Revenues and Earnings as Key Value Drivers in Various Contexts: Implications for Financial Management and Statement Analysis

Transcription

1 Revenues and Earnings as Key Value Drivers in Various Conexs: Implicaions for Financial Managemen and Saemen Analysis Iay Kama Graduae School of Business Adminisraion Tel Aviv Universiy Tel Aviv 69978, Israel Firs Draf: December 2003 Curren Draf: November 2004 I would like o hank my advisor, Eli Amir, for his commens and dedicaed guidance. I also wish o hank Roy Dilmanian, Ronen Mansury, Dayna Kesen, paricipans a he 2004 Annual Congress of he European Accouning Associaion, and seminar paricipans a Tel Aviv Universiy. Finally, I would like o hank he David Orgler Banking and Finance Research Fund and he Harry Simons Scholarship Fund for providing financial suppor.

2 Revenues and Earnings as Key Value Drivers in Various Conexs: Implicaions for Financial Managemen and Saemen Analysis Absrac This sudy examines he relaive informaion conen of revenues and earnings in various conexs. Alhough i is clear ha invesors' percepion of he informaion conen of accouning variables is colored by he conexual framework, he informaion conen lieraure has no focused on he conexual analysis of informaion in revenues. I find ha in conexs in which curren earnings are a weak indicaor of fuure earnings he role of earnings (revenues) is relaively less (more) imporan: High R&D firms, loss-reporing firms, and indusries wih oligopolisic compeiion. Due o relaively high earnings managemen, earnings have lower informaion conen in he fourh quarer han in he firs hree fiscal quarers. Overall, hese resuls have imporan implicaions for financial managemen and for financial saemen users, in undersanding he role earnings and revenues play in differen conexs, such as R&D inensiy, firm specificaion, and indusry characerisics.

3 Revenues and Earnings as Key Value Drivers in Various Conexs: Implicaions for Financial Managemen and Saemen Analysis 1. Inroducion On April 14, 2003, IBM announced quarerly earnings ha were lower han Wall Sree expecaions (as measured by Thomson Firs Call). The company s quarerly revenues, on he oher hand, were higher han expeced. IBM repored is earnings afer he marke close; he nex day is shares rose by 3.4%. The IBM example is an ap illusraion of he opic explored in his sudy: The relaive explanaory power of earnings and revenues in deermining sock reurns. Why did he company s sock price increase, even hough is earnings had fallen below expecaions? The reason could be, as one analys saed, ha "The sligh earnings miss was no significan and he company did well considering he economy and he war." On he oher hand, he 3.4% rise in share price may have been due o revenues exceeding he consensus analys esimae. During he preliminary earnings announcemen companies usually announce revenues and earnings. Therefore he research quesions explored in his sudy are focused on hese wo variables and can be summarized as follows: Wha is he role of revenues in explaining sock reurns afer conrolling for earnings? Moreover, does he role of revenues and earnings vary in differen conexs such as R&D inensiy, negaive earnings, magniude of earnings managemen and indusry concenraion? A relaed issue is he comparison beween revenues and earnings as explanaory variables of sock reurns. This comparison is also meaningful for company- and indusry-specific feaures. Revenues may have an incremenal influence on sock reurns over earnings, because hey generae curren earnings and cash flows, and hey may also serve as a 1

4 proxy for fuure performance. Jegadeesh and Livna (2004a,b) argue ha he imporance of revenues is refleced by he fac ha firms announce revenues, raher han oher financial variables, in he preliminary earnings announcemen and analyss provide revenues forecas in addiion o earnings forecas. Furher, as Erimur e al. (2003, p.185) poin ou "since firms announce boh revenues and earnings in he preliminary earnings announcemen, invesors can use he disclosure of revenues o beer assess and inerpre he qualiy of he disclosed earnings signal." Therefore, he relaive role of revenues as value driver and indicaor of fuure cash flows is expeced o be more imporan in conexs in which curren earnings are a weak indicaor of fuure earnings. These conexs may include inensive R&D and loss-reporing firms, indusries wih oligopolisic compeiion and fiscal quarers wih relaively high earnings managemen. These research quesions are relaed o he informaion conen lieraure iniiaed by Ball and Brown (1968) and Beaver (1968). During he las four decades, over one housand sudies have been published on his opic; many of hese have deal wih relaions beween accouning variables and sock reurns, and he primary focus has been on earnings (and cash flows), raher han on revenues, as explanaory variables of sock reurns. 1 The issue of informaion conen and value drivers is of obvious ineres o many groups: equiy invesors and analyss, deb holders, raing agencies, regulaors and sandard seers. While, as saed, he bulk of he informaion conen lieraure has concenraed on earnings and cash flows, one srand does invesigae he informaion conen of revenues and proxies for fuure revenues. Hopwood and McKeown (1985) and Hoskin e al. (1986) focus on he associaion beween revenues and sock reurns in a variey of indusries, and find ha revenues do no have an incremenal influence 2

5 on sock reurns. In conras, Swaminahan and Weinrop (1991), Rees and Sivaramakrishnan (2001), Erimur e al. (2003) and Jegadeesh and Livna (2004a,b) find ha revenues incremenally explain sock reurns beyond earnings. Anhony and Ramesh (1992) find ha revenues have higher explanaory power for firms in heir early sages; while Livna (2003) shows ha he magniude of he pos-earnings announcemen drif is posiively correlaed wih he sign of he revenues surprise. Recen sudies in he high-ech secor have focused on he informaion conen of proxies for expeced revenues insead of revenues. 2 Alhough i is clear ha invesors' percepion of he informaion conen of accouning variables is colored by heir conexual framework company specificaion, indusry srucure and macroeconomic condiions he informaion conen lieraure has no focused on he conexual analysis of informaion in revenues. Conexual analysis may assis inernal and exernal financial saemen users in undersanding he relaive imporance of revenues and earnings in various firm and indusry conexs. This sudy exends Erimur e al. (2003) and Jegadeesh and Livna (2004a,b) by providing conexual framework for he incremenal explanaory power of revenues (earnings) over earnings (revenues) in deermining sock reurns. Conexual framing makes i possible o undersand he role ha revenues and earnings play given he magniude of earnings managemen, in high R&D companies compared o low R&D companies, in loss-reporing firms relaive o profiable firms, and in indusries wih oligopolisic compeiion compared wih monopoly or indusries wih low concenraion. Moreover, he sudy exends previous sudies by invesigaing marke reacion o conflicing earnings and revenues surprise, in various firm- and indusry-specific circumsances. 3

