables are scaled by stock price on the pre-announcement forecast date and adjusted for stock splits and dividends. All correlations in Panel B as significant at the .001 level. Forecast errors in Panel C are scaled by 100. 46 TABLE 2 – Primary Regression Results ARit = a0 + a1 FEit + a2 FRQit + a3 FRYit + eit Panel A: All Variables FE FRQ FRY Est. t-stat Est. t-stat Est. t-stat N Adj. R2 Pooled 1.18 36.58 1.21 21.24 1.05 43.74 156,993 0.05 1984 0.44 3.06 –0.05 –0.13 0.61 3.16 1,013 0.04 1985 0.44 4.25 0.67 2.72 1.06 9.51 2,944 0.07 1986 0.71 6.31 1.04 4.30 0.89 7.79 3,659 0.07 1987 0.46 3.56 0.63 2.02 0.87 5.83 3,254 0.03 1988 0.44 3.99 0.70 2.95 1.02 8.76 3,555 0.06 1989 0.64 6.61 0.77 3.90 0.82 8.52 4,721 0.06 1990 0.77 6.36 1.14 5.34 0.91 8.45 5,036 0.07 1991 1.33 9.69 0.85 3.67 1.00 8.77 5,774 0.07 1992 1.10 7.78 1.04 4.47 0.78 6.83 6,664 0.05 1993 1.57 10.84 1.01 4.41 0.90 8.57 8,075 0.05 1994 1.57 12.33 0.99 4.68 1.16 12.43 10,021 0.07 1995 1.21 9.44 1.86 8.13 1.27 12.89 10,866 0.08 1996 1.43 11.56 1.61 6.86 1.03 10.68 12,265 0.06 1997 2.02 15.34 1.54 7.12 1.18 13.38 13,691 0.07 1998 1.33 10.42 1.03 4.57 0.95 10.84 13,632 0.05 1999 1.41 10.26 1.18 5.27 1.08 11.18 12,784 0.05 2000 1.29 7.98 0.81 2.83 1.02 8.79 10,650 0.04 2001 1.45 9.28 1.35 5.76 0.76 8.75 9,820 0.05 2002 1.47 9.44 1.24 5.24 1.24 14.22 9,877 0.07 2003 1.80 11.00 2.11 8.64 1.16 11.42 8,692 0.09 Mean 1.14 8.48 1.08 4.64 0.99 9.54 7,850 0.06 FM t 10.39 9.57 24.13 Z1 37.90 20.74 42.65 Z2 11.41 9.68 15.55 47 TABLE 2 – Continued ARit = a0 + a1 FEit + eit Panel B: Forecast Error only FE Est. t-stat N Adj. R2 Pooled 2.00 63.74 156,993 0.03 1984 0.59 4.35 1,013 0.03 1985 0.71 7.01 2,944 0.03 1986 1.13 10.67 3,659 0.04 1987 0.80 6.62 3,254 0.01 1988 0.92 8.86 3,555 0.03 1989 1.11 12.35 4,721 0.04 1990 1.39 12.16 5,036 0.03 1991 2.01 15.88 5,774 0.05 1992 1.74 13.23 6,664 0.03 1993 2.26 16.87 8,075 0.04 1994 2.47 21.24 10,021 0.05 1995 2.39 20.40 10,866 0.05 1996 2.36 20.73 12,265 0.04 1997 3.06 24.75 13,691 0.05 1998 2.04 16.81 13,632 0.03 1999 2.24 17.43 12,784 0.03 2000 1.90 12.26 10,650 0.03 2001 2.21 14.90 9,820 0.03 2002 2.34 15.55 9,877 0.03 2003 3.03 19.43 8,692 0.05 Mean 1.83 14.58 7,850 0.03 FM t 10.76 Z1 65.17 Z2 11.75 Variables are defined in table 1. Z1 = 1 / N ∑ N=1 t j / k j /(k j − 2) , where tj is the t-statistic for year j, kj is the degrees of j freedom, and N is the number of years. Z2 = t /( stddev(t ) / ( N − 1) ) , and the Fama-MacBeth t- statistic (FM t) = β /( stddev( β ) / ( N − 1) ) , where N is the number of years. 