um ......J.:: 5:. 1.... Meme lllllllllllllllllllllilllllllllllllllllllllHllllllllllllll 3 1293 00904 586 This is to certify that the dissertation entitled THE MAGNITUDE AND TIMING OF ANALYST FORECAST RESPONSE TO QUARTERLY EARNINGS ANNOUNCEMENTS presented by LISE N. GRAHAM has been accepted towards fulfillment of the requirements for Ph . D . degree in Bus . Adm . Xx: cfl’ZZ/M Kirt C. Butler Major professor Date July 21, 1993 MS U is an Affirmative Action/Equal Opportunity Institution 0-12771 LIBRARY Michigan State University PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. DATE DUE DATE DUE DATE DUE V-._\ t "0 min; MSU Is An Affirmative Action/Equal Opportunity Institution c:\cIrc\datedue.pm3—p.1 THE MAGNITUDE AND TIMING OF ANALYST FORECAST RESPONSE TO QUARTERLY EARNINGS ANNOUNCEMENTS BY Lise Newman Graham A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Finance and Insurance 1993 ABSTRACT THE MAGNITUDE AND TIMING OF ANALYST FORECAST RESPONSE TO QUARTERLY EARNINGS ANNOUNCEMENTS BY Lise Newman Graham Investors frequently use earnings per share as a proxy for estimated cash flows or as the starting point in the cash flow estimation process. Consequently, accurate and timely forecasts of corporate earnings are critical to security valuation and investment success. This study examines the magnitude and timing of revisions in analysts' forecasts of annual earnings around the time of quarterly earnings announcements. The sample includes earnings forecasts for 49 large firms with December fiscal year-ends for the years 1983 through 1986. The forecasts of annual primary earnings per share before extraordinary items come from the Institutional Brokers Estimate System (I/B/E/S) detail tapes of Lynch, Jones and Ryan, which contains forecasts made by individual analysts. Consensus forecasts for each firm are constructed for weekly, bi-weekly, and monthly intervals in the earnings anticipation period (the eight weeks preceding the earnings announcement), the announcement period (the week of and week following the announcement), and the post-announcement period (the seven weeks following the announcement period). A "market" average is also constructed using all firms in the sample. Tests of revisions from one interval to the next are then conducted using the both the unadjusted firm consensus forecasts and those forecasts adjusted for market- wide revisions occurring at the same time. The results provide little evidence that forecasters revise their forecasts in ways which anticipate annual earnings in the two months preceding quarterly earnings announcements. There is evidence that analysts underreact to the information in a quarterly earnings announcement and continue to revise their forecasts for as much as two months after the announcement. These findings are sensitive to the length of the period used to aggregate analyst forecasts, however. One implication for other studies of analyst forecasts and forecast revisions is that the choice of forecast aggregation period in forming a consensus forecast of earnings per share may affect the results. Also, studies of changes in analyst forecasts which do not adjust for changes in macroeconomic factors may be drawing spurious conclusions. This work is dedicated to Don, my husband, and to my children, Donald and Karen. Their love, support, understanding and patience throughout this process is greatly appreciated. iv ACKNO'LBDGMBNTB I would like to thank my dissertation chairman, Dr. Kirt Butler, for his expert and willing assistance and guidance. Thank you also to the other members of my committee, Dr. John Gilster and Dr. Dale Domian, for their helpful and insightful comments. The willingness of all the Finance faculty members to share their knowledge and expertise is greatly appreciated, as is the support system provided by the Ph.D. candidates. I would especially like to thank Janet Todd, Claudia Kocher and Rick Osborne for their friendship and support these past few years. Finally, I would like to acknowledge my family. My husband, children, parents, brothers, and sisters provided listening ears, helping hands and loving hearts as I endeavored to complete this dissertation, for which I will always be grateful. TABLE OF CONTENTS LIST OF TABLES . . . . . . . . . . . . . . LIST OF FIGURES . . . . . . . . . . . . . CHAPTER 1 CHAPTER 2 2.1 INTRODUCTION 0 O O O O O O O 0 LITERATURE REVIEW . . . . . . . Relationship Between Earnings and Security Returns . . . . . . . . Relationship Between Analyst Earnings Forecasts and Stock Returns . . Properties of Consensus Forecasts Properties of Individual Analyst Forecasts Research Issues . . . . . . . . PILOT STUDY ON THE NATURE OF ANALYST FORECASTS . . . . . . . HYPOTHESES AND METHODOLOGY . . Sample Selection . . . . . . . . Identifying Surprise/Nonsurprise Good/Bad News Samples . . . . . Testable Hypotheses . . . . . . 4.3.1 Forecast Revisions in the Post- Announcement Period . 4.3.2 Forecast Revisions in the Earnings Anticipation Period . 4.3.3 Forecast Revisions in the Earnings Announcement Period . 4.3.4 Relationship Between Forecast Revisions and Return Surprises . vi 11 12 16 21 43 43 46 50 54 57 58 59 CHAPTER 5 EMPIRICAL RESULTS . . Weekly Forecast Changes and Return Surprises . LIST OF RE APPENDIX A APPENDIX B APPENDIX C CONCLUSIONS AND EXTENSIONS Conclusions . . . . . Extensions . . . . . . 6.2.1 6.2.2 CAR Persistance and Analyst Revision Activity 6.2.3 FERENCES . . . . . . . INDIVIDUAL ANALYST FORECASTS OF 1986 EARNINGS PER SHARE FOR THIRTEEN COMPANIES (PILOT STUDY) . . . LIST OF COMPANIES AND FIRM-YEARS INCLUDED IN THE SAMPLE . . . STATISTICAL RESULTS vii Relationship Between Forecast Revisions Analyst Optimism or Pessimism Bi-Weekly and Monthly Forecast Changes . Sample and Methodology Changes 60 63 69 75 81 81 83 83 83 85 86 90 122 123 Table Table Table Table Table Table Table Table Table Table Table 10 11 LIST OF TABLES Airborne Freight 1986 Dates With Excess Return > 5% Absolute Value . . . . . . . . . . AFG Industries 1986 Dates With Excess Return > 5% Absolute Value . . . . . . . . . . Ahmanson (H F) & Co 1986 Dates With Excess Return > 5% Absolute Value . . . . . . . . . . Abbott Laboratories 1986 Dates With Excess Return > 5% Absolute Value . . . . . . . . . . Affiliated Publications 1986 Dates With Excess Return > 5% Absolute Value . . . . . . . . . . AGS Computers 1986 Dates With Excess Return > 5% Absolute Value . . . . . . . . . . AMR Corporation 1986 Dates With Excess Return > 5% Absolute Value . . . . . . . . . . AMCA International 1986 Dates With Excess Return > 5% Absolute Value . . . . . . . . . . ADT, Inc. 1986 Dates With Excess Return > 5% Absolute Value . . . . . . . . . . AVX Corporation 1986 Dates With Excess Return > 5% Absolute Value . . . . . . . . . . Adobe Resources 1986 Dates With Excess Return > 5% Absolute Value . . . . . . . . . . viii 24 25 26 27 28 29 30 31 32 34 36 Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table 12 13 14 15 17 18 19 20 21 22 23 24 25 26 Adams-Millis 1986 Dates With Excess Return > 5% in Absolute Value . Airborne Freight Analyst Forecast Revision Statistics . Earnings Announcements . Weekly Changes in Annual Earnings Forecasts/Price Ratios Without Adjustment for Market-Wide Changes Weekly Changes in Annual Earnings Forecasts/Price Ratios After Adjustment for Market-Wide Changes Bi-Weekly Changes in Annual Earnings Forecasts/Price Ratios Without Adjustment for Market-Wide Changes Bi-Weekly Changes in Annual Earnings Forecasts/Price Ratios After Adjustment for Market-Wide Changes Monthly Changes in Annual Earnings Forecasts/Price Ratios Without Adjustment for Market-Wide Changes Monthly Changes in Annual Earnings Forecasts/Price Ratios After Adjustment for Market-Wide Changes Correlation Between Weekly Market-Adjusted Earnings Forecast/Price Ratio Revisions and Residual Return Surprises Correlation Between Bi-Weekly Market-Adjusted Earnings Forecast/Price Ratio Revisions and Residual Return Surprises Correlation Between Monthly Market-Adjusted Earnings Forecast/Price Ratio Revisions and Residual Return Surprises I/B/E/S Analyst Forecasts for Abbott Laboratories (1986) I/B/E/S Analyst Forecasts for Adams Express (1986) I/B/E/S Analyst Forecasts for Adams-Millis (1986) ix 37 42 50 67 68 71 72 73 74 78 79 80 90 96 96 Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 I/B/E/S Analyst Forecasts for Adobe Resources (1986) O O O O O O O O O I O O O O O O O O O O 97 I/B/E/S Analyst Forecasts for ADT, Inc (1986) 98 I/B/E/S Analyst Forecasts for Affiliated Publications (1986) . . . . . . . . . . . . . 99 I/B/E/S Analyst Forecasts for AFG . . . . . . 101 I/B/E/S Analyst Forecasts for AGS Computers (1986) O O O O O O O O O O O O O O O O O O O O 104 I/B/E/S Analyst Forecasts for Ahmanson (H F) & co (1986) O O O O O O O O O O O I O O O O O 106 I/B/E/S Analyst Forecasts for Airborne Freight (1986) O O O O O O O O O O O O O O O O 110 I/B/E/S Analyst Forecasts for AMCA International (1986) . . . . . . . . . . . . . 114 I/B/E/S Analyst Forecasts for AMR Corporation (1986) . . . . . . . . . . . . . . 116 I/B/E/S Analyst Forecasts for AVX (1986) . . . 121 Weekly Mean Earnings/Price Ratios for the Entire Sample . . . . . . . . . . . . . . . . 124 Weekly Market Average Earnings/Price Ratios for the Entire Sample . . . . . . . . . . . . 124 Weekly Revisions in Earnings/Price Ratios for the Entire Sample Unadjusted for Market-Wide Revisions . . . . . . . . . . . . 125 Weekly Revisions in Earnings/Price Ratios for the Entire Sample Adjusted for Market-Wide Revisions . . . . . . . . . . . . 125 Weekly Revisions in Earnings/Price Ratios for the Good News & Bad News Samples Unadjusted for Market-Wide Revisions . . . . . 126 Weekly Revisions in Earnings/Price Ratios for the Good News & Bad News Samples Adjusted for Market-Wide Revisions . . . . . . 127 Weekly Revisions in Earnings/Price Ratios for the Good News & Large/Small Surprise Samples Unadjusted for Market-Wide Revisions . . . . . 128 Table Table Table Table Table Table Table Table Table Table Table Table Table Table 44 45 46 47 48 49 50 51 52 53 54 55 56 57 Weekly Revisions in Earnings/Price Ratios for the Bad News & Large/Small Surprise Samples Unadjusted for Market-Wide Revisions . . . . . 129 Weekly Revisions in Earnings/Price Ratios for the Good News & Large/Small Surprise Samples Adjusted for Market-Wide Revisions . . . . . . 130 Weekly Revisions in Earnings/Price Ratios for the Bad News & Large/Small Surprise Samples Adjusted for Market-Wide Revisions . . . . . . 131 Bi-Weekly Mean Earnings/Price Ratios for the Entire Sample . . . . . . . . . . . . 132 Bi-Weekly Market Average Earnings/Price Ratios for the Entire Sample . . . . . . . . . . . . 132 Bi-Weekly Revisions in Earnings/Price Ratios for the Entire Sample Unadjusted for Market-Wide Revisions . . . . . . . . . . . . 133 Bi-Weekly Revisions in Earnings/Price Ratios for the Entire Sample Adjusted for Market-Wide Revisions . . . . . . . . . . . . 133 Bi-Weekly Revisions in Earnings/Price Ratios for the Good News & Bad News Samples Unadjusted for Market-Wide Revisions . . . . . 134 Bi-Weekly Revisions in Earnings/Price Ratios for the Good News & Bad News Samples Adjusted for Market-Wide Revisions . . . . . . 135 Bi-Weekly Revisions in Earnings/Price Ratios for the Good News & Large/Small Surprise Samples Unadjusted for Market-Wide Revisions . . . . . 136 Bi-Weekly Revisions in Earnings/Price Ratios for the Bad News & Large/Small Surprise Samples Unadjusted for Market-Wide Revisions . . . . . 137 Bi-Weekly Revisions in Earnings/Price Ratios for the Good News & Large/Small Surprise Samples Adjusted for Market-Wide Revisions . . . . . . 138 Bi-Weekly Revisions in Earnings/Price Ratios for the Bad News & Large/Small Surprise Samples Adjusted for Market-Wide Revisions . . . . . . 139 Monthly Mean Earnings/Price Ratios for the Entire Sample . . . . . . . . . . . . 140 xi Table Table Table Table Table Table Table 58 59 60 61 62 63 64 Monthly Market Average Earnings/Price Ratios for the Entire Sample . . . . . . . . . . . Monthly Revisions in Earnings/Price Ratios for the Entire Sample Unadjusted for Market-Wide Revisions . . . . . . . . . . . Monthly Revisions in Earnings/Price Ratios for the Entire Sample Adjusted for Market-Wide Revisions . . . . . . . . . . . Monthly Revisions in Earnings/Price Ratios for the Good News & Bad News Samples Unadjusted for Market-Wide Revisions . . . . Monthly Revisions in Earnings/Price Ratios for the Good News & Bad News Samples Adjusted for Market-Wide Revisions . . . . . Monthly Revisions in Earnings/Price Ratios for the Good/Bad News & Large/Small Surprise Samples Unadjusted for Market-Wide Revisions Monthly Revisions in Earnings/Price Ratios for the Good/Bad News & Large/Small Surprise Samples Adjusted for Market-Wide Revisions . xii 140 141 141 142 143 144 145 Figure Figure Figure Figure Figure Figure 6: LIST OF FIGURES AIRBORNE FREIGHT - 1986 EARNINGS FORECASTS . AIRBORNE FREIGHT - 1986 EARNINGS FORECASTS . AFG INDUSTRIES - 1986 EARNINGS FORECASTS . . AHMANSON (H F) & CO - 1986 EARNINGS FORECASTS ABBOTT LABORATORIES - 1986 EARNINGS FORECASTS AFFILIATED PUBLICATIONS - 1986 EARNINGS FORECASTS I O O O O O O O O O C O O O O O O O AGS COMPUTERS - 1986 EARNINGS FORECASTS . . . AMR CORPORATION - 1986 EARNINGS FORECASTS . . AMCA INTERNATIONAL - 1986 EARNINGS FORECASTS ADT, INC - 1986 EARNINGS FORECASTS . . . . . AVX CORPORATION - 1986 EARNINGS FORECASTS . ADOBE RESOURCES - 1986 EARNINGS FORECASTS . ADAMS-MILLIS - 1986 EARNINGS FORECASTS . . . AIRBORNE FREIGHT MEAN EPS FORECASTS - PRE/POST OCTOBER 1986 EPS ANNOUNCEMENT . . . STOCK RETURN PERIODS O O O O O O O O O O O O EARNINGS FORECAST PERIODS . . . . . . . . . xiii 2 24 25 26 27 28 29 30 31 32 33 35 37 39 46 54 CHAPTER 1 INTRODUCTION Financial theory holds that the value of an asset is simply the present value of its expected future cash flows discounted at a rate appropriate to the risk of the asset. Investors frequently use earnings per share as a proxy for estimated cash flows or as the starting point in the cash flow estimation process. Thus, information (such as quarterly earnings announcements) which affects the investment community’s evaluation of a security receives a great deal of attention from investors and security analysts. Previous studies of the relationship between earnings per share and the market's valuation of a security have found that announcements of unexpected changes in earnings are positively correlated with stock price changes (e.g. Brown [1978] and Rendleman, Jones and Latane [1982]). A similar link has been established between revisions in security analysts' forecasts of those earnings and security returns (e.g. Givoly and Lakonishok [1979] and Benesh and Peterson [1986]). Thus, it appears that forecasts of corporate earnings are an important component of investment analysis and that accurate and timely forecasts of earnings 1 2 may be critical to security valuation and investment success. Consider the forecasts of annual earnings for 1986 made by 31 security analysts covering Airborne Freight and reporting to the Institutional Brokers Estimate System (I/B/E/S) shown in Figure 1. AIRBORNE FREIGHT 1986 EARNIAGS FORECASTS MN 00 00 Cd] 00 Dani-'10 o a 0 E363 80 000 00 an E FORECAST EARNINGS PER SHARE IllllllItTllllIlllill 000000000 OJMU’MMUDOJJMUSU‘OQOWMJN FORECAST DATE (MONTHS) Figure 1: AIRBORNE FREIGHT - 1986 EARNINGS FORECASTS This figure shows the time series of analyst forecasts of annual earnings per share made in January through December 1986. The two vertical lines indicate the April and October quarterly earnings announcement dates. Several questions arise upon inspection of the graph: a) In what manner do analysts anticipate earnings information prior to earnings announcements? b) Do analyst forecasts react unbiasedly and without delay to information such as earnings announcements? 3 c) If analyst forecasts do not immediately impound the information in earnings announcements, in what manner do market participants adapt their forecasts to new information? d) Does share price lead and/or lag earnings forecast changes? To the extent that individual investors rely on forecasts made by individual analysts, an understanding of the process by which earnings forecasts are formed may add to investors' ability to use those forecasts. Direct tests of individual analyst forecast accuracy have found that there is no significant difference in the overall accuracy of the various analysts (O’Brien [1990] and Butler and Lang [1991]). However, the intra—year behavior of analysts’ forecasts of annual earnings is the subject of some debate. Using consensus forecasts, Kerrigan [1984] and Arnott [1985] find that large upward or downward forecasts early in the year tend to be followed by further revisions in the same direction. Abdel-Khalik and Espejo [1978] and Brown and Rozeff [1979] model this behavior as an adaptive expectations process in which forecasts are revised to incorporate the "permanent" component of the most recent forecast error. Givoly [1985] finds that the coefficients of adaptation vary over time and across companies, but that different forecasters of the same company exhibit similar adaptive behavior. Each of these studies is consistent with analyst underreaction to new information. 4 In a study of analyst earnings forecasts and forecast revisions reported to I/B/E/S during April and December from 1976 to 1984, DeBondt and Thaler [1990] find that analysts are generally too optimistic in the beginning of the year and that forecast revisions between April and December tend to reverse this bias. Based on these results, DeBondt and Thaler conclude that analysts typically overreact to new information. Brown, Foster, and Noreen [1985] and O'Brien [1992] also find empirical evidence that analysts are too optimistic in their earnings forecasts. This is perhaps especially true of "sell-side" analysts working for brokerage houses. In addition, investors may be concerned not only with the accuracy of the forecaster, but also with the timing of the forecast revision. If investors rely on these forecasts in making investment decisions, less timely revisions may have an adverse impact on their decisions. Given the demonstrated link between financial analysts’ forecasts and security returns, a deeper understanding of the way in which forecasts are made becomes important. This study adds to our knowledge of the forecasting process by examining the timing and magnitude of revisions of analyst forecasts of annual earnings per share around the time of quarterly earnings announcements. Forecast revisions are examined separately for positive versus negative surprises, as well as large and small surprises. To the extent that analysts' forecasts act as proxies for market expectations, 5 we gain additional insight into the way in which the market processes unexpected information. The next chapter reviews related literature and develops hypotheses. The third chapter contains the results of an initial investigation of the data used in this study. The fourth chapter discusses sample selection and study methodology. Results of the empirical tests are discussed in the fifth chapter. Conclusions and extensions of the study are in chapter six. CHAPTER 2 LITERATURE REVIEW This chapter reviews relevant prior research and identifies research issues regarding the behavior of analysts’ revisions of earnings forecasts subsequent to the receipt of new information. 2.1 Relationship Between Earnings and Security Returns Financial theory holds that the value of an asset is simply the present value of its expected future cash flows discounted at a rate appropriate to the risk of the asset. Investors have often used earnings per share as a proxy for estimated cash flows and even now, with explicit estimation of cash flows receiving more attention, earnings per share is frequently the starting point for that calculation. Also, a survey of investment managers regarding their securities analysis and portfolio management techniques by Carter and Van Auken [1990] found that fundamental analysis was considered to be an important valuation technique and, within that group of techniques, price/earnings analysis was highest ranked. Accordingly, earnings announcements receive a great deal of attention from investors and security analysts. In fact, as Givoly and Lakonishok [1984] note, 6 7 "earnings per share emerges from various studies as the single most important accounting variable in the eyes of investors and the one that possesses the greatest information content of any array of accounting variables." A relationship between earnings and security returns has previously been documented in the literature. For example, in an investigation of the earnings characteristics of the 50 best and 50 worst-performing NYSE stocks in 1970-1971, Niederhoffer and Regan [1972] found that the most important characteristic separating the two groups was profitability. Of the 50 stocks experiencing the greatest percentage gains in price, 45 reported earnings per share greater than those of a year earlier and 20 of those 50 reported earnings gains of at least 25 percent. In contrast, 46 of the 50 worst performers reported earnings decreases and 44 of those decreases were in excess of 25 percent. In a similar study, Benesh and Peterson [1986] also noted a strong relationship between unexpected earnings changes and security returns. Brown [1978] considered announcements of changes in annual earnings per share (excluding extraordinary items) of at least 20 percent for the years 1963 to 1971. His results indicated that the market does not react instantaneously but rather takes about 45 market days to fully impound the new information. More importantly, even with transactions costs, significant excess returns could have been earned 8 simply by purchasing the stocks at the time the announcement appeared in The Wall Street Journa . Similarly, Rendleman, Jones and Latane [1982], using a sample from the years 1971 - 1981, found a strong relationship between unexpected quarterly earnings and excess returns on common stock in the period following the announcement. In their study, approximately 50 percent of the adjustment to the new information occurred in the 90 days following the announcement and the greater the earnings surprise, the greater the cumulative excess returns. 