6 In invesigaing he informaion conen of revenues and earnings, I follow he mehodology of Jegadeesh and Livna (2004a) by using porfolio and regression analysis. The sample includes 280,113 quarerly observaions for 10,729 companies over In he porfolio analysis all observaions are ranked according o earnings surprise and revenues surprise assigned ino quiniles. I hen examine marke reacion o earnings and revenues surprise for weny-five porfolios (i.e. marke reacion when revenues surprise is in is i h quinile and earnings surprise is in is j h quinile). Earnings (revenues) surprise is calculaed as he sandardized difference beween earnings (revenues) per share and he expeced earnings (revenues) per share, where expecaions are based on he same quarer las year, plus an average drif. Marke reacion is defined as size-adjused reurns (SAR) raw reurns, minus he reurn on he size porfolio ha conains he firm; however, I also use marke-adjused reurns (MAR) raw reurns, minus he reurn on he valueweighed NYSE index and find similar resuls. In he regression analysis an examinaion is performed of he explanaory power of earnings in deermining sock reurns afer conrolling for revenues, and vice versa, in differen conexs, such as R&D inensiy, degree of earnings managemen and he level of indusry concenraion. 3 I use a Fama-MacBeh (1973) ype regression by performing 30- yearly regressions. The resuls in he full sample show ha earnings and revenues boh have informaion conen for invesors, each having an incremenal effec on abnormal reurns afer conrolling for he oher. Moreover, porfolio analysis shows ha marke reacion o increases in earnings (revenues) surprise is posiive whaever he level of revenues (earnings) surprise. The comparison beween earnings and revenues as explanaory variables of sock reurns shows ha, in he full sample, marke reacion 4

7 o high (low) earnings surprise is posiive (negaive) regardless of he level of revenues surprise. Regression analysis leads o similar resuls, he coefficien esimae of earnings surprise being abou wice as large as ha of revenues surprise. The overall conclusion from hese analyses is ha, in he enire sample, he explanaory power of earnings is higher han ha of revenues. This resul is consisen wih Erimur e al. (2003) and Jegadeesh and Livna (2004a). As expeced, he conexual analyses show ha he informaion conen of revenues and earnings relae o earnings precision and earnings persisence as refleced by firm and indusry characerisics such as R&D inensiy, negaive earnings, indusry concenraion and he magniude of earnings managemen. Regarding R&D inensiy, porfolio and regression analyses show ha in high R&D companies he role of revenues (earnings) appears o be more (less) imporan han in low R&D companies. As a resul, and in conras o he resuls in he full sample, in high R&D companies he explanaory power of earnings is no higher han ha of revenues. This resul may arise from he imporan role of he fuure performance indicaor in high R&D companies, due o heir relaively high uncerainy. 4 As already menioned, previous sudies show ha revenues serve as an indicaor of earnings persisence and have relaively high auocorrelaion. Therefore, revenues play a more imporan role in high R&D companies. 5 The resul regarding high R&D companies is also consisen wih Anhony and Ramesh (1992), who find ha marke reacion o unexpeced revenues is higher for firms in he early sages of heir life cycle. The resul regarding he lower explanaory power of earnings in high R&D companies is a resul of low earnings persisence and precision, and is consisen wih Kohari e al. (2002), who find posiive relaions beween R&D expendiure and earnings variabiliy. Erimur e al. (2003) and Jegadeesh and Livna (2004a) find ha 5

8 he explanaory power of revenues and earnings is higher in growh companies han in value companies. However, R&D is an addiional facor beyond growh, he resuls in my sudy being robus o he inclusion of he marke-o-book raio as an explanaory variable. In loss-reporing firms, he explanaory power of earnings and revenues is smaller han in he full sample. However, he decrease in he explanaory power of earnings is higher han he decrease in he explanaory power of revenues. Therefore, he dominaion of earnings as an explanaory variable of sock reurns in companies wih negaive EPS diminishes. This finding is consisen wih Hayn (1995), who finds ha he explanaory power of earnings is lower in loss-reporing firms. As for indusry characerisics, revenues play a more imporan role when an indusry is ruled by wo o hree dominan companies, ha is, in oligopolisic compeiion. In indusries wih oligopolisic compeiion, marke sraegy plays a more imporan role, because a larger marke share is an invesmen ha migh yield an abiliy o dicae prices and erms of supply in he fuure. The explanaory power of revenues is no higher in a monopolisic environmen, because marke power already exiss. Ineresingly, he explanaory power of revenues is weaker when indusry concenraion is a is lowes level, due o he fac ha marke share does no change significanly in relaion o compeiive abiliy. Robusness esing indicaes ha hese resuls occur for he full sample and for manufacuring companies (4-digi SIC codes ). Anoher invesigaion is performed regarding he differences in he explanaory power of earnings and revenues beween fiscal quarers. I find ha he explanaory power of earnings is significanly lower in he fourh fiscal quarer han in he firs hree fiscal quarers. This resul is consisen wih Cohen e al. (2004) who 6