48 TABLE 3 – Compustat Actuals ARit = a0 + a1 FE_COMPit + a2 FRQit + a3 FRYit + eit Panel A: Compustat Actuals FE_COMP FRQ FRY Est. t-stat Est. t-stat Est. t-stat N Adj. R2 Pooled 0.27 22.31 1.48 26.26 1.14 47.85 156,993 0.05 1984 0.54 3.53 –0.08 –0.19 0.55 2.78 1,013 0.04 1985 0.56 6.56 0.59 2.42 1.04 9.38 2,944 0.08 1986 0.29 4.50 1.31 5.53 0.97 8.56 3,659 0.07 1987 0.36 3.45 0.62 2.00 0.91 6.20 3,254 0.03 1988 0.26 3.44 0.76 3.24 1.06 9.23 3,555 0.06 1989 0.29 5.09 0.88 4.53 0.91 9.65 4,721 0.06 1990 0.21 3.24 1.21 5.65 1.04 9.88 5,036 0.06 1991 0.37 6.43 1.10 4.78 1.21 10.87 5,774 0.06 1992 0.23 4.15 1.33 5.80 0.88 7.76 6,664 0.04 1993 0.30 5.59 1.41 6.20 1.09 10.47 8,075 0.04 1994 0.39 8.30 1.34 6.41 1.35 14.75 10,021 0.07 1995 0.22 4.93 2.22 9.89 1.42 14.61 10,866 0.08 1996 0.28 6.34 2.03 8.77 1.18 12.26 12,265 0.06 1997 0.30 7.34 2.00 9.26 1.38 15.75 13,691 0.06 1998 0.16 3.69 1.38 6.22 1.05 12.05 13,632 0.04 1999 0.31 6.32 1.49 6.78 1.18 12.23 12,784 0.04 2000 0.33 6.22 1.04 3.66 1.07 9.25 10,650 0.04 2001 0.16 3.74 1.74 7.53 0.82 9.38 9,820 0.04 2002 0.28 5.99 1.61 6.96 1.25 14.34 9,877 0.07 2003 0.18 3.16 2.59 10.71 1.30 12.76 8,692 0.08 Mean 0.30 5.10 1.33 5.81 1.08 10.61 7,850 0.06 FM t 12.39 9.29 21.66 Z1 22.81 25.97 47.43 Z2 14.48 9.22 14.74 49 TABLE 3 - Continued ARit = a0 + a1 FEit + a2 FRQit + a3 FRYit + a4 ADJit + eit Panel B: I/B/E/S Actuals, I/B/E/S Adjustments FE FRQ FRY ADJ Est. t-stat Est. t-stat Est. t-stat Est. t-stat N Adj. R2 Pooled 1.19 37.07 1.18 20.75 1.04 43.08 0.14 10.78 156,993 0.05 1984 0.62 3.72 –0.12 –0.29 0.54 2.73 0.40 2.12 1,013 0.04 1985 0.64 5.83 0.55 2.26 1.03 9.27 0.51 5.11 2,944 0.08 1986 0.74 6.56 1.05 4.32 0.89 7.71 0.14 2.03 3,659 0.07 1987 0.51 3.83 0.59 1.91 0.86 5.74 0.21 1.57 3,254 0.03 1988 0.48 4.21 0.68 2.86 1.00 8.63 0.14 1.61 3,555 0.06 1989 0.68 6.91 0.74 3.77 0.80 8.35 0.15 2.49 4,721 0.06 1990 0.77 6.37 1.14 5.30 0.90 8.43 0.03 0.43 5,036 0.07 1991 1.34 9.77 0.83 3.60 1.00 8.74 0.18 2.83 5,774 0.07 1992 1.11 7.82 1.02 4.40 0.77 6.75 0.09 1.43 6,664 0.05 1993 1.57 10.83 1.00 4.34 0.90 8.54 0.10 1.73 8,075 0.05 1994 1.57 12.34 0.95 4.47 1.16 12.40 