The Underperformance of IPOs: the Sensitivity of the Choice of Empirical Method

Transcription

1 Journal of Economcs and Busness Vol. XI 2008, No 1 & No 2 The Underperformance of IPOs: he Sensvy of he Choce of Emprcal Mehod Wald Saleh & Ahmad Mashal ARAB OPEN UNIVERSITY Absrac Ths paper nvesgaes he performance of Jordanan nal publc offerngs (IPOs) usng daa from Amman Sock Exchange (ASE) over he perod We suded he sensvy of he model used o esmae he cumulave average abnormal reurns. In dong ha, we employed hree dfferen models and used wo approaches o es he resul: Sandard even-me analyss and he calendar-me approach. The hree models used o esmae he cumulave average abnormal reurns produce sgnfcan negave abnormal reurns when we employ even-me analyss. However, he calendar-me approach concludes ha he long-erm performance of Jordanan IPOs s no dfferen han ha of he overall marke. Keywords: Inal publc offerngs (IPOs); long-run reurns; emergng markes. JEL classfcaon: G14 Inroducon Prevous research has concluded ha nvesng n recen nal publc offerngs (IPOs) s a poor nvesmen sraegy (e.g. Rer, 1991; Loughran & Rer, 1995; and Levs, 1993, amongs ohers) snce IPOs underperform a marke ndex. However, Bravo and Compers (1997) and Espenlaub e al. (2000) fnd ha he underperformance of IPOs s sensve o he approach used o esmae abnormal performance. 34

2 Up o dae, no publshed paper exss ha nvesgaes he performance of IPOs usng daa from Amman Sock Exchange (ASE). Thus, he am of hs paper s o nvesgae he performance of he IPOs usng daa from ASE over he perod In parcular, hs paper seeks o nvesgae he sensvy of he models used o esmae abnormal reurns. We employ hree dfferen models o esmae abnormal reurns. These are: he capal asse prcng model (CAPM), he Fama and French (1993) hree-facor model and a Fama-French ype model (hereafer, he mul-facor model) exended for leverage effec, lqudy effec, sock reurn volaly, and loserwnner effec. Furhermore, we apply wo dfferen approaches o es he sensvy of he above hree models: even-me analyss n whch abnormal reurns are calculaed for up o 60 monhs afer he IPO and calendar-me analyss n whch equally weghed 12-monh porfolos are consruced o nclude any frm whch has an IPO durng he prevous 12-monh perod. The laer approach has advanage over he former n whch conrols for crosscorrelaon n reurns (e.g. Loughran and Rer, 1995 & Espenlaub e al, 2000), hus, yelds o well-specfed es sascs (e.g. Lyon e al., 1999). Usng even-me analyss, he resuls ndcae ha here s no sgnfcan underperformance n he frs 18 (14) monhs afer he ssuance of IPOs for he CAPM (Fama-French model). Afer ha he wo benchmarks ndcae a sgnfcan underperformance up o 60 monhs. The mul-facor model yelds slghly o dfferen resuls. The -sascs are posve (bu no sgnfcan) n mos cases n he frs 27 monhs, bu negavely sgnfcan afer ha up o 60 monhs. Overall, we conclude ha he exen of he underperformance of IPOs s sensve o he model adoped. Usng he calendar-me analyss mehod or model o conrol for crosscorrelaon n reurns, he resuls provde less evdence of underperformance. Ths resul s conssen wh Espenlaub e al. (2000). Therefore, hs paper concludes ha, among he models used o esmae abnormal reurns, he CAPM model seems o be he bes approach ha descrbes he paerns of abnormal reurns, and ha he calendar-me analyss s more approprae n reflecng hs paern. Leraure Revew The problem of underperformance has been expansvely researched. Fnancal economss n recen years have closely examned and nensely debaed he performance of Inal Publc Offerngs (IPO s). 35