9 find ha he magniude of earnings managemen is higher in he fourh fiscal quarer. Insofar as earnings managemen decreases earnings precision, i also decreases he role of earnings as a value driver. 6 The comparison beween revenues and earnings as value drivers shows ha negaive earnings surprise accompanied by posiive revenues surprise leads o significanly negaive SAR. In high R&D companies and in indusries wih oligopolisic compeiion here are conflicing resuls, such ha negaive earnings surprise accompanied by posiive revenues surprise does no lead o significan changes in SAR. This finding is consisen wih he previously resuls of he sudy regarding he imporan role of revenues in high R&D companies and in indusries wih oligopolisic compeiion. However, posiive earnings surprise accompanied by negaive revenues surprise leads o posiive SAR for he full sample, as well as for high R&D companies and for companies operaing in oligopolisic compeiion. Overall, his sudy has imporan implicaions for financial managemen and for financial saemen users, in ha i highlighs he imporan role played by revenues as a proxy for fuure cash flows in conexs in which curren earnings are a weak indicaor of fuure earnings: high R&D companies, loss-reporing companies, indusries wih oligopolisic compeiion and quarers wih relaively high earnings managemen. The sudy proceeds as follows: In secion 2, I review he lieraure and moivae my analysis. Secion 3 develops esable predicions. Secion 4 discusses he sample and he variables. Secion 5 conains empirical resuls for he enire sample. Secions 6, 7 and 8 include deailed resuls for he conexual analysis. Secion 9 invesigaes he marke reacion o conflicing earnings and revenues surprise, and secion 10 provides concluding remarks and plans for furher research. 7

10 2. Background and Lieraure Review Kohari (2001) reviews he informaion conen lieraure for he lae 1980s and 1990s, and many exbooks analyze he connecion beween sock reurns (or firm value) and earnings (e.g. Palepu e al. (2000) and Penman (2001)). Earlier surveys for he 1970s and 1980s include Lev and Ohlson (1982) and Lev (1989). Holhausen and Was (2001) criically review he lieraure, arguing ha is conribuion o sandard seing is modes, as mos of he sudies do no seek o develop a descripive heory of accouning and sandard seing. 7 In a wider perspecive, Kohari (2001) refers o he invesigaion of he relaions beween accouning variables and sock prices as capial-marke sudies, which he divides ino four caegories: fundamenal analysis and valuaion, ess of capial marke efficiency, he role of accouning in conracs and in he poliical process, and disclosure regulaions. My sudy falls ino he caegory of fundamenal analysis and valuaion. 8 Since Ball and Brown (1968) and Beaver (1968), researchers have concluded ha, of all he accouning variables, earnings have he highes explanaory power over various ime inervals, 9 hough mos find ha earnings explain only 5% o 15% of he variaion in sock reurns. 10 Numerous sudies have ried o explain and improve he explanaory power of he variaion in sock reurns by: Adding he concep of earnings persisence (e.g. Subramanyam and Wild (1996) find ha he probabiliy of erminaion decreases he informaion conen of earnings); Invesigaing non-linear relaions (e.g. Das and Lev (1994) and Lipe e al. (1998)); Using he aggregaion of earnings across years as he explanaory variable (Ohlson and Penman (1992) and Eason e al. (1992)); 8

11 Aribuing he low explanaory power o he earnings lack of imeliness and o he hypohesis ha sock prices lead earnings changes (Beaver e al. (1980), Collins e al. (1994) and Ayers and Freeman (1997)); Including he earnings forecas as an explanaory variable for sock reurns (Liu and Thomas (2000)); Claiming ha he explanaory variable should be he accouning earnings predicion raher han sock reurns (Penman (1992) and Abarbanell and Bushee (1997)); Indicaing ha, when he sample includes only loss-reporing firms, he explanaory power of earnings decreases significanly (Hayn (1995)); Poining ou he informaion conen of managemen s earnings forecas; 11 Analyzing he pos-announcemen drif, meaning he under-reacion of sock prices o earnings surprise; 12 Referring o he relaionship beween sock reurns and oher accouning parameers. For example, Lev and Thiagarajan (1993) examine he informaion conen of some accouning parameers (fundamenals) in various macroeconomic siuaions, while Lipe (1986) indicaes he informaion conen of earnings componens. Several sudies find a posiive relaionship beween cash flows (and accrual componens of earnings) and sock reurns, alhough earnings seem o have a higher explanaory power. 13 A conflicing sudy by Bernard and Sober (1989) indicaes ha cash flows do no have a significan incremenal effec over earnings, while Das and Lev (1994) show ha he reurns/cash flows relaionship is non-linear. Alhough he informaion conen of revenues has no been he focus of he informaion conen lieraure, i has neverheless been invesigaed a several levels: 9

12 (1) Hopwood and McKeown (1985) and Hoskin e al. (1986) find ha revenues do no have an incremenal influence on sock reurns. (2) Anhony and Ramesh (1992) show ha unexpeced revenues growh and capial expendiure have more of an impac on reurns during he firm s growh sage, due o he effors made o achieve cos or demand advanages over compeiors. (3) Livna (2003) shows ha he magniude of he earnings pos-announcemen drif is sronger when revenues surprise and earnings surprise have he same sign. (4) Liu e al. (2000) evaluae several muliples by comparing he sock price o he price prediced by he muliples. In heir sudy, he use of revenues in muliples leads o he wors resul. (5) Several sudies invesigae proxies of expeced revenues as explanaory variables of sock reurns in high-ech indusries. Jurion and Talmor (2001), Davis (2001), Lazer e al. (2001) and Bagnoli e al. (2001) refer o Inerne companies, and Amir and Lev (1996) examine he wireless communicaions indusry. (6) Swaminahan and Weinrop (1991), Rees and Sivaramakrishnan (2001), Erimur e al. (2003) and, recenly, Jegadeesh and Livna (2004a,b) indicae ha revenues have incremenal explanaory power in deermining sock reurns. Jegadeesh and Livna (2004a) also indicae ha sock reurns are relaed o pas revenue surprises, while Jegadeesh and Livna (2004b) show ha alhough invesors reac o revenues surprises hey do no fully incorporae hem. In his sudy, I exend Erimur e al. (2003) and Jegadeesh and Livna (2004a,b) by analyzing he informaion conen of earnings and revenues in various conexs. Jegadeesh and Livna (2004a) find ha boh revenues and earnings have informaion conen, and he incremenal explanaory power of earnings is higher han 10