2.2 Relationship Between Analyst Earnings Forecasts and stock Returns A similar relationship between analysts’ forecasts of earnings and security returns has also been found. In their study of the 50 best and 50 worst performing stocks of 1970- 71, Niederhoffer and Regan [1972] observed that the stocks with the highest returns had earnings increases substantially greater than those forecast by analysts (as reported in the March 31, 1970 edition of the Standard and Poor’s Earnings Forecaster). For the worst performing stocks, analyst forecasts were generally too optimistic and the actual earnings were greatly below the projections. Givoly and Lakonishok [1979] studied the information content of revisions in financial analysts' forecasts by measuring abnormal returns in the months surrounding the revision month. Using revisions produced by the most active 9 forecaster (the one with the greatest number of revisions) reporting in the Standard and Poor’s Earnings Forecaster during the period 1967 to 1974, they found positive abnormal returns in the months surrounding an upward revision and negative abnormal returns in the period around a downward revision. These abnormal returns persisted for two months following the revision month and were sufficient to cover transaction costs. These results provide support for the hypothesis that these forecast revisions contain information and that the market is inefficient with respect to these revisions. Benesh and Peterson [1986] provide further support for this hypothesis in their study of the relationship between analyst forecasts and stock price fluctuations. Using consensus forecasts reported by the Institutional Brokers Estimate System (I/B/E/S) during 1980 and 1981, they found that when an earnings forecast was revised by 5 percent or more, the security tended to experience significant excess returns for the remainder of the year. Based on this result, they suggested that "investors may improve their performance by immediately purchasing stocks that have experienced an upward reVision in the consensus forecast and selling stocks for which the consensus forecast has been revised downward." Hawkins, Chamberlin, and Daniel [1984] constructed portfolios consisting of the 20 stocks with the largest one- month increases in the I/B/E/S consensus estimates of 10 earnings for each of the quarters from March 1975 through December 1980. These portfolios outperformed portfolios comprised of all stocks covered by I/B/E/S, the S&P 500, and various combinations of 20 stocks chosen at random from the I/B/E/S universe. Furthermore, these abnormal returns remained even after adjusting for risk and transaction costs. Dowen and Bauman [1989] found that this relationship between forecast revision and excess returns continued to exist even after the publication in 1984 and 1985 of articles reporting this phenomenon. Using the I/B/E/S consensus forecasts for the year 1977 through 1986, portfolios were constructed based on the value of a revision ratio calculated as follows: Revision Ratio = E4/E3 where E4 and E3 represented the April and March consensus estimates of EPS for the current year. A revision ratio greater than 1.00 indicated an upward revision and a downward revision resulted in a ratio less than 1.00. They observed a significant positive relationship between the direction of the April forecast revisions and the returns on the stock for the remainder of the year over the ten-year period, including 1986. In addition, this relationship could not be explained by either the small firm effect or analyst neglect (few analysts following selected stocks). Harris and Gultekin [1987], in a study of financial analysts' consensus forecasts of corporate earnings growth 11 (as reported by I/B/E/S for the time period 1982-1985), noted a strong positive relationship between the analysts’ growth forecasts and the value of the company’s stock. In particular, companies with high growth forecasts had higher price earnings ratios and market to book ratios than companies with low growth forecasts. 2.3 Properties of Consensus Forecasts Previous research on the properties of consensus forecasts of earnings has documented that the accuracy of these forecasts is greater than if one simply extrapolates past earnings trends. As Brown and Rozeff [1978] note, the earnings forecasts of security analysts should be superior to time series forecasts since financial analysts presumably employ a larger information set than simply a time series of past earnings. Also, the very fact that profit-maximizing firms continue to employ analysts rather than relying solely on less costly mechanistic time series models implies that the analysts’ forecasts must provide information of value. O'Brien [1988] examined the relative merits of three composite analyst forecasts and time series models as proxies for expected earnings. Consistent with prior research, she also found that the analysts’ forecasts were superior to time series models. Within the analyst forecast group, her results indicate that the most recent forecast is more accurate than either the mean or median forecast. If the consensus forecast is restricted to only those forecasts 12 made since the last earnings announcement, aggregating the forecasts to remove individual idiosyncratic error improves forecast accuracy. In a study using I/B/E/S consensus forecasts from 1977 to 1982, Kerrigan [1984] found that large upward or downward forecast revisions early in the year tended to be followed by further revisions in the same direction. Arnott [1985] achieved essentially the same result in a study of the 1976- 1982 period. More recently, Dowen and Bauman [1989] in their study covering 1977-1986 found that analysts were continuing to make revisions in the early part of the year that were in general too small. In contrast, DeBondt and Thaler [1990], using I/B/E/S consensus forecasts for the years 1976 to 1984, found that analysts' forecasts were generally too optimistic, that early-year revisions were too large (i.e. analysts "overreacted") and that forecasts of year-ahead earnings per share were even more extreme than current year EPS forecasts. In particular, they noted that actual changes in EPS averaged only 65 percent of the forecasted one-year changes, while the actual two-year change was only 46 percent of the forecasted change. 2.4 Properties of Individual Analyst Forecasts On the issue of whether some analysts are more accurate ' than others, O’Brien [1990] finds no evidence of systematic differences in forecast accuracy among individuals. (A 13 fixed effects model was used to control for average year and industry effects.) Butler and Lang [1991], using a different methodology, achieve essentially the same result. However, Butler and Lang also find that some analysts are consistently optimistic or pessimistic relative to the consensus forecast. Harris and Gultekin [1987] find evidence of analyst over-optimism in earnings forecasts for individual firms. A consistently optimistic estimate at a time when the consensus is consistently overestimating earnings can lead to inferior performance by that analyst relative to the group as a whole. Other research on the properties of individual analysts' earnings forecasts focused on the way in which analysts adjusted their forecasts to compensate for past errors. Abdel-Khalik and Espejo [1978] specified an adaptive expectations model which assumed that quarterly earnings announcements convey signals about the level of realizable earnings for the year. They tested the model using Value Line forecasts and actual earnings for 97 firms in 1976 and found a high degree of correlation between the announcement of interim earnings and the accuracy of the annual earnings forecasts. This provided empirical confirmation of the intuitively appealing theory that analysts use the information provided by those interim earnings reports. Brown and Rozeff [1979] also used revisions to Value Line earnings forecasts to examine the time series 14 properties of analyst forecasts. Using an ARIMA model, they found support for the hypothesis that analysts' forecast revisions follow an adaptive expectations model, in which expectations are revised to incorporate the "permanent" component of the most recent forecast error. However, the reaction coefficients, which summarize the forecast revisions by specifying the direction and size of response to the recent forecast error, imply a nonuniform reaction to forecast error by quarter. In addition, the explanatory power of their adaptive model is generally less than 50 percent, suggesting that information outside the time-series of earnings is also used in forecast revisions. This is also consistent with the idea that analysts use a richer information set than simply information about past errors. Givoly [1985] found further supporting evidence for the adaptive expectations model in a time series analysis of earnings forecasts reported in the S&P Earnings Forecaster. He noted that the coefficients of adaptation varied over time and across companies, but that different forecasters of the same company (for the 18 companies in his sample) exhibited similar adaptive behavior. Brown, Foster and Noreen [1985] also examined the relationship between security analyst multi-year forecast revisions and security price changes in the first year following the revision (i.e. fiscal year 1). For both consensus forecasts reported in the I/B/E/S data base and individual analyst forecasts reported in the Security Market 15 Line data base, there was a significant link between the security returns in fiscal year 1 and the forecast revisions for fiscal year 2 and later. They interpret this result as being "consistent with (i) the capital market having a multi-year earnings horizon and (ii) the forecast for fiscal year 1 not fully capturing the signal embodied in revisions to the earnings sequence over that multi-year horizon." Additionally, Brown, Foster and Noreen noted that "the sign and magnitude of security returns in the twelve month period preceding a revision in consensus security analyst forecasts are positively associated with the sign of the single year and multi-year forecast revisions." One potential explanation proposed by Brown, et al. is that consensus forecasts contain non—timely forecasts, thereby causing the consensus forecasts to appear to lag behind the security returns. O’Brien [1988] provides partial support for this alternative with her finding that the most current forecast is more accurate than either the mean or median forecast. Alternative explanations include (i) that security analysts process information less efficiently than does the market as a whole, (ii) that they use price changes as the signal to revise the earnings forecast, and/or (iii) that the analysts wait until they have had a chance to trade on the information before releasing the forecast. These explanations are all unappealing when applied to individual forecasts, particularly in light of the fact that individual 16 investors rely on the forecasts of individual analysts (often relayed through contact with a stock broker). If analysts wait until they have traded for their own accounts, that is at the least unethical, given that their clients pay for these earnings forecasts. Also, to assume that brokerage houses would continue to pay for analysts to simply recode the information contained in security prices is not consistent with economic theory. 2.5 Research Issues Given that it appears excess returns can be earned for some time subsequent to the announcement of unexpected information, and that investors rely on analysts’ earnings forecasts when making investment decisions, the accuracy and timeliness of analyst forecasts is critical. The relevant forecast for many individual investors is the forecast made by an individual analyst at a brokerage house, while professional investors use services such as I/B/E/S, Zack’s, and/or Value Line which aggregate individual forecasts in forming a consensus estimate. Therefore, knowledge of the behavior of individual forecasts may be beneficial. The evidence to date is that analyst forecasts do not differ in their degree of accuracy (O'Brien [1988] and Butler and Lang [1991]) based on a comparison of the forecasts to the realized earnings. Additionally, Givoly [1985] claims the coefficients of adaptation in an adaptive expectations model exhibit insignificant differences between 17 individual forecasters of the same company. However, as previously noted, individual investors are concerned with the timing as well as the accuracy of the earnings forecast. This second dimension of forecast revisions has not yet been tested. Therefore, the null hypothesis is that subsequent to the receipt of new information, analysts' revisions will not exhibit significant differences in either timing or magnitude (controlling for the firm and year effects noted by both O'Brien [1988] and Givoly [1985]). Also, consistent with rational expectations, the forecasts should be unbiased, efficient, and consistent. One alternative to the null hypothesis is that the magnitude and/or timing of the revision will be systematically different for "good news" vs "bad news" events. This alternative was suggested by Harris and Gultekin’s [1987] finding that there was significantly more revision activity for those firms for which earnings were initially overestimated than for those for which the earnings were underestimated. They speculate that this is the result of analysts’ reluctance to revise their published forecasts downward and so the forecasts are revised gradually in a series of small steps. If analysts overreact, as found by DeBondt and Thaler [1990], then security analysts should initially revise earnings forecasts upward subsequent to good news, followed by revisions downward to the true earnings level. For bad news, large downward revisions would be followed by smaller 18 upward revisions. It is also possible that early revisers may overreact and analysts which revise more slowly may be closer to the true earnings. Because analysts become more accurate as the year progresses (Butler and Lang [1989]), this requires that the revision period be carefully defined. If the underreaction noted by Kerrigan [1984], Arnott [1985], and Dowen and Bauman [1989] is the norm for analyst revisions of earnings per share forecasts, then revisions subsequent to good news should be followed by further upward revisions. Similarly, downward revisions subsequent to bad news should be followed by further downward revisions. Another factor may be that persistent analyst optimism or pessimism, as found by Butler and Lang [1991], has a systematic influence on the magnitude and/or timing of the revision. Persistently pessimistic (optimistic) analysts should overreact (underreact) to bad news and underreact (overreact) to good news. If, on average, there is as much good news as bad, no differences‘in the overall accuracy of the forecasters would be noted. If, however, there are systematic differences in the revisions of the two groups, this could be important during times of persistent good news or bad news. Even if analysts exhibit significant differences in the timing of the forecast revisions, in an efficient market this should have no significant influence on the return earned by individuals relying on those forecasts. However, if the overreaction hypothesis is correct, investors who 19 rely on "early revisers" should earn lower returns than those who rely on "late revisers". For good news, they will buy too soon, at too high a price, and for bad news they will sell too soon at too low a price. If the uncertain information hypothesis formulated by Brown, Harlow and Tinic [1988] is correct, however, investors who rely on "early revisers" should earn higher returns for good news events and lower returns for bad news events than those individuals who rely on "late revisers". For good news, positive excess returns are followed by more positive excess returns, so purchasing early allows one to capture more of the excess return. For bad news, negative excess returns are followed by positive excess returns, so selling early results in selling at too low a price (just as under the overreaction hypothesis). An additional question to be investigated is the degree to which the distinction between earnings announcements and other types of information influences the forecast revision process. The timing and/or magnitude of the revision may be less for certain types of information than for others. Similarly, there may be categories of information which have more influence on the forecast of next year's earnings than on the forecast of long term growth for a particular company or vice versa. Quarterly earnings announcements or management's earnings forecasts are expected to have a direct effect on the analysts' earnings forecasts. Subsequent to the release 20 of this information, we should observe analyst forecast revisions (at least, to the extent that the announcement contains new information). In general, firms are reluctant to announce dividend increases unless reasonably certain the higher dividend can be maintained and are equally reluctant to decrease dividends unless conditions force them to do so. Therefore, announcements of dividend changes may contain information about management's view of the firm’s earnings potential. If so, such announcements may be followed by revisions in analysts' earnings forecasts. Announcements of increased investment in plant and equipment or of an acquisition or merger may affect the estimated cash flows for a firm or its long-term growth prospects; however, the impact on current year earnings will probably be minimal. Therefore, such announcements would not be expected to lead to revisions of current forecasts of earnings. CHAPTER 3 PILOT STUDY ON THE NATURE OF ANALYST FORECASTS This pilot study was done in order to examine the individual analyst forecasts of earnings per share available on the I/B/E/S detail tape published by Lynch, Jones and Ryan. This is the database I/B/E/S uses in constructing its published consensus forecast. In addition to identifying potential problems with the data, the sample selection method and the study methodology were altered and/or refined as a result of this study. For the first 50 firms listed on the CRSP daily tapes, those dates with one day excess returns of 5 percent or greater (in absolute value) during the time period January 1983 through December 1986 constituted the initial sample of "events". Excess return is defined to be the difference between the return on the security and the return on the market for a given day. These returns are unadjusted for the risk of the security in this pilot study. This sample selection method is based on the methodology used by Brown, Harlow, and Tinic [1988] and is an attempt to avoid specifying in advance what information the market should consider in valuing securities or what analysts use in forecasting earnings. It has an additional benefit in that 21 22 it obviates the need to subjectively classify information as "good news" or "bad news"; the sign of the excess return proxies for the market consensus regarding the nature of the information. Of the firms so identified with excess returns in 1986, thirteen were then matched with forecasts available on the I/B/E/S data tapes. One of these firms, Adams Express, was a closed-end investment company followed by only one analyst, and so was not investigated any further. To develop a better understanding of the pattern of forecasts and revisions, the individual forecasts were plotted in order of forecast date. These plots can be seen in Figures 2 through 13 and a listing of the individual forecasts of 1986 earnings per share for the thirteen firms can be found in Appendix A. For some months, the tapes contain several instances of an analyst having more than one earnings forecast reported with the same forecast date. Upon further investigation, some of these cases were the result of the company splitting its stock. (See, for example, Figures 3, 4 and 5.) In order for the previous earnings forecasts to be consistent with future forecasts and with actual earnings, the existing forecast was adjusted for the split. In other cases, however, the tape seems to contain duplicate forecasts. For these twelve firms I then attempted to verify the nature of the underlying event for those days with an excess return of 5 percent or greater, using the Wall Street 23 Journal Index (WSJI). The results of this search are shown in Tables 1 through 12. 0f the twenty-eight identifiable events, eleven were announcements of quarterly earnings per share. However, for a number of the days on which the stock experienced an excess return, no event was recorded in the Wall Street Journal Index. One possible explanation for this is that these returns were unadjusted for risk. However, the magnitude of the excess return on some of these days for which no announcement was identified is such that an adjustment for risk is unlikely to be the entire explanation. An alternative explanation is suggested by the fact that the stock of several of these firms was trading at a very low price, meaning that small price changes in actual dollars could result in proportionally large returns. The results of this pilot study caused two major changes in the sample selection process. (For a detailed description of the process, see the following chapter.) Rather than drawing the sample from all firms on the CRSP tape, the sample was drawn from large firms which are widely-followed by security analysts and which represent a diverse set of industries. A second alteration to the sample selection process was the decision to specify quarterly earnings announcements which resulted in a return in excess of the market return for that period as the event of interest, rather than using the excess return criterion to define potential event dates and then identifying the FORECAST EARNINGS PER SHARE 24 AIRBORNE FREIGHT 1986 HRNINGS FOREASI'S IIIIIITTIIIIIIIITITII 00 on 335 Dcp qbth D L I. 00 c100 0 mad] at: 0.510 008 00 )— b p— u. y.— .— p.— y— i- - )— .— N FORECAST DATE (MONTHS) Figure 2: AIRBORNE FREIGHT - 1986 EARNINGS FORECASTS This figure shows the time series of analyst forecasts of annual earnings per share made in January through December 1986. Table 1 Airborne Freight 1986 Dates With Excess Return > 5% in Absolute Value 4/24/86 5/05/86 6/05/86 7/08/86 8/11/86 8/15/86 9/17/86 10/06/86 10/27/86 10/28/86 11/14/86 (“l (+) (“l ('l (+) (+) (+) (“l (+) (+) (+) Earnings & dividend announcement (4/29) Earnings announcement TNT proposes to offer $29 per share for Airborne .A + or - following the date indicates a positive or negative Those dates with excess returns which could excess return. the matched with a news article are so noted. 2.6 2.5 2.4 2.3 2.2 2.1 1.9 1.8 1.7 1.6 FORECAST EARNINGS PER SHARE 1.5 '- 1" 1.3 25 AFG INDUSTRIES 1986 EARNINGS FORWASTS 000 DD 0 0 cp [ID 0 U U D I!!! D C] D D D D U D D 7 8 9 10 11 12 13 roam DATE (MONTHS) Figure 3: AFG INDUSTRIES - 1986 EARNINGS FORECASTS This figure shows the time series of analyst forecasts of annual earnings per share made in January through December 1986. Table 2 AFG Industries 1986 Dates With Excess Return > 5% in Absolute Value 4/03/86 4/04/86 4/16/86 6/11/86 6/13/86 7/09/86 9/08/86 10/24/86 11/19/86 (+) (“I“) (+) (+) (4') (+) (’) (4') Earnings expected Earnings Redeemed Earnings Earnings Dividend expected to increase by 30-40%; to establish a cash dividend & dividend announcement (4/15) convertible debentures (6/9) expected to increase (6/16) announcement (7/10) announcement (10/28) AFG may seek to acquire Lear Siegler (11/20) A + or - following the date indicates a positive or negative Those dates with excess returns which could be matched with a news article are so noted. excess return. FORECAST EARNINGS PER SHARE 0) 26 AHMANSON (H F) 8: CO 1986 EARNINGS FORECASTS Q90 HUGE! ‘3 [3&8 0 [BED oaogaesamfiu FORKAST BATE (MONTHS) Figure 4: AHMANSON (H F) & CO - 1986 EARNINGS FORECASTS This figure shows the time series of analyst forecasts of annual earnings per share made in January through December 1986. Table 3 Ahmanson (H F) & Co 1986 Dates With Excess Return > 5% in Absolute Value 1/07/86 1/24/86 7/15/86 7/29/86 9/05/86 9/30/86 12/02/86 12/03/86 (+) (+) (+) (-) ('i (+) (+) (4') Ahmanson replaces Storer Communications as one of 127 stocks on which the ASE trades put and call options (1/8) 1985 earnings quadruple those of 1984 Earnings announcement (7/17) Named new president & CEO, plus a new chief operating officer (12/5) . A + or - following the date indicates a positive or negative Those dates with excess returns which could be matched with a news article are so noted. excess return. 27 ABBOTT LABORATORIES 1986 EARNINGS FORECASTS Ifiégii 3.8 P- 3.6 '- 3.4 '- 3.2 " 2.8 " 2.6 '— 24- :1 g aygg gm a In T e 2.2- 0 Ci, 1 2 FORECAST EARNINGS PER SHARE 1 1 1 1 1 1 1 1 1 3 4 5 6 7 8 9 10 u 12 13 z 1 1 FOREASI' DATE (MONTHS) Figure 5: ABBOTT LABORATORIES - 1986 EARNINGS FORECASTS This figure shows the time series of analyst forecasts of annual earnings per share made in January through December 1986. Table 4 Abbott Laboratories 1986 Dates With Excess Return > 5% in Absolute Value 11/10/86 (+) ’A + or - following the date indicates a positive or negative excess return. Those dates with excess returns which could be matched with a news article are so noted. 