Invesmen Managemen and Financial Innovaions, Volume 8, Issue 3, 011 Mohamed Albaiy (Malaysia), Rubi Ahmad (Malaysia) Reurn erformance, leverage effec, and volailiy sillover in Islamic sock indices evidence from DJIMI, FTSEGII and KLSI Absrac Emirical sudies on sock reurns and volailiy have no made serious aem o examine hese wo issues on he conex of Islamic sock marke indexes. This aer, herefore, invesigaes he behavior of reurns and volailiy of hree Islamic sock marke indices DJIMI, FTSEGII, and KLSI ha are lised in he USA, he Unied Kingdom, and Malaysia resecively. Our aer examines four main issues: (1) wheher here is a difference in reurns among hese Islamic sock marke indices; () wheher here is a risk remium in each sock exchange; (3) wheher hese indices face he leverage effec risk and lasly; (4) wheher here is a volailiy sillover among hese hree Islamic sock marke indices. The emirical invesigaion is conduced by means of he GARCH model (GARCH-M) using daily daa covering he eriod from January 1999 unil Ocober 007. No only our resuls show no significan difference in heir reurns, risk remium is found o be absen in each Islamic sock index. While KLSE reors no leverage effec, DJIMI and FTSEGII indicae oherwise. Finally, based on EGARCH and TARCH models here is a sillover from DJIMI and FTSEGII oward KLSI bu no vice versa. Keywords: Islamic index, volailiy, GARCH, Sillover, DJIMI, FTSEGII, KLSI. JEL Classificaion: G10, G11, G1, G15. Inroducion Over he las weny years here has been a coninuous develomen in he convenional banking and finance o roduce an Islamic counerar o caer for Muslim oulaion around he globe. One of hese develomens is he iniiaion of Islamic sock indices. An Islamic sock index measures he erformance of a cerain baske of securiies and hese securiies are ermissible for he Muslim o inves. The hree oular Islamic sock marke indices are Financial Times Sock Exchange Global Islamic index (FTSEGII) of he London Sock Marke, Dow Jones Islamic Marke Index (DJIMI) of he New York Sock Exchange and lasly, Kuala Lumur Syariah Index (KLSI) of he Bursa Malaysia inroduced beween January 1998 and December 1999. Similar o convenional sock indices, hese Islamic sock indices are designed o monior he erformance of some secors of he financial markes, which he invesmen follows closely o he enes of Islam. DJIMI and FTSEGII cover wide range of counries and socks while KLSI covers only local lised socks. Pas sudies have concenraed on he erformance of hese hree indices agains heir convenional counerars. Theoreically, he value of any invesmen is deermined by he resen value of he invesmen s execed fuure cash flows. Subseuenly, a raional invesor maximizes his uiliy by maximizing his wealh and minimizing risk. A raional invesor who wans o maximize his uiliy will choose he highes ossible reurn for a given level of risk ha can be achieved by consrucing a well- Mohamed Albaiy, Rubi Ahmad, 011. diversified orfolio. This alies o all orfolio invesmen decisions including screened invesmen funds such as he Islamic Muual Funds. Given ha no all socks lised on he sock exchanges are ermissible for he Muslims o inves, every fund manager of Islamic Muual Funds has o obain he aroval from his comany s Shariah Board before urchasing any new shares. The sricer screening crieria in screened invesmen as observed in he Islamic Muual Funds have been argued as one of he reasons why screened invesmen in general brings lower execed reurn han unscreened invesmen (Rudd, 1981; Teer, 1991; Johnson and Neave, 1996; and Langbein and Posner, 1980). The low diversificaion benefis by screened invesmen resuled o in higher orfolio risk. On o of ha, screened invesmen is also erceived o incur high adminisraion and monioring coss. Following he work by Abdul Rahim, Ahmad and Ahmad (009) ha exlores he volailiy of Islamic indices in Malaysia and Indonesia, in his aer we examine he sock reurns and volailiies in hree Islamic sock marke indices namely, FTSEGII, DJIMI and KLSI. This sudy is differen from Abdul Rahim e al. (009) sudy is four folds. Firs, his sudy uses hree differen sock marke indices while Abdul Rahim e al. (009) is sudying wo closely relaed markes Malaysia and Indonesia. Second difference is ha Indonesian Islamic marke index is raher small. I conains 30 lised comanies while DJIMI and FTSEGII indices conain more han 1000 lised comanies from many counries. Third, KLSI is lis comanies from Malaysia while FTSE- GII and DJIMI include local and inernaional firms from differen counries and regions. Forh, he Islamic sock indices in hese hree markes have dis- 161