13 he incremenal explanaory power of revenues. Jegadeesh and Livna also indicae ha persisence of earnings is posiively correlaed wih he sign of revenues surprise, and ha pos-announcemen abnormal reurns are posiively correlaed wih revenues surprise. My main conribuion o his branch of he lieraure lies in invesigaing he informaion conen of earnings and revenues according o firm-specific aribues (R&D inensiy, magniude of earnings managemen and sign of earnings) and indusry specificaion (degree of concenraion). In addiion, I use conexual analysis o examine marke reacion o negaive (posiive) earnings surprise accompanied by posiive (negaive) revenues surprise. The sudy conribues o financial saemen analysis in undersanding he role earnings and revenues play, according o he specificaion of he firm being analyzed and he indusry in which he firm operaes. Exernal financial saemen users (analyss, invesors and deb holders) and inernal users (managemen) can use he conexual resuls o predic marke reacion o earnings sraegy (sraegy ha is aimed a increasing earnings) and marke share sraegy (sraegy ha is aimed a increasing revenues). Such predicion is expeced o be more precise when i is based on firm and indusry characerisics, such as R&D inensiy, sign of earnings and indusry concenraion. 3. Predicions Prior lieraure suggess ha revenues and earnings each have incremenal explanaory power in deermining sock reurns afer conrolling for he oher. The economic inerpreaion for he informaion conen of revenues can be divided ino wo caegories: he firs and rivial explanaion is ha revenues generae curren 11

14 earnings and cash flows; he second explanaion is ha revenues serve as an indicaor of persisence and fuure performance. Jegadeesh and Livna (2004a,b) show ha revenues are an indicaor of persisence in earnings' growh, Gu e al. (2004) find ha he persisence of earnings surprise increases when i is driven by revenues surprise raher han by expenses surprise, while Erimur e al. (2003) show ha revenues have relaively high auocorrelaion. Ghosh e al. (2005) conribue o his line of research by showing ha earnings for companies wih susained increases in boh earnings and revenues are more persisen, and fuure operaing performances (measured as reurn on asses) are higher han for companies wih susained increase jus in earnings. Therefore, based on prior lieraure, I expec boh revenues and earnings o have incremenal explanaory power in deermining sock reurns, afer conrolling for each oher. Since revenues have relaively high auocorrelaion, and hey serve as an indicaor of fuure performance, my second predicion is ha he explanaory power of revenues is higher for companies wih relaively high uncerainy and lower earnings precision, such as high R&D companies. Amir e al. (2003) imply he imporance of a fuure performance indicaor in high R&D companies. They find ha he incremenal conribuion of analyss is higher in high R&D companies han in low R&D companies. Anoher reason for he imporance of revenues in high R&D companies may be driven by a life cycle argumen. Anhony and Ramesh (1992) find ha marke reacion o unexpeced revenues is higher for firms in he early sages of heir life cycle, due o he effors made o achieve cos and demand advanages over compeiors, while Jegadeesh and Livna (2004a) and Erimur e al. (2003) find ha revenues play a more imporan role in growh companies han in value companies. 12

15 Furhermore, regarding loss-reporing firms, I expec he relaive informaion conen of revenues o be higher, since hey are an indicaor of fuure aciviy. Hayn (1995) finds ha he explanaory power of earnings is indeed lower in loss-reporing firms, while Amir e al. (2003) find ha he conribuion of analyss is higher in lossreporing firms, hereby implying he imporance of a variable ha capures fuure aciviy in loss-reporing companies. Wih regard o indusry concenraion, I expec revenues o have a higher explanaory power in indusries wih oligopolisic compeiion. An oligopoly is a marke srucure generally consising of only a few compeing companies, where each company accouns for a significan porion of marke wide sales. In cases of oligopolisic compeiion, a larger marke share may be regarded as an invesmen ha can yield an abiliy o dicae fuure prices and supply erms. For his reason, I expec revenues o be a significan indicaor of fuure oucomes and o be more imporan for firms operaing in indusries wih oligopolisic compeiion. 14 This argumen may be srenghened by he discussion concerning excess capaciy as an invesmen in enry deerrence. In summarizing he usage of excess capaciy, Marin (2002, pp ) concludes ha invesmen in capaciy can be used o influence invesmen decision of rivals. This may deer or delay rival invesmen, and allow incumben o condiion marke srucure in a way ha allows he exercise of marke power. In our conex, revenues may serve as a proxy for capaciy, hereby condiioning marke srucure in a way ha allows he exercise of marke power. My las expecaion regarding he informaion conen of revenues and earnings relaes o he effec of fiscal quarer on marke reacion. Cohen e al. (2004) find ha earnings managemen is higher in he fourh fiscal quarer han in he firs hree. Since earnings managemen is negaively correlaed wih earnings precision, I 13

16 expec o find ha he role of earnings as a key value driver is less imporan in he fourh quarer han in he firs hree. To sum up, I expec boh revenues and earnings o have informaion conen, and revenues o have relaively higher explanaory power in high R&D companies, loss-reporing companies and in indusries wih oligopolisic compeiion. Furhermore, due o a higher likelihood of earnings managemen in he fourh quarer, I expec o find ha earnings have lower explanaory power in he fourh quarer. 4. Sample and Variables The iniial sample includes all public companies covered by he Compusa and CRSP daabases over he period Observaions wih missing daa needed o calculae size-adjused reurns (SAR), sandardized unexpeced earnings () and sandardized unexpeced revenues () are deleed. Also excluded are financial insiuions and public uiliies (4-digi SIC codes and ) because hese indusries are regulaed and hey may no be comparable o oher indusries in erms of financial reporing. To reduce he effecs of exreme observaions he sample is rimmed a 1% and 99% for and, and a 99% for SAR. Nearly 5% of observaions ae removed in his process. Table 1 presens he number of observaions in each year. The able is divided ino four caegories: he full sample (280,113 observaions for 10,279 companies), SIC codes (158,945 observaions for 5,080 companies), he R&D sample (152,582 observaions for 5,916 companies), and companies wih negaive earnings per share (72,129 observaions for 8,154 companies). As for he R&D sample, he exreme 3% of he highes observaions for he R&D-o-revenues raio (RDR) are deleed (RDR higher han 293.7%). In his process 14