28 AFFILIATED PUBLICATIONS 1986 EARNINGS FOREASIS 2.7 2.6 - o E 2.5 _ E 2.4 .. U} E 2.3 - 11. 8 2.2 - o E 112: 21 b 0 1m 11 0 1:1 0 1- 2 - c1 c1 2 c o :1 0 iii 1.9 — u c1 o o N a o D E 1.8 — 0 0:1 0 o o 1.7 - 0 1:1 0 0 n 1.6 L 1* L l l l l l 1 1 l l l 1 2 a 4 5 6 7 a 9 10 11 12 13 FORECAST DATE (MONTHS) Figure 6: AFFILIATED PUBLICATIONS - 1986 EARNINGS FORECASTS This figure shows the time series of analyst forecasts of annual earnings per share made in January through December 1986. Table 5 Affiliated Publications 1986 Dates With Excess Return > 5% in Absolute Value 2/28/86 (+) Newspaper stocks again start to draw notice (2/25) 11/24/86 (+) McCaw Communications affiliate puts cable television unit up for sale A + or - following the date indicates a positive or negative excess return. Those dates with excess returns which could be matched with a news article are so noted. 1.9 1.85 1.8 1 .75 1.7 1 .65 1.6 1.55 FORECAST EARNINGS PER SHARE 1.5 1 .45 1.4 29 AGS COMPUTERS 1986 EARNINGS FORECASTS D n O D D O D D D E] CI 0 D D U CI 0 1 1 1 L 1 1 1 1 1 1 1_1 1 3 4 5 6 7 8 9 10 1112 13 FORECAST DATE (MON'IHS) Figure 7: AGS COMPUTERS - 1986 EARNINGS FORECASTS This figure shows the time series of analyst forecasts of annual earnings per share made in January through December 1986. Table 6 AGS Computers 1986 Dates With Excess Return > 5% in Absolute Value 2/18/86 4/25/86 5/02/86 6/19/86 6/26/86 8/14/86 10/23/86 11/05/86 12/03/86 (+) (+) (') (+) (+) (+) (+) (+) (+) Earnings announcement (4/28) Earnings announcement (10/22) Predicts record earnings for 4th quarter (12/4) .A + or - following the date indicates a positive or negative Those dates with excess returns which could be matched with a news article are so noted. excess return. 30 AMR CORPORATION 1986 EARNINGS FORRASI‘S 7.5 :1 o 71- E1 0 o 2 6,51— 0 é o — E: 6' E o m 6 a a :1 g 1:1 c1 11. 5.5_ :1 0 :1 a: [a g m 35 1m 0 2 =1. 0 “1' 0 c1 0 Ct: 6' a o o a .5— a u 1% o D e c, o 5 8 3‘9 % a 4'- U D D g a U a a U D U E 3.5- U o 3- U n 25 l l l l 1 L1 1 LI l l l 1 2 a 4 5 6 7 6 9 10 11 12 13 FORmASl' DATE (MONTHS) Figure 8: AMR CORPORATION - 1986 EARNINGS FORECASTS This figure shows the time series of analyst forecasts of annual earnings per share made in January through December 1986. Table 7 AMR Corporation 1986 Dates With Excess Return > 5% in Absolute Value 1/16/86 (+) Agrees to sell Sky Chef unit (1/15) 1/24/86 (+) 4th quarter earnings up 3.5% 3/20/86 (+) 6/23/86 (+) A + or - following the date indicates a positive or negative excess return. Those dates with excess returns which could be matched with a news article are so noted. FORECAST EARNINGS PER SHARE 1 n U 31 AMCA INTERNATIONAL 1986 EARNINGS mmasrs DOD Dom c1 gar-ho u no D 000 c1c1l'JCJ 5 6 7 8 9 1O 11 12 13 mum DATE (MONTHS) Figure 9: AMCA INTERNATIONAL - 1986 EARNINGS FORECASTS This figure shows the time series of analyst forecasts of annual earnings per share made in January through December 1986. Table 8 AMCA International 1986 Dates With Excess Return > 5% in Absolute Value 1/29/86 1/30/86 3/05/86 3/06/86 3/21/86 3/24/86 4/22/86 10/23/86 10/24/86 10/28/86 10/29/86 (+) (+) (+) (+) (+) Algoma may sell 34% stake in AMCA (4/17) Earnings announcement (net loss) (4/23) Earnings announcement (net loss) (10/21) A + or - following the date indicates a positive or negative Those dates with excess returns which could be matched with a news article are so noted. excess return. 32 .ADT,INKL 1966 EARNINGS mamasrs 2.1 2 r— o E i 1.9 - U} E 12L 0 n. In 0 E 1.7 - E o 1- 1.6 — 0 2 o E 1.5 — 0 E a 1.4 - o 1.3 L l I l I l l l I I L l l FORlXZASl' DATE (MONTHS) Figure 10: ADT, INC - 1986 EARNINGS FORECASTS This figure shows the time series of analyst forecasts of annual earnings per share made in January through December 1986. Table 9 ADT, Inc. 1986 Dates With Excess Return > 5% in Absolute Value 1/23/86 (+) 4/14/86 (+) Odyssey Partners increases stake in ADT from 5% to 6.9% (4/15) 12/17/86 (+) A + or - following the date indicates a positive or negative excess return. Those dates with excess returns which could be matched with a news article are so noted. 33 AVX CORPORATION 1986 EARNINGS FOREASTS 0.8 0 0.7- m 0.6- 0: a 0.51- CI 0. 0.4- :1 U) o E 08- 1: o 171 0.2— :1 1.. :1 G) g 01- c1 :1 a: E 0 a c1 c1 -01F— In! -02 I I I I I I I I I I I I I 12345676910111213 POW DATE (MONTHS) Figure 11: AVX CORPORATION - 1986 EARNINGS FORECASTS This figure shows the time series of analyst forecasts of annual earnings per share made in January through December 1986. 34 Table 10 AVX Corporation 1986 Dates With Excess Return > 5% in Absolute Value 1/15/86 (+) 2/12/86 (+) 2/20/86 (+) 3/24/86 (-) 4/02/86 (+ 4/16/86 (-) 5/09/86 (+) 6/17/86 (-) 7/01/86 (+) 7/31/86 (+ 8/04/86 (-) 8/08/86 (+) 10/01/86 (-) 10/29/86 (+) 12/16/86 (-) Propose to acquire CTS Corp; expect 4th quarter net income to be higher than 3rd quarter's .03/share 12/23/86 (-) A + or - following the date indicates a positive or negative excess return. Those dates with excess returns which could be matched with a news article are so noted. 35 ADOBE RESOURCES 1986 EARNINGS I'ORHIASTS FORECAST EARNINGS PER SHARE [3 FORM DATE (MONTHS) Figure 12: ADOBE RESOURCES - 1986 EARNINGS FORECASTS This figure shows the time series of analyst forecasts of annual earnings per share made in January through December 1986. 36 Table 11 Adobe Resources 1986 Dates With Excess Return > 5% in Absolute Value 1/28/86 3/12/86 3/14/86 3/21/86 3/24/86 5/08/86 5/09/86 7/08/86 7/29/86 7/30/86 8/05/86 9/04/86 9/10/86 10/24/86 10/28/86 10/29/86 11/10/86 11/12/86 11/14/86 11/17/86 12/03/86 12/16/86 12/23/86 (') (+) (“l (+) (“i (+) (+) (+) (+) (+) (‘) (-) (+) (+) (-) (+) (-) (-) (+) (') Repurchased 1 million shares of stock Net loss for prior year Named a director A + or - following the date indicates a positive or negative Those dates with excess returns which could be matched with a news article are so noted. excess return. 37 .ADAAflS-NUIIJS 1986 EARNINGS FORECASTS 2.9 2.8 2.7 2.6 2.5 2.4 2.3 2.2 2. 1 D 1.9 1.0 1.7 1.6 1.5 1.4 1.3 1.2 1.1 ' FORECAST HRNINGS PER SHARE N IIIITITIIIIITIIIII FORECAST DATE (MONTHS) Figure 13: ADAMS-MILLIS - 1986 EARNINGS FORECASTS This figure shows the time series of analyst forecasts of annual earnings per share made in January through December 1986. Table 12 Adams-Millie 1986 Dates With Excess Return > 5% in Absolute Value 2/14/86 (+) 2/18/86 (+) 3/21/86 (+) 4/07/86 (-) 12/19/86 (+) A + or - following the date indicates a positive or negative excess return. Those dates with excess returns which could be matched with a news article are so noted.- 38 event. This change was motivated by the large number of days with excess returns for which no underlying event could be identified and by the fact that the largest class of identifiable events was earnings announcements. The individual forecasts of annual earnings per share available for Airborne Freight both before and after its announcement of quarterly earnings on October 23, 1986 were then examined in order to further understand the data available. Thirty-one different analysts provided forecasts to I/B/E/S throughout the year. The results of various methods of constructing a consensus forecast are illustrated in Figure 14. The choice of analysts to include in calculating the mean forecast prior to the event had very little effect on the consensus; there was less than a one cent difference among the three groups. The mean forecast for the 25 analysts providing a forecast prior to the event date was $1.472. Of those 25 analysts, only 21 had an I/B/E/S report subsequent to the event; the mean forecast for those 21 analysts was $1.473. The mean pre-event forecast was $1.467 for the eleven analysts that actually revised their forecast subsequent to the earnings announcement. The choice of analysts to include in post-event mean forecast calculations affected the results to a much greater degree than for the pre-event forecasts. The mean post- event forecast for the eleven analysts who revised their estimates of earnings per share was $1.709. In addition, 39 AIRBORNE FREIGHT MEAN EPS FORECASTS PRE/POST OCTOBER 1986 EPS ANNOUNCEMENT 2 . 00 1 . 9O — 1'80 _ 1 709 1.710 4444 - 44 1.60 - :::.:.: " 1 50 — 47 1.473 4V85 . . v .v v ‘ . O...... 4.4.4.4 4.4.4.4. Z 1H40 L 4%fifi» «QQfiR "‘ 4 4 4 4 4 4 4 4 w! 444 44 444 1,30 — 4““; q“; SEEP 3 ”gay ”a; gag, m 120 _ %&&9 4““; 4““, O 4444 4444 4444 I11 1 . 1 O ‘- 4:4:4:4 :4:4:4 4:4:4:4 4444 4444 4444 m 00 — 4VTS ivvb 4%flfi9 o 1' dTOS 4%35 dggw’ h 444 4 4 4 (L90 - :fifig 3&fi: 3%fi: m dVV5 iVVS iUVb 9' 0 . 80 -. 4:4:4:4 4:4:4:4 4:4:4:4 W 4*“; gfifip ”a“, 070 h 4444 4444 4444 Z ' 4%”% $555 #555 FONS deb 5V55 a (L60 T 4%VN éVVL 4WVN 444 444 444 2 gap; 433; ”4&9 (150 - 4““; ”any gag; qyg gfl‘4 NfifiP (L40 -' fififi’ :figfi :fifigfi 444 444 444 4444 4444 4444 o 30 — 4 4 4 4 4 4 4 4 4 - qgg qfifip %&&P F 4’4; ”a”; ”a“; 0 . 20 4 4 4 4 4 4 4 4 4 4 ° dag? <fifi§§ 0 1 0 — ’4’4’4’ ’4’4’4’ ' 4““; %&&P 0 0 0 4 4 4 4 4 - 7 1 ALL FEE-EVENT REVISERS l PRE AND POST I PRE AND POST REVISERS POST EVENT ONLY VARIOUS GROUPS OF ANALYSTS Figure 14: AIRBORNE FREIGHT MEAN EPS FORECASTS - PRE/POST OCTOBER 1986 EPS ANNOUNCEMENT. This figure illustrates the results of different methods of constructing the consensus forecast for Airborne Freight. The all pre-event group includes any analyst with a forecast dated prior to the announcement. The pre and post group includes those analysts with a forecast included on the I/B/E/S tape both before and after the announcement, even though the most recent self-reported date may have been prior to the announcement. The post event group is those analysts with a forecast date subsequent to the earnings announcement, including those reporting for the first time, while the revisers group includes only those analysts who actually revised their forecasts. 40 four analysts reported for the first time after the earnings announcement; the mean forecast for those four analysts was $1.712. For the fifteen analysts with a forecast date subsequent to the event, the mean forecast was $1.710. If the consensus is calculated using the 17 analysts with a report date both before and after the event (even though the most recent forecast date for six of the analysts was actually before the event), the mean forecast was $1.629. Including the four "first-timers" with these 17 resulted in a consensus forecast of $1.645. Thus, the manner in which the consensus was constructed resulted in a difference of over eight cents per share between the highest and lowest mean forecasts of annual earnings per share. Since reports of earnings that are even one or two cents per share more or less than expected can result in large price changes, the choice of consensus is not trivial. Of the eleven analysts who revised their forecasts, four were above and seven were below the prior mean, as seen in Table 13. Ten of the eleven analysts revised their forecasts upward, consistent with the positive excess return for the stock. 0f the four who were above the prior mean, two were above and two were below the mean subsequent to the announcement. For the seven who were below the prior mean, four were above the new mean and three were below. The average revision for the eleven analysts was $0.242; the four analysts which were above the prior mean had an average revision of $0.1125 and the seven analysts below the prior 41 mean revised their forecasts an average of $0.3157 per share. Regarding the timing of the forecasts subsequent to the announcement, no forecasts were made in the first week following the event. Four, five, and four analysts revised their forecasts or reported for the first time in weeks 2, 3 and 4 respectively. One analyst required over 5 weeks to revise his/her forecast, one reported for the first time more than 5 weeks after the event and two who revised shortly after the announcement revised their forecasts upward once again a month later. While this provides no evidence that individual analysts differ systematically in the length of time needed to revise their forecasts, it does suggest that there are delays in analyst reaction to unexpected information. 42 Table 13 Airborne Freight Analyst Forecast Revision Statistics For Those Analysts Who Revised Their Forecast If Above If Below Prior Mean Prior Mean # Above New Mean 2 4 # Below New Mean 2 3 Mean Revision $.1125 $.3157 Number of Weeks Until Revision or Initial Forecast 1 2 3 4 5 >5 # of Analysts 0 4 5 4 0 2 CHAPTER 4 HYPOTHESES AND METHODOLOGY The purpose of this chapter is to describe the sample selection process, identify testable hypotheses, and develop empirical models to be used in testing the hypotheses. Earnings announcements are associated with both price changes and earnings forecast revisions. In Section 4.1, large return surprises at the time of quarterly earnings announcements are identified by comparing the magnitude of market model abnormal return to the standard deviation of abnormal return estimated over the 50 trading days preceding the announcement. Section 4.2 identifies good/bad news and surprise/nonsurprise samples based on the magnitude and direction of abnormal return. In Section 4.3, the magnitude and timing of analyst forecast revisions are examined over the periods preceding and subsequent to the announcements. 4.1 Sample Selection A group of 49 large firms with fiscal years ending in December, selected from the top 65 firms in the December 1986 Fortune index as in Butler and Lang [1991], is examined in this study. These firms are widely followed by investors and analysts and represent a broadly diversified cross- 43 44 section of U.S. industries. A list of these 49 firms can be found in Appendix B. The firms so identified were then matched with the forecasts of annual primary earnings per share before extraordinary items available on the Institutional Brokers Estimate System (I/B/E/S) detail data tapes supplied by Lynch, Jones and Ryan. This data base contains forecasts made by individual analysts. Each analyst is assigned a number and can therefore be identified in different time periods or with forecasts of earnings for different companies. Each individual forecast has a forecast date associated with it, which is the date the analyst actually made the revision. These dates are self-reported by the analysts. To the extent that competition with other analysts causes them to backdate their forecasts, this information may not be fully reliable. Tests of the timing, direction and magnitude of earnings forecast revisions in this paper require a measure of consensus earnings forecasts in each period. O'Brien [1988] examines individual security analysts' forecasts as earnings expectations and concludes that consensus earnings forecasts should be constructed from timely analyst forecasts. She finds that the forecast accuracy of forecasts just three months old is inferior to the most recent forecast. Accordingly, Lynch, Jones and Ryan now provides a "flash" consensus estimate, in addition to its traditional consensus, based only on forecasts made in the 45 most recent six week period. Similarly, Zack's Investment Service recently began reporting a consensus forecast constructed of forecasts reported in the most recent month. Unfortunately, neither of these consensus forecasts is currently available for empirical research. Following Butler and Lang [1991], a proxy for this flash estimate is constructed on a weekly basis using the individual analyst forecasts available on the I/B/E/S detail tape. Because the study focuses on intra-year earnings forecast revision behavior, only those analysts who supplied at least 3 forecasts to I/B/E/S each year were included in the sample. The need for an accurate, current consensus required that duplicate forecasts be deleted from the sample. If a firm was undergoing a stock split, stock dividend or merger, the earnings forecasts for that year were omitted from the sample. The fact that several analysts forecast earnings for each company in the sample represents an improvement over the forecasts used by Abdel-Khalik and Espejo [1978] and Brown and Rozeff [1979] in that several forecasts are available for each company, rather than a single Value Line estimate. The quarterly earnings announcement dates were then identified using COMPUSTAT. For the years 1983-1986, there were 784 announcements made by the 49 firms in the sample. Because the study focuses on intra-year revision behavior, only the first three quarterly announcements for any given year are included in the sample; the fourth quarter 46 announcement does not occur until after the end of the fiscal year, at which point earnings for the prior year are known. This reduces the number of potential events to 588 for the 49 firms in the sample. Furthermore, some of the 49 firms split their stock in one or more of the four years in this study. Because I/B/E/S does not indicate whether the forecast is of pre- or post-split earnings per share, those firm years were excluded from the study, reducing the number of events to 481. 4.2 Identifying Surprise/nonsurprise and Good/Bad News Samples Consider the stock return periods depicted in Figure 15. Estimation Announcement Post—Announcement Period Period Period |;\1|11|1\1| I 7-51 7-2 I T-l T 7+1 I 1+2 T+51 I days days 1 = the announcement day Figure 15: STOCK RETURN PERIODS 47 A three-day event window centered on the earnings announcement date was chosen as the period over which to measure return surprise. First, a market model regression E[Rjt] = aj + pj amt + ejt (4.1) was estimated over the 50 trading days prior to each event window. There are typically 65 trading days between quarterly earnings announcements. The 50 day estimation period was chosen as a compromise between the properties of timeliness and precision. Estimating the market model over a greater number of days increases the precision of the estimate. However, the farther back in time one goes, the more likely it is that an earlier earnings event will "contaminate" the returns in the estimation period. The 50- day estimation period is unlikely to include a great deal of information from the previous quarterly earnings announcement. Using daily returns in a market-model regression is suspect for any stock with price-adjustment delays which influence daily returns. One source of delay arises from the fact that not all securities trade at market close (see Scholes and Williams [1977]), although Simonds, Butler and Atchison [1993] find that this nonsynchronous security trading has a relatively minor impact on market model betas estimated with daily returns, especially for actively-traded stocks which are more likely to trade near market close. 48 More problematic are price-adjustment delays in transaction prices themselves arising from supply-demand forces and dealer activities (see Cohen, Maier, Schwartz and Whitcomb [1986]). While a more robust test might employ the market model estimation techniques of Scholes and Williams [1977], Dimson [1979] and Fowler and Rorke [1983], it is likely that these influences on daily returns are minimal for the 49 actively traded firms in this sample. Let sej represent the standard deviation of firm j's residuals from the above regression. If instantaneous returns Rjt follow the stochastic differential equation Rik = ujdt + ajdz, then residual returns from the market model are white noise with instantaneous daily standard deviation Sej° Since variance in this return generating process is a function of time, the standard deviation of residual return over the three-day event window is Jt Sej = (3 sej. In order to discriminate between large and small return surprises across the sample firms, risk-adjusted excess return over the three-day event window is measured relative to the residual standard deviation as follows: Normalized 3-day = 3-day excess return scaled by excess its standard deviation return +1 = )3 R. -ER. (35 .. 4.2 = _1( 3, 1 3,1) / e3 ( 1 49 Firms with normalized 3-day excess returns greater than one in absolute value constituted the initial sample. This process resulted in classifying the original 481 earnings announcements into two groups of approximately equal size. One group, the "nonsurprise" sample, is comprised of 253 earnings announcements which were not associated with a 3- day excess return greater than the standard deviation. The second group, the "surprises", is comprised of 228 earnings announcements which were associated with a normalized 3-day excess return greater than 1 (see Table 14). This second group was then dichotomized on good/bad news (i.e. positive/negative return). This dichotomization is motivated by Brown and Harlow’s [1988] finding that the way in which the market reacts to extreme stock price movements depends upon the direction of the initial change. Of the 228 "surprise" announcements, 128 were positive surprises and 100 were negative surprises (see Table 14). As noted in Chapter 3, a potential benefit of this sample selection method is that it may obviate the need to subjectively classify the earnings announcements as "good news" or "bad news"; the sign of the excess return proxies for the market consensus regarding the direction of the surprise. 50 TABLE 14 Earnings Announcements 1983 1984 1985 1986 TOTAL Total Number of Earnings Announcements 114 114 124 129 481 Total 11/ Normalized Excess Return 3 1 58 44 63 63 228 timber of Positive Surprises 35 26 41 26 128 timber of Negative Surprises 23 18 22 37 100 4.3 Testable Hypotheses The theory of rational expectations has implications for both security prices and security analysts' earnings forecasts. If security analysts’ forecasts conform to rational expectations, then forecasts should react unbiasedly and without delay to information contained in earnings announcements. Forecasts subsequent to an earnings announcement should fully incorporate information contained in the announcement. Analyst forecasts should also anticipate announcements as information arrives during the periods preceding the announcements. Suppose an earnings announcement is made at event time t. In order to investigate the level of analyst forecast revisions preceding, at the time of, and subsequent to an earnings announcement, define the change in analysts’ consensus forecast over event period t+w (i.e. over the interval (t+w-1,t+w]) as 51 > A A A th+w = (th+w - th+w-1) / Pj “j + 9j xMt+w + ejt+w (4'3) where th+w = the mean consensus forecast revision for firm j (j=1,...,J) over period t+w (i.e. from time t+w-1 through t+w) scaled by beginning-of-year price Pj K't+w A Kjt+w-1 A = [(I/Kjt+w)k:1 ijt+w-1] " [(1/Kjt+w-1) 1:21 th+w-1] > ijt+w = analyst k’s forecast of EPS for firm 3 during period t+w (k = 1""'Kt+w)’ a. = the intercept term, 8. = percentage change in firm j's EPS forecast for a given percentage change in the total EPS forecast for all firms in the sample (where each EPS forecast is scaled by its beginning-of-year share price), 52 fiMt+w = average change in earnings/price ratios across the J sample firms during the period J l/J z (1? i=1 jt+w ' th+w-1) / Pj ' ejt+w = the firm-specific component of firm j's consensus forecast revision over the interval (t+w-1, t+w]. Earnings forecasts are scaled by beginning-of-year stock price in order to avoid confounding share price changes with information contained in the earnings events. Adjusting the earnings forecast changes for average forecast changes across all other firms during period t+w controls for economy-wide forecast changes during the period. As in the market model regression 4.1 applied to price changes, the specification in 4.3 separates firm-specific revisions from market-wide revisions. The choice of the sampling interval ‘w' is important because of the scarcity of analyst forecasts around earnings announcements. Too short an interval may capture too few forecast revisions to yield reliable results. Too long an interval will provide a less timely revision measure. Empirical tests in the remainder of this section will use weekly, bi-weekly, and monthly measurement intervals. This will provide some perspective on how robust the empirical results are to the sampling interval. 