3 Ibboson (1975) dscovered a saucer shaped paern upon sudy of a random seleced secury from each monh n he sxes whereby he offerng nally yelded posve reurns; below marke reurns ensued, whle he fourh year reurns ended owards normal. The nal 48 monhs exhbed below normal performance. As mos reurns were negave wh a few very hgh, he dsrbuon of he reurns was evdenly very skewed. Ths pons o ndvdually rsky nvesmens. Ths underperformance was no sascally sgnfcan due o he small sze of he sample and he hgh sandard devaons. Rer (1991) focused on he 1975 o 1984 me perod wheren he reurns from a conrol sample of 1,526 nal publc offerngs from frms smlar n ndusry and sze were analyzed. The hree-year reurn was 34.47%; he conrol sample reurned 61.86% over he same hree years. Loughran (1993) examnaon of he reurns from 3,556 IPO s durng uncovered an average sx year oal reurn of 17.29% n conras wh 76.23% for he NASDAQ ndex durng he same me span. These resuls are consderably worse han hose of Rer's durng hs hree year ess. Comparavely, frms mached n sze on boh he New York Sock exchange and on NASDAQ yelded resuls whch showed a much sronger performance. Despe conrol of he exchange and he book o marke rao, a regresson equaon for July December 1988 had a sascally sgnfcan negave coeffcen for havng had an IPO whn sx years. Laer, Loughran & Rer (1995) examned nal publc offerngs from Ther examnaon uncovered a 5% average rae of reurn per year for he fve years afer ssuance. Ths s compared o frms of smlar sze whch yelded a 12% average rae of reurn. A more sgnfcan underperformance was found n he nal publc offerngs of medan frms; here, afer fve years, her IPO's averaged a negave reurn of 39%; n sgnfcan conras, smlar szed frms yelded a posve reurn of 16% for he fve years. The auhors calculae ha he forgone reurn was $39 bllon dollars, makng he underperformance of economc sgnfcance. Servaes & Rajan (1997) examned nal publc offerngs from They uncovered a fve year raw reurn of 24%. In comparson wh he NYSE/AMEX ndex, hs represens a 47% underperformance; wh he smalles decle from he NYSE/AMEX, a 17% underperformance; and wh frms smlar n sze and ndusry, a 41% underperformance. 36

4 Oher researchers (Bravo & Gompers 1997; Gompers & Lerner, 1999) examned a slghly dfferen se of years and smlarly uncovered hgh level underperformance effecs, wh he underperformance greaer for non-venure capal-backed companes. Teo, Welch, and Wong(1998) used frms gong publc beween 1975 and 1984 also found underperformance; her sudy also shows ha underpeformance was greaer n frms whch used more aggressve accounng. Aggarwal & Rvol (1990) found ha 1598 IPO's offered beween 1977 o 1987 underperformed he US marke by 13.73% over he frs 250 radng days. Forbes magazne (Sern & Bernsen, 1985, as ced n Rer, 1991) found, afer analyzng 1,922 IPO s prced over $1.00 ssued from January 1975 o June 1985, ha "from s dae as gong publc o las monh, he average new ssue was down 22% relave o he broad Sandard & Poors 500 sock ndex". I s mporan o pon ou ha hese sudes of US IPO's overlap n me perods as well as IPO's. The resuls are hence relable because dfferen mehodologes whch use varyng sascal mehods ((ncludng conrollng for a range of oher varables) have yelded smlar conclusons. Two US sudes whch examned earler me perods also found underperformance: Smon (1989) found ha IPO's offered from 1926 o 1933 lsed on regonal exchanges showed subsanal underperformance over 60 monhs whle. Soll & Curley (1970) found underperformance for 205 small ssues n he ffes and sxes. Ths underperformance s no lmed o he Uned Saes. Levs (1993), n examnng he hree year performance of 712 UK IPO's ssued beween 1980 and 1988, and dependng on he chosen benchmark, uncovered underperformance ha vared beween 8.3% and 23.0%. Uhlr (1988) also uncovered an underperformance of 7.4% afer one year for German ssues Fnn & Hgham's (1988) examned 93 Ausralan IPO's for They found ha buyng a he end of he lsng monh and holdng o he end of he frs year earned 6.52% below he ndces, bu ha hs loss was no que sascally sgnfcan. Kunz & Aggarwal (1994) found ha 42 Swss IPO's beween 1983 and 1989 experenced an underperformance of 6.1%. Keloharju (1993) found ha he average Fnnsh IPO los 22.4% from he frs marke radng o hree years laer, versus 1.6% average declne for he marke ndex. Therefore, seems ha he US paern of underperformance exends o oher counres. Sudes of emergng markes also revealed a smlar underpeformance effec o ha exhbed by developed counres. Aggarwal, Leal, & Hernandez (1993) uncovered an underperformance of 47% afer hree years Brazlan IPO's. 37