17 4,955 observaions are removed. R&D expendiure and revenues used o calculae RDR are in annual erms. (Table 1 abou here) I calculae abnormal sock reurns as size-adjused sock reurns around quarerly earnings announcemens. Size-adjused reurns (SAR) are calculaed as raw reurns minus he reurn on he equally weighed reurn on he porfolio of all companies in he same size decile. The 4-day reurns window conains days -2 hrough +1, where day 0 is he earnings announcemen dae, as saed in Compusa. I also use marke-adjused reurns (MAR) as an alernaive measure of abnormal reurns. MAR are calculaed as raw reurns, minus he reurn on he value-weighed NYSE index. The resuls are similar across measures of abnormal reurns. Quarerly earnings per share (EPS) are calculaed as basic earning per share, excluding exraordinary iems, minus afer sauory federal ax special iems. In Compusa iems: EPS = daa32 (1 ax) daa19 daa15 daa17 Quarerly sales per share (SPS) are calculaed as ne sales divided by he number of common shares used o calculae EPS: SPS = daa2 daa15 daa17 In calculaing and I follow he mehodology of Jegadeesh and Livna (2004a) by calculaing as he sandardized difference beween EPS and he expeced EPS: EPS = E( EPS S ) 15

18 where EPS, is EPS for firm i in quarer, E( EPS, ) is he expeced EPS for firm i in i quarer, and S, is he sandard error of [ E EPS )] i i EPS (. E( i EPS, ) is calculaed as he EPS in he same equivalen quarer of he previous year, plus an average drif: where E ( EPS ) = EPS 4 + D D i, is he average drif of EPS over 8 quarers: D 8 = 1 = j ( EPS j EPS 8 j 4 ) As previously saed, is calculaed as: S i, is he sandard error of he unexpeced par of he EPS and S = ( EPS j E( EPS) j ) 7 j= 1 is calculaed using a similar procedure. Table 2 conains descripive saisics for SAR, MAR, and. For all caegories, mean and median SAR are equal o zero, as expeced by design. For all caegories, mean is negaive bu he median is close o zero. Ineresingly, s are similar across caegories. Mean and median are posiive for all caegories, bu is higher for companies wih high RDR han for hose wih low RDR, as expeced. (Table 2 abou here) Table 3 presens a deailed depicion of he R&D-o-revenues raio (RDR) disribuion. The exreme 3% of he observaions highes in RDR are deleed (RDR higher han 293.7%). In his process 4,955 observaions are removed. The mean RDR is higher han he median, due o exreme observaions of high RDR. Abou 16

19 30% of he observaions have an RDR of less han 1%. However, i should be noed ha i is he ordinary value raher ha he absolue value of RDR ha influences he research. The division of RDR ino quiniles indicaes ha mean and median SAR equal zero for he firs four quiniles, and are slighly negaive for high R&D companies. As o and, here is no monoonic change across quiniles. As expeced, in high R&D companies, here is a significan difference beween he mean and median. (Table 3 abou here) 5. General Empirical Resuls 5.1.Correlaion analysis Table 4 presens he Pearson and Spearman correlaions beween and. As expeced, hey are posiive and significanly differen from zero a he 1% level. They are only beween 0.28 and 0.36, suggesing ha earnings and revenues provide differen signals. The correlaion beween and is significanly higher over han over (Table 4 abou here) 5.2. Marke reacion o and : Porfolio analysis Table 5 presens he effec of earnings and revenues surprise on size-adjused reurns (SAR). Panels A and B presen SAR for variable-sized porfolios. All observaions are ranked according o earnings surprise and revenues surprise and assigned ino quiniles. Each cell presens SAR for porfolios conaining observaions ha are in quinile i of and in quinile j of. For example, Cell (1, 1) conains observaions ha are in he lowes quinile of boh and 17

20 . Panels C and D presen he number of observaions in each cell of panels A and B. Panels A1, A2 and A3 describe marke reacion o revenues and earnings surprise in he full sample, over and over , respecively. As expeced, higher or leads o higher SAR, implying ha boh earnings and revenues serve as explanaory variables of sock reurns. The differences in mean SAR beween he firs and he fifh quinile of in he full sample, over and over are 3.93%, 4.10% and 3.83%, respecively (significanly differen from zero a he 1% level). The differences in mean SAR beween he firs and he fifh quinile of in he full sample, over and over are 2.80%, 2.40% and 3.02%, respecively (significanly differen from zero a he 1% level). The differences beween exreme cells (beween cell (1, 1) and cell (5, 5)) in he full sample, over and over are 5.18%, 4.95% and 5.33%, respecively (significanly differen from zero a he 1% level). Moreover, he marke reacion o an increase in is posiive, uncondiional on he level of. Marke reacion o an increase in is also posiive, regardless of he level of for he full sample. However, over , here is a sligh decrease in abnormal reurns when moving from cell (3, 2) o cell (3, 3), and over , here is a sligh decrease in abnormal reurns when moving from cell (4, 2) o cell (4, 3). These wo excepions do no change he overall view, namely he posiive relaionship beween revenues surprise and abnormal reurns. The posiive relaionship beween SAR and earnings (revenues) over revenues (earnings) implies ha revenues and earnings are 18