Invesmen Managemen and Financial Innovaions, Volume 8, Issue 3, 011 incive screening crieria. Having differen screening crieria migh lead o difference in reurns. Therefore, he firs uesion of his sudy is wheher here is a significan difference beween he hree Islamic sock marke indices. Besides comaring heir reurns and volailiy, we also examine he leverage effec of a fall in he securiy rices lised in DJIMI, FTSEGII and KLSI. According o Black (1976), volailiies and asse reurns can be negaively correlaed and his relaionshi is oularly known as he leverage effec. Brooks (008) exlains ha leverage effec haens when a fall in he rice of a firm s sock causes he firm s deb o euiy raio o increase. When he large decline in he euiy rice is no mached by he decline in he value of deb, he firm s deb o euiy raio will rise ogeher wih he financial risk of he firm s invesors. Because of he higher risk, invesors would exec he volailiy of he sock reurn o rise also. Cheung and Ng (199), Poon and Taylor (199), Koumos (1996) Koumos and Booh (1995), Booh, Marikainen and Tse (1997) found ha here is a significan leverage effec and bad news (i.e., decrease in sock rices) seem o have a greaer influence on sock rices han good news (i.e., increase in sock rice). If he Islamic indices screen high deb o euiy raio firms such as DJIMI and FTSEGII hen hey should minimize he leverage effec comared o KLSI which does no have any screening ac agains deb o euiy raio. This is because a comany having a higher han he benchmark deb o euiy raio is excluded from he DJI- MI and FTSEGII. Ulrich and Marzban (008) ha boh Islamic and convenional finance agree ha lower deb is beer han higher deb because lower deb is inerreed as a osiive invesmen signal. Boh DJIMI and FTSEGII have a screening crieria based on he level of deb. Boh indices eliminae firms ha have deb raios exceeding 33%. However, KLSI does no have any crieria agains deb raio. Based on his reasoning, we osulae ha leverage effec o be rominen in KLSI bu no in DJIMI and FTSEGII. In addiion o ha, he Islamic indices ha yield low reurns are execed o have higher risk and will no be comensaed for he exra risk incur by screening. This sudy also examines wheher he inclusion of deb raio screen makes any difference. Finally, Kouoms (1996) srongly suggess ha sudies invesigaing he informaion ransmission in he firs momen and second momen can be done based on reurns and volailiy, resecively. In addiion o examining he sock marke indices volailiy, his sudy analyzes wheher here is informaion ransmission from KLSI o DJIMI and FTSEGII and vice versa. The informaion ransmission from one marke o anoher has been widely reored. Bu maoriy of hese sudies are based on develoed markes only (Anoniou, Pesceo and Violaris, 003; Baur and Jung, 006; Caorale, Piis and Sagnolo, 006; Kouoms, 1996; and Kasibhala, Sewar, Sen and Malindreos, 006). Only few sudies examine he emerging markes (Daly, 003; Lamba and Ochere, 001; Shachmurove, 005; and Soydemir, 000). Our aer herefore examines four main issues: (1) wheher here is a difference in reurns among hese Islamic sock marke indices, () wheher here is a risk remium in each sock exchange, (3) wheher hese indices face he leverage effec risk; and lasly, (4) is here a volailiy sillover among hese hree Islamic sock marke indices. The emirical invesigaion is conduced by means of he GARCH model (GARCH-M) using daily daa covering he eriod from January 1999 unil Ocober 007. No only our resuls show no significan difference in heir reurns, risk