5 Chleean IPOs also yelded underperformance afer hree years whch averaged 23.7%, whle for Mexco he underperformance afer one year was 19.6%. In sudes of Asan markes, Dawson (1987) found, upon examnaon of one year marke adjused reurns for nal publc offerngs n Hong Kong, Sngapore, and Malaysa durng , ha hose for Hong Kong were down 9.3% and Sngapore 2.7%. Noceably, neher declne was sascally sgnfcan. Conrasngly, Malaysa yelded a posve, sascally sgnfcan overperformance of 18.2%. The auhor mporanly pons ou ha he Malaysan ndex used n he sudy was no a marke wde one, bu an ndusral one and hence does no consue a sgnfcan excepon. The one sgnfcan excepon o he paern of underperformance n Asan markes s Inda where Shah (1995) fnds (n a large daa se wh 2056 IPO's from ), ha afer ypcally ouperformng he marke for he frs 200 radng days, IPOs hen declne where afer 400 days hey are approxmaely a he level of he frs radng. From he aforemenoned sudes, appears ha n mos counres, IPOs underperform he marke over perods of one o fve years. These sudes examne IPO's by comparng hem o oher socks and do no ypcally adjus for rsk as rsk s dffcul o measure when no radng hsory exss. However, IPO reurns are much more varable han mos socks, wh he mean reurn usually exceedng he medan snce he average can be rased by a few large wnners. Thus, because of her hgh rsk, one would expec IPO's o ouperform he ndces, especally for undversfed ndvdual nvesors who canno coun on a porfolo wh only a few secures ncludng any of he bg wnners among he IPOs. Even on a sysemac rsk bass (bea), IPOs appear o be rsker han average. Daa and Research Mehodology Daa The sample of hs sudy consss of all IPOs ssued by Jordanan companes over he perod The emprcal analyss of hs sudy uses monhly marke nformaon as well as annual accounng daa. Thus, he daase used n hs paper consss of monhly sock reurns 11, monhly 3-monh Treasury bll raes as a proxy for rsk-free reurns, and monhly reurns on he ASE ndex as a proxy for marke reurns. Daa on book-o-marke of equy, marke 11 These reurns are adjused for sock dvdends, sock spl, and dvdend yeld. 38

6 capalzaon, radng volume, equy o oal asse rao are colleced from ASE gudelnes. Cumulave Average Abnormal Reurns (CAARs) In order o analyze he long-erm performance of IPOs, we apply he sandard even-sudy mehodology. To calculae long-erm abnormal reurns, hree models are employed. These benchmarks are: he capal asse prcng model (CAPM), he Fama and French (1993) hree-facor model (FF), and a Fama- French ype model (hereafer, he mul-facor model; ) exended for leverage effec, lqudy effec, sock reurn volaly, and loser-wnner effec. For each model, monhly abnormal reurns are compued for up o 60 monhs afer he IPO excludng he nal reurn perod (he perod beween he offer dae and he lsng dae). Followng prevous research n hs area, we esmae he model parameers and he excess reurns jonly and use n-sample esmaes of abnormal reurns. Then, abnormal reurns, for each model, are cumulaed over me up o perod T afer he IPO, hus: 1 AAR = N N = 1 = 1 Ab, CAAR T = CAR T where AAR s he average abnormal reurns, Ab, s he abnormal reurn n monh afer he IPO from company, N s he number of frms n he sample. We adop -es sascs ha are generaed based on Rer (1991). Thus, he followng -es s used o es wheher CAART dffers from zero: 1 2 CAART ( n) ( CAART ) = (3) 1 2 (.var+ 2( 1) cov) where, n s he number of frms radng n each monh, s he even monh, var s he average cross-seconal varance over 60 monhs, and cov s he frsorder auo-covarance of he abnormal reurn seres. Three Models o Esmae Abnormal Reurns We employ hree dfferen models o esmae abnormal reurns. Manly, we use: (1) he capal asse prcng model (CAPM) as follows: CAPM CAPM Ab = R + β ( R )] (4),, [ f, m, f, (1) (2) 39