21 wo disinc signals affecing sock reurns. This resul is consisen wih Jegadeesh and Livna (2004a). I now examine marke reacion o earnings surprise vs. revenues surprise. As repored in panel A, SAR is posiive in he fifh quinile of, regardless of he level of (significan a he 1% level, excep for cell (5, 1) over ). Moreover, SAR is negaive in he firs quinile of, regardless of he level of (significan a he 1% level). According o hese resuls, earnings dominae revenues wih regard o marke reacion. Overall, he resuls in panel A are consisen wih Erimur e al. (2003) and Jegadeesh and Livna (2004a) and sugges ha boh earnings and revenues ac as key value drivers (over he full sample and in each of he wo sub-periods), bu ha he marke reacion o earnings is sronger han he reacion o revenues. Panel B presens he effec of R&D inensiy on marke reacion. Panel B1 describes he resuls in he R&D sample, and panel B2 (panel B3) describes he resuls when RDR is higher (lower) han 5%. As repored in panel B, higher or leads o higher SAR and, overall, marke reacion o an increase in () is posiive, uncondiional on he level of (). However, here is an imporan disincion beween high R&D companies (panel B2) and low R&D companies (panel B3). In low R&D companies, SAR is posiive (negaive) when is in is highes (lowes) quiniles, regardless of he level of (significan a he 1% level). The resul is differen for high R&D companies, where low wih high leads o negaive, bu no significan, SAR of -0.26% (-sa of -0.96), and high wih low leads o negaive SAR of -1.11% (significan a he 1% level). This finding suggess ha, in conras o he full sample and o low R&D companies, in high R&D companies, revenues are no 19

22 dominaed by earnings as a key value driver. The reason is ha, in high R&D companies, an esimaor of sabiliy and fuure performance has higher informaion conen due o heir high uncerainy and low earnings precision; in such siuaions, revenues may funcion as a credible esimaor pf seadiness, as discussed in secion 3. I should be menioned ha, in panel B2 (high R&D companies), cells (1, 5) and (5, 1) conain only 1,191 and 978 observaions, respecively, as repored in panel D2. To srenghen he conclusions regarding he effec of R&D on marke reacion, a regression analysis is performed and repored in secions 5.3 and 6.1. (Table 5 abou here) I now focus on he incremenal influence of and on SAR. Table 6 panel A (panel B) presens he incremenal effec of earnings surprise (revenues surprise). As saed in panel A (panel B) he difference in SAR beween he highes and he lowes quinile of () is posiive in all he () quiniles (significan a he 1% level). This resul is valid for all he differen samples (over , and he R&D sample). These resuls are consisen wih Jegadeesh and Livna (2004a) and imply, once again, ha revenues and earnings serve as key value drivers and each of hem has an incremenal effec over he oher. Ineresingly, he incremenal influence of revenues on SAR is higher over han over This finding may resul from he lower correlaion beween and over han over The economic explanaion for he differences could be he economic boom of he early 1990s, and he imporance of revenues as an indicaor for persisence and fuure performance may be higher in volaile periods. 20

23 Focusing on panel B, differences in he effecs of revenues over he samples can be idenified. The incremenal effec of in high R&D companies is, on average, double he effec observed in low R&D companies (abou 3.0% in comparison wih 1.6%, he difference being significan a he 1% level). This resul srenghens he conclusion ha revenues play a more imporan role in high R&D companies. To summarize he resuls presened in Tables 5 and 6, i can be seen ha boh earnings and revenues have informaion conen, he influence of earnings dominaing ha of revenues. However, in high R&D companies, he role of revenues is more imporan due o heir high uncerainy and he persisence parameer requiremen. As a resul, in high R&D companies he informaion conen of earnings is no higher han ha of revenues. (Table 6 abou here) 5.3. Regression analysis To examine marke reacion o earnings and revenues surprise, I also use regression analysis. I use a Fama-MacBeh (1973) ype regression by performing 30- yearly regressions. Table 7 presens regression resuls for equaion (1): SAR = 0 + α1 + α 2 + e α (1) To es robusness, I also used he marke-o-book raio as an addiional explanaory variable in he regression analysis process, and found ha he resuls are robus o he inclusion of a growh parameer. Table 7 conains he following caegories: (1) The full sample, divided ino wo sub-periods ( and ). (2) Firms wih negaive EPS. 21

24 (3) The R&D sample, divided ino high and low R&D companies. (4) The H index sample. The H index is he average Herfindahl-Hirshman Index (HHI) for each indusry, HHI being an index of indusry concenraion, calculaed as he sum of he squares of marke shares in revenues for each firm. 15 The value of HHI lies beween 0 and 10,000, where 10,000 means a single company in he indusry, and a value of zero indicaes ha he indusry is characerized by a number of infiniesimally small firms. When here are N equal-size companies, HHI equals 1/N. The H sample is divided ino hree subcaegories: (a) Indusries wih HHI under 4,000. (b) Indusries wih HHI beween 4,000 and 6,000. (c) Indusries wih HHI over 6,000. In all he caegories, boh and are posiive (significan a he 1% level). Adj-R 2 is in he range of as found in similar sudies. As expeced, adj-r 2 is lower for firms wih negaive EPS, where earnings and revenues have less explanaory power. The coefficien esimae of is abou for mos caegories, excep for firms wih negaive EPS, where he coefficien esimae is lower (0.0016), consisen wih Hayn (1995). In conras, he coefficien esimae of varies across caegories, being higher for han for (consisen wih Table 6, as discussed in secion 5.2.). The coefficien esimae of in high R&D companies is higher han observed in low R&D companies, consisen wih he resuls in Tables 5 and 6. The coefficien esimae of is lower in firms wih negaive EPS. The effec of he EPS sign on marke reacion is furher discussed in secion 6.2. As for he HHI index (he concenraion es), he coefficien esimae of is higher when HHI 22