0.20 4.10 10,021 0.07 1995 1.21 9.42 1.84 8.03 1.27 12.84 0.08 1.62 10,866 0.08 1996 1.43 11.55 1.60 6.80 1.02 10.58 0.11 2.34 12,265 0.06 1997 2.04 15.45 1.51 6.94 1.17 13.21 0.14 3.32 13,691 0.08 1998 1.34 10.42 1.02 4.54 0.95 10.82 0.02 0.38 13,632 0.05 1999 1.42 10.31 1.16 5.19 1.06 10.91 0.16 2.97 12,784 0.05 2000 1.30 8.00 0.79 2.73 0.99 8.54 0.21 3.88 10,650 0.04 2001 1.46 9.36 1.33 5.64 0.75 8.60 0.08 1.76 9,820 0.05 2002 1.49 9.60 1.22 5.16 1.22 13.93 0.18 3.71 9,877 0.08 2003 1.80 10.99 2.12 8.65 1.16 11.43 –0.03 –0.45 8,692 0.09 Mean 1.18 8.66 1.05 4.53 0.97 9.41 0.16 2.25 7,850 0.06 FM t 11.42 9.11 23.06 5.78 Z1 38.75 20.26 42.07 10.06 Z2 12.40 9.30 15.18 7.22 ADJ = Compustat actual (income before extraordinary items) – I/B/E/S actual, scaled by stock price on the pre-announcement forecast date. All other variables are defined in table 1. Z1 = 1 / N ∑ N=1 t j / k j /(k j − 2) , where tj is the t-statistic for year j, kj is the degrees of j freedom, and N is the number of years. Z2 = t /( stddev (t ) / ( N − 1) ) , and the Fama-MacBeth t- statistic (FM t) = β /( stddev ( β ) / ( N − 1) ) , where N is the number of years. 50 TABLE 4 – Fourth-Quarter Effects FE FRQ FRY* DFE DFRQ DFRY* Est. t-stat Est. t-stat Est. t-stat Est. t-stat Est. t-stat Est. t-stat N Adj. R2 Pooled 1.35 35.00 1.28 19.89 1.03 38.03 –0.59 –8.53 0.63 5.17 0.14 2.32 156,993 0.05 1984 0.61 2.52 –0.55 –0.89 1.40 4.86 –0.26 –0.86 0.86 1.10 –1.39 –3.64 1,013 0.05 1985 0.42 3.27 0.64 2.35 0.83 6.73 0.04 0.21 1.78 3.00 1.22 4.38 2,944 0.08 1986 0.83 5.93 1.09 3.89 0.85 6.66 –0.41 –1.75 0.95 1.70 0.20 0.68 3,659 0.07 1987 0.74 4.61 0.39 1.09 0.98 5.74 –0.69 –2.56 1.49 2.16 –0.48 –1.38 3,254 0.03 1988 0.52 3.57 0.56 2.00 1.01 7.48 –0.15 –0.65 1.50 2.91 0.05 0.21 3,555 0.06 1989 0.72 6.31 0.93 4.11 0.84 7.79 –0.26 –1.22 0.16 0.38 –0.03 –0.13 4,721 0.06 1990 0.87 6.05 1.39 5.79 0.86 7.26 –0.40 –1.51 –0.07 –0.15 0.34 1.23 5,036 0.07 1991 1.56 9.73 0.70 2.75 0.93 7.56 –0.88 –2.84 1.86 3.47 0.46 1.44 5,774 0.07 1992 1.23 6.99 1.10 4.13 0.58 4.46 –0.39 –1.30 1.19 2.32 0.88 3.24 6,664 0.05 1993 1.97 11.31 0.95 3.65 0.82 7.01 –1.32 –4.17 1.24 2.35 0.41 1.51 8,075 0.06 1994 1.72 11.16 0.99 4.04 1.26 11.78 –0.51 –1.89 0.64 1.41 –0.41 –1.87 10,021 0.07 1995 1.35 8.78 2.33 8.84 1.36 11.99 –0.61 –2.19 –0.98 –2.10 –0.23 –1.02 10,866 0.09 1996 