7 Where R, s he reurn on frm n even monh, marke n even monh, f R m, s he reurn for he R, s he rsk-free rae n even me as measured CAPM by he reurn on he Treasury blls, β s he CAPM bea of frm, esmaed by he ordnary leas square (OLS) regresson up o 60 monhs afer he IPO. (2) he Fama and French (1993) hree-facor model as follows: FF FF FF Ab = R + β ( R ) + s SMB + h FF,, [ f, m, f, HML ] (5) SMB (small mnus bg) s he dfference, each monh, beween he average of he reurns on he hree small-sock porfolos (S/L, S/M, and S/H) and he average of he reurns on he hree bg-sock porfolos (B/L, B/M, and B/H). HML s he dfference, each monh, beween he average of he reurns of he wo hgh-book-o-marke porfolos (S/H and B/H) and he average of he reurns on he wo low-book-o-marke porfolos (S/L and B/L). Followng Fama and French (1993) he mmckng porfolos for he sze (SMB) and book-o-marke (HML) facors are consruced as follows. A he end of Aprl 12 of each year socks are allocaed o wo groups (bg and small) based on wheher her marke value s above or below he medan of he marke 13. Moreover, socks are allocaed ndependenly no hree book-o-marke groups (hgh, medum, and low) based on he breakpons for he op 30 percen, mddle 40 percen, and boom 30 percen of he book-o-marke values. We es wheher he hree-facor model explans he dfference n reurns beween wnners and losers by esng wheher he nercep n each regresson s equal o zero usng convenonal -sascs. (3) he mul-facor model. In addon o he above wo models, we employ a Fama-French ype model exended for leverage effec (LMU), lqudy effec (LMI), sock reurn volaly (HMLSTD), and loser-wnner effec (LMW) as follows: 12 Noe ha he fscal year-end for all companes lsed on Amman Sock Exchange s he end of December. Thus, we perm for four-monh gap o ensure ha he daa s avalable a he formaon dae. 13 We use he medan of he marke snce he number of he companes over he sample perod s jus

8 Ab, = R, [ Rf, + β ( Rm, f, ) + s v HMLSTD+ w LMW] SMB + h HML + u LMU + l LMI + LMU s he dfference, each monh, beween he average of he reurns on he wo hgh-leverage porfolos and he average of he reurns on he wo lowleverage porfolos. LMI 14 s he dfference, each monh, beween he average of he reurns on he wo hgh-radng-volume porfolos and he average of he reurns on he wo low-radng-volume porfolos 15. LMW s he reurn dfference beween porfolos of pas losers and pas wnners based upon reurns over he pas 12 monhs. HSVMLSV 16 s he dfference, each monh, beween he average of he reurns on he wo hgh-sock-volaly porfolos and he average of he reurns on he wo low-sock-volaly porfolos. Even and Calendar me-analyses For each model and for each IPO, we regress 60-monh excess reurns agans he respecve benchmarks. The average nercep value from hese regressons wll be a measure of average long-run abnormal reurns followng an IPO. Followng Loughran and Rer (1995), we also adop a calendar-me analyss. (6) 14 To consruc such a facor, I do he followng: A he end of Aprl of each year socks are allocaed o wo groups (bg and small) based on wheher her marke value s above or below he medan of he marke. Furher, socks are allocaed n an ndependen sor o hree radng-volume groups (hgh, medum, and low) based on he breakpons for he op 30 percen, mddle 40 percen, and boom 30 percen of he radng-volume values. From he nersecon of he wo sze groups (S and B) and he hree radng-volume groups (L, M, H), sx sze-radng-volume porfolos are consruced. 15 Noe ha radng volume was scaled by marke value of equy. 16 HSVMLSV facor s consruced as follows: A he end of Aprl of each year socks are allocaed o wo groups (bg and small) based on wheher her marke value s above or below he medan of he marke. Furher, socks are allocaed n an ndependen sor o hree sock volaly groups (hgh, medum, and low) based on he breakpons for he op 30 percen, mddle 40 percen, and boom 30 percen of he sandard devaon of he pas 12-monh reurns. From he nersecon of he wo sze groups (S and B) and he hree sandard devaon groups (L, M, H), sx sze-volaly porfolos are consruced. 41