25 lies beween 4000 and 6000, as expeced. This resul implies ha revenues play a more imporan role in indusries wih oligopolisic compeiion, where wo o hree dominan companies rule he marke. In his siuaion, marke sraegy plays an imporan role, because, in an oligopoly, a larger marke share is an invesmen ha can yield an abiliy o dicae fuure economic parameers in he marke. The effec of indusry concenraion is discussed in deail in secion 7. Focusing on he differences in he explanaory power of and, i can be seen ha in mos caegories he associaion of wih SAR is higher han he associaion wih. However, in high R&D companies, coefficien esimaes of and are similar. The resul regarding R&D companies is consisen wih Tables 5 and 6, and furher discussed in secion 6.1. In sum, he resuls in Table 7 suppor he previous conclusions ha boh earnings and revenues have informaion conen and he influence of earnings dominaes ha of revenues. However, in high R&D companies and in indusries wih wo o hree dominan companies, I find ha he role of revenues is more imporan. Regarding R&D companies, he resuls show ha he dominaion of earnings over revenues as an explanaory variable for sock reurns diminishes in high R&D companies. The dominaion of earnings as an explanaory variable of sock reurns is also diminished for firms wih negaive EPS, due o he small explanaory power of and. (Table 7 abou here) 6. The Effec of Firm s Specificaion on Marke Reacion 6.1. The effec of R&D on marke reacion Table 8 presens he regression analysis for equaion (2): 23

26 SAR β 0 + D + β1 + β 2 + β 3 D + β 4 D + ε = (2) D is a dummy variable ha obains he value of 1 if RDR is higher han 5%, and 0 oherwise. As found in secion 5.3, he coefficien esimae of boh and is posiive and saisically significan a he 1% level. The resuls also confirm my predicion regarding he influence of earnings and revenues in firms wih high R&D. The coefficien esimae of D ( D ) is posiive (negaive) and significan a he 1% level (2% level). These resuls imply ha revenues (earnings) have higher (lower) explanaory power in deermining sock reurns in high R&D companies. However, he change is greaer, in absolue erms, for revenues. As previously saed, hese resuls sem from he fac ha he environmen of high R&D companies is characerized by higher uncerainy regarding fuure performance and low earnings persisence, where Kohari e al. (2002) find posiive relaions beween R&D expendiure and earnings variabiliy. I is an environmen ha evokes a need for parameers ha serve as proxies for sabiliy and furher persisence, and one such parameer is revenues. This is consisen wih Erimur e al. (2003), Jegadeesh and Livna (2004a,b), Gu e al. (2004) and Ghosh e al. (2005) regarding persisence of revenues, and wih Anhony and Ramesh (1992), who find ha marke reacion o unexpeced revenues is higher for firms in he early sages of heir life cycle. Erimur e al. (2003) and Jegadeesh and Livna (2004a) also show ha revenues and earnings have higher informaion conen in growh companies han in value companies. To es robusness, I also used he marke-o-book raio as an addiional explanaory variable in he regression process, including i as a growh parameer, and found ha i does no change he resuls, implying ha R&D is an addiional facor beyond growh. In addiion, he moivaion for esing he R&D effec is differen as explained above, and he resuls regarding earnings surprise are 24

27 herefore opposie o he resuls in Erimur e al. (2003) and Jegadeesh and Livna (2004a) regarding growh companies. (Table 8 abou here) 6.2. The marke reacion o negaive EPS Table 9 presens he regression analysis for equaion (3): SAR + γ D 4 = γ + D L 0 L + γ 1 + η + γ 2 + γ D 3 L (3) D L is a dummy variable ha obains he value of 1 if EPS is negaive, and 0 oherwise. The sample is rimmed a 1% and 99 % for and, and a 99% for SAR. Nearly 5% of he observaions (12,017) are removed in his process. The coefficien esimae of D L and D L is negaive and significan a he 1% level, implying ha earnings and revenues have lower informaion conen in loss-reporing firms. The resuls regarding are consisen wih Hayn (1995). However, he influence of he EPS sign on he explanaory power of earnings is greaer han he influence on revenues, because in loss-reporing firms invesors look a revenues as a "feasible" measure of fuure oucomes. (Table 9 abou here) To sum up he effec of firm's specificaion on marke reacion, i can be concluded ha in high R&D companies and in loss-reporing companies, due o lower earnings precision and earnings persisence, revenues raher han earnings have higher informaion conen. Amir e al. (2003) indicae ha analyss play a more imporan role in high R&D companies and in loss-reporing companies, implying ha in hese kinds of companies, indicaors of fuure performance are more valuable. In he conex of his sudy, revenues are indicaors of fuure oucomes, as discussed in secion 3. 25

28 7. The Effec of Indusry Specificaion on Marke Reacion: Indusry Concenraion Table 10 presens regression resuls for equaions (4) (6). These regressions are used o analyze he effec of indusry concenraion on he informaion conen of revenues. In panel A, I presen he regression resuls for equaion (4): SAR = D D (4) δ 0 + HHI + δ1 + δ 2 + δ 3 HHI + ξ D HHI is a dummy variable ha obains he value of 1 if HHI is beween Y and Z, and 0 oherwise. The sample is rimmed a 1% and 99 % for and, and a 99% for SAR. Nearly 5% of he observaions (14,038 for he full sample, and 7,469 for SIC codes ) are removed in his process. For robusness esing, panel A is divided ino wo caegories: he full sample and SIC codes SIC codes represen he manufacuring division. 16 Consisen wih Table 7, he resuls in panel A show ha he informaion conen of is significanly higher when HHI lies beween 4000 and The coefficien esimae of in an oligopolisic compeiion environmen is abou 33% higher han in oher forms of compeiions. In indusries wih oligopolisic compeiion, marke sraegy plays an imporan role because a larger marke share can yield an abiliy o influence fuure economic variables, such as prices. Ineresingly, he explanaory power of revenues is weaker when HHI lies beween 0 and 2000 (significan a he 4% level, in he full sample). This resul can be explained by he fac ha, a he lowes edge of HHI, marke share does no change ha much since i does no lead o an abiliy o dicae prices or supply erms. 17 Several sudies regarding he influence of indusry concenraion on profiabiliy have found ha, when HHI is relaively low (under 2500), differences in HHI do no change 26