remium is found o be absen in each Islamic sock index. While KLSE reors no leverage effec, DJIMI and FTSEGII indicae oherwise. Finally, based on EGARCH and TARCH models for KLSI here is sillover from DJIMI and FTSEGII oward KLSI bu no vice versa. The remainder of he aer is organized as follows. Secion 1 oulines he lieraure review while secion discusses he daa and mehodology emloyed. Secion 3 analyzes he resuls and finally, he las secion highlighs he maor conclusions and imlicaions of he sudy. 1. Lieraure review The invesigaion of volailiy is a rominen issue in financial ime series analysis. Many aers have been wrien using differen mehodology and variable o invesigae differen issues abou volailiy. This secion will review some of hese sudies. Yalama and Sevil (008) emloyed seven differen GARCH models o sudy he sock marke volailiy in 11 differen markes using daily daa from 1995 o 007. They found ha he bes model o exlain marke volailiy differ from one marke o he oher. Meanwhile, Yeh and Lee (000) invesigaed he resonse of invesors o unexeced reurns and he informaion ransmission in China, Hong Kong and Taiwan sock markes. Using GARCH model o analyze he asymmeric reacion of reurn volailiy o good and bad news, hey found ha he imac of bad news of volailiy is greaer han he imac of good news in Taiwan and Hong Kong bu no in China. Koulakiois, Paasyriooulos and Molyneux (006) invesigaed wheher he here is a relaion- 16
Invesmen Managemen and Financial Innovaions, Volume 8, Issue 3, 011 shi beween volailiy and sock reurns in 8 develoed markes. Using weekly daa and imlemening GARCH-M and EGARCH-M, hey found ha here is a relaionshi beween risk and reurns in he GARCH- M model for he UK. Liao and Qi (008) using daily daa comared he risk and reurn in NYSE comosie index and Shanghai sock index (SSI). They used ARCH, GARCH, TARCH, and EGARCH on boh markes and found ha he bes model ha fi SSI was EGARCH while TARCH was he bes fi for NYSE comosie index. In addiion, hey found ha here is leverage effec in NYSE comosie index bu no in SSI. Moreover, hey found ha SSI volailiy causes NYSE comosie index bu no vice versa. A recen sudy by Abdul Rahim e al. (009) uses develoing counries sock marke daa. They analyze he informaion ransmission in boh reurn and volailiy beween Jakara Islamic index (JII) and Kuala Lumur Syariah index. They reor ha here is informaion ransmission ha flows from KLSI o JII. However, he wo sock indices are no highly correlaed. The low correlaion could be because hese wo sock exchanges do no cross lis. Tesing for leverage effec in boh markes also roved insignifican. The unidirecionaliy in he ransmission migh be due o KLSI s higher marke caializaion given ha he number of shares included in KLSI is weny imes greaer han JII. Caorale e al. (006) examined he inerrelaionshis among he US, Euroean and Jaanese markes wih he Souh Eas Asian markes by using hree bivariae GARCH-BEKK models. Their findings show ha Souh Eas Asian volailiy deends osiively on shocks from Euroean markes and Jaanese markes. Rashid and Ahmad (008) evaluaed he erformance of linear and non-linear model of volailiy in Karachi Sock Exchange (KSE) using daily daa from 001 o 007. They found ha GARCH-M is beer han EGARCH in exlaining he volailiy in KSE. In addiion, hey found ha here is risk remium or relaionshi beween risk and reurns in GARCH-M model. Regarding leverage effec in EGRACH, i was found ha here is a leverage effec in KSE. Ozun (007) examined he effec of develoed sock markes on he reurns of emerging markes using daily daa from 00 o 006 and EGARCH model for volailiy. The emerging markes used are Brazil and Turkey and he develoed markes are Jaan, he UK, France, Germany and he US. I was found ha Brazil is affeced by he lagged reurns of all he markes exce he US while France, he US and Jaan, affeced Turkey reurn. In erm of leverage effec boh indices have leverage effec. Kovai (008) invesigaed he leverage effec as well as he risk remium in he Macedonian Sock Exchange using daily daa from 005 o 007. I was found ha risk