9 Ths approach has an advanage over he even-me analyss n whch conrols for cross-correlaon n reurns (Loughran and Rer, 1995) and hus, yelds o well-specfed es sascs (Lyon e al., 1999). Therefore, for each calendar monh we form equally weghed n monh porfolos se up o nclude any frm whch has an IPO durng he prevous n monhs, n= 12, 24, 36, 48, and 60. Then, porfolo reurns are calculaed by equal weghng of he 5 years porfolo reurns n calendar me. Tha s, we nves 20% n he frs year, 20% n he second year, 20% n he hrd year, 20% n he fourh year, and 20% n he ffh year. Nex, we examne he performance of he 5 years porfolo by runnng he followng regressons: The CAPM = α + β ( R + e (7) Rp f, m, f, ), Rp f, = + β ( Rm, f, ) + ssmb + h HML + e, Rp α (8) f, = α + β ( Rm, f, ) + s SMB + h HML + u LMU + l LMI + v HMLSTD + w LMW + e, (9) where, R p f s he excess reurn n monh on he 5 years porfolo of IPO. Oher varables are defned above. Emprcal Resuls Table 1 presens he cumulave average abnormal reurns for he frs monh up o he 60 h monh afer he IPOs of alernave models used n hs sudy. The resuls confrm ha he CAAR s no sgnfcan n he shor and medum erms rrespecve of he model employed. For example, he performance of he Jordanan IPOs s no sascally and economcally dfferen han zero n he frs 17, 12, and 26 monhs, when we use he CAPM, FF, and model, respecvely, o esmae he CAARs. However, he resuls also provde evdence of long-erm IPO underperformance regardless of he model used o esmae he CAARs. For example, he hree models (CAPM, FF, and ) confrm he exsence of sascally and economcally sgnfcan long-erm IPO underperformance. Overall, he above resuls confrm he sgnfcan underperformance of Jordanan IPOs n he long-erm rrespecve of he model used o esmae he CAARs. However, he magnude of underperformance dffers based on he model used. For nsance, he CAAR over he 36 (48 and 60) monhs s -25% (-29% and -34%), -35% (-40% and -50%), and -14% (23% and -36%) for he CAPM, FF, and model, respecvely. 42

