‘The Information Content of Annual Earnings Announcements William H. Beaver Journal of Accounting Research, Vol. 6, Empirical Research in Accounting: Selected Studies 1968, (1968), pp. 67-92. Stable URL: bhtp:flinks,jstor-org/sici?sici~0021-8456%28 1968 %296%3C67%3ATICOAB3E2.0,CO%3B2-K Journal of Accounting Research is currently published by The Institute of Professional Accounting, Graduate School of Business, University of Chicago. ‘Your use of the ISTOR archive indicates your acceptance of ISTOR’s Terms and Conditions of Use, available at htp:sseww jstor org/aboutiterms.html. ISTOR’s Terms and Conditions of Use provides, in part, that unless you hhave obtained prior permission, you may aot download an entie issue of a journal or multiple copies of articles, and ‘you may use content in the ISTOR archive only for your personal, non-commercial use Please contact the publisher regarding any further use ofthis work, Publisher contact information may be obained at ftps jtor.orwournal/erad-uchicago hil. Each copy of any part ofa JSTOR transenission must contain the same copyright tice that appears on the screen or printed page of such transtnission, ISTOR isan independent not-for-profit organization dedicated to creating and preserving a digital archive ot scholarly journals. For more information regarding ISTOR, please contact support@jstor.org. hup:thrwwjstor.orgy Wed Aug 30 13:24:37 2006The Information Content of Annual Earnings Announcements WILLIAM H, BEAVER “The information content of esmings is an issue of obvious importaneo aad is a focal point for many measuremout controversies in accounting. This paper empirically examines the extent to whieh common stoek investors peresive earnings to possess informational value, The study dirocts its at- tention to investor rezetion to earings announcements, as reflected in the volume and price movements of eommon stocks in the wooks surrounding the announcement date. Valuation theory has long nosited a relationship hetween earnings and the value of common stock. Miller and Modigtiani postulate that one im- pottant element in determining the value of common stock is the product of earnings times the appropriate earnings multiplier for thaé risk class.‘ Graham, Deda, and Cottle take s,similar position with respect to the com- putation of their “infrinsie value” of eommon stock securities? MM also provide empirical evidence that suagests if roported eamings are adjusted for measurement errors through the use of instrumental variables, the ad- justed earnings are usefut in the prediction, of the matiet value of electric utility firms. Tn fact, the evidence indteated that the earnings term was the roost important explanatory variable in the valuation equation The rela~ tionship is « necessary condition for exsnings to have information content, * Apsistant Professor, University of Chictgo, + Marion H, Miller and Franeo Madiglisai, “Some Batimates of the Cost of Capital ‘to the Hlecivic Uibty Industry, 1954-87," Umarisan Fecaomie Resiow, LVI (Sune, 1900), 281. 1 Henjamin Graham, David L, Dodd, and Sidney Cottle, Securily Analysis (New York: MoGraw Hill, 962), 48H. ‘Miler and Modialiani, op. eit p. 878. Tho thevst of their study wae directed to ‘value of the frm rather then value of aammnon stack. However, the MM adel con- {sino an explisit relationship between the ‘so, and the statematta made above apply swith equal foree to the value of common stack, o68 RAPIAICAL ANSEARCH IN ACCOUNTING: SELECTED STUDIES, 1968 Laut the evidence does not preclude the possibility that the opposite may be true, ‘Although there are many reasons for adopting the position that earnings luck informational value, to are frequently offered. (1) Measurement errors in earnings are so large that it would be hetter to estimate the value ‘of common stock directly from the instrumental variables rather than use earnings as an intermediate step. (2) Even though earuings might convey information, there are other sources available to investors that contain essentially the same information but, are more timely. By the time annuel eamings are released, any potential information content has already been processed by investors and is impounded in the market price. The implica- tion of both arguments is eamings reports have little or no information content, The ise is of major concern to the accounting profession because its outeome directly refleots upon the utility of the accounting activity. One approach to examining this issue is to specify an expectations model af how investors relate reported eamings to market prices. ‘The paper presented bby Benston at last yeat*s Conference fallosced such an approach * Benston found price changes wete largely insensitive to earnings, which taken at face value is unfavorable to the utility of earnings deta, But such results ae always difficult to interpret hecanse the lack of an observed relationship may be due to cither one or both of twro factors. Either no relationship existe or the expectations madel was improperly specified. Pt fs impossible +o determine the extent to which the negative findings are due to the letter rather than the former. ‘The approach taken here is to anply tests that require no assumption, about the expectations models of investors, Note that the issue under con- sideration is of a positive rather than a notwative nature—that is, the question of concer is not whether investors should react to earnings but rather whether investors do react to earnings. Definitions of Information Content Information has been defined as a change in expectations about the out come of an event.* Within the context of this study, a firm’s earings report is said to have information content if it leads to a change in investors’ asesments of the probability distribution of future retums (or prices), such that there is a change in equilibrium value of the current market price.* “George J. Benston, “Published Corporate Accounting Data and Stock Prioss,” Prpivical Research tu Acccunting: Selected Studies, 1067, Supplement to Val. 5, ournal of Accounting Research, Dp. 1-96 ‘Fienri Taeil, Heonontics and Iyformation Theory (Chicago and Amsterdam: Rand MeNally and North Holland Publishing Company, 1967}, Cb. 1. ‘A further stipulation is often mada tbat information concoras changes in expecta. ons about an event that isa paramater ofa decision sods, Defining earnings informa ‘an in corms af its Impact an fate rettms (ar press) ie consistent with that fertherCONPERE OF ANNUAL EARNINGS ANNOUNCEMENTS 60 Although neither the direction nor the magnitude of the price change can bbe specified without knowing the expectations modells) of investors, the variability of price changes is likely to be greater when earnings are an- nouneed than et othor times during the year? “Another definition of information states that not only must there be 2 ‘change in expectations but the change must be sufficiently large to induce ‘a change in the deeision-maker’s behavior. According to this definition, a firm’s earings report possesses informational value only if it leads to an altaring of the optimal holding of that firm’s stock in the portfolias af in- dividual investors. The optimal adjustment. might be to buy more shares or to sell some or al of the shares already held. In either event, the shift in portfolia position would be reflected in the volume. If earnings reports have information content, the number