53 The forecast sensitivity coefficient Bj could be estimated for each firm over the anticipation period. However the anticipation period has at most 12 weeks since the previous quarterly earnings announcement date, so the standard errors would be large and statistical precision low. For convenience, all Bj are instead assumed equal to one and all a- are assumed equal to zero. Under these 3 assumptions, firm-specific revisions in 4.3 are given by: ejt+w = th+w - xMt+w (4'4) With this alternative specification, each test is a joint test of the null hypothesis and the assumption that aj equals zero and Bj equals one for each sample firm. While there will be some firm-specific mis-estimation under this assumption, it does allow a simple adjustment for market- wide changes in the level of analyst forecasts. Tests of hypotheses in the next three sections focus on the behavior of the residuals eflfiw in the above model. Each test is based on the expected value of this residual: E[e = E[§ jt+w1 jt+w ' “j'pijt+w] [ EEth+w-th+w-1J-E[§jt+w-§jt+w-1] ] / Pj= 0' (4'5) where wgo, w=1 and wzz for the anticipation, announcement, and post-announcement periods, respectively, as shown in Figure 16. 54 Anticipation Announcement Post-Announcement Period Period Period l1\1|1 l 1 \1 I I t-8 t-l l t t+l week I t+2 t+8 I weeks weeks t = the announcement day FIGURE 16: EARNINGS FORECAST PERIODS 4.3.1 Forecast Revisions in the Post-Announcement Period Rational expectations requires that analysts react quickly and without bias to the information contained in earnings announcements. While analysts may not report forecasts until some time after announcement dates, forecasts in the period (e.g., week) immediately following the announcement should fully reflect the information contained in the event. Subsequent forecast revisions should not systematically alter revisions reported immediately after an earnings announcement. As a test of rational expectations, the following null hypothesis proposes that forecast revisions during the post- announcement period are unrelated to forecast revisions immediately after an announcement: H10: 5 = 0 for wzz. jt+w 55 Under rational expectations, the first forecast reported after an earnings announcement should fully and unbiasedly reflect the information in the announcement. Subsequent forecasts should then be unrelated to the change in the forecast during the announcement period. In testing H10, firms will enter the sample only after the first analyst to report subsequent to an announcement establishes the new level of earnings expectations. For each firm, if the first analyst to report a forecast subsequent to an announcement does not do so until week t+w (w > 1), then the firm is included only in t+w+1 and later tests. This ensures that consensus forecasts are only included in the empirical tests after the post-announcement level of earnings forecasts is established. Several authors have noted that security prices continue to react to earnings information well after earnings announcement dates. For example, Brown [1979] (see also Latane, Rendleman and Jones [1982]) finds that prices take up to 45 days to incorporate earnings information. If this is true for security prices, then analyst forecasts may also exhibit delayed reaction to earnings information. The literature suggests several alternatives to null hypothesis H10. Brown, Harlow and Tinic’s [1988] uncertain information hypothesis, for example, predicts that the magnitude of share price response is greater for negative 56 than for positive events. Brown, Harlow and Tinic find empirical results which are consistent with this hypothesis. A second alternative to H10 is DeBondt and Thaler’s over-reaction hypothesis. DeBondt and Thaler suggest and then empirically find that both prices (DeBondt and Thaler [1988]) and analyst earnings forecasts (DeBondt and Thaler [1990]) "overreact" to recent information. In contrast, several authors (e.g. Kerrigan [1984], Arnott [1985] and Dowen and Bauman [1989]) find empirical results suggesting that analysts underreact to earnings announcements. Still other authors, including Givoly [1985] and Klein [1990], fail to find either stock price over- or underreaction to earnings announcements. Under the null hypothesis of rational forecasts, and assuming equation (4.3) is correctly specified for all sample firms, H10 should apply to every subsample of earnings forecasts. T-tests of each hypothesis are run on the full set of earnings announcements, on the surprise and nonsurprise groups, and on the good news (positive return) and bad news (negative return) subsets of the surprise group. With respect to DeBondt and Thaler's [1990] overreaction hypothesis, for the good news sample ejt-I-w > 0 implies an underreaction to the earnings announcement and ejt+w < 0 implies an overreaction. For bad news, ejt+W > 0 and ejt+W < 0 imply an overreaction and underreaction, respectively. Similarly, a t-test comparison of means based 57 on ejtw for the good and bad news samples provides a test of Brown, Harlow and Tinic’s [1988] uncertain information hypothesis. 4.3.2 Forecast Revisions in the Earnings Anticipation Period Earnings forecasts reported before an announcement date may at least partially anticipate the event, especially if there are many analysts following a particular company. As a test of this premise, the following null hypothesis proposes that forecast revisions during the anticipation period are unrelated to forecast revisions at the time of an announcement: H20: e = 0 for w < 1. jt+w Hypothesis H20 provides a test of the extent to which analyst forecasts anticipate (i.e. converge toward) announcement-period earnings forecasts. Comparing ejt+w between good/bad and surprise/nonsurprise samples will provide a test of whether analysts anticipate earnings differentially in these various samples. 58 4.3.3 Forecast Revisions in the Earnings Announcement Period The last hypothesis proposes that forecast revisions during the announcement period are unrelated to information contained in the earnings announcement: H30: e = 0 for w 1. jt+w In the unlikely event that analysts are able to completely anticipate the information in earnings announcements, e3t+1 will be zero according to H3O. In most cases, it is expected that earnings announcements bring with them new information. This is especially true of the good and bad news samples with normalized excess return greater than one (the surprise samples). For these samples, analyst forecasts should react to the sign of price (and presumably earnings) surprise. Computing era” for the surprise/nonsurprise and good/bad samples will allow a comparison of the magnitude of firm-specific forecast changes between these samples. 59 4.3.4 Relationship Between Forecast Revisions and Return Surprises The correlation of firm-specific forecast revisions with the size of return surprise is another interesting measure. The resulting "forecast revision coefficient" can be considered a variation of the "earnings response coefficient" of, for example, Collins and Kothari [1987]. If there is no relationship between analyst revisions and market prices, then Corr(ejt+w, Rjt) = 0 for w = -1,0,1,2. Alternatively, if the announcements contain information, then Corr(ejt+1, Rjt) will be greater than zero. If the analysts are able to anticipate this information, then the correlation coefficient at time t-l would be greater than zero. And if the analysts exhibit a delayed reaction to the announcement, then the correlation at time t+2 will be greater than zero. CHAPTER 5 EMPIRICAL RESULTS Empirical results are presented in this chapter for each of the three hypotheses regarding analysts’ revisions of forecasts of annual earnings per share developed in the previous chapter. Those hypotheses are as follows: H10: No post-announcement revisions: eflflw'= 0 for wzz. Under rational expectations, the first forecast reported after an earnings announcement should fully and unbiasedly reflect the information in the announcement. This hypothesis proposes that forecast revisions immediately after an announcement are unrelated to forecast revisions during the subsequent post-announcement period. H20: No pre-announcement anticipation: eflflw.= 0 for w < 1. Earnings forecasts reported before an announcement date may at least partially anticipate the event, especially if there are many analysts following a particular company. Hypothesis H20 provides a test of the extent 60 H30: 61 to which analyst forecasts anticipate (i.e. converge toward) announcement-period earnings forecasts. Announcement period reaction: ejt+W = 0 for w = 1. In the unlikely event that analysts are able to completely anticipate the information in earnings announcements, éfiwd will be zero according to H3O. In most cases, it is expected that earnings announcements bring with them new information. This is especially true of the good and bad news samples with normalized excess returns greater than one (the surprise samples). For these samples, analyst forecasts should react to the sign of price (and presumably earnings) surprise. Tests of each of the hypotheses focus on the behavior of the residuals ejtw,in.the following model: A A A A th+w = (th+w ’ th+w-1) / Pj aj + Bj xMt+w + ejt+w (4’3) where 62 A th+w = the mean consensus forecast revision for firm j (j=1,...,J) over period t+w (i.e. from time t+w-1 through t+w) scaled by beginning-of-year price Pj z ijt+w-1 Kjt+w A > ] k=1 Kjt+w-1 A [(llxjt+w-1) k: th+w-1] = [[(llxjt+w 1 > ijt+w = analyst k’s forecast of EPS for firm 3 during period t+w (k = 1'°°°'Kt+w)' a. = the intercept term, 3. = percentage change in firm j's EPS forecast for a given percentage change in the total EPS forecast for all firms in the sample (where each EPS forecast is scaled by its beginning-of-year share price), fiMt+w = average change in earnings/price ratios across the J sample firms during the period J A l/J j:1(yjt+w - th+w-1) / Pj ' ejt+w = the firm-specific component of firm j’s consensus forecast revision over the interval (t+w-1, t+w]. 63 Each test is based on the expected value of this residual: = [E[th+w-th+w_1]'E[th+w-th+w_1] ] / P3,: 0. (4.5) where wgo, w=1 and wzz for the anticipation, announcement, and post-announcement periods, respectively. For each hypothesis, tests are run on the entire sample, the good/bad news samples, and the joint good/bad news and surprise/no surprise samples. Tests are run for weekly, bi-weekly, and monthly sampling intervals. Empirical results are presented for pre-announcement earnings anticipations (w<0), announcement period reactions (w=1) and post-announcement revisions (w>1) in Tables 15 through 23. (Full results may be found in Appendix C.) 5.1 Weekly Forecast Changes For the entire sample, the results in Tables 15 and 16 suggest that analysts do not anticipate the information contained in earnings announcements, nor do they react to that information quickly (i.e. within the first week following the announcment.) Only those forecasts reported in the third week following an earnings announcement are significantly different than those reported in the previous week at a 5% level of confidence. This is the case for both 64 the unadjusted forecast revisions (Table 15) and the revisions adjusted for sample-wide changes (Table 16). The significant coefficient in week t+3 is not consistent with the null hypothesis that the mean change is zero in the post-announcement period. Because earnings announcements may be either more or less than expected, analysts may revise their forecasts either upward or downward. Thus, the changes for all firms within the sample might cancel one another out, leading to the appearance of no change in the consensus forecasts. The sample was therefore dichotomized on the basis of positive and negative abnormal excess return as a proxy for whether the information was good or bad news. If the announcement contained negative information from the market’s perspective, we might reasonably expect that analysts also found the information to be negative and would accordingly revise their forecasts downward. The sample is dichotomized on the basis of the direction (positive or negative) of normalized excess stock return in the second and third panels of Tables 15 and 16. For announcements associated with positive returns, a statistically significant revision appears in both weeks t+2 and t+3 in Table 15. Interestingly, the revision is negative in week two and positive in week 3. After adjusting for forecast revisions occurring across all sample firms, only the third week following the announcement is significant. This delayed upward revision is consistent 65 with underreaction to the information in the earnings announcement for positive surprises, although the revisions of opposite sign in weeks 2 and 3 are puzzling. For the announcements associated with negative returns, none of the weeks in the anticipation, announcement, or post- announcement periods are significant. Large forecast revisions would be expected to be most prevalent around announcements associated with large price reactions. Dichotomizing on the basis of whether the return surprise was small or large revealed that large positive surprises were associated with statistically significant changes in week 4 after adjusting for market-wide changes (Table 16). Negative return surprise samples and small return surprise samples exhibited no statistically significant forecast revisions. The negative coefficient in week four suggests that analysts revise their forecasts downward despite the positive reaction of share price to the announcement. While one might suspect that this is evidence of an overreaction to the earnings announcement followed by a subsequent backward revision, there is no evidence of large upward revisions preceding this as one would expect with overreaction. The sum of the coefficients for positive surprise over the four-week post-announcement period is in fact close to zero in both Tables 15 and 16. These results are troubling in that the weeks with significant results and the signs of the revisions are not consistent across the subsamples. Also, it implies that 66 analysts and investors are somehow processing the information differently, in that information which the market appears to feel justifies larger than usual stock price changes does not seem to lead to a corresponding large change in the annual earnings forecast. A possible explanation for this is that the market is reacting to a single piece of information, while the analysts incorporate a broader information set into their forecast revisions. Under the assumption that the information in the quarterly earnings announcement is a component of the broader information set analysts use, an additional test was done with the data dichotomized by whether the earnings forecast was an overestimate or underestimate of the actual earnings per share at year-end. For earnings overestimates, some evidence of statistically significant reactions appears in Tables 15 and 16. However, as in the positive return sample, the direction of the forecast revisions changes from week to week. For underestimates, the revisions are positive in each post-announcement week using either unadjusted or adjusted forecast revisions, with several of the revisions being statistically significant. Dichotomizing the sample based on the ex-post direction of analyst forecast error does induce an ex-post selection bias. Nevertheless, these results suggest that for earnings forecasts which ultimately turn out to be too low (i.e. underestimates), analysts are 67 .He>uouca 883003 coco new one» on case: eoaueu m\a usacoucoo on» ad acade8>ou Gnu .eoaonuoman Ads: on» .monEennse oeueoavca on» «0 zone you one 080:: e as nausea on» new He>uousq aaxooz «0 none nous: some new ooueasoaeo can» mum: nodueu ooaum\uueomu0u mmcacueo ususoucoo on» ad newsman uuoa mo mcqccaoon on» an season ouo3 auueOGHOh .ooaum ounce .3003 co>dm one :8 aces muons non emcacueo finance uo uuueoou0u on» 88:0 made: Shaw some how couosuuucoo one: nodueu eoaum\ueooou0u umcqcueo usmcoucoo oUDOUIU I GO 000.00 H0>0H QOGOUHMGOO flm d ”I 0N0" EOHM ”Gmhwwudn— c 8.888.8 88888.8 88888.8 88.88.8 88888.8 888888.8- 8..88.8 88888.8- 88888.8. 88. 88.88.888.888: 88888.8. .8888 88.8.8- 88888-8- 8888.8- 8888.8 8888.8. .8888- .8888 888 832389.888 88.88.8- 88888.8 8..88.8- 88888.8- 88888.8- 88888.8 88888.8. 8..88.8- 8.888.8- .8. 888.88888 ..828 88888.8 ~..88.8 ~..88.8 88888.8- 88888.8 888888.8. 88~88.8 8..88.8- 88888.8 88. 888.8888» 8.8 88888.8. 88.88.8 ~.888.8- 88888.8- 8.888.8- 88888.8. 88888.8. 8..88.8- 88888.8. .88 88.888. o>..88oz 88.88.8 ..8888.8 88888.8- 88.88.8- 88888.8- 88.88.8- ~8.88.8 88888.8- ...88.8 88. 888.88.88 8.888 c.8888.8- 88.88.8 ~..88.8- .8888.8- 88888.8- 88888.8 88888.8 88888.8- 8.888.8 88. 888.8888. 8.8 88888.8. .88888.8 .88.88.8. 88888.8- 88888.8. 88888.8. 88888.8 .8.88.8- 88888.8 888 88.8.8. «8.8.888 888 888 888 ~88 888 .88 .88 888 888 .888.88>88888 .8 888882. 88888.8. .¢.8~88-8 88888.8. 88.88.8- 88888.8- 88888.8- 88888.8 88.88.8- ..888.8. .88 888.888.8888 .88 8.. 8+. 8+. ..8 8 .-8 8.. 8-. 8.. 888.888.8888 8.8888 88.2152 88>888c. >.xoo: moduem coaum\uuneoou08 mundane» ascend ca nooceco haxoes 8888888 8883-88888: 808 8882888888 88088.: ma manna 68 .He>ueuc« uaxees nose new ones en 08:03 nodueu m\m aseceecoo en» :8 ecoqew>eu en» .eeeesuom>n Has: ecu we gone neon: .eeameeense oeueoaosa one «0 some now one 080:3 a me eumeee one now He>ueucw xaxee3 gone new veueasoHeo sen» ewe: eameee ecu ca usuau an: uenuo ecu now eoaueu m\u useceecoo ecu ca emseno ecu need Shaw some now modueu oedum\ueeoeu0m emcqcuee useceecoo ecu c8 eemcano nee» no acaccqmen ecu ha oeaeoe eue: eueeoeuOh .eoqum euene .xees ce>wm ecu :8 eves euene hem emcacuee nuance mo eueeoeROu ecu 88:0 made: away some how ceuosuuecoo euez nodueu eouum\ueeoeu0w eucacuee useceecoo .ueeuln e no woman He>ea eocecwmcoo am e am one» Bonn uceueuuwo c ~nooo.o 8~ooo.o 88888.8 as-oc-o cso~88.o amomoo.o- ~o.8o.o 8.888.8- 8~888.8- no. 88.85.880888cs 8..oo.8- «88888.8 88.88-8. no~oo.o. 8.~oo.o- .88~oo.o ~noco.8- 88.88.8- n.ooc.o 88~ 80885.88oto>o 88.88.8- m-88.8 o~.oo.o- 88.88.8- nn~oo.o- .8~88.8 88.88.8- 88888-8. 8.~oo-8- .~. 808.8a838 .8828 88888.8 8n.oo.o 88.88-8 8-oo.o. 88~oo-o 88.88.8- oo~oo.o 88888.8- .8888.8 88. 808.8a838 8.8 88888-8. 88.oo.8 88888.8 88.88-8- 8.888.8- 88888.8 88888.8 ~8888.8- .o.oo.8. .- 8:83.08 o>.uoeoz ~8.oo.8 8N888.8 .8~88-8- 88888.8. 8.888.: 8K888.8- 88.oo.8 88.88.8- 8~.88.o N8. 808.88838 8.858 c88~oo.o- mn.oo.8 88888.8- 88888.8 88888.8- 8K.8o.o 88888.8 88888.8- .nooo.o 8m. 888.8eas8 8.8 88888.8- cmm~oo.o 88.88.8- o~coc.o ~nooo-o- 88888.8 88888.8 ...88.o. 88808.8 oo~ measuoa e>.8.8oa 888 888 ~88 ~88 nun .88 .88 888 888 88cc.ua>aomno .8 tongue. emcee-o- c-~oo.o ~sooo.o- .mcoo.o- 8~ooo.o- 88888.8 88888.8 88888.8- 88888.8- .88 8co_ue>888ae ..< 83 mt. ~48 .3 a ..u -u 8.. 8.. 85.882838 39888 .88x152 .8>8ouc. >.xoo: eemceso evazuvexuez new uceEpesflcd neuud eoauem eoqum\eueeoeuom emcqcuem amazed :8 eemcenu haxeez 0H wanna 69 likely to revise their forecasts progressively upward in the weeks following a quarterly earnings announcement. In the week of the announcement, results for the sample of firms with overestimated and with underestimated earnings are as one would expect. Analysts tend to revise their estimates downward for overestimated firms and upward for underestimated firms, with p-values in the range of .01 to .10. However, forecast revisions immediately preceding the announcement are puzzling. Revisions during week t-l are in the opposite direction than in week t and are in most cases significant at 5%. These pre-announcement forecast changes are nearly completely reversed in the week of the announcement.1 5.2 Bi-Wookly and Monthly Forecast Changes With a weekly sampling interval as in Tables 15 and 16, the entire sample and some of the subsamples reveal forecast revisions which vary unpredictably in sign. To test the sensitivity of these conclusions to the choice of forecast aggregation period, the individual forecasts were also aggregated into bi-weekly and monthly consensus forecasts. With bi-weekly aggregation (Tables 17 and 18), we again 1The baffling directional changes in forecast revisions across subsequent weeks may be due to the timing of firms' earnings announcements. Since nearly all sample firms report quarterly earnings in a four-week period beginning approximately one month after the fiscal end-of-quarter, each sample was further dichotomized based on the week of the announcement within this four-week period. Since no consistent pattern emerged from this dichotomization, results are not reported here. 70 conclude that there is no significant earnings anticipation in the forecast revisions prior to the announcement period for the sample as a whole or the positive/negative return subsamples. For the announcement period itself, it initially appears there is significant revision activity. However, when we adjust for revisions occurring across all other firms at that time, this revision activity appears to be due to macroeconomic and hence sample-wide information rather than to firm-specific information, as the results are no longer significant for any of these samples. So, although we would reject H1o using weekly aggregations, we cannot reject any of the first three hypotheses if a bi- weekly aggregation is used. For the samples based on over/underestimates of actual earnings, we now observe downward revisions of overestimates as well as upward revisions of underestimates in the announcement period. Again, this is consistent with analysts focusing on the long-term information component of the current announcement, rather than reacting to the short- term results. With monthly aggregations (Tables 19 and 20), the null hypothesis regarding the anticipation period cannot be rejected for the small negative surprises, nor for the announcement and post-announcement periods for the large positive surprises. However, t=2 for the monthly aggregation includes forecasts made from 5 to 8 weeks following the earnings announcement. Therefore, if some 71 emcucuee eoeceecoo ecu cu eemceco eueeoeuOh 88:0 ocues Euuu come now ceuoouuecoo euez eouumu ecuum\uuuoeuow mocucuee useceecoo .uueuau e co .ae>ueucu come now oueu en 08:03 eouueu m\m concencoo ecu cu ecoueu>eu ecu .eeeecuomac Has: ecu no come ueccb .ueamaeucae oeumouocu ecu no cone uoH ocu eaoc3 e we eameee ecu uOm Ho>ueucu aaxeezluc come how veumHsoaeo cecu euez nodueu eouum\ueeoeuo« .eouum euece nee» mo ocuccuoec ecu 8c oeaeoe eue3 .He>ueucu ce>um ecu :u eer euece hem emcucuee Hencce mo eueeoeuou ecu oeeec He>e~ eoceouucoo am e um oueu Eouu uceueuuuo 8 cn8.oo.o 8..oo.8- 8..oo.c 8.8.88.8 88.oo.o 88~8o.o- 88888.8. 88. neues_umeceuc: 888-8- 888-8- 5.8.88.8 «238.8- 88886- -88.o- 8. .88.... 8mm 8386380826 8.888.8 ~8~88.8- 88888.8 88.88.8- 88888.8. 88.88.8- 88.88.8- .8. 888.88888 .8828 ~8888.8 8~888.8- .8888.8 «88888.8. 88.88.8 88888.8 88888.8. 88. «88.88.88 8.8 8~888.8 .8.88.8. 88888.8 .88~88.8. 88888.. 88888.8. 88.88.8- .- 8:88.88 o>..auoz 88888.8- 88888.8. 888888.8 a8.888.8- 88888.8 88~88.8- 8~888.8- ~8. 888.8888» .8858 88888.8. 88888.8 88888.8 88888.8- .8888.8- 88888.8 88888.8- 8~. «88.88888 8.8 88~88-8- 88888.8- a8m~oo.o «88.88-8- 8.ooo.o 8..oo.o- 88888.8- 88~ megaueu o>_u_8ou .88 888 888 888 888 .88 888 .mco.uo>ceoco wo census. 8..oo.o- 88.88.8- ~8.oo.o a-~oo.8- 88888.8 ~8ooo.o- 88888.8- .88 8co.u8>comco ..< 8+. 8+. ~+u .+. u .