10 The resuls from Table 1 are llusraed graphcally n Fgure 1. Comparng he CAARs across he hree models used n hs paper, we can observe ha he mul-facor model () produces less underperformance, whls he Fama- French model (FF) produces hgh underperformance, and he capal asse prcng model (CAPM) les n beween. Table 2 repors he cross-seconal averages of he even-me regresson resuls for he alernave models used. Recall ha for each IPO, we run a me seres regresson of he 60-monh excess reurns on he varous models and hen average he coeffcens across he IPOs. Thus, α n he regresson represens a measure of he average abnormal reurns. If α s no sgnfcanly dfferen han zero, he null hypohess of no abnormal reurns can be acceped. The resuls from Table 2 confrm he resuls from Fgure 1. Tha s, he CAPM, FF, and models produce sgnfcanly negave nerceps whch mply long-run underperformance 17. However, here s a consderable body of sudes whch argues ha he longerm ess are ms-specfed and he here s sgnfcan over-rejecon of he null hypohess of no posve abnormal performance for he CAAR approach (e.g. Kohar and Warner; 1997 and Barber and Lyon; 1997). Moreover, Loughran and Rer (1995) argue ha he -sascs assessng he sgnfcance of abnormal reurns are lkely o be overesmaed due o he presence of crosscorrelaon n conemporaneous reurns. Lyon, Barber, and Tsa (1999) sugges a calendar-me approach nsead of he even-me analyss as conrols for cross-correlaon. Furher, hey show ha he calendar-me approach yelds well-specfed es sascs. Therefore, we employ he calendar-me approach. Table 3 presens he calendar-me regressons of he CAPM, FF, and FF models, respecvely. The resuls show ha he nercep coeffcens (α ) are nsgnfcanly posve, 0.14%, 0.09%, and 0.13% wh -sascs of 0.62, 0.37, and 0.46 for he CAPM, FF, and models respecvely. These resuls sugges ha he performance of Jordanan IPOs s posve, bu no sgnfcan compared o he marke performance. In addon o ha, Table 3 shows ha neher he sze facor nor he book-omarke facor has an effec on he performance of he Jordanan IPOs. The only facor ha loads sgnfcanly s he loser-wnner facor (LMW). The adjused for he CAPM, FF, and models s 15.6%, 4.5%, and 6.8%, respecvely. 17 Noe ha n he case of usng model o esmae he CAARs, he nercep s margnally sgnfcan a 10% level. 43

11 Ths suggess ha he CAPM s he bes approach ha descrbes he paerns of abnormal reurns n ASE. Fgure 2 shows and confrms he resuls llusraed n Table 3. The mplcaon of he resuls n Table 3 and Fgure 2 s ha a long-erm nvesmen sraegy based on holdng Jordanan IPOs yelds reurns close o he marke reurn. Summary and Conclusons Up o dae no publshed paper exss ha examnes he performance of Jordanan IPOs. Therefore, he objecve of hs paper was o nvesgae he long-erm reurns of Jordanan IPOs, up o fve years afer he ssuance. The sample consss of all IPOs over he perod. We use hree models o esmae he cumulave average abnormal reurns. Namely, hese are: he capal asse prcng model (CAPM), he Fama and French (1993) hree-facor model (FF), and a mul-facor model () ha exends he Fama-French model o nclude he leverage effec, he lqudy effec, he sock reurn volaly, and he loser-wnner effec. Furher, we use wo dfferen approaches o es he resuls: sandard even-me analyss and he calendar-me approach. The hree models used o esmae he cumulave average abnormal reurns produce sgnfcan negave abnormal reurns when we employ he even-me analyss. However, he calendar-me approach concludes ha he long-erm performance of Jordanan IPOs s no dfferen han ha of he overall marke. 44

12 Table 1: Cumulave average abnormal reurns for alernave models Monh CAPM resuls FF resuls resuls CAAR monh CAAR CAAR monh CAAR CAAR monh % -sa % -sa % -sa % -sa % -sa CAAR % -sa

13 Noe: fgures n Table 1 represen he cumulave average abnormal reurns for he capal asse prcng model (CAPM), he Fama-French hree facor model (FF), and a Fama-French ype model exended o nclude leverage effec, lqudy effec, sock reurn volaly, and loser-wnner effec. The -sascs are compued based on he mehod of Brown and Warner (1980). 46