of shares traded is likely to be higher when the earnings report is released than at other times during the years Relationships between Price and Volume Tasts ‘The relationships posited above are consistent with the economist’s notion that; volume reflects a lack of consensus regarding the priee, The lack of consensus is indueed by a new piece of information, the earings report, Since investors anay differ in the way thoy interprot the report, sora time may elapse before a consensus is reached, during which time increased volume would be observed. If consensus were reached on the first transae- tion, there would he a price reaction but no valine reaction, assuming homogeneous risk preferences among investors. If risk preferences differ, there still could be a volume reaetion, even after the equilibrinm price had bbeon reached. An important distinction between the price and volume tests is that the former reflects changes in the expectations of the market as a whale while the latter reflects changos in the expeetstions of individual investors. A stipulation. For support, sea the literature on portfolio thscry, expecially Harry M. Matkowite, Perifolio Selection: E@lciont Disereifeation of Envecimanls (New York: John Wiley & Sone, 1992). The change [n equilibrium price i in addition co any priee change that would normally acer in the aheence of ary earmlng? announcement, The aaesmption ia ‘bat. the to price changes are positively correlated, independent, oF mildly core- lated. If there were atrong negative correlation, the prioe cbange variability might hot be gevater at the ancouncement date, In the light of Previous vesearch in the Dehavior of sosurty prices, the assumption af independence iz moct likely to be the ‘eurrect ane, See Buigene F. Ramos, “The Bebavinr of Stock Market Prices," Journal of Business, XXXVIL (lanuary, 1965), 24-108, Au final parenthetical comment on defisitions of information, note thet radu ion of xacerainty was not one of the definitions chosen. It should be apparent chat Ina dynamiceituation (., where probability dietribution aseozements are changing ‘aver time}, a decision maker may be more uncertain abou a given evant after racaiv- ing 2 mescage about the event than he was hefore be reatived the message. To Use ‘Theil’s terminoloay, the entropy may inesesse as a rasull-of a message, yet the mes ‘sage bas information eaptent. See Theil, cp. ell, Ch. 270 WiMuAM 1. Raven pieve of information may be neutral in the sense of not changing the expec- tations of the market as a whole but it may greatly alter the expectations of individaals, In this situation, there would be no priee reaction, but there ‘would be shifts in portfolio positions refleated in the volume. Because the price reflects the expectations of many investors, it may iraply a very eff cient forecast of earnings for several weoks prior to the announcement? If 50, the price tast may be leas sensitive than volume to earnings reports. ‘The foregoing discussion suggests that a resetion may be observed in only one of the tests or thal the two tests may not. respond equally. IF neither test responds, the utility of earnings data and the study's sample design will be suspect. Sample Design Selection of Sample, The study is based upon a sample of anaual earnings announcements released by 148 firmas using the years 1961 through 1966. Six criteria were used in the selection of the sample firms, () The firm must be on the Compustat: tape; (2) the finm must be a ‘member of the New York Stock xchange; (3) the fiseal year anust end on a date other than December 31; (#) no dividends were announced in the same week as the eamings announcement; (5) no stock splits were announced during the 17 week period surrounding the announcement of camings; and (6) there were less than 20 news announcements per year appearing in the Wall Stret Journal. Table | indicates the extent to which each eriterion affectot the sample size, Criterion (I) was selected because the Compustat population represents over 90 per cent of the total market value of the common stocks of publicly held corporations and hence is a relevant population for study. A secondary reason is the ease with which financial statement items ean be obtained for ‘the Compustat firms relative to firms not-on the tapes, Although no finan ial statement date are needed for the earlier phases of this study, even ‘ually the eoope will be extended to relating market prices to the financial statement items, namely the earings numbers. Criterion (2) was used because weekly nriee and volume data on NYSE firms aro relatively eaay to obtain, The Center for Research in Security Prices (CRSP) provided tape: which contain daily price, volume, and transaction information on all fleas on tho NYSE for the years 1961 through 1965.2 ‘Criterion (3) was celocted in order to avoid a clustering of announcement * Bicieucy ia delved in terme of E(2 — 2*)8, where $ is the forecasted value of reparied earnings and 2 ie agtusl value. The closer the expostation is t 20r0, the ‘mare efficent the forenset is. Note that & foresast may be unbiased but very inal cient. The distinetion betwosn oficloney and unbiasedness is more important ta the interpretation of the Sodings presented Iotar "Without the cooperation of ORSP, the data collection chare sould have heen overwhelmingCONTENT OP ANNUAL, FARNINGS ANNOUNCEMENTS 71 dates during any time period. Without this eriter‘on, the sample data would exhibit a farge clustering of announcements in the months of Feb- rusry, March, and April because two out of three Compustat firms are 12/81 firms, In subsequent analysis, an attempt will be made to remove the effocis of market-wide events from the individual soourity's volume ‘ad price data. When camings announcements cluster, they become form of market-wide price indexes and the volume statistics. Hence, any at- tempt to remove the effeets of market-wide events would eliminate the cffeots of the earnings repart as well. ‘The purnoces of eriteria (4) and (5) ate similar in that they attempt to minimize sny ambiguity asaciated with an ohserved reartion in the week: of the esraings samouncement. If these criteria. were not. applied, there would be a joint effect, and it would be extremely difficuls to separate the announcement effeats af dividends or stnok splits from thase of the earnings report." Criterion (6) was chosen 60 that there would be weeks where few, if any, announcements were released. To the extent that: news items are announced in wesks other than the earings snnouneement week, compar ing those weeks with the earnings snnouincement weels compares the infor ‘mation content of the earnings reports with that of other types of news annoncements, which is nat the issue under study. Both the direction and magnitude af any potential bias introduced by the selection criteria are difienlt to as:ens. Criteria (1) and (2) led to the selec tion of the larger firms in the economy. The average total axsets (per finen- cial statements) for the 148 firms in 1965 was 167 million dollars, and their ‘average market value of common stoek outstanding in 1965 was 189 million dollars. The effect of selecting lamger firsns would tend to induce a bias aguinet earnings reports because the larger firms are generally associated with 2 greater flow of additional information than stualler firins ‘The offect of criterion (8) was twofold: (a) Out of the