remium effec, is saisically weakly significan in all models wih a negaive sign indicaing ha as reurns increase risk decreases. Similarly, in erms of leverage effec i was found ha leverage effec is weakly significan. Based on he above sudies, his aer uilizes he models from he GARCH family. GARCH-M EGARCH-M and TARCH-M are used o es he risk remium, he mean and volailiy sillover, and leverage effec in hese hree sock marke indices. The deailed exlanaion of he mehodology used is discussed in he nex secion.. Daa and mehodology Rosly (005) indicaed ha here are four main mehods of screening. The firs mehod is roducion aroach where he aciviies of he comany are he focus of he screening. The second mehod is he caial srucure aroach where he modes of finance of he comany will be under Shariah screening. The hird mehod is he income aroach where he income of he comany is scruinized. The las mehod is he asse aroach where comany s asses are o be screened. Mos of he Islamic indices do no follow a single mehod bu a mixure of almos all of hem. The difference is only in he exen of he focus. Some indices focus more on income and roducion bu migh be flexible in modes of finance. Ohers migh emhasis more on he roducion han on income. Unlike he revious sudies, his aer examines he reurns and volailiy of hree Islamic sock marke indices in hree differen counries, he US, he UK and Malaysia. While he DJMI and FTSE screened indices follow he same screening crieria, KLSI in Malaysia follows differen screening crieria. DJ and FTSE screened indices focus more on he income aroach han he aciviy aroach while KLSI end o give greaer weigh on he aciviies of he comany raher han heir incomes. Therefore, i is no surrising ha DJ Islamic marke index and FTSE Islamic Global index follow he same se of screening crieria 1. The firs crierion is ha he comany s rimary business mus be ermissible according o Islamic laws. Therefore, comanies ha engage in gambling, alcohol, armamens, obacco, ornograhy, or ork are excluded from he lis. Second crierion is ha he comany mus mee secific financial consrains ha include a deb raio of eual or less han 33%, accoun receivables euals or less han 45% for FTSEGII and 33% for DJIMI. Finally, he comany s ineres income mus be less han 5% for FTSEGII and 33% for DJIMI of is oal revenue. On he oher hand, he screening crieria for Malaysia s KLSI excludes comanies ha have non-ermissible 1 h://www.dindexes.com/mdsidx/downloads/rulebooks/imi_rulebook.df. 163
Invesmen Managemen and Financial Innovaions, Volume 8, Issue 3, 011 aciviies under Islamic laws such as gambling, gaming, alcohol, ineres, ec. For comanies wih aciviies comrising boh ermissible and non-ermissible elemens, he Syariah Advisory Council (SAC) considers wo addiional crieria. Firs, he ublic erceion or image of he comany mus be good. Second, he core aciviies of he comany are imoran and considered beneficial o Muslims and he counry, and he nonermissible elemen is very small and involves maers such as common ligh, cusom and he righs of he non-muslim communiy. To deermine he olerable level of mixed conribuions from ermissible and non-ermissible aciviies, he SAC has esablished several benchmarks based on reasoning from ualified Syariah scholars. If he conribuions from nonermissible aciviies exceed he benchmark, he comany is classified as non-syariah comlian 1. Time series daa usually exhibi hree main characerisics. Firs, hey exhibi volailiy clusering or volailiy ooling. In oher words, eriods of high volailiy is followed by eriods of high volailiy and he same alies for eriods of low volailiy. Second, heir disribuion is leokurosis, which mean ha he disribuion is fa-ailed. Third characerisic is he leverage effec. The leverage effec is he fac ha bad news affecs reurns more han good news. In oher words, changes in he rices end o be negaively correlaed wih changes in volailiy. Therefore modeling such series needs o be exended using oher models. The firs wo characerisics have been successfully modeled using ARCH (Auoregressive Condiional Heeroscedasiciy) by