-8 ~-8 8:8..o>.omno 8.8888 .o.22:= 88>touc. >.xoe:-.m 8888880 wcuznuoxuax uou 8805888888 8808883 eouucm eouum\euuuoeu08 emcucuen 888888 c. eeocuco haxeezsum 88 ednea 72 .Hu>ueucu come haw ones en 88:03 moduau m\m cancencoo ecu :8 ecoauu>eu ecu .uemecuomac Adan ecu 80 come neon: .ueameumcsu veuaouvcu ecu 80 come you add eaocz 8 88 eamsdn ecu how ao>ueuc8 aaxeeslac come now oeuaaaocdo cecu one: e8m588 ecu :8 88888 at: uecuo ecu now mowuuu m\m usucemcoo ecu cu eonnco ecu awed EH88 come yaw mouumu enqum\uumoeuoN nocacuue usucemcoo ecu c8 nemcmco .euqum eumcm ume8 mo unaccumec ecu 8c veamom ewe: uummoeuom .Hd>ueucu ce>au ecu cu e888 euucu uem umcucuae stccm no uuunoeuo« ecu 88:0 means Sham come new veuoauuucoo euez mouuau eouum\uumoeu0u umcacuoe usucencoo .uueulu 8 co veuuc He>ea eocequcoo am 8 um oueu eouu uceueuuua 8 88.88... 8888.8- 88886 «8888.8 8.8888... .8886- 8.8oo.8- 88. 8883.888883: 88888.8- 88888.8. 88.8.8 88888.8- 88888.8- 88888.8 88888.8- 888 888-8.888888 8.88.8 8888.8- .8888.8 88888.8. 8888.8- ..888.8- 88.88. .8. 88.88888 :88 88.8.8 8.888.8- 88.8.8 8.888.8- 88.888 88.888 88888.8- 88. 88.88888 8.8 8:88 88.88.8- 88888.8 88.88.8- 88888.8 88888.8 8:88.8- .88 8888888 88.8888: 88888.8- .8888.8- 8888.8 88.88.8- 88.8.8 88.8.8. 88888.8 88. 83.88888 :88 88888.8 88888.8 88888.8 88888.8 88888.8 88.88 8888.8- 88. «88.88888 8.8 88888.8 8888.8 ..888.8 88888.8- 88888.8 88888.8 8.88.8- 888 8888888 88.8.88 .88 888 888 888 888 .88 888 38.82.8888 .8 88.858. 8.88.8 .8888- 88.88 88888.8- 88888.8 88888.8 88888.8- .88 88.838888 :8 8+. 8+. 8+. .+. 8 .-8 8-8 888.88>88888 8.8888 .8 8885.. .8288... 88:883.; wemcuco evuzluexuaz new uceEuuancc ueuuc mouucm eoaum\uuuuoeuou macucuus Husccc cu nemcaco waxeezlum 8H ednea 73 Table 19 Monthly Changes in Annual Earnings Forecasts/Price Ratios Without Adjustment for Market-Wide Changes Month ly interval Nuvber of Sauple observations t-1 t t+1 t+2 All observations 480 -0.00142* -0.00418 -0.00183' -0.00073 (nurber of observations) 678 373 680 680 Positive returns 2S9 -0.00155" -0.00020 -0.00092 -0.00033 Big surprises 127 -0.00041 -0.00002 0.00022 -0.00013 Small surprises 132 -0.00265* -0.00038 -0.00202‘ -0.00052 Negative returns 221 -0.00127 -0.00066 -0.00290" -0.00119 Big surprises 100 0.00001 0.00073 -0.00415* 0.00081 Small surprises 121 -0.00232' -0.00187 -0.00187 -0.00286* Overestimates 332 -0.00166* -0.00110 4.00370" -0.00102 Underestilnates 168 -0.00088 0.00112 0.00237' -0.00008 * Different from zero at a 5% confidence level based on a t-test. Consensus earnings forecast/price ratios were constructed for each firm using only the forecasts of annual earnings per share made in the given interval. share price. Forecasts were scaled by the beginning of year Changes in the consensus earnings forecast/price ratios were then calculated for each monthly interval for the sample as a whole and for each of the indicated subsamples. Under each of the null hypotheses, the revisions in the consensus E/P ratios would be zero for each interval. 74 Table 20 Monthly Changes in Annual Earnings Forecasts/Price Ratios Without Adjustment for Market-Wide Changes Month ly interval quber of Bengal e observations t -1 t t+1 t+2 All observations 680 -0.00026 0.00069 0.00016 0.00105 (umber of observations) 678 373 680 680 Positive returns 259 -0.00025 0.00078 0.00096 0.00261* Big surprises 127 0.00065 0.00089 0.00230. 0.00659. Smell surprises 132 -0.00113 0.00068 ~0.00032 0.00030 Negative returns 221 -0.00022 0.00016 -0.00078 -0.00055 Big surprises 100 0.00130 0.00166 -0.00177 0.00136 Small surprises 121 -0.00167' -0.00112 0.00006 -0.00212* Overestilnates 332 -0.00066 -0.00021 -0.00163* -0.00097 Underestimates 1413 0.00025 0.002039 0.0041r 0.00123' * Different from zero at a 5% confidence level based on a t-test. Consensus earnings forecast/price ratios were constructed for each firm using only the forecasts of annual earnings per share made in the given interval. Forecasts were scaled by the beginning of year share price. Changes in the consensus earnings forecast/price ratios for each firm less the change in the consensus E/P ratios for the other n-l firms in the sample were then calculated for each monthly interval for the sample as a whole and for each of the indicated subsamples. Under each of the null hypotheses, the revisions in the consensus 8]? ratios would be zero for each interval. 75 delayed reaction to the announcement were occurring, aggregating the forecasts into a monthly consensus forecast may obscure those changes. 5.3 The Relationship Between Analyst Forecast Revisions and Return Surprises Another way to investigate analyst forecasts of annual earnings around the time of quarterly earnings announcements is by comparing them to revisions in share prices at the time of the announcement. If there is no relation between analyst revisions and market prices, then Corr(ejt+w, Rjt) = 0 for w = -1,0,1,2. If investors’ and analysts’ responses to the information in earnings announcements are indeed statistically independent, then this correlation coefficient will be zero. To examine the relationship between abnormal stock returns at the time of the announcement and changes in analysts’ forecasts, correlations between the two were calculated for the sample as a whole and for each of the large and small surprise subsamplesz. These results are summarized in tables 21 through 23. For the sample as a whole, there were no significant correlations in the week preceeding the announcement, the 2Correlations between both the forecast revisions and the forecast revisions adjusted for "market-wide" changes and both the excess return and normalized excess return were computed. Because the results were essentially the same regardless of the combination of revision measure and excess return measure, only the correlation between the market-adjusted forecast revisions and the normalized excess return is reported here. 76 announcement week, or the two weeks following the announcement. When the data is aggregated into bi-weekly forecast revisions, there again appears to be no significant correlation between the excess returns and the forecast revisions. It is only when the data is aggregated on a monthly basis that any correlation is observed; the correlation is particulary strong in the month following the announcement. When the sample is dichotomized into its large and small surprise components, for the small surprise group of announcements there are no significant correlations between analyst revisions and stock return at the time of the announcement. However, for the large surprise group of announcements, there is a strong positive correlation between the excess return on the stock and the forecast revisions in the announcement period. So, the larger the surprise to the market as measured by the abnormal excess return on the stock, the larger the changes in analyst forecasts for that firm.3 Since these forecast changes are computed relative to changes reported in periods at or before the week of the announcement, this evidence is consistent with rational expectations. It is only when the data is aggregated on a monthly basis that any correlation is observed for the post- 3These tests were also performed on the positive and negative return subsets of the large and small surprise samples. The results are qualitatively the same as for the large/small surprise samples as a whole and so are not reported here. 77 announcement period. Recall that forecast revisions in month t+2 are computed relative to a forecast computed in the first month after an announcement. The positive sign on month t+2 in Table 23 indicates that analysts continue to adjust their earnings forecasts in the direction of excess return for 2 months after the announcement. This is evidence of systematic analyst underreaction to the information in quarterly earnings announcements. n 1% __ 78 .usoaoauuooo soauoaouuoo xsou m.:oEuoomm moans pouusosoo one: coauoaouuou no mumoa .moameounsn vouooaocd 0:9.uo zoom uOu use egos: o no cameos on» ham Ho>u0usa some now oeumasoaou con» one: cameos any ad neuqu Hus nosuo on» uOu mowuou m\u momsmmsoo may :4 mucosa on» wood away some new moduou ouwum\ummuou0u mocasuom momsmmsoo on» ad neocono .ooaum wanna use» «0 modscammn 0:» kn vmaoum ou03 mumououOh .Ho>uous« sm>am 0:» sq some muons non masseuse Hansen uo mumoOoHOu on» haso moans Shaw some u0u ouuosuumsoo owes moauou muaum\uuoowuow mmcwcumo nsmsomcoo am an unmonuncOAm . m- o- as. as. Fe, -~ h- o- m- Ao~o¢.oo a~ooo.oo ..hmo.cs .mono.o. .oo.n.ov Akosw.oC .oaoa.oC Assom.oo no¢~o.oo .o:.a>-av o.oeo.c. oosoo.o -n.o.o cop—m..o cospo.o ~omoo.o ~o._o.o ~m««o.o. ~s~no.o mu»_tataa autos os~ neN ¢o~ so~ on. .m~ Pm~ .m~ omm Am~om.o6 a~o~».o3 acmeo.cv .mms~.os -~o~.o. .m.g~.oo .Psms.oo xkoqo.oo .on~m.oo .o:.a>-ao :88... @386. 6386 088.? 82.6.? «83.? ~38... «$8.? 33:... 33283 :98 «sq new ~on ~on ohm use use use mus Amcomua>aouno so honesty “spam.oo R~o~o.oo .omoo.o~ Aom.~.oo .omn~.o. .an..oo .sooo.oo .~o¢n.oo .an..ov .o:..>-ao 0:85- 628.? «88.? 8.0.86 32...? 833. $23... 83...? ”.886 933.2030 .2 3+» n+u ~+u .+u a .-u N-“ n-~ e-“ o.asom do>aouc_ >.xoo: nouqumusm cusuom descauom use acowmq>om oavom monum\munooouom umswcuen ooumsfloaiuoxuo: haxmoz soozuom soauoaouuoo an manna 79 Table 22 Correlation Between Bi-Weekly Market-Adjusted Earnings Forecasts/Price Ratio Revisions and Residual Return Surprises Bi-ueekly interval Sample t-2 t-1 t t+1 t+2 All observations 0.01722 0.03986 -0.02135 0.06001 0.03566 (p-value) (0.7081) (0.3865) (0.6823) (0.1976) (0.6662) 6E5 638 3N3 MB 653 Small surprises 0.09666 0.02926 -0.02088 -0.06195 0.01826 (p-value) (0.1365) (0.6668) (0.7755) (0.5151) (0.7772) 250 251 189 263 263 Large surprises -0.06167 0.06010 -0.00908 0.12758 0.06619 (p-value) (0.3588) (0.5678) (0.9035) (0.0589) (0.3633) 225 227 181 220 220 * Significant at 5% Consensus earnings forecast/price ratios were constructed for each firm using only the forecasts of annual earnings per share made in the given interval. Forecasts were scaled by the beginning of year share price. Changes in the consensus earnings forecast/price ratios for each firm less the change in the consensus E/P ratios for the other n-l firms in the sample were then calculated for each interval for the sample as a whole and for each of the indicated subsamples. Tests of correlation were conducted using Spearman's rank correlation coefficient. 80 Table 23 Correlation Between Monthly Market-Adjusted Earnings Forecasts/Price Ratio Revisions and Residual Return Surprises Monthly interval Sauple t-1 t t+1 t+2 All observations 0.01922 -0.00766 0.12397" 0.10177" (p-value) (0.6761) (0.8865) (0.0067) (0.0262) M5 3%) 4H7 677 Small surprises 0.05632 -0.00235 -0.05783 0.10762 (p-value) (0.3926) (0.9766) (0.3616) (0.0896) 2H) 189 an 251 Large surprises -0.01589 -0.00376 0.22350. 0.10191 (p-value) (0.8126) (0.9602) (0.0007) (0.1266) 225 181 226 226 * Significant at 5% Consensus earnings forecast/price ratios were constructed for each firm using only the forecasts of annual earnings per share made in the given interval. Forecasts were scaled by the beginning of year share price. Changes in the consensus earnings forecast/price ratios for each firm less the change in the consensus 2]? ratios for the other n-l firms in the sample were_then calculated for each interval for the sample as a whole and for each of the indicated subsamples. Tests of correlation were conducted using Spearman's rank correlation coefficient. CHAPTER 6 CONCLUSIONS AND EXTENSIONS In this chapter, conclusions from the results of the test of hypotheses are discussed, along with several possible extensions of this study. 6.1 Conclusions In tests of the correlation between excess stock return and analyst revisions of forecasts of earnings per share, we find significant correlation for the large return sample in the announcement period. This result is consistent with our expectations regarding analyst revision activity around the time of earnings announcements. If the announcement is a surprise to the market as a whole, it seems reasonable that the analysts would be surprised as well and would revise their forecasts. If the analysts had information regarding the firm's quarterly earnings and used that to revise estimates of annual earnings prior to the announcement, one would expect they would communicate that to their clients and that price adjustments would take place prior to the announcement as well. Hence, there is not a statistically significant correlation between stock returns and forecast 81 82 revisions for the small surprise firms, but there is for the large surprise firms. However, when the hypotheses are tested with the model A A th+w (th+w ' th+w-1) / Pj = “j + Bj xMt+w + ejt+w (4'3) we cannot draw any consistent conclusions regarding the three hypotheses. The results change when the forecast aggregation period is changed from weekly to bi-weekly to monthly. Also, revisions which appear to be significant when the raw or absolute changes are examined are often no longer significant when the revisions are adjusted for market-wide changes in that time period. Because so many of the results are not robust to the choice of aggregation period, the conclusions to be drawn from this regarding analyst forecast revisions are limited. This does, however, have implications for other studies of analyst forecasts; it demonstrates that the choice of the forecast aggregation period in forming a consensus can dramatically affect the results obtained. Similarly, studies of analyst forecasts which do not adjust for changes in macroeconomic (and hence sample-wide) factors may be drawing spurious conclusions. 83 6.2 Extensions 6.2.1 Sample and Methodology Changes The sample used in this study was earnings announcements made by 49 large firms in the first three quarters of the years 1983 through 1986. An immediate extension which suggests itself is to repeat the study using more recent forecasts made for a larger sample of firms. A second issue is the way in which the "market-wide" forecast revisions were calculated. Currently, market-wide revisions are defined as revisions of forecasts for all firms excluding firm j in week t+w. Thus, the market-wide revisions are calculated within-sample. Given the relatively small number of firms in the sample, this may not be a good estimate of the changes due to macroeconomic forces. Therefore, in conjunction with selecting a sample of more recent forecasts, an out-of-sample proxy for market- wide forecast revisions might yield results from which more definitive conclusions could be drawn. 6.2.2 CAR Persistence and Analyst Revision Activity In an efficient market, security prices will instantaneously and unbiasedly reflect new information. Similarly, if analysts respond to new information in an efficient manner, forecasts of earnings per share will fully incorporate new information unbiasedly and without delay. This should be true regardless of the size of the firm or 84 the number of analysts forecasting earnings per share. Therefore, there should exist no systematic relationship between the length of time over which investors earn excess returns, the length of the earnings forecast revision ' period, firm size and/or analyst following. Various authors have found that the security price response to earnings announcements is not immediate. For example, Brown [1978] concluded that the market takes about 45 trading days to fully impound information regarding changes in annual earnings per share. Rendleman, Jones and Latane found that only about 50% of the adjustment to unexpected quarterly earnings announcements occurred in the 90 days following the announcement. A potential explanation for the above is that, as analysts revise forecasts of annual earnings per share, information continues to trickle into the market. If so, we may note a correlation between the length of time needed for analyst revision activity to return to normal levels and the persistence of abnormal returns, such that the longer the forecast revision period, the longer the persistence of these excess returns. Company size, earnings predictability and analyst following are additional facets of the information environment of a firm which may affect market and analyst response to new information. Thus, there may be a relationship between the number of analysts following a particular firm, that firm’s market capitalization and/or 85 earnings predictability, the forecast revision period, and/or the abnormal return persistence period. 6.2.3 Analyst Optimism or Pessimism Butler and Lang [1991] defined relative optimism as being above the consensus forecast on average for the year each year in their four year time period. Thus, an analyst may have been significantly above the consensus for only one month and been below the consensus for the remainder of the year and still have classified as an optimist in their study. An extension of this study which builds on their work would examine the intra-year individual forecasts relative to the consensus to determine if analyst optimism or pessimism persists over a series of revisions. In addition, the effect of that relative optimism or pessimism on the revision process could be further examined. Persistently pessimistic (optimistic) analysts may overreact (underreact) to bad news and underreact (overreact) to good news. If, on average, there is as much good news as bad, no differences in the overall accuracy of the forecasters would be noted. If, however, there are systematic differences in the revisions of the two groups, this could be important during times of persistent good news or bad news. If analysts overreact, as found by DeBondt and Thaler [1990], then security analysts should initially revise earnings forecasts upward subsequent to good news, followed 86 by revisions downward to the true earnings level. However, persistently pessimistic analysts should overreact to good news to a lesser degree. For bad news, large downward revisions would be followed by smaller upward revisions. In this case, the optimists would have smaller downward revisions than would the pessimists. If the underreaction noted by Kerrigan [1984], Arnott [1985], and Dowen and Bauman [1989] is the norm for analyst revisions of earnings per share forecasts, then revisions subsequent to good news should be followed by further upward revisions. Similarly, downward revisions subsequent to bad news should be followed by further downward revisions. However, the number and magnitude of the revisions could be different between optimists and pessimists. LIST OF REFERENCES LIST OF REFERENCES Abdel-Khalik, A. Rashad and J. Espejo, "Expectations Data and the Predictive Value of Interim Reporting", Journa of Agcgnnting Research, Spring 1978, pp. 1-13. Arnott, Robert D., "The Use and Misuse of Consensus Earnings", Tna gonrnal gr Portfolio Management, Vol. 11, No. 3, Spring 1985, pp. 18-27. Benesh, Gary A. and Pamela P. Peterson, "On the Relation Between Earnings Changes, Analysts’ Forecasts and Stock Price Fluctuations", Financial Analysrs Journal, November-December 1986, pp. 29-39. Brown, Keith C. and W. V. Harlow, ”Market Overreaction: Magnitude and Intensity", Ina Jonrnal of Borrfolio Management, Winter 1988, pp. 6-13. Brown, Keith C., W.V. Harlow, and Seha M. Tinic, "Risk Aversion, Uncertain Information, and Market Efficiency", Journal of Financial Economigs 22, 1988, pp. 355-385. Brown, Lawrence D. and Michael S. Rozeff, "The Superiority of Analyst Forecasts as Measures of Expectations: Evidence from Earnings", Journal of Finance, March 1978, pp. 1-16. Brown, Lawrence D. and Michael S. Rozeff, "Adaptive Expectations, Time Series Models, and Analyst Forecast Revision", Journal of Accounting Research, Autumn 1979, pp. 341-351. Brown, Philip, George Foster, and Eric Noreen, "Security Analyst Multi-Year Earnings Forecasts and the Capital Market", American Accounting Association Studies in Accounting Research #21, 1985. Brown, Stephen J. and Jerold B. Warner, "Using Daily Stock Returns: The Case of Event Studies", Jonrnal or Einangial £22n2m12§_1&. 1985. pp- 3-31. Brown, Stewart L., "Earnings Changes, Stock Prices, and Market Efficiency", Journal a: Ernange, March 1978, pp. 17-28. 87 88 Butler, Kirt C. and Larry H.P. Lang, "Differences Among Analysts in Earnings Forecast Performance", Michigan State University Working Paper, November 1989. Butler, Kirt C. and Larry H.P. Lang, "The Earnings Forecasts of Individual Analysts: Evidence of Systematic Optimism and Pessimism", Journal of Accounting Research, Spring 1991. Carter, Richard B. and Howard E. Van Auken, "Security Analysis and Portfolio Management: A Survey and Analysis", The Journal of Portfolio Management, Spring 1990, pp. 81-85. Crichfield, T., T. Dyckman, and J. Lakonishok, "An Evaluation of Security Analysts' Forecasts", Accounting Rev'ew, July 1978, pp. 651-668. Cohen, K.J., S.F. Maier, R.A. Schwartz and D.K. Whitcomb, Tne Microstructure of Security Markats: Theory and Implications, 1986, Englewood Cliffs, New Jersey, Prentice Hall. Collins, Dan and S.P. Kothari, "An Analysis of Intertemporal and Cross-sectional Determinants of Earnings Response Coefficients", Journal of Accounting and Econgmics 11, 1989, pp. 143-181. De Bondt, Werner F. M. and Richard H. Thaler, "Do Security Analysts Overreact?", AEA Papers and Proceedings, Vol. 80, No. 2, May 1990, pp. 52-57. DeBondt, Werner F. M. and Richard Thaler, "Does the Stock Market Overreact?", Journal of Finance, July 1985, pp. 793- 808. -Dimson, Elroy, "Risk Measurement When Shares Are Subject to Infrequent Trading," Jonrnai of Finangial Ecgngmigs 7, 1979, pp. 197-226. Dowen, Richard J. and W. Scott Bauman, "Revisions in Corporate Earnings Forecasts and Common Stock Returns", Northern Illinois University working paper, 1989. Fowler, David J. and C. Harvey Rorke, "Risk Measurement When Shares Are Subject to Infrequent Trading:Comment", Jgurnai of Financial Economics 12, 1983, pp. 279-283. Givoly, Dan, "The Formation of Earnings Expectations", Accounring Review, July 1985, pp. 372-386. Givoly, Dan and Josef Lakonishok, "The Information Content of Financial Analysts’ Forecasts of Earnings: Some Evidence on Semi-Strong Efficiency", Journa; of Accounting and Eggngmis§_1. 1979. pp- 165-185. 89 Givoly, Dan and Josef Lakonishok, "The Quality of Analysts' Forecasts of Earnings", Financial Anal sts Journal, September-October 1984, pp. 40-47. Harris, Robert S. and Mustafa Gultekin, "Financial Analysts' Forecasts of Corporate Earnings Growth", University of North Carolina Working Paper, 1987. Hawkins, Eugene H., Stanley C. Chamberlin and Wayne E. Daniel, "Earnings Expectations and Security Prices", Finansial_Analxsts_Jgurnal. September-October 1984. pp- 24- 38, 74. Kerrigan, Thomas J., "When Forecasting Earnings, It Pays to Watch Forecasts", The Journal of Portfolio Mana ement, Summer 1984, pp. 19-26. Klein, April, "A Direct Test of the Cognitive Bias Theory of Share Price Reversals", Journal of Accounting and Economics 1;. 