subpopulation of ‘Compustat, NYSE firms, the criterion tended to select the smaller firms, ‘oven though they axe probably still larger than average for the economy as a whole, In 1965, the average total assets far Computsat, NYSE firms wore 441 million dollars, and the average market value of common stock outstanding was 564 million dollars. (b) A grester proportion of retailers and food processors appears in the caraple than would have been obiained if firms had not been reséricted non-12/31 firms. Retailers comprise 14.6, per cent of the sarnple, while 17.5 per cent of the frmns are food processors. ‘Phe expected percentages would have been 6.8 and 10.0, respectively, if 44 pilot study with similar objectives did not exclnde frms with dividend an- nouncements ia the same week ss esrnings, The investor resction in terms of volume was almost tied as larga ae tho reaction abserved in this etudy- Stock splits were cxchided! hecatse previous eeonazeh has found that stock epita posmeae information conten, Sor Eugene F. Fama, ol, “The Adjustment of Stock Priess to New Tn formation,” Report 6705 (Canter for Mathematical Studien In Businars and co nomics, Graduate Sebool of Business, University of Trioago, 1957), forteamaing im the International Heoonec Rosi,72 wituam , peaven 4 random sample were drawn from the Compustst, NYSE eubpopulation. With respect, to the implications for information content, there are to be zo obviows reasons why there firms would constitute a biased sample, with the ane exception that retailers tend to report financial statement: dats, monthly whieh would tend to induce bias against finding informational value in annul eamnings reports. In fact, in the analysis described Tater, both the price and volume reactions were less dramatic for the retailers and food processors than for the other firms in the saraple, Tt is possible that the selection eriteria, especially eriterion (6), ray i duce sotne bias in the opposite direction. As iong as the etiteria are visible ‘ex anie, the population for which the study's findings are relevant can be easily identified, Also, the sample criteria can he relaxed in future studies to discover the generality of the findings presented here for other popula- tions. Data, Collection, The fret step was the identification of firms that would comprise the sample, Meeting the eriteria in any one of the five years (1961-1945) was suficient condition for a firn’s inclusion for that year. ‘The result was a sample of 143 firms, Because all firms did not meet the criteria in every year, the 143 firms gave rise to 606 annusl earnings an- nouncements, The date of the earnings announcement was obtained frox the Wall Street donrnal Inder. ‘The distributions of financial statement dates and announcement dates appear in Table 2, Restrieting the sainple to non-12/31 firms was suocess- ful in reducing the clustering of dates. ‘The most frequent month in which the fiseal year ended (June} represents only 23.8 per cent of the sample, while an unrestricted eaanple would have resulted in 67 per cent in single month (December). With respect to announcement dates, the highest three-month period (September, October, and November) contains 47.6 per cont of the announcements, while under an unrestricted eampling procedure the highest percentage would have been approximstely 87 per cent (during February, March, and Aptil). The most frequettt month of announcement october) represents 13.4 per cent of the announcements, which, is only slightly higher than the perventage that, would be obtained under 2 com- pletely mniform distribution throughout the year (9.1 per cent). ‘One by-product of the data gathering was some insight into the time Jag between the financial statement date and the announcement date (see ‘Table 3). The median lag was 9 weeks, only 8 per cent of the announce ments were made by the end of 4 weeks, and 93 per cent of the earings hhad been reported by the ond of 18 wacks. A posible avenue for future research would be to study the information content of the time lag itself (e.g, is “bad?” news reported less repidty than ‘ good” news?}. Definition of Variahlet. The noxt step was to compute the following ‘variables for each firm on a weerly basis for the 261 weeks (from January 1, 1961 to December 31, 1985):CONTA OF ANNUAL BARNINGS ANNOUNCEMENTS 73 no. of shares of firm i traded in week ¢ 1 Vag = ———— adel in week yg __1___ no. of shares outstanding ~~ no. of trading for frm iin wook ! days in week t no. of shares traded for all NYSE firms tue in week t 1 ‘no, of shaves outstanding ~ no. of trading * for all NYSE firms days in week ¢ jin week ¢ cash dividend “paid” on share af firm 1 in week & losing price for share of firm # at end of week f, Pies = closing price at end of week ¢ — 1, adjusted for capital changes (eStock splits and stock dividends), {SP}, = closing value of Standard and Poor's Prive Index ai end of week t, (SP). = closing value at end of week ¢ — 1, Vn is a woeldy average of the daily percentage of shares traded. Weekly ‘volume was divided by the number of shares outstanding so that the results, ‘would not be dominated hy those firms with the largest mumber of shaves outstanding, The percentage of shares traded per week were then divided by the number of trading days in order to adjust for the fact that not all vwooks have the same number of trading days Viysroflets the level of volume for all NYSE firms, The weighting, scheme ‘implicit in this volume index assigns greater weight to nercentaue of shares inadod of firms with the larger number of shares outstanding. While this feature is not entirely satisfying, its use is defended an the grounds that this index is much easier to obtain than an index that assignt equal weight in all firms and because there is no ressan to believe the we of this index leads to either an upward or a. downward bias in the findings regrding the. information content of earnings report, Rais the naizrat logarithra of the priee relative and ean be viewed as ‘measure of price change or as the rake of return of the security assuming continuous compounding.* Ry, is x similar theame for 425 industrial "The propertion af Ry ate further deterTued in Fama, op. eft; Benjamin ¥. King, “Market and Industry Pactars in Stock Price Behsvior,” Journal of Busesees, XXXLK (anvary, 1988), 159-40; James H. Lorie and Lawrence Fisher, “Bales of Return an Investment in Common Siooks," Journal of Buvinese, SSVI Gaeunry, 994), 1-2174 WIDKAM #, BEAVER NYSE firms. This statistic has some limitations ay a tarket-wide index of priee change and in many respects is less preferable than some recently developed indexes, notably Fishet’s Link Relative! However, again, its tse is defended on the satti grounds as those for the macket-wide index of volume, The Fisher Link Relative has been computed for monthly data only. To construct » sintilar index on a weekly basis is a research project in itself. Not only is the 8 & P index easier to obiain but it was found in ‘other studies that results were insensitive to which index is used. Within the context of this study, there is no reason to believe that the use of the 'S & P based index will lead to an overstatement or understatement: of the information content of earings reports. Volume Analysis — Unadjusted for Markel Influences Vs. was computed for each week fin the report period for each of the 506, ‘earnings announcetnent j, The