14 Table 2: Even-Tme Regresson Resuls for Alernave Models. Coeffcen Panel A CAPM 47 Panel B FF Panel C α (α ) β ( β ) SMB (SMB) HML (HML) LMU (LMU) LMI (LMI) HMLSTD (HMLSTD) LMW (LMW) Noe ha he followng regressons were run n even me for each IPO, hen he coeffcen values were averaged over all IPOs. Rp f, = + β ( Rm, f, ) + e, Rp f, = + β ( Rm, f, ) + ssmb + h HML + e, Rp Rf, = α + β ( Rm, Rf, ) + ssmb + hhml + u LMU + llmi + + wlmw+ e R, R α (Panel A) α (Panel B) v HMLSTD (Panel C) where, p f s he excess reurn n monh on he 5 years porfolo of IPO. R, s he reurn for he marke n even monh, m R, s he rsk-free rae n even me as measured by he reurn on he Treasury blls. SMB (small mnus bg) s he dfference, each monh, beween he average of he reurns on he hree small-sock porfolos (S/L, S/M, and S/H) and he average of he reurns on he hree bg-sock porfolos (B/L, B/M, and B/H). HML s he f

15 dfference, each monh, beween he average of he reurns of he wo hghbook-o-marke porfolos (S/H and B/H) and he average of he reurns on he wo low-book-o-marke porfolos (S/L and B/L). LMU s he dfference, each monh, beween he average of he reurns on he wo hgh-leverage porfolos and he average of he reurns on he wo low-leverage porfolos. LMI s he dfference, each monh, beween he average of he reurns on he wo hghradng-volume porfolos and he average of he reurns on he wo lowradng-volume porfolos. LMW s he reurn dfference beween porfolos of pas losers and pas wnners based upon reurns over he pas 12 monhs. HSVMLSV s he dfference, each monh, beween he average of he reurns on he wo hgh-sock-volaly porfolos and he average of he reurns on he wo low-sock-volaly porfolos. T-sascs are calculaed wh sandard errors usng Whe (1980). Table 3: Calendar-Tme Regresson Resuls for Alernave Models. Coeffcen Panel A CAPM Panel B FF Panel C α (α ) β ( β ) SMB (SMB) HML (HML) LMU (LMU) LMI (LMI) HMLSTD (HMLSTD) LMW (LMW) R Noe ha he followng regressons are esmaed usng 252 monhly observaons: = α + β ( + (Panel A) Rp R f, Rm, R f, ) e, 48

16 Rp f, = + β ( Rm, f, ) + ssmb + h HML + e, Rp f, = α + β ( Rm, f, ) + ssmb + h HML + u LMU + l LMI + w LMW + e R, R α (Panel B) + v HMLSTD (Panel C) where, p f s he excess reurn n monh on he 5 years porfolo of IPO. R, s he reurn for he marke n even monh, m R, s he rsk-free rae n even me as measured by he reurn on he Treasury blls. SMB (small mnus bg) s he dfference, each monh, beween he average of he reurns on he hree small-sock porfolos (S/L, S/M, and S/H) and he average of he reurns on he hree bg-sock porfolos (B/L, B/M, and B/H). HML s he dfference, each monh, beween he average of he reurns of he wo hgh-book-o-marke porfolos (S/H and B/H) and he average of he reurns on he wo low-book-o-marke porfolos (S/L and B/L). LMU s he dfference, each monh, beween he average of he reurns on he wo hghleverage porfolos and he average of he reurns on he wo low-leverage porfolos. LMI s he dfference, each monh, beween he average of he reurns on he wo hgh-radng-volume porfolos and he average of he reurns on he wo low-radng-volume porfolos. LMW s he reurn dfference beween porfolos of pas losers and pas wnners based upon reurns over he pas 12 monhs. HSVMLSV s he dfference, each monh, beween he average of he reurns on he wo hgh-sock-volaly porfolos and he average of he reurns on he wo low-sock-volaly porfolos. T-sascs are calculaed wh sandard errors usng Whe (1980). f 49

17 Fg 1- Cumulave average abnormal reurns on Jordanan IPO for alernave benchmarks CAAR-CAPM CAAR-FF CAAR CAAR Monhs afer IPO Fg. 2: Cumulave Reurns on IPO Porfolos, Marke Porfolo, and Rsk-Free rae Rp Rf Rm Rp Rm % 1 Rf Monh From Jan 1981-Jan