1.63 10.69 1.63 6.05 1.12 10.10 –0.62 –2.36 0.39 0.81 –0.28 –1.21 12,265 0.06 1997 1.99 12.52 1.73 7.08 1.20 12.11 0.08 0.29 0.37 0.80 0.04 0.20 13,691 0.08 1998 1.49 9.77 1.13 4.30 0.91 9.13 –0.57 –2.03 0.56 1.25 0.14 0.69 13,632 0.05 1999 1.37 8.65 1.13 4.59 1.14 10.77 0.19 0.61 1.41 2.75 –0.36 –1.39 12,784 0.05 2000 1.35 7.16 0.97 2.93 1.05 7.72 –0.28 –0.76 0.27 0.46 –0.06 –0.22 10,650 0.04 2001 1.86 10.09 1.38 5.34 0.59 6.06 –1.49 –4.30 0.85 1.60 0.85 3.92 9,820 0.05 2002 1.54 8.58 1.37 5.10 1.22 12.58 –0.37 –1.03 0.64 1.30 0.19 0.85 9,877 0.08 2003 1.87 11.00 2.22 8.82 1.10 10.53 –1.18 –1.77 –0.25 –0.30 1.27 2.83 8,692 0.09 Mean 1.28 7.93 1.10 4.30 1.00 8.42 –0.50 –1.60 0.74 1.36 0.14 0.52 7,850 0.06 FM t 10.60 9.06 19.08 –4.43 6.44 0.54 Z1 35.48 19.22 37.63 –7.17 6.09 2.31 Z2 11.72 7.89 14.53 –5.38 4.45 1.11 51 FRY* = FRY for Q1 – Q3, = FRY – FRQ for Q4 All other variables are defined in table 1. Z1 = 1 / N ∑ N=1 t j / k j /(k j − 2) , where tj is the t-statistic for year j, kj is the degrees of freedom, and N is the number of years. j Z2 = t /( stddev (t ) / ( N − 1) ) , and the Fama-MacBeth t-statistic (FM t) = β /( stddev ( β ) / ( N − 1) ) , where N is the number of years. D is suppressed in Panels B and D. 52 TABLE 5 – Extended Return Window ARit = a0 + a1 FEit + a2 FRQit + a3 FRYit + eit Panel A: I/B/E/S Summary Data FE FRQ FRY Est. t-stat Est. t-stat Est. t-stat N Adj. R2 Pooled 1.32 22.47 2.01 19.27 2.00 45.13 153,974 0.05 1984 –0.01 –0.05 –0.07 –0.10 1.78 5.27 1,011 0.08 1985 0.81 4.46 0.87 2.01 1.97 10.01 2,943 0.10 1986 1.26 6.02 1.27 2.78 1.45 6.65 3,658 0.06 1987 0.44 2.12 1.07 2.14 1.90 7.79 3,247 0.05 1988 0.69 3.53 1.62 3.85 1.77 8.73 3,548 0.06 1989 1.35 7.27 0.93 2.49 1.71 9.30 4,710 0.06 1990 1.09 4.65 1.15 2.80 1.84 9.05 5,036 0.06 1991 1.16 4.50 1.83 4.22 2.34 10.93 5,762 0.07 1992 1.37 5.29 1.50 3.51 1.81 8.59 6,664 0.05 1993 1.71 6.52 2.50 5.98 1.98 10.25 8,072 0.06 1994 1.46 6.47 2.32 6.19 1.55 9.34 10,031 0.06 1995 1.63 7.38 2.36 5.91 2.42 14.04 10,872 0.08 1996 1.57 7.38 2.69 6.62 1.88 11.26 12,281 0.06 1997 2.01 8.03 3.48 8.53 2.05 12.36 13,549 0.06 1998 1.06 4.78 3.43 8.68 1.57 10.12 13,416 0.04 1999 1.48 5.58 1.65 3.79 2.24 12.01 12,526 0.06 2000 1.53 5.35 1.94 3.83 1.84 9.02 10,615 0.08 2001 1.74 5.47 1.85 3.74 1.22 6.61 7,979 0.04 2002 1.94 7.14 0.85 2.06 2.18 14.20 