report period is defined as the 17 week period surrounding the announcement, date. (8 weeks before the announcement: week, and 8 weeks aiter). Then the 7, (averaging across j) was computed jor each of the 17 wees, and the results appear in Figure i, ‘The dotted line denotes the value of V+ in the nonreport period (i.e. that portion of the 261 weeks not included in the I? week report periods). ‘Phe evidence indiontes a rather dramatic inerease in volume in the an- nnouncement: weal (week 0). In fuct, the mean volume in week 0 is 33 per cent larger than. the mean volume during the nonreport period, and it is by far the largest value observed during the 17 weeks. Investors do shift portfolio positions at the time of the eamings announcement, and this shift is consistent with the contention that earnings reports have informnation content. ‘The eontention is further supported by the behavior of investors in the other werks, Bight weeks prior to the announcement, volume is below normal, which suggests that investors may postpone their purchases ed sales of the eccucity until the earnings report is released. The four weeks after the announcement, wher the annual reports are reecived, exhibit, slightly above normal volume and henee permit a mote thorough evaluation of tho earnings data. ‘Tho investor response appoars to be very rapid, for almost all of the above-normal activity occurs during week 0. This finding supports previous studies that also show investors respond quickly (as sefleoted in price changes) to new pieces of information (see Fama, et al,). Perhaps some comment is in ordor regarding the averall level of volume throughout: the year. The volume statistics reported in Figure 1 are multi- “For & dleaussion of the 8 & P Index vi.A-vie Fidherte Tndex, aoe Levwrence Fisher, "Soma Now Stock Market Indices,” Journal af Buaiaase, XXXIX Ganuary, 1906), 191-225, Pama, otal, op. ellCONTENT OP ANNUAL EARNINGS ANNOUNCEMENTS — 75 plied by the fuetor of 10°. The average volume in the nonreport, period is -OOLL2~that ig, the average daily pereentage of shares traded is slightly greater than one-tenth of ane per cent of the shares outstanding, ‘This imt- plies att annual turnover of approximately 25 per cent and a weekly turn- ‘over of 5 of ome per cent. If corporation X has 10 million shares outstanding, during a normal week 50,000 shares will be traded with aa. expected volurne ‘of 66.667 shares during the eamings report woek. Volume Analysis—Adjusted for Market Influences ‘Tho section will present aa analysis whieh attempts ta comove the effects ‘of market-wide events upon the individual security's volume, The motiva~ tion for the analysis is two-fold. (1) This possible that the abrormally high volume may be caused in part: by market-wide pieces of information that aro released at the same time as the earings announcements, Since the ‘earnings announcements are released almost uniformly throughout the year, this is not s very plausible explanation of the findings. Nevertheless, re- moving the maricet wide effects should allay any fears that this unlikely situstion doas account for the results. (2) More importantly, the analysis will serve to reduee “noise” in the volume data, Noise is any movements in volume due to unspecified factors, one of which is market-wide events that woald cause incresses in the volume. Analysis for Nonseport Period. The following model wes used to abstract from market-wide factors:!5 Ver = ach bac + en Batimatos of ay and b, were obtained from linear regressions basod upon observations frazn the nonrepor’ period. ‘The obsecvations from the report pesiod wore deleted from the regression because i earaings annotneements have information content, the assumptions of the elsssieal regression modet are violated duting the report period (e.g £Gi) # 0). Some summary statisties relating to the regressions appear in Table 4. ‘The mean volume of the sample firms is much higher than that, of the market index. One reason is the diferent weighting scheme implicit in each measure. The market index assigns greater weight io firms with the grester number of shares oustanding. I these firms have Tower volume (expressed as a percentage of shares outstanding), then the market index would be expected fo have 1 lower mean. Another explanation is that the sample "hg rationale for using this particular model is two-fold: (1) Tt is a simple relationship, and there is no otiious reason why moro complex model would be ‘more appronviace, @) Ics anslognus to the model that wil be used to remove effects ‘of market-ide evunts upon the price chonges of individual securities. The paper will Tater indioate that such a model ems to be & reazorable way ta eharse'or'se prlee changes, Hence, it would seem reasonable to aasume a similar process ia generating volume aver time a well76 winiaM H, weaver selection criterion implicitly favored higher tumover securities, But it is not obvious why that should be true nor witat implication it has for inferences regarding information content. ‘The average correlation coeficient was low, implying that removing the influence of Pure should have little effect upon the analysis. In spite of the Jow association, the sign of the comelation coefficient, was positive for 139 firms and negative for only 4, These two findings taken together sugnest that the market influence on an individual firm’s volume is significantly different from zero but that its magnitude is small.!* ‘The residual, uc is that portion of an individual seourity’s volume that cannot be expisined by market-wide events as reflected in Vie. The mean of (averaging anross time for given firm i) is forced to be zero by the ‘mechanies af the regression camputations. However, the mean of ¢, {aver~ age actos firms for a given week £) may be nonzero. An inspection of its distribution for the 261 weeks provides some interesting insights (see Figme 2). ‘The distribution is skewed to the right, as indicated by the fact thst 88 per cont of & are negative and 42 per cent are positive. The tedian of is — 02 and its mean is zero (again this must be true becsuse of the me- chanics of the regression compntations). ‘The eqs are even more asym- metrical, with 646 per cent negative snd 35.4 per cent positive. One inter- pretation of the asymmetzy is that information is provided to investons in discontinuous “humps” rather than smoothly or continuously over time, Residual Analysis for the Repost Period. ‘Tho vosidusl, ess, was computed for each wook of the eeport period for each of the 608 earnings announce- ments j in the following manner: f= eee — be gu 506 t= Vie : PB eB where as and by were obtained from the rogressions in the nanreport period.” ‘Then the é, was eomputed for each of the 17 wooks, and tho results appear in Figure 3, A positive residual implies above normal volume; negative, below normal; and zero, normel volume. ‘The bebavior of the volume residual is the same as that of the previous smalysis, Thove is a large poak in week 0, where the mean volute is ab- proximately 30 yer eent higher than during the nonreport petiod (i.e, -33/1.12, mean residual in week O/mean volume in the nonseport pesiod} ‘and is about 40 por cent higher than the mean volume in the weeks prior WCThe probability that the expacted value ofthe earrlation coefficient is leas than ‘om equal ta sera ie ese than 2 ezance in 100,00, Note that the subscript {refers to rm