1990, pp.155-l66. Klemkosky, Robert C. and William P. Miller, "When Forecasting Earnings, It Pays To Be Right", The Journal of Portfolio Management, Summer 1984, pp. 13-18. Niederhoffer, Victor and Patricia J. Regan, "Earnings Changes, Analysts' Forecasts and Stock Prices", Financial Analysts Journal, May-June 1972, pp. 65-71. O'Brien, Patricia C., "Analysts' Forecasts as Earnings Expectations", Journal or Acgounting ang Economics i0, 1988, pp. 53-83. O’Brien, Patricia C., "Forecast Accuracy of Individual Analysts in Nine Industries", Journal of Accounting Research, Autumn 1990, pp. 286-304. Rendleman, R.J., C.P. Jones and H.A. Latane, "Empirical Anomalies Based on Unexpected Earnings and the Importance of Risk Adjustments", gournal of Financial Economigs l0, 1982, pp. 269-287. Scholes, Myron and Joseph Williams, "Estimating Betas from Nonsynchronous Data", Jgnrnal of Financial Economics 5, 1977, pp. 309-327. Simonds, Richard R., Kirt C. Butler and Michael D. Atchison, "Nonsynchronous Trading and OLS Beta Bias", Working Paper, 1993. APPENDICES APPENDIX A INDIVIDUAL ANALYST FORECASTS OF 1986 EARNINGS PER SHARE FOR THIRTEEN COMPANIES (PILOT STUDY) 90 APPENDIX A Table 24 I/B/E/S Analyst Forecasts for Abbott Laboratories (1986) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 2824 308 4.45 11686 21386 12 2824 308 4.45 11686 41786 12 2824 308 2.20 70886 71086 12 2824 308 2.20 70886 81486 12 2824 308 2.20 70886 111386 12 2824 308 2.20 70886 111386 12 2824 363 4.50 22786 32086 12 2824 363 4.60 41086 41786 12 2824 363 2.35 51286 71786 12 2824 363 2.35 51286 81486 12 2824 392 4.48 21186 21386 12 2824 392 4.50 31286 41786 12 2824 392 2.31 51486 71086 12 2824 392 2.31 51486 81486 12 2824 392 2.32 90986 111386 12 2824 392 2.32 90986 111386 12 2824 397 4.50 22586 22786 12 2824 397 2.35 51486 71086 12 2824 397 2.35 51486 81486 12 2824 397 2.35 51486 112086 12 2824 397 2.35 51486 121186 12 2824 535 4.70 41686 41786 12 2824 535 2.35 41686 71786 12 2824 535 2.35 41686 81486 12 2824 535 2.35 41686 111986 12 2824 535 2.35 41686 121886 12 2824 543 4.50 12886 21386 12 2824 543 4.50 12886 32086 12 2824 543 4.50 12886 41786 12 2824 543 2.33 70986 71786 12 2824 543 2.32 72486 81486 12 2824 543 2.32 72486 112086 12 2824 543 2.32 72486 121186 12 2824 550 2.27 60486 61986 12 2824 550 2.27 60486 61986 12 2824 550 2.35 82186 102386 12 2824 550 2.32 120886 121186 12 2824 597 2.35 52986 71086 12 2824 597 2.35 52986 73186 12 2824 597 2.35 52986 103086 12 2824 597 2.35 52986 121886 12 2824 891 4.50 10986 21386 12 2824 891 4.55 41686 41786 12 91 APPENDIX A Table 24 (cont'd) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 2824 891 2.35 51286 71086 12 2824 891 2.35 51286 81486 12 2824 891 2.35 51286 111386 12 2824 891 2.35 51286 111386 12 2824 975 2.30 52286 71086 12 2824 975 2.30 52286 81486 12 2824 975 2.30 52286 111386 12 2824 975 2.30 52286 111386 12 2824 1023 2.30 80586 80786 12 2824 1422 2.25 51486 71086 12 2824 1422 2.30 81386 81486 12 2824 1422 2.30 81386 112086 12 2824 1422 2.30 81386 112086 12 2824 1738 2.25 61886 71786 12 2824 1738 2.30 81286 81486 12 2824 1738 2.45 91786 112086 12 2824 1738 2.30 81286 121886 12 2824 1775 4.35 40986 41086 12 2824 1775 2.25 71586 71786 12 2824 1775 2.25 71586 71786 12 2824 1775 2.25 71586 111386 12 2824 1775 2.25 71586 121186 12 2824 1826 2.35 51286 112086 12 2824 1826 2.35 51286 121886 12 2824 1974 4.45 21286 21386 12 2824 1974 4.45 21286 32086 12 2824 1974 4.55 32786 41086 12 2824 1974 2.35 51286 71786 12 2824 1974 2.35 51286 80786 12 2824 1974 2.30 111386 111386 12 2824 1974 2.30 111386 111886 12 2824 2026 2.30 70286 71786 12 2824 2026 2.30 70286 81486 12 2824 2026 2.30 70286 111386 12 2824 2026 2.30 70286 121886 12 2824 2066 2.25 42986 71786 12 2824 2079 4.50 31986 31986 12 2824 2079 4.50 31986 31986 12 2824 2079 2.25 31986 61286 12 2824 2079 2.25 31986 73186 12 2824 2133 2.27 51486 71086 12 2824 2133 2.27 51486 81486 12 2824 2133 2.27 51486 111386 12 2824 2133 2.27 51486 121886 12 92 APPENDIX A Table 24 (cont'd) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 2824 2156 4.45 12886 22086 12 2824 2156 4.45 12886 41786 12 2824 2156 2.30 71586 71786 12 2824 2156 2.30 71586 81486 12 2824 2156 2.30 71586 112086 12 2824 2156 2.30 71586 121886 12 2824 2162 4.45 40386 41086 12 2824 2162 2.33 60586 71786 12 2824 2162 2.33 60586 80786 12 2824 2162 2.30 102286 112086 12 2824 2162 2.30 102286 121186 12 2824 2227 4.40 40986 41785 12 2824 2227 2.25 50886 71086 12 2824 2227 2.35 81286 81486 12 2824 2227 2.35 81286 81486 12 2824 2227 2.35 81286 121186 12 2824 2240 2.35 71086 71786 12 2824 2240 2.35 71086 80786 12 2824 2240 2.35 71086 110586 12 2824 2240 2.35 71086 121886 12 2824 2280 2.25 71686 71786 12 2824 2280 2.25 71686 81486 12 2824 2280 2.30 111986 112086 12 2824 2280 2.30 111986 121286 12 2824 2283 2.25 42986 81486 12 2824 2283 2.35 90986 112086 12 2824 2283 2.35 90986 121886 12 2824 2379 4.40 12086 13086 12 2824 2379 4.40 12086 32086 12 2824 2379 4.40 12086 32086 12 2824 2379 2.20 12086 62686 12 2824 2379 2.20 12086 62686 12 2824 2379 2.30 82886 103086 12 2824 2379 2.30 82886 103086 12 2824 2435 4.50 20586 20586 12 2824 2435 4.50 20586 31386 12 2824 2435 4.60 41586 41786 12 2824 2435 2.35 61186 71086 12 2824 2435 2.33 80786 80786 12 2824 2435 2.32 101586 111386 12 2824 2435 2.32 101586 121886 12 2824 2442 4.50 10986 21386 12 2824 2442 4.50 10986 31386 12 2824 2442 4.50 10986 41786 12 93 APPENDIX A Table 24 (cont’d) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 2824 2442 2.30 71686 71786 12 2824 2442 2.30 71686 81486 12 2824 2442 2.30 71686 112086 12 2824 2442 2.30 71686 121886 12 2824 2465 4.50 41086 41786 12 2824 2465 2.25 41086 71786 12 2824 2465 2.25 41086 71786 12 2824 2465 2.30 90986 100286 12 2824 2465 2.30 90986 100286 12 2824 2534 4.55 20486 22086 12 2824 2534 4.60 31386 31386 12 2824 2534 4.60 31386 41786 12 2824 2534 2.35 42286 71786 12 2824 2534 2.35 42286 80786 12 2824 2534 2.35 111986 112086 12 2824 2534 2.35 111986 121886 12 2824 2574 2.30 80586 80786 12 2824 2574 2.30 80586 110686 12 2824 2574 2.30 80586 110686 12 2824 2577 4.48 21386 21386 12 2824 2577 4.47 22786 32086 12 2824 2577 4.47 22786 41086 12 2824 2577 2.27 50886 71786 12 2824 2577 2.27 50886 80786 12 2824 2577 2.27 50886 111386 12 2824 2577 2.27 50886 121886 12 2824 2618 4.45 21986 22086 12 2824 2618 4.45 21986 32086 12 2824 2618 4.45 21986 41786 12 2824 2618 2.32 51386 71786 12 2824 2618 2.32 51386 81486 12 2824 2618 2.32 51386 112086 12 2824 2618 2.30 121786 121886 12 2824 2679 4.43 21286 22086 12 2824 2679 4.43 21286 32086 12 2824 2679 4.43 21286 41086 12 2824 2679 2.22 51386 71786 12 2824 2739 2.30 61086 71786 12 2824 2739 2.30 61086 72486 12 2824 2739 2.30 61086 100986 12 2824 2739 2.30 61086 121886 12 2824 2809 4.50 21486 21486 12 2824 2809 4.50 21486 21486 12 2824 2809 2.32 50986 50986 12 94 APPENDIX A Table 24 (cont'd) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 2824 2809 2.32 50986 80886 12 2824 2809 2.32 50986 80886 12 2824 2809 2.32 50986 80886 12 2824 2822 4.50 30686 32086 12 2824 2822 4.50 30686 41786 12 2824 2822 2.32 42286 71786 12 2824 2822 2.35 82886 100986 12 2824 2822 2.35 82886 100986 12 2824 2881 2.33 72386 112086 12 2824 2881 2.33 72386 121886 12 2824 2997 4.48 12286 22086 12 2824 2997 4.48 12286 32086 12 2824 2997 4.48 12286 41786 12 2824 2997 2.33 61886 71786 12 2824 2997 2.33 61886 81486 12 2824 2997 2.33 61886 112086 12 2824 2997 2.33 61886 121886 12 2824 3012 4.50 22586 32086 12 2824 3012 4.50 22586 41086 12 2824 3012 2.35 42486 71786 12 2824 3012 2.35 42486 80786 12 2824 3012 2.35 42486 111386 12 2824 3012 2.35 42486 121886 12 2824 3057 4.60 12386 21386 12 2824 3057 4.60 12386 31386 12 2824 3057 4.60 12386 41086 12 2824 3057 2.38 51286 70286 12 2824 3057 2.38 51286 70286 12 2824 3057 2.35 81986 111386 12 2824 3057 2.35 81986 121886 12 2824 3085 4.35 21386 22086 12 2824 3085 4.50 31386 31386 12 2824 3085 4.50 31386 41786 12 2824 3085 2.25 31386 71086 12 2824 3085 2.25 31386 71086 12 2824 3085 2.25 31386 111386 12 2824 3085 2.25 31386 120486 12 2824 3200 4.55 11486 20686 12 2824 3200 4.55 11486 32086 12 2824 3200 4.55 11486 41786 12 .2824 3200 2.35 51286 71786 12 2824 3200 2.33 73186 81486 12 2824 3200 2.33 73186 112086 12 2824 3200 2.33 73186 121886 12 95 Table 24 (cont'd) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 2824 3337 2.32 51386 71786 12 2824, 3337 2.30 81386 81486 12 2824 3337 2.35 111386 111386 12 2824 3337 2.35 111386 121886 12 2824 4157 4.50 40286 41786 12 2824 4157 2.30 70186 71786 12 2824 4157 2.30 70186 81486 12 2824 4409 2.25 92586 40386 12 2824 4409 2.25 92586 40386 12 2824 4489 4.30 22086 32086 12 2824 4489 2.15 22086 60586 12 2824 4489 2.15 22086 60586 12 2824 4489 2.15 22086 60586 12 2824 4489 2.15 22086 60586 12 2824 4644 4.40 20686 20686 12 2824 4644 4.50 22686 32086 12 2824 4644 4.50 22686 41786 12 2824 4644 2.32 42286 71786 12 2824 4644 2.33 72386 81486 12 2824 4760 4.55 41086 41086 12 2824 4760 2.27 41086 71786 12 2824 4760 2.27 41086 71786 12 2824 4760 2.32 103086 111386 12 2824 4760 2.32 103086 111386 12 2824 5038 4.50 41086 41786 12 2824 5082 4.30 21786 22086 12 2824 5083 4.45 20686 20686 12 2824 5083 4.45 20686 20686 12 2824 5083 4.45 20686 20686 12 2824 5083 2.22 20686 71786 12 2824 5304 2.32 51386 61286 12 2824 5304 2.35 81386 81486 12 2824 5367 2.30 101586 112086 12 2824 5367 2.35 112686 121886 12 2824 5368 2.22 20686 71786 12 2824 5368 2.30 103086 112086 12 2824 5368 2.30 103086 121186 12 2824 5421 2.30 70186 112086 12 2824 5421 2.30 70186 121886 12 Forecasts are sorted by analyst number. The forecast date is self-reported by the analyst as the date on which the forecast was made. The I/B/E/S date is the date on which the forecast was entered in the database. 96 APPENDIX A Table 25 I/B/E/S Analyst Forecasts for Adams Express (1986) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 6212 2864 0.70 32686 71786 12 6212 2864 0.70 32686 81486 12 6212 2864 0.67 91986 112086 12 6212 2864 0.67 91986 121886 12 Forecasts are sorted by analyst number. The forecast date is self-reported by the analyst as the date on which the forecast was made. The I/B/E/S date is the date on which the forecast was entered in the database. Table 26 I/B/E/S Analyst Forecasts for Adams-Millis (1986) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 6284 2311 2.70 12386 12386 12 6284 2311 2.90 31986 32086 12 6284 2311 2.90 31986 32786 12 6284 2311 2.85 71686 71786 12 6284 2311 2.85 71686 71786 12 6284 2311 2.40 111986 112086 12 6284 2311 1.17 121686 121886 12 6284 4587 2.25 21986 22086 12 6284 4587 2.25 21986 32086 12 6284 4587 2.25 21986 41786 12 6284 4587 2.60 61286 71786 12 6284 5369 2.60 61286 81486 12 6284 5369 2.65 90986 112086 12 6284 5369 1.20 121586 121886 12 Forecasts are sorted by analyst number. The forecast date is self-reported by the analyst as the date on which the forecast was made. The I/B/E/S date is the date on which the forecast was entered in the database. 97 APPENDIX A Table 27 I/B/E/S Analyst Forecasts for Adobe Resources (1986) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 7240 1308 0.68 121686 121886 12 7240 1508 -1.00 11486 11586 12 7240 1508 -1.00 11486 31386 12 7240 1508 -1.00 11486 41086 12 7240 1508 0.50 11486 71786 12 7240 1508 0.50 11486 71786 12 7240 1508 -0.70 81986 111386 12 7240 1508 -0.70 81986 121186 12 7240 2505 -2.00 21386 21386 12 7240 2505 -2.00 21386 31386 12 7240 2505 -2.00 21386 41786 12 7240 3083 0.25 70286 71786 12 7240 3083 0.25 70286 81486 12 7240 3083 -0.05 92486 112086 12 7240 3083 -0.20 121086 121886 12 7240 3471 -0.45 82186 112086 12 7240 3471 -0.45 82186 121886 12 7240 4167 -0.95 61886 71786 12 7240 4167 -0.95 61886 81486 12 7240 5425 -1.00 91786 112086 12 7240 5425 -1.10 121686 121886 12 Forecasts are sorted by analyst number. The forecast date is self-reported by the analyst as the date on which the forecast was made. The I/B/E/S date is the date on which the forecast was entered in the database. 98 APPENDIX A Table 28 I/B/E/S Analyst Forecasts for ADT, Inc (1986) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 1007 748 1.80 111986 112086 12 1007 748 1.80 111986 121886 12 1007 1958 1.65 101586 111386 12 1007 1958 1.65 101586 121886 12 1007 2224 1.45 51486 71786 12 1007 2224 1.45 51486 80786 12 1007 2224 1.55 111586 112086 12 1007 2224 1.55 111586 121886 12 1007 3018 1.40 21286 71786 12 1007 3018 1.40 21286 81486 12 1007 3018 1.40 21286 112086 12 1007 3018 1.60 120486 121886 12 1007 4423 2.00 92586 40386 12 1007 4423 2.00 92586 40386 12 1007 5192 1.50 ‘51386 71786 12 1007 5382 1.50 51386 81486 12 1007 5382 1.50 51386 112086 12 1007 5382 1.50 51386 121886 12 Forecasts are sorted by analyst number. The forecast date is self-reported by the analyst as the date on which the forecast was made. The I/B/E/S date is the date on which the forecast was entered in the database. 99 APPENDIX A Table 29 I/B/E/S Analyst Forecasts for Affiliated Publications (1986) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 8261 595 1.85 61286 71086 12 8261 595 1.85 61286 81486 12 8261 595 1.90 91186 112086 12 8261 595 1.90 91186 121186 12 8261 699 1.90 20686 21386 12 8261 699 1.85 31386 32086 12 8261 699 1.85 31386 41786 12 8261 699 1.80 70986 71786 12 8261 699 1.90 81286 81486 12 8261 699 2.05 111386 112086 12 8261 699 1.95 121086 121186 12 8261 2288 2.05 22586 32086 12 8261 2288 2.05 22586 32086 12 8261 2288 1.70 52986 62686 12 8261 2288 1.70 52986 72486 12 8261 2288 1.86 91186 101686 12 8261 2288 2.10 112686 121886 12 8261 2441 1.90 21286 21386 12 8261 2441 1.90 21286 31186 12 8261 2441 1.90 21286 41086 12 8261 2441 1.90 21286 71786 12 8261 2441 1.90 21286 81486 12 8261 2441 1.90 21286 111386 12 8261 2441 1.90 21286 121886 12 8261 2460 1.80 11486 20686 12 8261 2460 1.90 31986 32086 12 8261 2460 1.90 31986 41786 12 8261 2460 1.90 31986 71786 12 8261 2460 2.00 80586 81486 12 8261 2460 2.00 80586 112086 12 8261 2460 2.00 80586 121886 12 8261 2508 1.80 60486 61986 12 8261 2508 1.80 60486 61986 12 8261 2508 1.95 102086 102386 12 8261 2508 2.20 120886 121186 12 8261 3217 1.65 22586 32086 12 8261 3217 1.65 22586 41786 12 8261 3217 1.65 62486 71786 12 8261 3217 1.95 80786 80786 12 8261 3217 2.03 82886 110686 12 8261 3217 2.10 120886 121886 12 8261 3531 1.75 40386 40986 12 8261 3531 1.80 62686 71786 12 100 APPENDIX A Table 29 (cont'd) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 8261 3531 1.80 62686 81486 12 8261 3531 2.00 92586 111386 12 8261 3531 2.60 121686 121886 12 8261 4193 1.70 31886 32086 12 8261 4193 1.70 31886 41786 12 8261 4193 1.75 61886 71786 12 8261 4193 1.75 61886 81486 12 8261 4193 1.95 91786 112086 12 8261 4193 2.10 121686 121886 12 8261 5428 1.70 101786 111386 12 8261 5428 1.70 101786 111386 12 Forecasts are sorted by analyst number. The forecast date is self-reported by the analyst as the date on which the forecast was made. The I/B/E/S date is the date on which the forecast was entered in the database. 101 APPENDIX A Table 30 I/B/E/S Analyst Forecasts for AFG (1986) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 1054 1361 1.90 71086 71086 12 1054 1361 1.90 71086 81486 12 1054 1361 1.90 71086 112086 12 1054 1361 1.90 71086 121186 12 1054 6451 1.75 51486 71086 12 1054 6451 1.95 81386 81486 12 1054 6451 1.85 111186 111386 12 1054 6451 1.85 111186 111386 12 1054 7481 2.35 11486 21386 12 1054 7481 2.35 11486 32086 12 1054 7481 1.57 11486 41786 12 1054 7481 2.00 71686 71686 12 1054 7481 2.11 71686 81486 12 1054 7481 2.11 71686 112086 12 1054 7481 2.11 71686 121886 12 1054 8661 1.43 60486 61986 12 1054 8661 2.43 60486 61986 12 1054 8661 2.27 120886 121186 12 1054 12111 2.50 13086 22086 12 1054 12111 2.50 13086 32086 12 1054 12111 1.67 13086 41786 12 1054 12111 2.00 61886 71786 12 1054 12111 2.00 61886 81486 12 1054 12111 2.00 61886 112086 12 1054 12111 2.00 61886 121886 12 1054 13611 2.40 22786 22786 12 1054 13611 1.60 22786 22786 12 1054 13611 2.00 70286 70286 12 1054 13611 2.00 70286 70286 12 1054 13611 1.95 91786 102386 12 1054 14251 2.00 120486 120486 12 1054 16351 2.35 31386 31386 12 1054 16351 1.57 31386 41786 12 1054 16351 1.80 51486 71086 12 1054 16351 1.80 51486 71086 12 1054 16351 1.80 51486 111386 12 1054 16351 1.80 51486 120486 12 1054 18851 2.30 21386 21386 12 1054 18851 2.30 21386 21386 12 1054 18851 1.53 21386 21386 12 1054 19401 1.58 41086 41086 12 1054 19401 1.64 51386 71786 12 1054 19401 1.85 81386 81486 12 102 APPENDIX A Table 30 (cont'd) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 1054 19401 1.95 111386 111386 12 1054 19401 2.00 112686 121886 12 1054 22821 1.75 71186 71186 12 1054 22821 1.80 80886 80886 12 1054 22821 1.80 80886 80886 12 1054 22821 1.80 80886 80886 12 1054 23781 2.40 31986 32086 12 1054 23781 1.69 31986 41786 12 1054 23781 1.74 52186 62686 12 1054 23781 2.00 72486 72486 12 1054 24941 1.65 62686 62686 12 1054 24941 1.65 62686 62686 12 1054 24941 2.00 82086 112086 12 1054 24941 2.00 82086 121886 12 1054 25751 2.00 121186 121186 12 1054 26101 2.25 21286 21386 12 1054 26101 2.25 21286 32086 12 1054 26101 1.60 41086 41086 12 1054 26101 1.80 71586 71786 12 1054 26101 1.80 71586 80786 12 1054 26101 1.95 111386 111386 12 1054 26101 1.95 111386 111886 12 1054 26341 1.90 62686 71786 12 1054 26341 1.95 80586 81486 12 1054 26341 2.10 110586 112086 12 1054 26341 2.10 110586 121886 12 1054 31521 1.80 61786 61986 12 1054 31521 1.80 61786 61986 12 1054 33451 1.53 31386 31386 12 1054 33451 2.30 31386 31386 12 1054 33451 1.75 70986 71086 12 1054 33451 1.75 70986 71086 12 1054 33451 2.20 101586 101686 12 1054 33451 2.20 101586 112686 12 1054 33661 2.00 61886 71786 12 1054 33661 2.00 61886 80786 12 1054 33661 2.08 102286 110586 12 1054 33661 2.08 102286 121886 12 1054 33901 1.95 70186 71786 12 1054 33901 1.95 70186 81486 12 1054 33901 1.95 70186 112086 12 103 APPENDIX A Table 30 (cont'd) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 1054 33901 2.11 112686 121886 12 1054 54291 2.00 111986 112086 12 1054 54291 2.00 111986 112086 12 Forecasts are sorted by analyst number. The forecast date is self-reported by the analyst as the date on which the forecast was made. The I/B/E/S date is the date on which the forecast was entered in the database. 104 APPENDIX A Table 31 I/B/E/S Analyst Forecasts for AGS Computers (1986) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 1240 149 1.45 21286 21386 12 1240 149 1.75 31886 32086 12 1240 149 1.75 31886 41786 12 1240 643 1.45 10986 21386 12 1240 643 1.85 10986 32086 12 1240 643 1.75 40186 41786 12 1240 643 1.65 50886 71786 12 1240 643 1.65 50886 81486 12 1240 643 1.65 50886 112086 12 1240 643 1.65 50886 121186 12 1240 748 1.75 31886 70286 12 1240 748 1.75 31886 81486 12 1240 748 1.75 31886 112086 12 1240 748 1.60 120886 121886 12 1240 1276 1.60 102386 112086 12 1240 1276 1.60 102386 121886 12 1240 1403 1.75 31986 32086 12 1240 1403 1.75 31986 41786 12 1240 1403 1.75 31986 52986 12 1240 1403 1.75 31986 80786 12 1240 1403 1.65 111986 112086 12 1240 1403 1.65 111986 120886 12 1240 1493 1.55 100986 112086 12 1240 1493 1.55 100986 121886 12 1240 1985 1.75 31286 41786 12 1240 1985 1.60 50186 71086 12 1240 1985 1.55 80586 80786 12 1240 1985 1.55 80586 112086 12 1240 1985 1.55 80586 121886 12 1240 2280 1.45 11386 21386 12 1240 2280 1.75 21286 31386 12 1240 2280 1.75 21286 41786 12 1240 2280 1.60 61086“ 71786 12 1240 2280 1.60 61086 72486 12 1240 2280 1.60 61086 100986 12 1240 2280 1.60 61086 121886 12 1240 2911 1.70 62686 71786 12 1240 2911 1.70 62686 81486 12 1240 3318 1.50 13086 22086 12 1240 3318 1.65 80586 81486 12 1240 5306 1.65 61786 61986 12 1240 5306 1.65 61786 61986 12 105 APPENDIX A Table 31 (cont'd) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 1240 5306 1.65 61786 102386 12 1240 5306 1.65 61786 121186 12 1240 5387 1.60 92386 112086 12 1240 5387 1.60 92386 121886 12 Forecasts are sorted by analyst number. The forecast date is self-reported by the analyst as the date on which the forecast was made. The I/B/E/S date is the date on which the forecast was entered in the database. 106 APPENDIX A Table 32 IIB/E/S Analyst Forecasts for Ahmanson (H F) & Co (1986) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 8677 287 9.30 31286 41786 12 8677 287 3.00 60586 71086 12 8677 287 3.00 60586 81486 12 8677 287 3.20 90986 111386 12 8677 287 3.20 90986 111386 12 8677 377 8.00 12486 13086 12 8677 377 3.08 40986 41086 12 8677 377 9.25 40986 41086 12 8677 377 3.08 40986 81486 12 8677 377 3.08 40986 111986 12 8677 377 3.08 40986 111986 12 8677 581 10.00 20686 21386 12 8677 581 10.00 20686 32086 12 8677 581 10.00 20686 41786 12 8677 581 3.35 70986 71786 12 8677 581 3.35 70986 81486 12 8677 581 3.35 70986 112086 12 8677 581 3.35 70986 121186 12 8677 981 9.00 10986 21386 12 8677 981 9.00 10986 41786 12 8677 981 3.10 70886 71086 12 8677 981 3.10 70886 81486 12 8677 981 3.20 100986 111386 12 8677 981 3.20 100986 111386 12 8677 1100 3.15 120886 121186 12 8677 1123 9.65 20686 21386 12 8677 1123 9.65 20686 32086 12 8677 1123 3.17 42386 71786 12 8677 1123 3.20 81386 81486 12 8677 1123 3.25 100186 111386 12 8677 1123 3.25 100186 121886 12 8677 1537 3.00 71686 71786 12 8677 1537 3.00 71686 71786 12 8677 1537 3.35 90486 111386 12 8677 1537 3.35 90486 111386 12 8677 1564 7.55 12086 22086 12 8677 1564 9.00 40186 40386 12 8677 1564 2.95 71586 71786 12 8677 1564 2.95 71586 80786 12 8677 1564 3.20 92586 110586 12 8677 1564 3.20 92586 121886 12 8677 1689 9.40 30386 32086 12 8677 1689 9.40 30386 41786 12 107 APPENDIX A Table 32 (cont’d) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 8677 1689 3.15 62586 71786 12 8677 1689 3.15 62586 81486 12 8677 1689 3.30 90986 111386 12 8677 1689 3.30 90986 121886 12 8677 2039 3.00 60486 61986 12 8677 2039 3.00 60486 61986 12 8677 2039 3.00 60486 102386 12 8677 2063 7.40 10286 11686 12 8677 2063 7.40 10286 31386 12 8677 2063 8.75 41586 41786 12 8677 2063 3.00 52786 71786 12 8677 2063 3.00 52786 71786 12 8677 2063 3.10 90486 111386 12 8677 2063 3.10 90486 111386 12 8677 2074 3.10 70286 71786 12 8677 2074 3.10 70286 81486 12 8677 2074 3.10 70286 112086 12 8677 2074 3.20 120486 121886 12 8677 2102 3.08 61186 71786 12 8677 2102 3.08 61186 80786 12 8677 2145 3.17 111386 111386 12 8677 2145 3.17 111386 121886 12 8677 2168 9.20 22786 31386 12 8677 2168 9.00 41686 41786 12 8677 2168 3.25 71086 71786 12 8677 2168 3.25 71086 80786 12 8677 2168 3.20 91186 112086 12 8677 2168 3.20 91186 121886 12 8677 2209 9.50 21986 22086 12 8677 2209 9.50 21986 32086 12 8677 2209 9.50 21986 41786 12 8677 2209 3.10 71686 71786 12 8677 2209 3.15 73186 81486 12 8677 2209 3.15 73186 112086 12 8677 2209 3.15 73186 121886 12 8677 2238 4.90 31386 41086 12 8677 2238 1.63 31386 71086 12 8677 2238 3.00 81386 81486 12 8677 2238 3.20 111986 112086 12 8677 2238 3.20 111986 112086 12 8677 2659 3.35 61886 81486 12 8677 2659 3.35 61886 91886 12 8677 2659 3.20 112686 112686 12 8677 2763 2.67 50886 71086 12 8677 2763 2.67 50886 81486 12 108 APPENDIX A Table 32 (cont’d) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 8677 2763 2.67 50886 81486 12 8677 2886 10.00 31886 32086 12 8677 2886 10.00 31886 40286 12 8677 2886 3.17 70286 71786 12 8677 2886 3.20 73086 81486 12 8677 2886 3.20 73086 112086 12 8677 2886 3.20 73086 121886 12 8677 3088 9.75 10786 22086 12 8677 3088 9.75 10786 32086 12 8677 3088 9.75 10786 41786 12 8677 3088 3.25 10786 71786 12 8677 3088 3.25 10786 81486 12 8677 3088 3.25 111686 112086 12 8677 3088 3.25 111686 121886 12 8677 3157 8.50 10986 41786 12 8677 3157 3.00 61986 71086 12 8677 3157 3.35 81386 81486 12 8677 3157 3.35 81386 111386 12 8677 3157 3.35 81386 111386 12 8677 3362 7.70 10286 21386 12 8677 3362 8.00 22786 32086 12 8677 3362 8.00 22786 40986 12 8677 3362 3.00 42586 71786 12 8677 3362 3.00 42586 81486 12 8677 3362 3.15 91786 111386 12 8677 3362 3.15 91786 121886 12 8677 3414 3.15 71086 71086 12 8677 3414 3.15 71086 81486 12 8677 3414 3.40 111986 111986 12 8677 3414 3.40 111986 121886 12 8677 3478 8.50 10986 21386 12 8677 3493 3.20 121686 121686 12 8677 3549 8.75 11486 22086 12 8677 3549 9.00 22786 32086 12 8677 4188 9.15 10986 20586 12 8677 4188 9.15 10986 31386 12 8677 4188 9.15 10986 41786 12 8677 4188 3.12 61186 71086 12 8677 4188 3.20 80786 80786 12 8677 4188 3.25 111386 111386 12 8677 4188 3.15 121786 121886 12 8677 4415 8.50 12786 22086 12 8677 4415 3.08 30586 40386 12 8677 4415 3.08 30586 40386 12 109 APPENDIX A Table 32 (cont'd) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 8677 4415 3.08 30586 40386 12 8677 4415 9.25 30586 40386 12 8677 4415 3.08 30586 40386 12 8677 4645 9.25 12386 20686 12 8677 4645 9.25 12386 31386 12 8677 4645 9.00 41686 41786 12 8677 4645 3.00 41686 71786 12 8677 4645 3.00 41686 81486 12 8677 4645 3.20 92586 112086 12 8677 4645 3.20 92586 120486 12 8677 4836 10.00 21986 22086 12 8677 4836 10.00 21986 32086 12 8677 4836 10.00 21986 41786 12 8677 4913 8.60 41686 41786 12 8677 5116 3.20 70286 71786 12 8677 5116 3.20 70286 81486 12 8677 5279 3.35 52286 71786 12 8677 5279 3.35 52286 81486 12 8677 5279 3.25 111986 112086 12 8677 5279 3.25 111986 121886 12 8677 5312 3.10 70286 71786 12 8677 5312 3.10 70286 81486 12 8677 5312 3.15 102386 112086 12 8677 5312 3.15 102386 121886 12 8677 5313 3.35 61886 71786 12 Forecasts are sorted by analyst number. The forecast date is self-reported by the analyst as the date on which the forecast was made. The I/B/E/S date is the date on which the forecast was entered in the database. 