9,528 0.07 2003 1.99 6.53 1.93 4.23 2.29 12.02 8,526 0.07 Mean 1.31 5.42 1.76 4.16 1.89 9.88 7,699 0.06 FM t 10.91 8.70 26.60 Z1 24.24 18.62 44.16 Z2 12.15 8.12 18.20 Variables are defined in table 1. AR is measured through the end of the postannouncement period. Z1 = 1 / N ∑ N=1 t j / k j /(k j − 2) , where tj is the t-statistic for year j, kj is the degrees of j freedom, and N is the number of years. Z2 = t /( stddev (t ) / ( N − 1) ) , and the Fama-MacBeth t- statistic (FM t) = β /( stddev ( β ) / ( N − 1) ) , where N is the number of years. 53 TABLE 5 - continued ARit = a0 + a1 FE_DETit + a2 FRQ_DETit + a3 FRY_DETit + eit Panel B: I/B/E/S Detail Data FE_DET FRQ_DET FRY_DET Est. t-stat Est. t-stat Est. t-stat N Adj. R2 Pooled 1.41 11.48 1.94 12.65 0.85 15.71 47,612 0.03 1984 –0.65 –1.45 0.41 0.51 0.40 1.04 292 0.02 1985 0.45 1.17 0.82 1.45 0.12 0.45 673 0.03 1986 –0.09 –0.17 –0.35 –0.49 0.24 0.90 565 0.03 1987 0.70 1.37 0.67 0.92 0.14 0.50 574 0.02 1988 0.37 1.06 0.00 0.00 0.48 2.24 839 0.04 1989 0.63 1.67 1.41 2.51 0.61 2.84 1,148 0.03 1990 0.52 1.25 2.12 3.75 0.48 2.14 1,502 0.05 1991 1.48 3.78 0.86 1.71 0.83 3.83 1,887 0.04 1992 1.05 2.20 1.34 2.06 0.74 2.95 1,904 0.03 1993 0.75 1.50 1.33 1.87 0.86 3.34 1,912 0.08 1994 1.86 4.28 1.23 2.22 0.99 4.89 3,077 0.05 1995 2.25 4.62 0.89 1.67 1.15 5.99 3,328 0.04 1996 2.65 5.24 1.91 3.43 0.98 5.22 3,481 0.04 1997 2.48 4.70 2.69 4.10 0.47 2.22 3,878 0.03 1998 1.50 2.85 3.36 5.40 0.84 4.07 4,047 0.04 1999 2.11 3.46 3.11 4.29 0.70 2.99 3,898 0.04 2000 1.51 2.14 2.29 3.03 0.99 3.93 3,599 0.09 2001 2.50 4.41 1.72 2.87 0.57 2.87 3,955 0.02 2002 3.89 7.38 2.12 3.63 1.18 5.96 3,769 0.05 2003 1.25 2.78 3.97 6.88 0.90 4.22 3,284 0.06 Mean 1.36 2.71 1.59 2.59 0.68 3.13 2,381 0.04 FM t 5.46 6.20 9.41 Z1 12.13 11.58 13.99 Z2 5.74 6.25 8.16 FE_DET = forecast error using detail data; FRQ_DET = subsequent fiscal quarter forecast revision using detail data; FRY_DET = subsequent fiscal year forecast revision using detail data. AR is measured through the end of the postannouncement period. Z1 = 1 / N ∑ N=1 t j / k j /(k j − 2) , where tj is the t-statistic for year j, kj is the degrees of j freedom, and N is the number of years. Z2 = t /( stddev (t ) / ( N − 1) ) , and the Fama-MacBeth t- statistic (FM t) = β /( stddev ( β ) / ( N − 1) ) , where N is the number of years. 54