for secsfty j but, eters ta an earnings announcement. Hence a sad dy may be used a maximum of five times; (és Frequency of wwe will depend upon. the momber of earrings anvouncamonts of Bim # or security Cineluded in the sample of 506 snnouncemensCONTENT OF ANNUAL EARNINGS ANNOUNCEMENTS — 77 to the announcement. Agnin the volume during; these. weeks is abnormally low, while slightly shove norma! valume persists for four wees following the announcement, week. The interpretation of these findings is the samo. a5, ‘hat of the previous analysis In short, tho results aro very eonsistont with the contention that earnings announcements possess information content. Because & comparison of mean values ear aften be misleading, two addi tional compsrisons were made to see how unususl #s an @, of 88, which was the value observed in week 0, The first comparison examined the values of é in the report period and those in the nonreport period (see Figure 2). Out of the 261 nonreport petiad values af 2, only 4 had values exceeding 33, Although such a comparison is admittedly 2 erude approxi- ‘mation, it does suggest that the valve in week 0 is unusually high. Moreover this comparison tends to understate the unusual nature of the week 0 residual, The é, during the nonreport period is based upon a maximum of 143 observations per mean, while the éy in week 0 (2s well as the rest of the report period) was hased upon 506 observations. Since the e's ere less shan perfectly correlated, the dispersion of the distribution of & would desteass as the number of observations per mean. increases, Hence, if the distribution in the nonreport period were also based upon 506 observations per mean, its dispersion would be smaller and the number of values ebove .38 would be fewer. Another factor leading to an uader- statement is that & in the nonreport period was based upon residuals taken from the same week, while the mean residual in week 0 was bared upon observations taken from different. weeks. If contemporaneous residuals fre mote highly correlated then nomeontemporeneous residuals (and the evidence suggests they are), then a distribution of in the nonreport period based upon noncontemporaneaus observations would have a smaller dispersion and fewer values shove 33. Tho msjor paint is the eamparison indicates that the mean residual in week 0 is unusually high, inspite of the fret that the comparison tends to understate how unusual it really i. ‘A second comparison involved the analysis af the frequency af positive residuals in each veport petiod week as compared with the number duving the noureport period (see Figure 4). The behavior of the positive residuals is consistent with the previous relationships abserved in Figures 1 and 3. Prior to the announcement, the frequency of positive residuals is below that of the nonreport petiod, while the frequency is slightly ehove normal after the announcement, By far the largest frequency accurs in week 0, and there is an extremely small probability that such 2 high number of positive residuals could have occurred by chance."* This second comparison sux- ‘gpsts the same inference drawn frotn the firsi-—natnely, the volume in week Os an unusually high value, In eutn, the behavior of volume uniformly ‘Ske meal correlation in reflected in the positive astooorrelation cooficient of the rexidusla (ace Table 4). Another indicatian ia th the fave wales of @, exceeding 48 occured in.a fiveaweck period, The probability i less than 1 chance in 100,00,78 wroecase 3, emAVER, supporis the contention thet eainings have information content for individ uel investors, Ta some respects, these findings do not reloot ho entire extent to which activity is above normal in weok 8. Not all of the earnings anaouncemonts of the 143 firms were used—in faet, only £05 out of a possible 715. Tho 17 ‘weok periods surrounding the remaining 209 are included in the nonreport, potied. This will tend to induce a bizs agzinst earnings reports since volume setivity is inercased in the nonreport period by the inclusion of the 209 “report periods.” The extent of this bias could be serious because one of the ressons for placing # report in the 200 group was the announcement: of oamings and dividends ia the same week which would produes avon ‘more price and volume aetivity than the 606 announcements studied. How. ever, there are also compensating factor. Although the setivity in week 0 ‘was shove normal, the activity i the weeks prior were below normal for the 506 observations. If this tends to be true of the deleted announcements, as well, the bias may not be 40 great. If the 209 observations were deleted rom the nonteport period, to be completely eonsistent, ather types of news announcements would also have to be deleted for the same ressons. The result would be virtually no observations in the nonreport period, Since the noureport period does include these events, ifs important ta stress the fact that comparing the eeraings report petiods with the nanzeport period involves a compatison of the information coatent of earnings reports with the average amount of information being released duting the nonrepart pesiod. By necessity, this is = bias againat earnings reports éince the sppro- priate comparison would bea noureport period with no information at all, Price Analysis— Adjusted for Influence of Markel-Wide Buents 1 caraings reports convey information in the sense of leading to changes in the ogulibrium vaiue of the current market price, the magoitade of the price change (without respect to siga) should be larger in week 0 than during the nonxeportperind, Tho first step in making this predietion opera- tions! is Yo remove the affect of markat-wide events upon the individual security's price change. The reasons for wishing to abstrast fram these events are similar to those cited in the volume analysis” ‘The model used here was first sugested by Sharpe, and it provided the motivation for using ‘an analogous modal for volume?" The Sharpe modal states: eye = ast bullae + vice Riis ameasure of the price change of seourity i during time period t, and Rye is a measure of average price change during time period ¢ for 425 in- dustrial NYSE firtns, Both variables wete defined earlier. The residual, Bee p. 75. A William F, Sharpe, “A Simplified Mode for Porttlio Analysis,” Boaagenient Science, TX Genuary, 198), 27-88.CONTENT OF ANNUAL RARNINGS ANNOUNCEMENTS — 79 ‘un, represents that portion of the individual seeurity’s price ehange that cannot be accounted for by the effects of market-wide events as reflected fin Race ‘The Sharpe model has been investigated by Fama, of a. and by Scholes and was helpful in abstracting from the influence of market-wide factors. King’s study of monthly price changes found that, an the average, 81 per ‘ent of the variation in an individial security’ price change can be ex plained by market-wide factors ca releoted in 2 market-wide index of price change. For thote reasons, price change analysis, tinadjusted for the influence of maxket-wide factors, was not conducted. The evidence will Tater indicate that if such an analyeis hed been conducted, the results would be essentially the same as thase reported here. Since the direction of the price change cannot be specified, a knowledge of the investors" expectation medel(s), some transformation of sa that abstracts from ite siga, is needed. One such transformation is the square of the