29 profiabiliy (Sigler (1964), Collins and Preson (1966)), implying ha when HHI is relaively low, marke share sraegy is less imporan. Panels B and C presen he resuls of equaions (5) and (6), respecively: SAR = χ + χ χ D χ + χ D χ D χ D χ D χ D ψ (5) SAR = κ + κ κ D κ + κ D κ D κ D κ D µ + (6) Dxw is a dummy variable ha obains he value of 1 if HHI is beween x and w, and 0 oherwise (e.g. D0020 is a dummy variable ha obains he value of 1 if H is beween 0 and 2,000, and 0 oherwise). As repored in panels A and B, D 4060 * is saisically significan for he full sample and SIC codes The resuls of regressions (5) and (6) confirm he conclusion ha revenues surprise plays a more imporan role when HHI lies beween 4000 and 6000, ha is, in an oligopolisic compeiion environmen. (Table 10 abou here) 8. Effec of Fiscal Quarer on Marke Reacion Table 11 presens he explanaory power of revenues and earnings across differen quarers. This invesigaion is performed since earnings managemen is expeced o be a sronger facor in he fourh quarer han in he firs hree quarers (Cohen e al. (2004)). To he exen ha earnings managemen decreases earnings precision and does no affec revenues precision, earnings are expeced o have a weaker explanaory power relaive o revenues in quarers wih more inensive earnings managemen. 18 Panel A presens regression resuls for equaion (7): 27

30 SAR = D D D (7) ν ν 1 + ν 2 + ν ν ζ D 4 is a dummy variable ha obains he value of 1 for he fourh quarer, and 0 oherwise. As expeced, he resuls of panel A indicae ha he explanaory power of earnings is significanly decreasing in he fourh quarer. The change in he explanaory power of revenues is no significanly differen from zero, which reduce he dominaion of earnings as key value driver. Panel B presens resuls for equaion (8): SAR = θ 0 + Di + θ1 + θ 2 + θ 3 D1 + θ 4 D2 i= 1 (8) + θ D + θ D + θ D + θ D + ϕ D i is a dummy variable ha obains he value of 1 for he i h quarer, and 0 oherwise. Consisen wih panel A, he resuls of panel B indicae ha he explanaory power of earnings is higher in firs hree fiscal quarers relaive o he fourh quarer, due o he fac ha mos conracs are less incenives o manage accouning numbers before knowing he annual resuls. However i should be noed ha earnings managemen may no be he only reason for lower explanaory power of earnings in he fourh quarer. Alernaive explanaions, in his conex, may relae o inerim reporing requiremen and he inegral approach o quarerly earnings ha influence ime-series properies of quarerly earnings. This migh leads o higher forecas error and lower informaion conen of earnings in he fourh quarer. 19 (Table 11 abou here) 9. Marke Reacion o Conflicing Earnings and Revenues Surprises Panel A (panel B) of Table 12 presens marke reacion o negaive (posiive) and posiive (negaive). The objec of he able is o use anoher mehod o es invesor reacion o earnings vs. heir reacion o revenues, in differen conexs 28

31 (R&D inensiy and indusry concenraion). As repored in Table 12, for he full sample, earnings have higher informaion conen han revenues, SAR being posiive (negaive) when is posiive (negaive), significan a he 1% level. However, over SAR is abou zero when is negaive and is posiive. This resul is consisen wih my previous resuls regarding he higher informaion conen of revenues for , as discussed in secion 5.2. For high R&D companies, negaive wih posiive does no lead o significan negaive SAR. This resul is sronger regarding indusries wih HHI beween 4000 and 6000, where mean SAR is posiive, bu no significanly differen from zero, when is negaive and is posiive. However, when is posiive and is negaive, SAR is posiive and saisically significan in high R&D companies and in an oligopolisic compeiion environmen. These resuls confirm my previous conclusions ha, relaive o earnings, revenues play a more imporan role in high R&D companies, loss-reporing companies, and in oligopolisic indusries. The reason for his could be ha in high R&D companies, when is negaive and is posiive, here is opimism regarding fuure oucomes because R&D may serve as invesmen, and revenues may be seen as a proxy for fuure reurns. Regarding indusries wih oligopolisic compeiion, when is negaive bu is posiive, marke reacion is no unfavorable because posiive implies higher marke share. Higher marke share is an invesmen ha migh yield higher earnings in he fuure, as discussed in secion 3. The resuls in Table 12 may assis managemen in deciding beween an earnings sraegy and a marke share sraegy. (Table 12 abou here) 29

32 10. Concluding Remarks and Furher Research This sudy examines he informaion conen of revenues and earnings and shows, using several mehods, ha each parameer has an incremenal effec on abnormal reurns afer conrolling for he oher. The general conclusion from comparing he informaion conen of hese wo variables is ha he explanaory power of earnings in deermining sock reurns is higher. The main conribuion of his sudy is is examinaion of he informaion conen of earnings and revenues according o firm and indusry specificaion. In addiion, I use conexual analysis o examine marke reacion o opposie signs of earnings and revenues surprise. I find ha in high R&D companies, he role of revenues is more imporan han in low R&D companies, due o high uncerainy, low earnings precision and persisence, and he requiremen for an indicaor of fuure oucomes, previous sudies having indicaed ha revenues may serve as a suiable indicaor of persisence. Thus, in high R&D companies he explanaory power of earnings is no higher han he explanaory power of revenues. The dominaion of earnings over revenues is also diminished in firms ha repor losses. In loss-reporing firms, he explanaory power of earnings and revenues is lower han in profiable firms, bu he decrease in he explanaory power of earnings is higher. This resul is consisen wih Hayn (1995), who finds ha he informaion conen of earnings is lower in companies wih losses. As expeced, I also find ha he explanaory power of earnings is significanly lower in quarers wih relaively high earnings managemen. Turning o indusry concenraion, he influence of revenues on sock reurns is shown o be higher when indusry is ruled by wo o hree dominan companies in an environmen of oligopolisic compeiion. In indusries wih oligopolisic compeiion, marke share sraegy is imporan, because a larger marke 30