Engle (198) and GARCH (Generalized Auoregressive Condiional Heeroscedasiciy) develoed by Bollerslev (1986). The idea of ARCH and GARCH is o model he variance of he error erm from he mean euaion on he revious suared error erms. If he mean euaion is as follows. Y i 1X, (1) where Y is he deenden variable or reurns in his case, X is he indeenden variable and is he error erm and i and 1 are he coefficiens. The error erm ~ N 0, is assumed o have zero mean and a consan variance or homoscedasic. However, i is unlikely in he financial ime series ha he variance of he error erm be homoscedasic. Ignoring he fac ha he variance of he error erm is heeroskedasic will resul in eiher over/under 1 (1) The five-ercen benchmark is used o assess he level of mixed conribuions from he aciviies ha are clearly rohibied such as Riba, gambling, liuor and ork. () The 10-ercen benchmark is used o assess he level of mixed conribuions from he aciviies ha involve he elemen of common ligh which is a rohibied elemen affecing mos eole and difficul o avoid. (3) The 5-ercen benchmark is used o assess he level of mixed conribuions from he aciviies ha are generally ermissible according o Syariah and have an elemen of benefi o he ublic, bu here are oher elemens ha may affec he Syariah saus of hese aciviies. esimaion of he sandard error and herefore bias inferences. To overcome his roblem ARCH model is used. The arch model is as follows:, () i1 i 1 where is he condiional variance, 1 is he lagged erm of he suared error erm from he mean euaion, and and i are he coefficiens. This model indicaes ha he variance of he error erm is deenden on he lagged suared error erm. Such model is referred o as ARCH (), where indicaes he lag order of he suared error erm in he variance euaion. Alhough ARCH model is caable of eliminaing he heeroscedasiciy in he mean euaion, i sill has some drawbacks ha led o he develomen of GARCH model. GARCH model was develoed by Bollerslev (1986) who indicaed ha a GARCH model wih smaller number of erms can erform as well as or even beer han ARCH model wih many lags. The idea of he GARCH model is simly o include he lagged value of he variance in he variance euaion. The GARCH model is as follows:, (3) i 1 i 1 i1 The firs erm in he righ hand side is he ARCH erm exlained earlier, while he second erm is he lagged variance ha is GARCH. This model is referred o as GARCH (,) where () is he lagged ARCH erm and () is he GARCH lagged erm. The above model indicae ha is he long-erm average variance, i is he informaion abou he volailiy in he revious eriod, and he bea is he coefficien of he lagged condiional variance. Alhough GARCH model is beer han ARCH secificaion since i is more arsimonious and less likely o breach he non-negaive consrain i is sill does no accoun for he leverage effec in he aaren in financial ime series and does no allow for any direc feedback beween he condiional variance and he condiional mean. Anoher exension of GARCH by Engle, Lilien and Robins (1987) is GARCH-M where eiher he sandard deviaion or he variance is included in he mean euaion in order o es wheher here is a risk remium or a radeoff beween risk and reurns. This model is reresened as follows: Y 0 1X 1, (4) where Y is he deenden variable or reurns in his case, X is he indeenden variable, is he condiional variance or he risk remium, and is he error erm and 0, 1 and 1 are he coefficiens. The 164
GARCH-M model allows ime-varying volailiy o be relaed o execed reurns. An increase in risk, given by he condiional sandard deviaion leads o a rise in he mean reurn. The value of gives he increase in reurns needed o comensae for a give increase in risk. Therefore, i is a measure of risk aversion. One of he roblems in GARCH is ha i reas any shocks o he volailiy as symmerical. Tha is good news and bad news has he same effec. One of he mehods used o overcome hese issues in GARCH is Invesmen Managemen and Financial Innovaions, Volume 8, Issue 3, 011 asymmeric GARCH. However, i was argued by revious sudies such as Black (1976), Chrisie (198), Engle