residual (ie, 1). If earnings reports possess information content, ‘ul, should be greater during week 0 than dueing the nonreport period. The mean of sf, during the nonreport period is simply the varianee of thai variable (2) ‘The relationship between the squared residual in week 0 and the average squared residual during the nonreport period enn be expressed in the form of the ratio, ce, whote the numerator is wf. and the denominator is 8, LF the ratio is greater than one, the residual price ehange is larger than normal, and canversely for a ratio of less than one. ‘The prediotion is the mean of U (averaging actoas announcements) will be greater than one during week 0, if earnings reports possess information content, Analy of Nonseport Period. Hetimates of as, bs, and 8 were obtained from regressions based upon the nonreport, period. ‘The observations from the report period (Le., the 17 woeks surrounding each snnounesmient) were deleted from the regression hecsuse if esrnings have information eontent, the sssumptions of the classical regression model sre vielated during the report period (e, the varianee of the residuals during the report. petiod is not equal to the variance during the nonreport period}. Some sunmary ststisties relating to the regressious appear in Table 5. ‘The mean price changes iand to be lower for the semplo firme than for the runrket index. Since the Ri can. also be interpreted as a rate of return, the Jower retums for the sample firms would suagest that they are loss risky ® Fama, eal. op. cits Myron Scholes, “Too Bifects of Secondary Distributions fagon the Marke! Price (paper presented at the November, 1967 seasion of the Conference for the Study of Security Prices held af the Gradanta Schoo af Busines, University of Chistgo); and King, op. ci. The porsentage rors ca the potid AumuRt, 1952 chravgh December, 1000. The varianoe aw Flas, — Elus) #62 is she estimate of 72, gomputed from sample datas? [Solana VP, where — number of weakly cheervatlons for fhe nonteport peri for sextrity«80 WILLIAM H, BEAVER than the firms comprising the index. An inspection of the disteibution of by also lends support to thst contention, Shaxpe states that b, ean be viewed as an operational measure of a security's riskiness, with larger values of be implying greater riskiness** A by of ome denotes a security of “averaze”” riskineae, The average bi for the sample firms is less than one ($9), which suggests that the sample firms are less risky. However, the diseussion in the section on definition of variahles indicated that the definition of Ry, hased upon the S dP index may be subject to measurement error. An erratii variables model suggests that measurement error in the independent varia ble, even if it has a zero expectation, will induce o downward bias in the estimates of the rearossion coefliciont associated with tho indapendent variablo (.e,, 54). Eidoris woro undorteken to assess the extent, of the downward bias by computing 0: for the sample firms, using monthly data and Fishor’s Link Relative as « definition of Rs. The median b; was 993, suggesting the semple firms are of average riskiness relative to NYSE firms (L.e., the firms that eomprise the Fisher Index). On the avorage, the association betwoon Ass and Rus was low. Only 6 por cent of the variation in Rs esm be explained hy tho varintian in Ruy ag measured by the square af the average correlation coefficient. ‘The im- plication is two-fold: (1) Removing the influence of Rys should have ltile ‘effect, upon the results, rlative to what, would have been obiained if Fe ‘were analyzed rather than ut. (2) The explanatory power is much lower than that obtained by King, suggesting that ether weekly data have more noise than monthly data of that Ry, was not properly defined, oF both, The presence of either faator will make it more dificult to detect any price effects of the exrainge reports. ‘The distribution of C, (averaging seross 143 firms, t = 2, --+,261) during the nonreport period is shown in Figure 5. Te will be used as a bs for assessing the significance of the Us observed during the report neviod.** Price Residual Analysis for Report Pericd. The residual, ue, was com- puted for each week ¢ of the report period and for each of the 606 earnings announceraenta jin the following manner: ie ie 8 uy = Ry — as ~ dR Pe le 506 Ba AB FS. ‘The residual was then squared and divided by the varianeo of the residuals for its firm during the nonreport period, as fallows: ‘William P. Sharpe, “Capital Anset Prices: A Theory af Market Equilibrium ‘under Conditions af isk,” Journal of Pinance, XIX (Beptember, 1964), 425-42. 264, Johnston, Feonomatrie Mathody (New York: McGraw-Hill, 1962), 148 The distribution ia skewed to tha right. Qne explanation for this phenomenon ia tye Ieptoleretio nate af the underlying we’'s oe Fama, op. cit). The diatribation of ass al skesad in the seme dlesetion. Although the mean of wis one for each secutity during the nonreport period, only 26 per cent of the ahservationa exceed one.CONTENT OF ANNUAL RARNINGS ANNOUNOEMENTS — 81 ie Dy = Wiel? i in -8, 1G, (averaging across j) was computed for each of the 17 woeks of the report perfod, end the results apposr in Figure 6. ‘Dhe magnitude of the priee changes in week Ois much largoe (G7 per cent higher) than the average during the nonreport period, The shove novmal price activity is what would be expected if changes in equilibrium peices are more likely {0 oceur when oarnings reports were released, and henee the evidence is very consistent with earnings reports possessing informa- tions. value, Although the price activity is highest in week O, the next largest, values ceeur in the weeks immodiately contingent to week 0. Prico changes are above average in the week immodiately prior to the announcement, which ‘may reflect information leakage or the fact that the Wald Street Journal ‘was not the first souree to report. the earnings in some cases. Above normal activity is alsa present for two weeks after the annomeement, during which, time the annutal reports are released and are evaltated by investors, ‘The below price activity in weeks —8 through —2 is open to at least two interpretations: (1) There is a below normal amount of information coming onto the market at this time. (2) The below aormal price activity is a result of the below normal volume also obwerved during the same petiod. More will be said about both (1) and (2) later. ‘The behavior of the mean residual, i aléo indicates greater price activ~ ity in week 0 (see Table 6}. The mean in week 0 is 00600, which is the largest value observed during the 17 weeks and is four ties larger than the aver- age value of Ree during the nonreport, petiod (.00125, see Table 5). The means give the impression that serial correlation may be present in the data, However, the average autocorrelation of the price residuals was quite low (—.08) during the report period. ‘The low degree of austovortelation supports the similar findings of Fama and his conclusion that the market moves to new equilibrium poritions quickly. Further evidence of this ‘s reflected in the fact that the bulk of the price resetion does occur in week 0 (see Figure 8). The low sutoeorvelation also suggests that the price changes were permanent in nature and wete not reversed in subsequent weeks. In. fact, the autocorrelation of the residuals in the weeks immediately after the snnouncement