18 References Aggarwal, R. &Rvol, P Fads n he nal publc offerng marke? Fnancal Managemen 22, Agarwal, R. Leal, R. &Hernandez, L The afermarke performance of nal publc offerngs n Lan Amerca. Fnancal Managemen 22, Barber, B. M. and Lyon, J. D., Deecng Long-Run Abnormal Sock Reurns: The Emprcal Power and Specfcaon of Tes Sascs. Journal of Fnancal Economcs, 43, 3, Bravo, A.& Gompers, P.A Myh or realy? The long-run underperformance of nal publc offerngs: Evdence from venure and nonvenure capal-backed companes. Journal of Fnance, 52, Brown, S. and Warner, J Measurng Secury Prce Performance. Journal of Fnancal Economcs, Vol. 8, PP Dawson, S. M Secondary sock marke performance of nal publc offers, Hong Kong, Sngapore and Malaysa: Journal of Busness Fnance and Accounng 14, Espenlaub, S., Grergory, A., and Tonks, I Re-Assessng he Long-Term Under-Performance of UK Inal Publc Offerngs. European Fnancal Managemen, 6,3, pp Fama, Eugene F. and Kenneh R. French, 1993, Common rsk facors n he reurns on socks and bonds, Journal of Fnancal Economcs 33,3-56. Fnn, F. J. & Hgham, R The performance of unseasoned new equy ssues-cum-sock exchange lsngs n Ausrala. Journal of Bankng and Fnance 12, Gompers, P. and J. Lerner 1999, The Venure Capal Cycle, (Cambrdge, MA, MIT Press). Ibboson, R. G Prce performance of common sock new ssues. Journal of Fnancal Economcs 2, Keloharju, M The wnner s curse, legal lably, and he longrun prce performance of nal publc offerngs n Fnland. Journal of Fnancal Economcs, 34. Kohar, S.P. and Warner, J Measurng Long-Horzon Secury Prce Performance. Journal of Fnancal Economcs, Vol. 43, PP Kunz, R.M. & Aggarwal, R Why nal publc offerngs are underprced: Evdence from Swzerland. Journal of Bankng and Fnance 18, Levs, M The long run performance of nal publc offerngs. The UK experence Fnancal Managemen 22,

19 Lougharan, T NYSE vs NASDAQ reurns: Marke mcrosrucure or he poor performance of nal publc offerngs. Journal of Fnancal Economcs 33, Loughran, T. and Rer, J The New Issues Puzzle. The Journal of Fnance, Vol. 50, PP Lyon, J., Barber, B. and Tsa, C Improved Mehods for Tess of Long-Run Abnormal Sock Reurns. The Journal of Fnance, Vol. 54, PP Rer, J.R The long-run performance of nal publc offerngs. Journal of Fnance, XLVI, No. 1,3-27. Servaes, H. Rajan, R Analys followng of publc offerngs. Journal of Fnance, 52, No. 2, Smon, C The effec of he 1933 secures ac on nvesor nformaon and he performance of new ssues. Amercan Economc Revew, 79, Shah, A The Indan IPO marke: Emprcal facs. Workng paper. Cenre for Monorng Indan Economy, Bombay, Inda. Sern, R. L.& Bernsen, P Why new ssue are lousy nvesmens. Forbes 136, Soll, H.R. & Curley, A.J Small busness and he new ssues marke for eques, Journal of Fnancal and Quanave Analyss 5(3) Teo, S. H., Welch, I, and Wong, T.J Earnngs managemen and he long-run marke performance of nal publc offerngs. Journal of Fnance, 53, Uhlr, H. (1988). Gong publc n he F.R.G. n A reapprasal of he Effcency of Fnancal Markes. In Gumaraes, R.M.C., Kngsman, B,. G. & Taylor, S.J. (Eds). New York: Sprnger-Verlag. Whe H, A heeroscedascy conssen covarance marx esmaor and a drec es of heeroscedascy. Economerca