110 APPENDIX A Table 33 I/B/E/S Analyst Forecasts for Airborne Freight (1986) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 9266 784 1.90 22786 32086 12 9266 784 1.90 22786 41786 12 9266 784 1.34 62686 71786 12 9266 784 1.40 80586 81486 12 9266 784 1.85 110686 112086 12 9266 784 1.85 110686 121186 12 9266 1010 2.00 11386 21386 12 9266 1010 2.00 11386 41786 12 9266 1010 1.50 51486 71086 12 9266 1010 1.50 51486 81486 12 9266 1010 1.50 51486 111386 12 9266 1010 1.50 51486 111386 12 9266 1067 0.15 82186 102386 12 9266 1152 1.85 12386 31386 12 9266 1152 1.85 12386 41086 12 9266 1152 1.70 60586 70286 12 9266 1152 1.70 60586 70286 12 9266 1152 1.65 110686 111386 12 9266 1152 1.85 121186 121886 12 9266 1180 1.55 40386 41786 12 9266 1423 1.70 110686 112086 12 9266 1423 1.70 110686 121886 12 9266 1488 1.55 40386 71086 12 9266 1488 1.55 40386 81486 12 9266 1488 1.75 111686 112086 12 9266 1488 1.75 111686 121886 12 9266 1516 1.94 32486 41086 12 9266 1516 1.37 61886 71786 12 9266 1516 1.37 61886 73186 12 9266 1516 1.70 111986 112086 12 9266 1516 1.70 111986 121886 12 9266 1557 1.75 31986 31986 12 9266 1557 1.75 31986 41786 12 9266 1557 1.00 50686 71786 12 9266 1557 1.00 50686 80786 12 9266 1557 1.40 102386 111386 12 9266 1557 1.40 102386 121186 12 9266 1627 1.75 111986 121786 12 9266 1952 1.32 70986 71786 12 9266 1952 1.32 70986 80786 12 9266 1952 1.65 111086 111386 12 9266 1952 1.75 120486 121886 12 9266 2182 1.75 120486 120486 12 111 APPENDIX A Table 33 (cont'd) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 9266 2305 1.95 21386 21386 12 9266 2305 1.95 21386 32086 12 9266 2305 1.80 41686 41786 12 9266 2305 1.45 52186 62686 12 9266 2305 1.45 52186 72486 12 9266 2305 1.80 111386 111189 12 9266 2305 1.80 111386 111386 12 9266 2385 1.85 12386 21386 12 9266 2446 1.65 111986 112086 12 9266 2446 1.65 111986 121886 12 9266 2539 2.00 21786 22086 12 9266 2539 2.00 21786 41086 12 9266 2539 2.00 21786 71786 12 9266 2539 2.00 21786 71786 12 9266 2539 2.00 21786 111386 12 9266 2539 2.00 21786 111386 12 9266 2594 1.85 21286 22086 12 9266 2594 1.85 21286 30686 12 9266 2594 1.85 21286 41786 12 9266 2594 1.60 50886 71786 12 9266 2594 1.60 50886 81486 12 9266 2594 1.65 111586 111586 12 9266 2594 1.65 111586 121886 12 9266 2814 2.00 31386 31386 12 9266 2814 2.00 31386 41086 12 9266 2814 1.25 61086 71786 12 9266 2814 1.25 61086 81486 12 9266 2814 1.40 101486 111386 12 9266 2814 1.75 121786 121886 12 9266 2967 1.80 20586 21386 12 9266 2967 1.80 20586 31386 12 9266 2967 1.60 41086 41786 12 9266 2967 0.45 71686 71786 12 9266 2967 0.45 71686 81486 12 9266 2967 1.75 111686 112086 12 9266 2967 1.75 111686 121886 12 9266 3266 2.10 12986 12986 12 9266 3266 2.10 12986 31386 12 9266 3266 1.65 41586 41786 12 9266 3266 1.10 52786 71786 12 9266 3266 1.10 52786 71786 12 9266 3266 1.10 52786 111386 12 9266 3266 1.10 52786 111386 12 9266 3303 1.55 31286 41786 12 112 APPENDIX A Table 33 (cont’d) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 9266 3303 1.40 71086 71086 12 9266 3303 1.40 71086 81486 12 9266 3303 1.50 111186 111386 12 9266 3303 1.50 111186 111386 12 9266 3386 1.50 71086 71086 12 9266 3386 1.35 81286 81486 12 9266 3386 1.35 81286 81486 12 9266 3386 1.35 81286 121186 12 9266 3607 1.75 12986 22086 12 9266 3607 1.75 12986 31386 12 9266 3607 1.75 12986 41786 12 9266 4496 1.95 20486 22086 12 9266 4496 1.95 20486 32086 12 9266 4496 1.95 20486 41786 12 9266 4496 1.50 61186 71086 12 9266 4496 1.50 80586 80786 12 9266 4496 1.75 111986 111986 12 9266 4497 1.80 20686 20686 12 9266 4497 1.80 20686 32086 12 9266 4497 1.70 32686 41786 12 9266 4497 1.35 62686 71786 12 9266 4754 1.50 21986 22086 12 9266 4754 1.50 21986 32086 12 9266 4754 1.75 40886 41786 12 9266 4754 1.50 70986 71786 12 9266 4856 1.80 20386 20686 12 9266 4856 1.80 20386 22786 12 9266 4856 1.70 32786 32786 12 9266 4856 1.50 70286 70286 12 9266 4856 1.50 70286 70286 12 9266 4856 1.50 70286 102386 12 9266 5053 2.00 21286 22086 12 9266 5053 1.80 22786 32086 12 9266 5053 1.80 22786 41786 12 9266 5053 1.25 61286 71786 12 113 APPENDIX A Table 33 (cont'd) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 9266 5053 1.25 61286 81486 12 9266 5372 1.50 70986 81486 12 9266 5373 1.35 62686 81486 12 9266 5430 1.55 100886 112086 12 9266 5430 1.55 100886 121886 12 Forecasts are sorted by analyst number. The forecast date is self-reported by the analyst as the date on which the forecast was made. The I/B/E/S date is the date on which the forecast was entered in the database. AI 114 APPENDIX A Table 34 I/B/E/S Analyst Forecasts for AMCA International (1986) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 1640 1247 0.40 31386 32086 12 1640 1247 0.40 31386 41786 12 1640 1247 0.40 31386 71786 12 1640 1247 0.40 31386 81486 12 1640 1247 -1.90 111786 111786 12 1640 1247 -1.90 111786 121886 12 1640 1254 0.75 41086 41086 12 1640 1254 0.50 61986 71786 12 1640 1254 0.50 61986 81486 12 1640 1254 0.50 61986 111386 12 1640 1254 0.50 61986 121186 12 1640 1354 0.80 31986 31986 12 1640 1354 0.80 31986 40886 12 1640 1354 0.50 70986 71086 12 1640 1354 0.50 70986 81486 12 1640 1354 -0.50 111086 111086 12 1640 1354 -0.50 111086 121186 12 1640 1526 0.60 50886 71086 12 1640 1526 0.60 50886 81486 12 1640 1526 0.30 91786 91886 12 1640 1526 -0.45 121186 121886 12 1640 1849 0.25 60586 71786 12 1640 1849 0.25 60586 73186 12 1640 1849 0.50 103086 103086 12 1640 1932 -3.36 121186 121186 12 1640 2321 0.60 52786 71786 12 1640 2321 0.60 52786 80786 12 1640 2748 0.72 21386 21386 12 1640 2748 0.21 22786 22786 12 1640 2748 0.22 22786 41786 12 1640 2748 -0.65 121786 121886 12 1640 2898 0.35 12286 12386 12 1640 2898 0.35 12286 12386 12 1640 2898 0.35 12286 12386 12 1640 2898 0.50 41886 61286 12 1640 2898 0.50 41886 80786 12 1640 2898 0.50 41886 102486 12 1640 2898 -3.00 121586 121886 12 1640 2984 0.60 41686 41786 12 1640 2984 0.35 52786 71786 12 1640 2984 0.35 52786 71786 12 1640 2984 -1.07 112686 121886 12 115 APPENDIX A Table 34 (cont’d) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 1640 3393 0.85 13086 21386 12 1640 3393 0.85 13086 32086 12 1640 3393 0.85 13086 41786 12 1640 3393 0.54 60486 71786 12 1640 3393 0.54 60486 73186 12 1640 3393 -0.80 111786 112086 12 1640 3393 -0.80 111786 121886 12 1640 3409 0.40 40386 41786 12 1640 3409 0.40 40386 71786 12 1640 3409 -1.50 80686 81486 12 1640 3409 -2.30 91786 112086 12 1640 3409 -2.30 91786 121886 12 1640 3550 0.65 21986 22086 12 1640 3550 0.65 21986 22086 12 1640 3550 0.65 21986 41786 12 1640 3550 0.25 103086 103086 12 1640 3550 -0.92 121786 121886 12 1640 4592 0.60 22686 32086 12 1640 4592 0.60 22686 41786 12 1640 4592 -0.25 52786 71786 12 1640 4860 0.61 10986 21386 12 1640 4860 0.87 31986 31986 12 1640 4860 0.62 31986 32086 12 1640 4860 0.63 31986 41086 12 1640 5084 0.65 41086 41086 12 1640 5084 0.90 41086 71786 12 1640 5084 0.90 41086 71786 12 1640 5376 -0.25 52786 81486 12 1640 5434 0.20 82886 112086 12 1640 5434 -1.30 112686 121886 12 Forecasts are sorted by analyst number. The forecast date is self-reported by the analyst as the date on which the forecast was made. The I/B/E/S date is the date on which the forecast was entered in the database. 116 APPENDIX A Table 35 I/B/E/S Analyst Forecasts for AMR Corporation (1986) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 1765 191 6.30 11386 21386 12 1765 191 5.25 40386 40386 12 1765 191 5.25 40386 40386 12 1765 191 5.25 40386 40386 12 1765 191 5.25 40386 40386 12 1765 191 5.25 40386 40386 12 1765 191 5.25 41686 41786 12 1765 191 5.00 51486 71086 12 1765 191 4.70 81386 81486 12 1765 191 5.20 72486 81486 12 1765 191 4.70 81386 111386 12 1765 191 4.70 81386 111386 12 1765 191 4.80 92586 112086 12 1765 191 4.75 121786 121886 12 1765 541 7.20 121186 111386 12 1765 3771 6.50 12486 13086 12 1765 3771 6.50 12486 41086 12 1765 3771 6.50 12486 41086 12 1765 7201 6.00 11686 21386 12 1765 7201 5.00 31386 41786 12 1765 7201 5.00 31386 71086 12 1765 7201 5.00 31386 81486 12 1765 7201 5.00 31386 111386 12 1765 7201 5.00 31386 111386 12 1765 7841 5.60 20586 21386 12 1765 7841 5.10 31986 32086 12 1765 7841 5.10 31986 41786 12 1765 7841 3.05 70986 71786 12 1765 7841 4.46 73186 81486 12 1765 7841 4.75 103086 112086 12 1765 7841 4.75 103086 121186 12 1765 10091 4.00 81486 112086 12 1765 10091 4.00 81486 121186 12 1765 10671 5.25 61786 61986 12 1765 10671 5.25 61786 61986 12 1765 10671 4.00 82186 102386 12 1765 11521 5.30 41086 41086 12 1765 11521 5.30 41086 70286 12 1765 11521 5.30 41086 70286 12 1765 11521 4.54 110686 111386 12 1765 11521 4.04 121186 121886 12 1765 11801 5.30 12786 21386 12 1765 11801 5.30 12786 32086 12 117 APPENDIX A Table 35 (cont'd) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 1765 11801 5.30 12786 41786 12 1765 13691 5.05 41086 41786 12 1765 13691 4.42 70286 70286 12 1765 13691 4.42 70286 80786 12 1765 13961 4.47 111986 111986 12 1765 13961 4.47 111986 121886 12 1765 15161 5.30 32486 41086 12 1765 15161 4.84 61886 71786 12 1765 15161 4.00 72986 73186 12 1765 15161 5.00 91786 112086 12 1765 16621 7.10 41586 41786 12 1765 16621 6.70 61186 71086 12 1765 16621 5.70 80786 80786 12 1765 16621 5.70 80786 111386 12 1765 16621 4.40 121786 121886 12 1765 17051 5.05 20586 22086 12 1765 17051 5.05 20586 32086 12 1765 17051 4.95 41686 41786 12 1765 17051 4.54 61286 71786 12 1765 17051 4.80 111986 112086 12 1765 17051 4.80 111986 121886 12 1765 17481 3.53 32786 41086 12 1765 17481 3.53 32786 71786 12 1765 17481 3.53 32786 80786 12 1765 17481 4.54 111386 111386 12 1765 17481 4.54 111386 121886 12 1765 19811 5.55 40186 40386 12 1765 19811 5.55 40186 71786 12 1765 19811 4.24 73186 80786 12 1765 19811 4.60 102886 110586 12 1765 19811 4.60 102886 121886 12 1765 20191 5.30 12986 22086 12 1765 20191 6.06 22786 32086 12 1765 20191 6.06 22786 41786 12 1765 20191 5.05 ’70286 71786 12 1765 20191 4.29 73086 81486 12 1765 20191 4.80 92486 112086 12 1765 20191 4.80 92486 121886 12 1765 20201 7.00 21886 22086 12 1765 22571 6.77 41686 41786 12 1765 22571 5.86 62486 71786 12 1765 22571 5.86 62486 81486 12 1765 22571 5.86 62486 112086 12 1765 22571 5.86 62486 121886 12 118 APPENDIX A Table 35 (cont’d) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 1765 24461 4.54 91186 112086 12 1765 25391 5.55 11086 22086 12 1765 25391 5.55 11086 41086 12 1765 25391 4.39 71686 71786 12 1765 25391 4.39 71686 71786 12 1765 25391 4.19 103086 111386 12 1765 25391 4.19 103086 111386 12 1765 25691 5.05 21286 21386 12 1765 25691 5.05 21286 31386 12 1765 25691 5.05 21286 41786 12 1765 25691 4.85 61086 71786 12 1765 25691 4.85 61086 72486 12 1765 25691 4.85 61086 100986 12 1765 25691 4.85 61086 121886 12 1765 26091 4.22 81386 81486 12 1765 26091 4.22 81386 112086 12 1765 26091 4.22 81386 112086 12 1765 26251 6.06 11586 22086 12 1765 26251 5.55 31986 32086 12 1765 26251 5.55 31986 41786 12 1765 26251 4.04 71086 71786 12 1765 26251 4.04 71086 81486 12 1765 26251 4.04 71086 112086 12 1765 26251 4.04 71086 121886 12 1765 26461 5.30 11586 22086 12 1765 26461 5.30 11586 31386 12 1765 26461 5.05 40286 41786 12 1765 26461 4.29 61886 71786 12 1765 26461 4.29 61886 80786 12 1765 26461 4.29 61886 112086 12 1765 26461 4.29 61886 121886 12 1765 26501 5.55 41686 41786 12 1765 26501 4.54 61886 71786 12 1765 26501 4.54 61886 81486 12 1765 26501 4.04 102986 112086 12 1765 26501 4.04 102986 121886 12 1765 28091 4.00 10986 10986 12 1765 28091 5.00 31486 31486 12 1765 28091 3.50 71186 71186 12 1765 28091 3.00 80886 80886 12 1765 28091 4.20 111486 111486 12 1765 28091 4.00 121286 121286 12 1765 29671 4.54 71686 71786 12 1765 29671 4.54 71686 81486 12 119 APPENDIX A Table 35 (cont'd) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 1765 29671 4.80 101486 112086 12 1765 29671 4.80 101486 121886 12 1765 30421 5.86 52186 71786 12 1765 30421 4.14 80786 80786 12 1765 30421 4.29 110586 112086 12 1765 30421 4.29 110586 121186 12 1765 30801 6.06 41686 41786 12 1765 30801 5.55 61886 71786 12 1765 30801 4.54 81386 81486 12 1765 30801 4.54 81386 112086 12 1765 30971 6.00 41086 41086 12 1765 30971 5.00 51386 71786 12 1765 30981 4.50 72386 81486 12 1765 30981 4.75 111386 111386 12 1765 30981 4.75 111386 121886 12 1765 31381 4.70 12386 22086 12 1765 31381 4.44 32786 41786 12 1765 31381 3.94 71686 71786 12 1765 31381 3.94 71686 81486 12 1765 31381 4.60 102886 112086 12 1765 31381 4.04 121786 121886 12 1765 33031 4.60 21086 21386 12 1765 33031 4.04 41686 41786 12 1765 33031 3.79 61986 71086 12 1765 33031 3.69 81386 81486 12 1765 33031 3.84 100886 111386 12 1765 33031 3.84 100886 111386 12 1765 33731 5.25 62686 71786 12 1765 33731 4.65 72486 81486 12 1765 33731 3.95 112686 121886 12 1765 33861 4.80 71086 71086 12 1765 33861 5.05 81286 81486 12 1765 33861 5.05 81286 81486 12 1765 33861 4.80 121686 121686 12 1765 34551 6.00 13086 22086 12 1765 34551 5.20 31986 32086 12 1765 34551 5.20 31986 41786 12 1765 34551 3.99 61886 71786 12 1765 34551 3.95 81286 81486 12 1765 34551 3.75 111786 111786 12 1765 34551 4.00 121186 121886 12 1765 41611 4.80 10686 22086 12 1765 41611 4.80 10686 32086 12 1765 41611 4.80 10686 41786 12 120 APPENDIX A Table 35 (cont’d) Forecast I/B/E/S Fiscal CUSIP Analyst Forecast . Date Date Year-End 1765 41611 4.00 70986 71786 12 1765 41611 4.00 70986 81486 12 1765 44971 5.80 20686 20686 12 1765 44971 5.86 20686 32086 12 1765 44971 5.86 20686 41786 12 1765 44971 5.45 62686 71786 12 1765 49521 4.80 20686 22086 12 1765 49521 4.80 20686 41086 12 1765 51171 6.00 60486 60586 12 1765 51171 6.00 60486 60586 12 1765 51171 6.00 60486 60586 12 1765 51171 6.00 60486 60586 12 1765 51181 6.00 21386 21386 12 1765 51181 6.00 21386 32086 12 1765 51181 5.71 41686 41786 12 1765 52641 6.00 21886 22086 12 1765 53731 5.20 72386 81486 12 1765 54511 4.00 70986 112086 12 1765 54511 4.15 120386 121886 12 Forecasts are sorted by analyst number. The forecast date is self-reported by the analyst as the date on which the forecast was made. The I/B/E/S date is the date on which the forecast was entered in the database. I/B/E/S Analyst Forecasts for AVX (1986) 121 APPENDIX A Table 36 Forecast I/B/E/S Fiscal CUSIP Analyst Forecast Date Date Year-End 2440 255 0.50 22586 22786 12 2440 255 0.50 22586 71086 12 2440 255 0.10 81486 112086 12 2440 255 0.01 121186 121186 12 2440 2023 0.40 31986 32086 12 2440 2023 0.40 31986 41786 12 2440 2023 0.20 51486 71086 12 2440 2694 0.75 31386 31386 12 2440 2694 0.75 31386 41786 12 2440 2694 0.25 71686 71786 12 2440 2694 0.25 71686 81486 12 2440 2694 -0.05 111686 112086 12 2440 2694 -0.05 111686 121886 12 2440 2983 -0.10 110586 111386 12 2440 2983 -0.10 110586 121886 12 2440 4167 0.10 21286 22086 12 2440 4167 0.10 21286 32086 12 2440 4167 0.10 21286 41786 12 2440 4167 0.15 51386 71786 12 2440 4167 -0.05 81286 81486 12 2440 5425 -0.10 111386 112086 12 2440 5425 -0.10 111386 121886 12 Forecasts are sorted by analyst number. The forecast date is self-reported by the analyst as the date on which the forecast was made. The I/B/E/S date is the date on which the forecast was entered in the database. APPENDIX B LIST OF COMPANIES AND FIRM-YEARS INCLUDED IN THE SAMPLE 122 Appendix B List of Companies & Firm-Years Included in Sample McDonnell Douglas Corp 580169 Minnesota Mining & Mfg Co 604059 COMPANY NAME CUSIP 1983 1984 1985 1986 Allied Signal Inc 19512 x Amoco Corp (Std Oil-IND) 31905 x Anheuser-Busch Cos Inc 35229 x Atlantic Richfield Co 48825 x Baxter Labs 71892 x Boeing Co 97023 x Caterpillar Inc 149123 X Chevron Corp (Std Oil-CA) 166751 x Chrysler Corp 171196 x Coca-Cola Co 191216 X Dow Chemical 260543 x Du Pont (E.I.) De Nemours 263534 x Eastman Kodak Co 277461 x Exxon Corp (Std Oil-NJ) 302290 X Ford Motor Corp 345370 x General Dynamics Corp 369550 X General Electric Co 369604 X General Motors Corp 370442 x Georgia-Pacific Corp 373298 X Goodyear Tire and Rubber 382550 X Grace (W.R.) & Co 383883 x Honeywell 438506 x Intl Business Machines Co 459200 X Intl Paper Co 460146 x Johnson & Johnson 478160 x Lockheed Corp 539821 x X X X x X X X x X X X X X X X x >4h1¥1fi><fl3 :x ><><>cx:x:<><><>cu:x:n:¢><>< N >¢x:n>:x:~><>¢x:s>:xea><~:xan><>ex Mobil Corp 607059 Monsanto Co 611662 Motorola Inc 620076 Northrop Corp 666807 Occidental Petroleum Corp 674599 Pepsico Inc 713448 Philip Morris Cos Inc 718154 Phillips Petroleum Corp 718507 RJR Nabisco Inc 74960L Raytheon 755111 Sohio (Std Oil-Ohio) 853734 Sun Co Inc 866762 TRW Inc 872649 Tenneco Inc 880370 Texaco Inc 881694 USX Corp (US Steel) 902905 Unisys Corp (Burroughs) 909214 United Technologies Corp 913017 Unocal Corp 915289 Westinghouse Electric Cor 960402 Weyerhaeuser Co 962166 xas>:x:<>¢x=n>exa<>exi<>0xa<>0xas>ex0400x0<>0xaa>cx:n>:x:s>ex:<>:x:<>:xas>cx ><>:T: .A1 481 0.08451 0.06951 -0.07759 0.569777 26.66457 0.0001 .A2 481 0.08576 0.06781 -0.09701 0.42969 27.73978 0.0001 .A3 481 0.08937 0.07306 -0.09631 0.58546 26.82851 0.0001 A4 481 0.08833 0.07425 -0.08964 0.66968 26.09183 0.0001 .A5 481 0.08747 0.07545 -0.09811 0.70588 25.42380 0.0001 .A6 481 0.08958 0.07109 -0.03226 0.65158 27.63532 0.0001 A7 481 0.09330 0.07279 -0.02593 0.49472 28.11163 0.0001 A8 481 0.09513 0.07489 -0.10127 0.65158 27.85808 0.0001 A. 481 0.08308 0.06813 -0.08598 0.55747 26.74595 0.0001 A9 481 0.08704 0.06580 -0.09969 0.45701 29.00931 0.0001 .A10 481 0.09589 0.06856 -0.12658 0.51077 30.67357 0.0001 .A11 481 0.10064 0.07608 -0.11249 0.70890 29.01003 0.0001 .A12 481 0.09296 0.07368 -0.10516 0.67511 27.66890 0.0001 .A13 481 0.08637 0.07634 -0.09130 0.66968 24.81244 0.0001 .A14 481 0.08083 0.07007 -0.10648 0.43439 25.29777 0.0001 .A15 481 0.08392 0.07391 -0.15211 0.51900 24.90341 0.0001 .A16 481 ‘0.07573 0.07117 -0.09631 0.66968 23.33473 0.0001 Table 38 Weekly Market Average Earnings/Price Ratios for the Entire Sanple Variable N Mean Std Dev Mininum Maximum T Prob>:T: M]. 481 0 . 11164 0 . 01844 0 . 07420 0 . 13732 132 . 81465 0 . 0001 3MB 481 0.11105 0.01814 0.07420 0.13437 134.28001 0.0001 IMB 481 0.11104 0.01829 0.07420 0.13144 133.13860 0.0001 IN! 481 0.11126 0.01847 0.07420 0.13144 132.12270 0.0001 IMS 481 0.11120 0.01850 0.07318 0.13144 131.84261 0.0001 1M5 481 0.11065 0.01857 0.07245 0.13242 130.67673 0.0001 MW 481 0.11023 0.01903 0.07204 0.13242 127.06059 0.0001 IMB 481 0.10935 0.01899 0.07204 0.13242 126.32032 0.0001 1! 481 0.10903 0.01906 0.07204 0.13242 125.47141 0.0001 3MB 481 0.10819 0.01937 0.07131 0.13133 122.48268 0.0001 IM10 481 0.10797 0.01986 0.07131 0.13428 119.22750 0.0001 RDJ. 481 0.10814 0.02019 0.07131 0.13428 117.44760 0.0001 1412 481 0 . 10831 0 . 02014 0 . 06718 0 . 13428 117 . 92558 0 . 0001 m3 481 0 . 10820 0 . 02001 0 . 06718 0 . 13428 118 . 57121 0 . 0001 IM14 481 0.10756 0.01992 0.06718 0.13144 118.40064 0.0001 IM15 481 0.10691 0.01980 0.06718 0.13144 118.39543 0.0001 1116 481 0 . 10578 0 . 0 . 0 . 13144 102 . 95990 0 . 0001 125 APPENDIX C Table 39 Weekly Revisions in Earnings/Price Ratios for the Entire Sanple Unadjusted for Market-Wide Revisions Variable N Mean Std Dev Mininum Maxim T Prob> : T} C2 371 0.00025 0.01362 -0.12097 0.10937 0.35629 0.7218 C3 450 -0.00099 0.01815 -0.21972 0.15107 -1.15423 0.2490 C4 469 0.00119 0.01634 -0.08952 0.15068 1.57603 0.1157 C5 478 -0.00011 0.01618 -0.09231 0.14661 -0.14717 0.8831 C6 480 -0.00148 0.02272 -0.31674 0.11831 -1.42856 0.1538 (7' 481 0.00004 0.02177 -0.30769 0.14178 0.04479 0.9643 C8 481 -0.00032 0.02166 -0.15190 0.30769 -0.32650 0.7442 C 373 -0.00038 0.01911 -0.19909 0.11946 -0.32208 0.7476 C9 392 -0.00189 0.02133 -0.22299 0.10028 -1.75533 0.0800 C10 392 -0.00099 0.02066 -0.18028 0.29575 -0.95089 0.3422 C31. 465 0.00241 0.01955 -0.09296 0.25188 2.66333 0.0080 C12 477 -0.00035 0.01631 -0.11014 0.18228 -0.51570 0.6063 C13 480 -0.00013 0.01395 -0.10534 0.13099 -0.20676 0.8363 C14 481 -0.00094 0.01286 -0.12097 0.06526 -1.60512 0.1091 C15 481 -0.00099 0.01870 -0.27873 0.08597 -1.16493 0.2446 016 481 0 . 00009 0. 01850 -0 . 21972 0. 15107 0. 11150 0. 9113 Table 40 Weekly Revisions in Earnings/Price Ratios for the Entire Sanple Adjusted for Mrket-Wide Revisiors Variable N Mean Std Dev Mininum mixinum T Prob>:T: E2 371 0.00080 0.01350 -0.11778 0.105415 1.13389 0.2576 E3 450 -0.00098 0.017936 -0.21345 0.149248 -1.15629 0.2482 EM 469 0.00098 0.01616 -0.08970 0.147132 1.30640 0.1921 E5 478 -0.00004 0.01604 -0.08944 0.142939 -0.05720 0.9544 E6 480 -0.00093 0.02222 -0.30483 0.114226 -0.91830 0.3589 E7’ 481 0.00046 0.02154 -0.30013 0.140456 0.47171 0.6373 E8 481 0.00056 0.02134 -0.14828 0.298446 0.57284 0.5670 E 373 -0.00026 0.018486 -0.19349 0.118607 -0.26807 0.7888 E9 392 -0.00081 0.02084 -0.21545 0.089485 -0.76894 0.4424 E10 392 -0.00072 0.02001 -0.16504 0.285987 -0.70783 0.4795 E11 465 0.00222 0.01938 -0.09009 0.247855 2.47226 0.0138 E12 477 -0.00056 0.01610 -0.11085 0.178614 -0.76159 0.4467 E13 480 -0.00002 0.01379 -0.10014 0.126762 -0.03048 0.9757 E14 481 -0.00031 0.01272 -0.11778 0.062011 -0.52714 0.5983 E15 481 -0.00034 0.01846 -0.27314 0.089161 -0.40612 0.6848 E16 481 0.00122 0.02114 -0.21345 0.149248 1.26821 0.2053 126 APPENDIXCC Table 41 Weekly Revisions in Earnings/Price Ratios for the Good.News & Bad.News Samples Unadjusted for Market-Wide Revisions Variable N Mean Std Dev Mininum Maxinum T Prob>:T Good News C2 196 -0.00031 0.00837 -0.03125 0.03664 -0.51669 0.6060 C3 243 0.00014 0.01556 -0.05779 0.15107 0.14338 0.8861 C4 254 0.00034 0.01344 -0.08952 0.07615 0.40805 0.6836 C5 258 0.00063 0.01625 -0.06212 0.12670 0.62573 0.5320 C6 260 -0.00180 0.02714 -0.31674 0.11831 -1.07229 0.2846 (9’ 260 0.00040 0.01647 -0.09776 0.14062 0.39358 0.6942 C8 260 -0.00036 0.01968 -0.15190 0.15255 -0.29453 0.7686 C 196 -0.00050 0.01428 -0.06774 0.06897 -0.48630 0.6273 C9 206 -0.00093 0.01443 -0.08013 0.05417 -0.92184 0.3577 C10 206 -0.00180 0.01110 -0.07762 0.02957 -2.30385 0.0222 C11 250 0.00299 0.01802 -0.02617 0.20043 2.62001 0.0093 C12 257 -0.00039 0.01700 -0.11014 0.18228 -0.37119 0.7108 C13 259 -0.00002 0.01003 -0.05080 0.05484 -0.02856 0.9772 C14 260 0.00012 0.01098 -0.04464 0.06526 0.17937 0.8578 C15 260 -0.00231 0.02118 -0.27873 0.07143 -1.75726 0.0801 C16 260 -0.00008 0.01197 -0.07857 0.06103 -0.10257 0.9184 Bad News C2 175 0.00088 0.01776 -0.12097 0.10937 0.65560 0.5129 1C3 207 -0.00232 0.02075 -0.21972 0.04932 -1.60503 0.1100 C4 215 0.00219 0.01919 -0.07023 0.15068 1.67096 0.0962 C5 220 -0.00098 0.01608 -0.09231 0.14661 -0.90283 0.3676 C5 220 -0.00110 0.01604 -0.10413 0.07421 -1.01597 0.3108 C7 221 -0.00038 0.02674 -0.30769 0.14178 -0.20907 0.8346 C8 221 -0.00028 0.02382 -0.06039 0.30769 -0.17399 0.8620 C 177 -0.00012 0.02337 -0.19909 0.11946 -0.06972 0.9445 C9 186 -0.00296 0.02699 -0.22299 0.10028 -1.49499 0.1366 C10 186 -0.00012 0.02764 -0.18028 0.29575 -0.05825 0.9536 C11 215 0.00175 0.02121 -0.09296 0.25189 1.20953 0.2278 C12 220 -0.00038 0.01550 -0.10437 0.14178 -0.35895 0.7200 C13 221 -0.00027 0.01748 -0.10534 0.13099 -0.22542 0.8219 C14 221 -0.00219 0.01469 -0.12097 0.03429 -2.21756 0.0276 C15 221 0.00055 0.01518 -0.07782 0.08597 0.54191 0.5884 C16 221 0.00029 0.02404 -0.21972 0.15107 0.18196 0.8558 Adjusted for Market-Wide Revisions 127 AETENDIXCC Table 42 Weekly Revisions in Farm'ngs/Prioe Ratios for the Good News & Bad News Sanples Variable N Mean Std Dev Mininum Maxinum T PrOb>:T Good News E2 196 0.00019 0.00862 -0.03151 0.03443 0.30588 0.7600 E3 243 0.00012 0.01538 -0.05752 0.14924 0.12319 0.9021 EM 254 0.00007 0.01332 -0.08970 0.07268 0.08576 0.9317 E5 258 0.00078 0.01621 -0.05843 0.12376 0.77458 0.4393 E6 260 -0.00111 0.02644 -0.30484 0.11422 -0.67958 0.4974 E7 260 0.00079 0.01640 -0.09402 0.13671 0.77333 0.4400 E8 260 0.00048 0.01954 -0.14828 0.14817 0.39574 0.6926 E 196 -0.00032 0.01350 -0.06626 0.05455 -0.33236 0.7400 E9 206 0.00020 0.01414 -0.07816 0.05427 0.20445 0.8382 E10 206 -0.00143 0.01111 -0.07840 0.03043 -1.84643 0.0663 E11 250 0.00255 0.01784 -0.02122 0.19331 2.25762 0.0248 E12 257 -0.00063 0.01680 -0.11085 0.17861 -0.60343 0.5468 E13 259 0.00022 0.01025 -0.04870 0.05311 