week was slightly positive ‘Two additional comparisons (analogous to those made in the volume analysis) were condueted to see how unusual an U, of 1.67 is. The first com- nrarison exatnined Ur in the nonzeport period (cee Figure 9). Out of 260 ‘values, only 11 execeded 1.67, The comparison suggests that the price a¢- * Fama, op. cit "The autocorrelation was examited on 8 weck-by-woek, crose-scational basi eae TEE teste LI Galll, b= Bes $8.82 wiusane maven tivity in week 0 is unusually high, in spite of the fact that such a compari- son tends to understate how unusual {t really is. The second comparison examined the frequency of U's larger than one relative to the frequency ‘that, occurred during the nonrepart. petiod (see Figure 7), The frequency of values above one is preatest. in week 0, with the next highest values oc- curring in the weeks adjacent. to the announcement week, There is an ex- tremely small probability that such a bigh number (181 in week 0) could have occurred by chance. The interpretation is the same as that of the mean analysis—namely, there is above normal prise activity when earnings reports are released. What this analysis reveals that the mean analysis does not is the fact that the abnormally high mean is not caused by s few ob- servations dominating the results but rather by a substantial proportion of the sample data. Tn summary, the behavior of the prios changes uniformly supports the contention that earnings reports possess information content, Observing a price reaction as well as a. volume reaction indicates that not only are expectationa of individual investors altered by the eamings report but also the expectations of the market as a whole, a6 reflected in the changes in equilibrium prices. Relationahin between the Volume and the Price Findings. ‘The previous sentence raises the issue, “how rauieh of the increased price activity ean be attributed merely to the fact that there is more ‘action’ in the security, rather than to changes in equilibrium prices?” One way to approseh this question is to view the price change during ‘2 given time period as a sum of prico changes on each transection that oc- curred during that period. In a world of uncertainty, the price change from each transaction can be treated as an observation from © probability dis- tribution af the investors assessment of what the price change should be. ‘The price change por period, then, is a sum of random variables. If trans- sotions occur as if they are independent over time (evidence on daily, weekly, and monthly price changos suggest they do), the varianeo of the weekly prico change will inerease in dinoet, proportion to the number of transactions that oceur during the time period. ‘The reasons for underatatoment ars similar to thoes steted in the volume analy sis, See p. 77. The probability is less than 1 chance in 100,000. 1 The evidence reganding aerial correlation af daily and monthly prise changes an be found in Fama, op. cit. and Fama, ef a, o. el, reqpectively. The average utacorrelation eoeffeient for weekly’ charges in this sample was ~.08, which would ‘sauge the variauge ta inereane loan than proportionately with the number of trancae- ‘ions. Within « givon trading day, the autceorrelation may be higher (e. boeavee of certain institutional feevom, such as clustering of Limit orders oF stnp loa orders). “Howaver, the oxistones of ocbleragers should prevent the autocorwation Crom being very large. In order or the price setvity to be explained entirely by increagnd tran action activity, the antocortlation would have to he oe. This would be highly tinlikely becaste of the empirical evidence cited and the opportunities for arbitrage.CONTENT OP ANNUAT RARNINGS ANNOUNCEMENTS — 83 ‘The issue now is what is the appropriate measure of the nuvaber of transactions occurring during a given period. If the volume is used as a measure, then Us in week 0 would be expected to be 1.30 merely because of more action in the security. The remaining portion would be attributed. to changes in tho equilibrium prices of the securities. However, it is not tall clear that volume (or even number of transactions) is the appropriate measure, heeause it reflects only the explicit transactions that; occur. Tt could be argued, with considerable support from eronomie theory, that the expectations of all investors influence the market price, whether cor nat they engago in & purchase or a sale. If the market aets in this manner, the total number of transaetions, explicit and implicit, sre the same per time period. Hence all of the above average price activity ean be attributed to changes in equilibrium prices, ‘Additional empirieal research is needed before this sue will he resolved. ‘The research would consist of studying increased volume activity due to reasons other than information coming onto the market. An initial analysis of the seasonal variation in volume (V2) from 1946 through 1956 revealed that the volume is greatest during the months December and January. The explanation seems (o stem from tax considerations rather than an above normal flow of information, Research also indicated that the price variabil- ity of Ruy during these months (.e., U.) was only .996, indicating no shove ‘average price variability during these months. This finding lends support to the position that none of the price activity in week 0 is due merely 10 ‘more motion. Before leaving this topic, note that isolating the volume effects on priee ‘changes is of conser only to the extent one wishes to distinguish between information that alters the expectations of the market. as a whole from in- formation that alters only the expectations of individual investors, AIL of the price activity oan be attributed to information in the latter sense, Frequency of Other News Announcements during Report Period ‘The purpose of this analysis was to discover if there was any clustering of other news snnouncernents around week 0 that might possibly aecount for the volume and price resetions. As indicated earlier, the sample design excluded any firms that announced dividends in the same weele as earnings ‘or any firms that split the'r stock during the report period. However, itis conceivable that dividends announcements might duster in weeks immedi- atoly prior to and after week 0 or that other types of announcements (eg, management earnings foreaasts) might cluster in week 0. To examine this possibility, the occurrence of other news snnounceimonts in the Wal! Street -Fousnat during the 506 report periods was examined (see Table 7), By far the mast frequent. type of announcement was dividends, which exceeded the frequency of all other types of announcstnenta by a fastor of84 WoMMIAM It, RAVER 9 to 1. With respect to the purpose of this analysis, there is no clustering of dividend announcements in wooks —1 or +1; in fuet: the opposite seems to. bo true. Also there is no clustering of any other type of announcements at any time during the period, including week 0. The volume and price rese- tion in week 0 does not appear to be attributable to the clustering of other news announcements. Suggestions for Putture Research ‘The dratnatio price and volume resotion indicates that investors do look. Aiveatly at reported camnings and do not use ather variables to the exclusion of reported earninge. The evidenee also indicates that news announcements ‘occurring prior to the earnings report do not entirely preempt the informa~ tion content of reported earnings. Given these findings, one of the fist extensions of the study will be to explore the possibility of constructing expectations models that will permit a prediction of the direction and magnitude of the price residual ‘The results of @ resent study by Ball and Brown in this area are very encouraging They used an earnings mode! sienilar in form to the price and volume models described in this study (e.g, changes in the earnings of ‘an individual security were viewed ab a linear funetion of market-wide index: of earnings changee). "The saniple was divided into two groups: instances ‘where the earnings residual was pasitive (actual earningn were higher than “expeeted”) and instances where the eamings residual was negative (actual earnings lower than “expected”. ‘Fhe behavior of the price residuals for theso two groups was examined, and the findings were: (1} The sian of the ceumulative price residual (summed over a 12 month period including the aznouocement month) was highly ascoclated with the sige of the eamings residual, (2) There was a persistent upward drift in the cumulative mean price residuals for the positive earnings residual group. This drift started 11 ‘months prior to the earnings announcement, and over 90 per cent of the sift had taken place by the heginning of the announcement month, The negative earnings group exhibited an analogous behavior pattern. ‘The findings indicate that reported earnings are aasoviated with underly- ing evenis that, are pereeived hy investors to affect the market price, Be- cause earlier news announcetnetts convey some of the sare information as the earings reports, investors are able to use this inforination to revise their forecssts of earnings and to adjust the prive accordingly. In fact, by the beginning of the announeemen’ month, investors fortn largely unbiased forecasts of reported eamings, even though the reported exmnings are above 1A meaaced in rerms of number of news tmouncenente per week, the flow of information during the weeks prior ia the annouosment. does not appeas toe below poral and bone would wot account for the below normal price aetvity during weeks ~2ihrough ~8. * Ray Ball and Philip Brow Nomiser,” Journal of Account .n Empirical Bvalwation of Accounting Income Ressarch, 6 (Autumn, 1988), pp, 150-78,CONTENT OF ANNUAL HARNINGS ANNOUNCEMENTS — 85 or below normal relative to their historical relationship with market-wide earnings. Although the forecasts are unbiased, they are not very efficient, for if they were, there would e ro volume or price reaction when earnings reports ‘were relessed,** The Ball and Brown findings and the findings presented hhete are mutually supportive with raspeet to the information content of camings reports and also are uniformly consistent with the Endings of previous studies in the behavior of security prices. One extension of the re search presented here will be to replieate the Ball and Brown study on this, sample of non-12/31 firms (the Ball and Brown study dealt exclusively with 12/31 firms) and then to attempt to pradiot the magnitude, as wall as the sign, of the price residual. ‘A second ates of further research is the application of this methodalony to other types of news announcements, At an earlier meeting of the Con- ference, Green and Segall explored the information content: of interim re- ports, An analysis of volume and price changes during the annouricement of interim earnings would provide » different. approach to thie same issue, Tho information content of dividend announcements is another topie that, has reozived much attention snd stills in need of additional empiriesl investiga- tion, Such research will indicate the importance of annual earings sn. nouncements relative to other kinds of information. Perhaps the most important extension of this study would be desling with the normative issue, “Should decision makers perceive eaminss reports to possess informational value?” ‘The normative question can be approzched by selecting an event of interest to decision makers (preferably as tree a8, possible from the influence of their perceptions) and by investigating the ability of earnings data to predict that event, A few studies of this type have been presented at earlier meetings of the Conference, but much more work is needed in this area.®* Hopefully, the findings presented here with respect to the positive question will provide greater insight into the normative issue as well. The datinetlon betwoan unblatednass aed effsiency wa dlzouaed in footnote 9. ¥ Jemes O. Horrigan, ‘The Determination of Long-Term Credit. Standing with, Fivancial Ratios," Rmpirieal Research in Accounting: Selected Studia, 1068, Supple: ment ta Vol. 4, Journal of Accounting Kecearch, pp. 44-62, and William Beaver, “Pinancisl Ration at Prediosors of Failure,” id, pp. 71-102,86 wma a. seaven TABLE 1 Effect of Selection Criteria upon Semple Size ete Reval one ‘Compustat Breas (ote 1) 886 Looe: 12/81 Armee esctececseeeectieceetiieecnieeseneee] $99 Won-12/81 firms dep 2). cote 27 Lose: Now-NYBE rms “ 85 NYSE and won-12/B1 (ep 3) 20 Lesa “More than 20 announcementa per year 8 Dividends in earnings annotincement week, 0 Bak pi during report pod oe z Other eee pitisevtneesssssesees 8 % Semple size (step 4) 18 “Sample criteria were applied sequentially is four stages. The semple sive after ‘eagh stage is denoted by parenthetical comment (e.g, stepe I, etc). Mizcallangous reasons such as firm's earaiuge were nat reported in Wat! Strat Journal. TABLE? Distribution of Pinansial Statonent and Announcement Daler Sanuary, February March April May... une. uly. Avguat, September... October... : November. December Total. ‘Total number of firma equals 148, and (otal number of aonouncemente equals 6,CONTENT OF ANNUAL EAENINGS ANNOUNCEMENTS 87 TABLE 3 Number of Weeks between Feent Vout Bnd and Date of Annocncement “"Potal number of announcements is 50, TABLE 6 Biuanmacy of Regression Statistics Volume Analysis BeeBe Summary of Regression Statistics Price Analyeie een] ego ate sain fran‘ vatbic MASE fs, | idee oan xa ‘Mean 187 12 | 138 80 * Fractile "10 1s | 2.13 96 2 13 2 m6 — me | 135 a 2 50 196 io | Lat 8 ” 5 210 28s | 208 ua 32 0 2 sas | 208 vat a788 WILLIAM #. nRaver TAMLE 6 Armia of eon Price Residual Xigp onto we cEeuses “3 onise = — "ones o = ‘om 3 oates 4 Conte 3 ‘wo = ~ "ome = tt 0 ‘500 : ‘want 2 ‘nie 3 ‘wiz + ‘2100 5 “ones, ‘ ~ e009 > ‘02s? & os TABLE? Oocurrence of Other Mews Announcenants wa | satavine | Moule -4 8 8 -3 a 2 -2 a ‘ a 1 5 0 0 4 1 16 4 2 Fa 4 a 2 a 4 a 2 ‘Total 282, aLCONTENT OF ANNUAL EARNINGS ANNOUNCEMENTS — 89 nw te (Bn) fa eb 8 -) 0G eS Wate monet Fro, 1, Volume Analysis al welSe)mnseto rs * + sehen kee 4 ro» » & Value, Fra, 2, Distribution of &, in the Nonrepiart Period90 wrtua , BEAVER uw » a(E a) whe, “8 one 9 “Skea Woreater snaseanet Fro, & Residual Volume Anelysie Reet ae 4 oe POH a ae 4 Wetks Bra, 4. Rrequeney of Positive 4's Residual Volume Analysis,CONTENT OF ANNUAL BARNINGS ANNOUNCEMENTS — OL Balative feeqieny on 5 chmcrvtions ee veut, Bua. 5. O¢ in Nonroport Period Mean daring eonreprs ug. Ev) Zu i an a 4 pest =, sR Be 14 mysriane of sesidua inthe Peasepurt peel, ” 2 Pa 4 0 2 ow 8 ek ater sancanoenent Fre, 6, Price Residual Analysis 892 Wi ¥. auavER 0 Expose nao U5 100 bad tm tele erat sept v0 aad ee Fra, 7. Rrequoney of Usi's > 1.00—Residual Price Analysis