0.34612 0.7295 E14 260 0.00076 0.01085 -0.04388 0.06201 1.12573 0.2613 E15 260 -0.00140 0.02083 -0.27314 0.06945 -1.07723 0.2824 E16 260 0.00193 0.01875 -0.07564 0.11743 1.65823 0.0985 Bad.News E2 175 0.00147 0.01742 -0.11778 0.10541 1.11932 0.2645 E3 207 -0.00227 0.02049 -0.21345 0.05400 -1.59155 0.1130 E4 215 0.00204 0.01895 -0.06638 0.14714 1.57953 0.1157 E5 220 -0.00101 0.01584 -0.08944 0.14293 -0.94390 0.3463 E6 220 -0.00072 0.01592 -0.10073 0.07299 -0.66655 0.5058 E7 221 0.00008 0.02636 -0.30013 0.14045 0.04659 0.9629 E8 221 0.00065 0.02333 -0.06117 0.29844 0.41359 0.6796 E 177 -0.00019 0.02280 -0.19349 0.11860 -0.10835 0.9138 E9 186 -0.00193 0.02634 -0.21545 0.08948 -0.99865 0.3193 E10 186 0.00007 0.02662 -0.16504 0.28598 0.03822 0.9696 E11 215 0.00184 0.02107 -0.09009 0.24785 1.28345 0.2007 E12 220 -0.00048 0.01527 -0.10283 0.13853 -0.46453 0.6427 E13 221 -0.00030 0.01705 -0.10014 0.12676 -0.26153 0.7939 E14 221 -0.00156 0.01454 -0.11778 0.03443 -1.59171 0.1129 E15 221 0.00089 0.01515 -0.07384 0.08916 0.87691 0.3815 E16 221 0.00039 0.02366 -0.21345 0.14924 0.24618 0.8058 128 APPENDIX C Table 43 Weekly Revisions in Earnings/Price Ratios for the Good News & large/Small amiss Sanples Unadjusted for Market-Wide Revisias Variable N Mean Std Dev Mininum Maximum T Prob>1T Good Nays/mall Surprise C2 106 -0.00046 0.01005 -0.03125 0.03664 -0.46806 0.6407 C3 126 -0.00089 0.01765 -0.05779 0.15107 -0.56325 0.5743 C4 129 0.00089 0.01198 -0.02223 0.06221 0.84612 0.3991 C5 131 0.00111 0.01608 -0.06012 0.12670 0.78992 0.4310 C6 132 -0.00332 0.03402 -0.31674 0.09786 -1.12042 0.2646 C7 132 0.00142 0.01700 -0.04857 0.14062 0.95662 0.3405 C8 132 -0.00157 0.02448 -0.15190 0.15255 -0.73626 0.4629 C 96 -0.00022 0.01596 -0.06774 0.06897 -0.13245 0.8949 C9 104 -0.00163 0.01539 -0.08013 0.05417 -1.08252 0.2816 C10 104 -0.00243 0.01261 -0.07762 0.02966 -1.96812 0.0517 C11 126 0.00440 0.02293 -0.01717 0.20043 2.15649 0.0330 C12 130 0.00168 0.01909 -0.05779 0.18228 1.00261 0.3179 C13 132 -0.00128 0.01070 -0.05080 0.03308 -1.37577 0.1712 C14 132 -0.00043 0.01098 -0.04464 0.04545 -0.45308 0.6512 C15 132 -0.00267 0.02706 -0.27873 0.07143 -1.13524 0.2583 C16 132 -0.00096 0.01331 -0.07857 0.06071 -0.83275 0.4065 Good News/large airprise 1C2 90 -0.00013 0.00586 -0.01720 0.01509 -0.21813 0.8278 C3 117 0.00125 0.01291 -0.05645 0.08597 1.04804 0.2968 C4 125 -0.00022 0.01482 -0.08952 0.07615 -0.16744 0.8673 C5 127 0.00014 0.01647 -0.06212 0.09450 0.09677 0.9231 C6 128 -0.00025 0.01744 -0.10215 0.11831 -0.15928 0.8737 C7 128 -0.00064 0.01589 -0.09776 0.07097 -0.45830 0.6475 C8 128 0.00089 0.01300 -0.04838 0.07512 0.77208 0.4415 C 100 -0.00077 0.01253 -0.04258 0.05373 -0.61037 0.5430 C9 102 -0.00021 0.01342 -0.05780 0.04508 -0.15501 0.8771 C10 102 -0.00112 0.00932 -0.05634 0.01488 -1.20926 0.2294 C11 124 0.00154 0.01091 -0.02617 0.09606 1.57662 0.1175 C12 127 -0.00252 0.01432 -0.11014 0.01938 -1.98004 0.0499 C13 127 0.00129 0.00916 -0.02782 0.05484 1.59374 0.1135 C14 128 0.00069 0.01099 -0.04384 0.06526 0.71478 0.4761 C15 128 -0.00193 0.01261 -0.07793 0.05229 -1.73245 0.0856 C16 128 0.00084 0.01038 -0.05779 0.06103 0.91594 0.3614 129 AETTNDIXIC Table 44 Weekly Revisims in Earnings/Price Intios for the Bad News & Iarge/Snall Smprise Sanples Unadjusted for mket-Wide Revisions Variable N Mean Std Dev Mininum Maxinum T Prob>1T Bad News/Stall Surprise C2 93 0.00222 0.00968 -0.01568 0.04743 2.21311 0.0294 C3 111 -0.00240 0.01287 -0.06122 0.03027 -1.96149 0.0523 C4 117 0.00227 0.01611 -0.04422 0.15068 1.52202 0.1307 C5 121 -0.00218 0.01271 -0.09231 0.03362 -1.88260 0.0622 C6 121 -0.00110 0.01338 -0.05959 0.07421 -0.90374 0.3679 C7 121 -0.00237 0.02942 -0.30769 0.02667 -0.88652 0.3771 C8 121 0.00243 0.02964 -0.02675 0.30769 0.90115 0.3693 C 95 -0.00252 0.02497 -0.19909 0.05000 -0.98278 0.3282 C9 100 -0.00307 0.02458 -0.19457 0.06798 -1.24872 0.2147 C10 100 -0.00118 0.01152 -0.04676 0.07089 -1.02318 0.3087 C11 119 0.00226 0.02434 -0.04582 0.25189 1.01312 0.3131 C12 121 -0.00146 0.01246 -0.10437 0.02370 -1.28625 0.2008 C13 121 -0.00130 0.01102 -0.08145 0.02996 -1.29662 0.1973 C14 121 -0.00129 0.01280 -0.08869 0.03056 -1.10516 0.2713 C15 121 -0.00036 0.01055 -0.04563 0.04550 -0.37490 0.7084 C16 121 0.00079 0.01596 -0.05068 0.15107 0.54625 0.5859 Bad News/Large Surprise C2 82 -0.00064 0.02381 -0.12097 0.10937 -0.24374 0.8080 C3 96 -0.00222 0.02724 -0.21972 0.04932 -0.79902 0.4263 C4 98 0.00209 0.02241 -0.07023 0.13371 0.92418 0.3577 C5 99 0.00048 0.01940 -0.03982 0.14661 0.24815 0.8045 C6 99 -0.00110 0.01887 -0.10413 0.07389 -0.57926 0.5637 C7 100 0.00204 0.02299 -0.05913 0.14178 0.88629 0.3776 C8 100 -0.00355 0.01325 -0.06039 0.04762 -2.68240 0.0086 C 82 0.00265 0.02117 -0.05042 0.11946 1.13481 0.2598 C9 86 -0.00283 0.02970 -0.22299 0.10028 -0.88376 0.3793 C10 86 0.00112 0.03881 -0.18028 0.29575 0.26647 0.7905 C11 96 0.00112 0.01663 -0.09296 0.08826 0.65742 0.5125 C12 99 0.00095 0.01853 -0.04681 0.14178 0.50863 0.6122 C13 100 0.00099 0.02300 -0.10534 0.13099 0.42848 0.6692 C14 100 -0.00329 0.01670 -0.12097 0.03429 -1.96845 0.0518 C15 100 0.00166 0.01937 -0.07782 0.08597 0.85618 0.3940 C16 100 -0.00031 0.03122 -0.21972 0.15068 -0.09897 0.9214 130 APPENDIX C Table 45 Weekly Revisions in Earnings/Price Ratios for the Good News & large/Stall Surprise Sanplas Adjusted for Market-Wide Revisions Variable N Mean Std Dev Mininum mxinum T Prob>:T: Good News/Shall Snprise E2 106 0.00006 0.01012 -0.03151 0.03443 0.06141 0.9511 E3 126 -0.00096 0.01746 -0.05752 0.14924 -0.61605 0.5390 E4 129 0.00045 0.01174 -0.01910 0.06099 0.43335 0.6655 E5 131 0.00124 0.01589 -0.05730 0.12376 0.89455 0.3727 E6 132 -0.00187 0.03304 -0.30484 0.09747 -0.64980 0.5170 57' 132 0.00147 0.01681 -0.04733 0.13671 1.00725 0.3157 E8 132 -0.00076 0.02438 -0.14828 0.14817 -0.35733 0.7214 E 96 0.00014 0.01475 -0.06626 0.05455 0.09405 0.9253 E9 104 -0.00045 0.01542 -0.07816 0.05427 -0.29780 0.7663 E10 104 -0.00231 0.01273 -0.07840 0.03043 -1.85115 0.0670 E11 126 0.00373 0.02256 -0.01622 0.19331 1.85441 0.0660 E12 130 0.00142 0.01872 -0.05478 0.17861 0.86653 0.3878 E13 132 -0.00068 0.01087 -0.04870 0.03518 -0.71884 0.4735 E14 132 0.00017 0.01096 -0.04388 0.04642 0.17500 0.8613 E15 132 -0.00173 0.02656 -0.27314 0.06944 -0.74890 0.4553 .E16 132 -0.00074 0.01326 -0.07564 0.06099 -0.64437 0.5205 Good News/Large Surprise E2 90 0.00034 0.00646 -0.01638 0.01719 0.49790 0.6198 E3 117 0.00128 0.01272 -0.04860 0.08916 1.09223 0.2770 E4 125 -0.00032 0.01481 -0.08970 0.07268 -0.23911 0.8114 E5 127 0.00031 0.01657 -0.05843 0.09149 0.20840 0.8353 E6 128 -0.00034 0.01727 -0.09746 0.11422 -0.22029 0.8260 E7 128 0.00008 0.01601 -0.09402 0.06968 0.05531 0.9560 E8 128 0.00176 0.01273 -0.04923 0.07108 1.56042 0.1211 E 100 -0.00076 0.01223 -0.04480 0.03932 -0.62473 0.5336 E9 102 0.00087 0.01276 -0.05431 0.03679 0.68576 0.4944 E10 102 -0.00053 0.00913 -0.05139 0.01710 -0.58512 0.5598 E11 124 0.00135 0.01112 -0.02122 0.09419 1.34836 0.1800 E12 127 -0.00274 0.01435 -0.11085 0.02389 -2.14771 0.0337 E13 127 0.00116 0.00950 -0.02699 0.05311 1.37145 0.1727 E14 128 0.00137 0.01075 -0.03933 0.06201 1.43876 0.1527 E15 128 -0.00104 0.01254 -0.07416 0.05067 -0.94003 0.3490 E16 128 0.00468 0.02282 -0.05752 0.11743 2.32225 0.0218 131 APPENDIX'C Table 46 Weekly Revisions in Earnings/Price Ratios for the Bad News 5: Large/Stall Surprise Sanples Adjusted for Market-Wide Revisions Variable N Mean Std Dev Mininum mxinum T Prob>:T: Bad News/Shall Surprise E2 93 0.00294 0.00929 -0.01865 0.04749 3.04815 0.0030 E3 111 -0.00209 0.01324 -0.06140 0.03188 -1.66272 0.0992 E4 117 0.00163 0.01613 -0.04638 0.14714 1.09615 0.2753 E5 121 -0.00217 0.01251 -0.08944 0.03014 -1.90718 0.0589 E6 121 -0.00086 0.01339 -0.05589 0.07205 -0.70914 0.4796 E7 121 -0.00156 0.02890 -0.30013 0.02547 -0.59196 0.5550 E8 121 0.00281 0.02890 -0.02239 0.29844 1.06852 0.2874 E 95 -0.00233 0.02440 -0.19349 0.03559 -0.93209 0.3537 E9 100 -0.00166 0.02424 -0.19003 0.06774 -0.68454 0.4952 E10 100 -0.00126 0.01187 -0.04579 0.07017 -1.06426 0.2898 E11 119 0.00225 0.02417 -0.04632 0.24785 1.01532 0.3120 E12 121 -0.00157 0.01238 -0.10283 0.02383 -1.39048 0.1670 E13 121 -0.00115 0.01108 -0.07916 0.02945 -1.14395 0.2549 E14 121 -0.00088 0.01286 -0.08520 0.02759 -0.75590 0.4512 E15 121 0.00035 0.01062 -0.04740 0.04869 0.36307 0.7172 E16 121 0.00076 0.01588 -0.04691 0.14924 0.52649 0.5995 Bad News/Large Surprise E2 82 -0.00018 0.02342 -0.11778 0.10541 -0.07105 0.9435 E31 96 -0.00247 0.02660 -0.21345 0.05400 -0.91052 0.3649 E4 98 0.00253 0.02193 -0.06638 0.12969 1.14105 0.2567 E5 99 0.00041 0.01910 -0.04098 0.14293 0.21488 0.8303 E6 99 -0.00054 0.01863 -0.10073 0.07299 -0.28583 0.7756 E7 100 0.00206 0.02291 -0.05943 0.14045 0.90095 0.3698 E8 100 -0.00196 0.01361 -0.06117 0.04672 -1.44160 0.1526 E 82 0.00230 0.02066 -0.04681 0.11860 1.00895 0.3160 E9 86 -0.00224 0.02874 -0.21545 0.08948 -0.72366 0.4713 E10 86 0.00163. 0.03706 -0.16504 0.28598 0.40804 0.6843 E11 96 0.00134 0.01656 -0.09009 0.08705 0.79422 0.4290 E12 99 0.00085 0.01817 -0.04729 0.13853 0.46566 0.6425 E13 100 0.00073 0.02225 -0.10014 0.12676 0.32888 0.7429 E14 100 -0.00237 0.01638 -0.11778 0.03443 -1.44794 0.1508 E15 100 0.00155 0.01930 -0.07384 0.08916 0.80348 0.4236 E16 100 -0.00005 0.03062 -0.21345 0.14714 -0.01765 0.9860 132 APPENDIX C Table 47 Bi-Weekly Lhan Earnings/Price Intios for the Entire Sanple Variable N than Std Dev Mininum Maximum T Prob> : T: A1 481 0 . 10370 0 . 06338 -0 . 08869 0 . 56977 35 . 88405 0 . 0001 A2 481 0 . 10372 0 . 06865 -0 . 09323 0 . 58232 33 . 13522 0 . 0001 A3 481 0 . 10620 0 . 06702 -0 . 08789 0 . 61448 34 . 75103 0 . 0001 A4 481 0 . 10524 0 . 06707 -0 . 10127 0 . 53823 34 . 41115 0 . 0001 A 481 0 . 08308 0 . 06813 -0 . 08598 0 . 55747 26 . 74595 0 . 0001 A5 481 0 . 10354 0 . 06478 -0. 12658 0 . 49911 35 . 05565 0 . 0001 A6 481 0 . 10679 0 . 07266 -0 . 11005 0 . 69955 32 . 23194 0 . 0001 A7 481 0 . 10170 0 . 07097 -0 . 10648 0 . 66968 31 . 42718 0 . 0001 A8 481 0. 09759 0. 07088 -0. 15211 0.58546 30. 19543 0. 0001 Table 48 Bi-Weekly Market Average Earnings / Price Ratios for the Ehtire Sanple Variable N Mean Std Dev Mininum Maxinum T Prob> : T: Ml 481 0 . 11131 0 . 01815 0 . 07277 0 . 13743 134 . 53641 0 . 0001 MZ 481 0 . 11114 0 . 01840 0 . 07277 0 . 13106 132 . 45065 0 . 0001 MB 481 0 . 11085 0 . 01848 0 . 07277 0 . 13106 131 . 53419 0 . 0001 M4 481 0. 10968 0 . 01883 0 . 07211 0 . 12984 127 . 74914 0 . 0001 M 481 0 . 10903 0 . 01906 0 . 07204 0 . 13242 125 . 47141 0 . 0001 MS 481 0 . 10818 0 . 01942 0 . 07160 0 . 13169 122 . 16357 0 . 0001 M6 481 0. 10832 0 . 02006 0 . 07026 0 . 13200 118 . 44262 0 . 0001 M7 481 0 . 10793 0 . 01984 0 . 06855 0 . 13200 119 . 34018 0 . 0001 MB 481 0 . 10576 0 . 02248 0 . 0 . 13106 103 . 19689 0 . 0001 133 APPENDIX C Table'49 Bi-Weekly Revisions in Earnings/Price Ratios for the Entire Sanple Unadjusted for Market-Wide Revisions Variable N than Std Dev Minimum Maximum T Prob> 1 T C2 450 0.00000 0.01393 -0.07108 0.14793 0.00474 0.9962 1C3 478 -0.00090 0.01406 -0.07070 0.14661 -1.40108 0.1618 C4 481 -0.00082 0.01597 -0.15190 0.14005 -1.13171 0.2583 C 373 0.00007 0.01371 -0.04809 0.08543 0.09716 0.9227 C5 465 -0.00222 0.01362 -0.10804 0.05634 -3.50927 0.0005 C6 465 0.00172 0.02010 -0.08380 0.21014 1.84986 0.0650 C7 480 -0.00100 0.01379 -0.16624 0.12606 -1.58172 0.1144 C8 481 -0.00118 0.01772 -0.27873 0.08455 -1.45828 0.1454 Table 50 Bi-Weekly Revisions in Earnings/Price Ratios for the Entire Sanple Adjusted for Market-Wide Revisions Variable N khan Std Dev Mininum Maxinum T Prob> 1T E2 450 0.00021 0.01387 -0.07091 0.14420 0.31553 0.7525 E3 478 -0.00060 0.01389 -0.07092 0.14229 -0.94886 0.3432 E4 481 0.00035 0.01581 -0.14553 0.13892 0.48405 0.6286 E 373 0.00049 0.01310 -0.04467 0.08525 0.71716 0.4737 E5 465 -0.00074 0.01353 -0.10447 0.05898 -1.18265 0.2376 E6 465 0.00157 0.01976 -0.08122 0.20596 1.71276 0.0874 E7 480 -0.00061 0.01365 -0.16170 0.12242 -0.97767 0.3287 E8 481 0.00100 0.02044 -0.26913 0.11567 1.07250 0.2840 134 APPENDIXiC Table 51 Bi-Weekly Revisions in Earnings/Price Ratios for the Good News a Bad News Sanple: Unadjusted for Market-Wide Revisions Variable N lhan Std Dev Mininum Maxinum T Prob>1T Good News C2 243 0.00028 0.01433 -0.05742 0.14793 0.30021 0.7643 C3 258 -0.00056 0.01189 -0.04674 0.07638 -0.76146 0.4471 C4 260 -0.00114 0.01476 -0.15190 0.06323 -1.24164 0.2155 C 196 0.00013 0.01285 -0.03560 0.06897 0.14391 0.8857 C5 250 -0.00186 0.01334 -0.10804 0.05634 -2.19939 0.0288 C6 250 0.00254 0.01934 -0.04221 0.20043 2.07687 0.0388 C7 259 -0.00047 0.00826 -0.04189 0.03015 -0.91173 0.3628 C8 260 -0.00243 0.01990 -0.27873 0.03864 -1.96646 0.0503 Bad News C2 207 -0.00032 0.01346 -0.07108 0.05242 -0.33897 0.7350 C3 220 -0.00130 0.01626 -0.07070 0.14661 -1.18281 0.2382 C4 221 -0.00046 0.01732 -0.11674 0.14005 -0.39194 0.6955 C 177 -0.00000 0.01463 -0.04809 0.08543 -0.00086 0.9993 C5 215 -0.00264 0.01396 -0.08959 0.04460 -2.76925 0.0061 C6 215 0.00078 0.02096 -0.08380 0.21014 0.54284 0.5878 C7 221 -0.00161 0.01825 -0.16624 0.12606 -1.31410 0.1902 C8 221 0.00029 0.01465 -0.08486 0.08455 0.29560 0.7678 Adjusted for Market-Wide Revisions 135 APPENDIXCC Table 52 Bi-Weekly Revisions in Earnings/Price Ratios for the Good News a Bad News Sanples Variable N bhan Std Dev Minimum Maximum T Prob>1T Good News E2 243 0.00040 0.01414 -0.05349 0.14420 0.43677 0.6627 E3 258 -0.00016 0.01193 -0.04760 0.07721 -0.20963 0.8341 E4 260 0.00006 0.01455 -0.14553 0.06332 0.06849 0.9454 E 196 0.00063 0.01222 -0.03687 0.05868 0.72302 0.4705 E5 250 -0.00033 0.01330 -0.10447 0.05898 -0.38798 0.6984 E6 250 0.00211 0.01902 -0.04012 0.19349 1.75371 0.0807 E7 259 0.00004 0.00858 -0.04197 0.02851 0.06821 0.9457 E8 260 0.00092 0.02434 -0.26913 0.11567 0.60665 0.5446 Bad Nels E2 207 -0.00002 0.01357 -0.07091 0.05259 -0.01763 0.9860 E3 220 -0.00113 0.01590 -0.07092 0.14229 -1.05153 0.2942 E4 221 0.00069 0.01719 -0.11430 0.13892 0.59367 0.5533 E 177 0.00033 0.01404 -0.04467 0.08525 0.30901 0.7577 E5 215 -0.00122 0.01380 -0.08477 0.04467 -1.30137 0.1945 E6 215 0.00094 0.02061 -0.08122 0.20596 0.66942 0.5040 E7 221 -0.00137 0.01784 -0.l6170 0.12242 -l.13808 0.2563 E8 221 0.00110 0.01464 -0.08058 0.08400 1.11555 0.2658 136 APPENDIXCC Table 53 Bi-Weekly Revisions in Earnings/Price Ratios for the Good News & large/Snell Sunrise Sanplee Unadjusted for Market-Wide Revisions Variable N Mean Sthev Mininum Maximum T Prob>1T Good News/Snail Surprise 88888988 126 -0.00075 0.01770 -0.05742 0.14793 -0.47665 131 -0.00026 0.01296 -0.04674 0.07638 -0.23102 132 -0.00263 0.01756 -0.15190 0.06323 -1.72205 96 0.00059 0.01275 -0.03560 0.06897 0.45699 126 -0.00313 0.01500 -0.10804 0.04062 -2.34385 126 0.00455 0.02573 -0.04221 0.20043 1.98377 132 -0.00095 0.00987 -0.04189 0.03015 -1.11150 132 -0.00383 0.02633 -0.27873 0.03742 -1.67149 0.6344 0.8177 0.0874 0.6487 0.0207 0.0495 0.2684 0.0970 News/Large Snprise 88880988 3 117 0.00138 0.00940 -0.02411 0.04938 1.59070 127 -0.00088 0.01071 -0.03982 0.04369 -0.92038 128 0.00040 0.01104 -0.07541 0.05723 0.41485 100 -0.00031 0.01299 -0.03398 0.05771 -0.24019 124 -0.00056 0.01133 -0.04088 0.05634 -0.54988 124 0.00050 0.00872 -0.04093 0.03117 0.63980 127 0.00004 0.00617 -0.02840 0.01895 0.06832 128 -0.00098 0.00942 -0.04147 0.03864 -1.17771 0.1144 0.3591 0.6790 0.8107 0.5834 0.5235 0.9456 0.2411 Bi-Weekly Revisiorh in Earnings/Price Ratios for the 137 APPENDIXIC Table 54 Bad News & large/mu an'prise Sanples Unadjusted for Market-Wide Revisions Variable N )han Std Dev Mininum Maxinum T Prob>1T1 Bad News/anall Sunrise C2 111 -0.00063 0.01174 -0.07108 0.02981 -0.56433 0.5737 C3 121 -0.00163 0.01042 -0.04199 0.04525 -1.72364 0.0873 C4 121 -0.00129 0.01322 -0.11674 0.03417 -1.07313 0.2854 C 95 -0.00093 0.01183 -0.04525 0.05071 -0.76274 0.4475 C5 119 -0.00179 0.01048 -0.04072 0.04460 -1.86738 0.0643 C6 119 0.00067 0.02111 -0.06676 0.21014 0.34618 0.7298 C7 121 -0.00272 0.01659 -0.16624 0.01692 -1.80052 0.0743 C8 121 0.00018 0.01215 -0.04563 0.08455 0.16583 0.8686 Bad News/Large Surprise C2 96 0.00004 0.01528 -0.06380 0.05242 0.02749 0.9781 C3 99 -0.00089 0.02139 -0.07070 0.14661 -0.41236 0.6810 C4 100 0.00055 0.02127 -0.08181 0.14005 0.25945 0.7958 C 82 0.00107 0.01734 -0.04809 0.08543 0.55897 0.5777 C5 96 -0.00368 0.01734 -0.08959 0.04252 -2.08027 0.0402 C6 96 0.00091 0.02090 -0.08380 0.16891 0.42571 0.6713 C7 100 -0.00028 0.02009 -0.07702 0.12606 -0.13946 0.8894 C8 100 0.00042 0.01727 -0.08486 0.05452 0.24455 0.8073 138 AETENDIXCC Table 55 Bi-Weekly Revisions in Earnings/Price Ratios for the Good News & large/Snell Surprise Sanples Adjusted for Market-Wide Revisions Variable N lhan Std Dev Mininum MaJthum T Prob>1T1 Good News/mall Surprise E2 126 -0.00069 0.01737 -0.05349 0.14420 -0.44752 0.6553 E3 131 0.00040 0.01283 -0.04760 0.07721 0.35640 0.7221 E4 132 -0.00142 0.01707 -0.14553 0.06332 -0.95670 0.3405 E 96 0.00122 0.01175 -0.03259 0.05540 1.01372 0.3133 E5 126 -0.00162 0.01490 -0.10447 0.04207 -1.21770 0.2256 E6 126 0.00377 0.02512 -0.03799 0.19349 1.68324 0.0948 E7 132 -0.00021 0.01014 -0.04197 0.02851 -0.24029 0.8105 E8 132 -0.00236 0.02571 -0.26913 0.03541 -1.05323 0.2942 Good News/large Surprise E2 117 0.00157 0.00943 -0.02142 0.05287 1.79935 0.0746 E3 127 -0.00073 0.01095 -0.04158 0.03916 -0.74973 0.4548 E4 128 0.00159 0.01126 -0.07141 0.05780 1.59849 0.1124 E 100 0.00007 0.01270 -0.03687 0.05868 0.05521 0.9561 E5 124 0.00099 0.01135 -0.03824 0.05898 0.96643 0.3357 E6 124 0.00043 0.00923 -0.04012 0.03083 0.51274 0.6091 E7 127 0.00029 0.00662 -0.02707 0.01787 0.50192 0.6166 E8 128 0.00429 0.02245 -0.04026 0.11567 2.16286 0.0324 Adjusted for Market-Wide Revisions 139 APPENDIXIC Table 56 Bi-Weekly Revisions in Earnings/Price Ratios for the Bad News & large/Small Surprise Samples Variable N lhan Std Dev Minimum Maximum T Prob>1T1 Bad News/Small Surprise E2 111 -0.00029 0.01203 -0.07091 0.02853 -0.24965 0.8033 E3 121 -0.00173 0.01011 -0.04020 0.04452 -1.88025 0.0625 E4 121 -0.00011 0.01313 -0.11430 0.03471 -0.09255 0.9264 E 95 -0.00062 0.01141 -0.04467 0.03715 -0.52615 0.6000 E5 119 -0.00045 0.01072 -0.04026 0.04467 -0.45369 0.6509 E6 119 0.00061 0.02086 -0.06607 0.20596 0.31885 0.7504 E7 121 -0.00238 0.01649 -0.16170 0.01639 -1.58896 0.1147 E8 121 0.00117 0.01233 -0.04764 0.08400 1.03983 0.3005 Bad News/Large Sirptrise E2 96 0.00029 0.01521 -0.06102 0.05259 0.18926 0.8503 E3 99 -0.00039 0.02095 -0.07092 0.14229 -0.18639 0.8525 E4 100 0.00165 0.02112 -0.08099 0.13892 0.78169 0.4363 E 82 0.00142 0.01659 -0.04280 0.08525 0.77367 0.4414 E5 96 -0.00219 0.01687 -0.08477 0.03855 -1.27236 0.2064 E6 96 0.00135 0.02039 -0.08122 0.16355 0.64921 0.5178 E7' 100 -0.00014 0.01937 -0.07505 0.12242 -0.07067 0.9438 E8 100 0.00102 0.01708 -0.08058 0.05287 0.59530 0.5530 for the Entire Sample Variable N khan Std Dev Minimum Maximum T Prob>1T1 A1 481 0.11057 0.06218 -0.09164 0.56931 38.99640 0.0001 A2 481 0.10959 0.06352 -0.08789 0.55611 37.83874 0.0001 A 481 0.08308 0.06813 -0.08598 0.55747 26.74595 0.0001 A3 481 0.10789 0.06637 -0.09632 0.60711 35.65087 0.0001 A4 481 0.10739 0.06663 -0.12169 0.56931 35.34787 0.0001 Table 58 Mrmthly Market Average Eamirgs/Prioe Ratios for the Entire Sample Variable N khan Std Dev Minimum Maximum T Prob>1T1 M1 481 0.11126 0.01829 0.07453 0.13141 133.41120 0.0001 M2 481 0.11009 0.01864 0.07289 0.12919 129.54917 0.0001 M 481 0.10903 0.01906 0.07204 0.13242 125.47141 0.0001 MB 481 0.10810 0.01964 0.07109 0.13033 120.69128 0.0001 M4 481 0.10633 0.02251 0. 0.12897 103.58512 0.0001 141. APPENDIXCC Table 59 Monthly Revisiom in Earnings/Price Ratios for the Entire Sample Unadjusted for Market-Wide Revisions Variable N khan Std Dev Minimml Maximum T Prob> 1T1 C2 478 -0.00142 0.00985 -0.04446 0.08681 -3.15336 0.0017 C 373 -0.00042 0.01287 -0.10362 0.06958 -0.62789 0.5305 C3 480 -0.00183 0.01140 -0.08141 0.07987 -3.51829 0.0005 C4 480 -0.00073 0.01031 -0.06997 0.11314 -1.54188 0.1238 Table 60 Monthly Revisions in Earnings/Price Ratios for the Entire Sample Adjusted for Market-Wide Revisions Variable N khan Std Dev Minimum Maximum T Prob> 1T1 E2 478 -0.00024 0.00960 -0.04180 0.08332 -0.54340 0.5871 E 373 0.00049 0.01237 -0.10182 0.06305 0.76269 0.4461 E3 480 0.00016 0.01146 -0.07743 0.08163 0.30715 0.7589 E4 480 0.00105 0.01520 -0.06744 0.12676 1.51055 0.1316 142 APPENDIXCC Table 61 Monthly Revisions in Earnings/Price Ratios for the Good News & Bad News Samples Unadjusted for Market-Wide Revisions Variable N khan Std Dev Minimum Maximum T Prob>1T Good News C2 258 -0.00155 0.00877 -0.04446 0.04422 -2.83096 0.0050 C 196 -0.00020 0.01177 -0.04395 0.06958 -0.23451 0.8148 C3 259 -0.00092 0.01115 -0.07307 0.07987 -1.32281 0.1871 C4 259 -0.00033 0.00813 -0.04128 0.07207 -0.65282 0.5145 Bad News C2 220 -0.00127 0.01100 -0.03412 0.08681 -1.71768 0.0873 C 177 -0.00066 0.01402 -0.10362 0.06206 -0.62956 0.5298 C3 221 -0.00290 0.01161 -0.08141 0.05100 -3.71488 0.0003 C4 221 -0.00119 0.01239 -0.06997 0.11314 -1.42707 0.1550 Adjusted far Market-Wide Revisions 143 APPENDIXIC Table 62 Monthly Revisions in Earnings/Price Ratios for the Good News & Bad News Samples Variable N khan Std Dev Minimum Maximum T Prob>1T Good News E2 258 -0.00025 0.00863 -0.04180 0.04474 -0.47311 0.6365 E 196 0.00078 0.01114 -0.03966 0.06305 0.98609 0.3253 E3 259 0.00096 0.01128 -0.06942 0.08163 1.37429 0.1705 E4 259 0.00241 0.01723 -0.04010 0.12676 2.24880 0.0254 Bad News E2 220 -0.00022 0.01065 -0.03070 0.08332 -0.30702 0.7591 E 177 0.00016 0.01363 -0.10182 0.06155 0.15665 0.8757 E3 221 -0.00078 0.01163 -0.07743 0.05008 -0.99738 0.3197 E4 221 -0.00055 0.01225 -0.06744 0.11312 -0.66212 0.5086 144 AEETNDEXIC Table 63 Monthly Revisions in Earnings/Price Ratios for the Good/Bad News & Large/Small Sirprise Samples Unadjusted for Market-Wide Revisions Variable N khan Std Dev Minimum Maximum T Prob>1T Good News/Small Sirprise C2 131 -0.00265 0.00955 -0.04446 0.04422 -3.17109 0.0019 C 96 -0.00038 0.01184 -0.04395 0.06958 -0.31652 0.7523 C3 132 -0.00202 0.01172 -0.07307 0.05684 -1.97495 0.0504 C4 132 -0.00052 0.01003 -0.04128 0.07207 -0.59498 0.5529 Good News/large erprise C2 127 -0.00041 0.00777 -0.02183 0.03720 -0.59829 0.5507 C 100 -0.00002 0.01175 -0.02737 0.06117 -0.01633 0.9870 C3 127 0.00022 0.01045 -0.03136 0.07987 0.24219 0.8090 C4 127 -0.00013 0.00552 -0.02964 0.01378 -0.26985 0.7877 Bad News/Small Slrprise C2 121 -0.00232 0.00770 -0.03412 0.01330 -3.31709 0.0012 C 95 -0.00187 0.01457 -0.10362 0.04793 -1.24855 0.2149 C3 121 -0.00187 0.01077 -0.05282 0.05100 -1.90877 0.0587 C4 121 -0.00284 0.00946 -0.06997 0.01623 -3.29897 0.0013 Bad News/large Sirprise C2 99 0.00000 0.01396 -0.02723 0.08681 0.00408 0.9968 C 82 0.00073 0.01331 -0.03662 0.06206 0.49730 0.6203 C3 100 -0.00415 0.01248 -0.08141 0.03365 -3.32286 0.0012 C4 100 0.00081 0.01500 -0.03787 0.11314 0.53771 0.5920 145 APPENDIXIC Table 64 Monthly Revisiorh in Farrfings/Prioe Ratios for the Good/Bad News & large/Small SIrprise Samples Adjusted for Market-Wide Revisions Variable N khan Std Dev Minimum Maximum T Prob>1T Good News/Snell SIrptrise E2 131 -0.00113 0.00927 -0.04180 0.04474 -1.40020 0.1638 E 96 0.00068 0.01090 -0.03966 0.05865 0.60963 0.5436 E3 132 -0.00032 0.01182 -0.06942 0.05890 -0.31584 0.7526 E4 132 0.00030 0.00998 -0.04010 0.06939 0.34825 0.7282 Good News/large Srrprise E2 127 0.00065 0.00784 -0.02116 0.03967 0.93886 0.3496 E 100 0.00089 0.01143 -0.02925 0.06305 0.77654 0.4393 E3 127 0.00230 0.01057 -0.02960 0.08163 2.45442 0.0155 E4 127 0.00459 0.02224 -0.02944 0.12676 2.32823 0.0215 Bad News/Stall Surprise E2 121 -0.00147 0.00747 -0.03070 0.01454 -2.16016 0.0327 E 95 -0.00112 0.01432 -0.10182 0.03700 -0.76192 0.4480 E3 121 0.00004 0.01063 -0.05035 0.05008 0.04197 0.9666 E4 121 -0.00212 0.00957 -0.06744 0.01546 -2.43428 0.0164 Bad News/Large Snprise E2 99 0.00130 0.01344 -0.02589 0.08332 0.96522 0.3368 E 82 0.00164 0.01270 -0.03025 0.06155 1.17129 0.2449 E3 100 -0.00177 0.01271 -0.07743 0.03749 -1.39532 0.1660 E4 100 0.00136 0.01469 -0.03552 0.11312 0.92403 0.3577 "71111111111111.1110?