and Ng (1993) ha volailiy resonds asymmerically o news esecially bad news. Therefore, asymmeric GARCH is develoed o overcome his roblem. Two main models deal wih asymmeric informaion EGARCH (Exonenial GARCH) and TARCH (Threshold GARCH). Nelson (1991) develoed he following euaion o rea he asymmery in he volailiy: i i log i i log. (5) i1 i i1 i 1 The lef-hand side is he log of he condiional variance. This imlies ha he leverage effec is exonenial, raher han uadraic, and ha forecass of he condiional variance are guaraneed o be nonnegaive. The resence of leverage effecs can be esed by he hyohesis ha < 0. While TARCH model was inroduced by Zakoian (1994) and Glosen, Jagannahan and Runkle (1993). This model is designed o es wheher here is asymmeric imac of news and wheher here is a leverage effec. The secificaion of he TARCH model is as follows: i d 1, (6) 1 1 1 i 1 i 1 where d -1 = 1 if 1 < 0 and 0 oherwise. In his model, good news 1 1, <, and bad news is ( 1 < 0), have differen imac on he condiional variance whereby good news has he imac of, while bad news has he imac of, for he leverage effec if > 0 here is leverage effec on he oher hand if 0 hen he news imac is asymmeric. Therefore, bad news causes more volailiy in he marke hen good news. In his aer, he EGARCH and TARCH are used o es wheher here is any leverage effec in he hree screened marke. Tha is wih here is an asymmery in informaion. The daa used for his sudy will cover hree Islamic indices namely, DJIMI, FTSEGII, and KLSI. The eriod of he sudy sar from Aril 1999 o November 007 on daily basis. Reurns are calculaed r using he comounded reurn formula. The calculaion is done as follows: P i, R ln, (7) i Pi, 1 where R i is he reurn for index i a ime, P i, is he rice for index i a ime and P i, 1 is he rice of index i a ime 1. Therefore, four euaions will be esed here o answer his aer uesions. Firs euaion is he mean reurns euaion where each marke reurns will be regressed on is own lag and he oher wo marke reurns lags. Second euaion is a GARCH-M (1,1) o es wheher here is any rade off beween risk and reurns and he effec of he volailiy of each index on iself. The hird and forh euaions are wo differen mehods of es he leverage effec in each sock marke indices. The euaion is as follows: 0 irdjimi, 1 1 KLSI DJIMI,, (8) DJIMI, i 1, (9) DJIMI, i 1 i 1 KLSI DJIMI, 1 i 1 1 i i DJIMI, i i log 1 KLSI DJIMI, 1, log (10) i1 i i1 i 1 DJIMI, i 1 1d 1 1 KLSI DJIMI, 1, (11) i1 1 165
Invesmen Managemen and Financial Innovaions, Volume 8, Issue 3, 011 Euaion (8) is he reurn euaion where r is he daily reurn for DJIMI regressed on is lagged, is he variance of DJIMI index, which reresen he risk and reurn rade off, and is he error erm. Euaion (9) is he variance euaion where is he condiional variance, i is he lagged erm of he suared error erm from he mean euaion, is he lagged condiional variance, and,, and i, are he coefficiens as in euaion (3). Euaions (10) and (11) are EGARCH and TARCH models ha are used in his sudy. The same four euaions will be run for each marke. 3. Resuls and analysis Figure 1 shows he reurns of he hree indices. From he reurn grahs, i is clear ha he mean reurns are consan, however he variance change overime for hese indices. I is eviden ha volailiy ends o cluser, i.e., changes in volailiy wheher big or small ends o ersis. I is eviden ha DJIMI and FTSEGII moves ogeher almos during he whole eriod of he sudy which exlains he srong or almos erfec correlaion. I also shows ha here was a lo of volailiy beween 1999 and 003. On he oher hand, KLSI seems no relicae he movemen on hose wo indices however in erm of volailiy i has he same eriod of higher volailiy as hose wo indices. Fig. 1. Plo of closing rices and reurns for DJIMI, FTSEGII and KLSI 166
Invesmen Managemen and Financial Innovaions, Volume 8, Issue 3, 011 Figure los hisogram of reurns for each marke index agains he normal disribuion. I shows ha various reurns fall beyond four sandard devaions which is unlikly in normal disribuion. This kind of disribuion is called o have heavy ails. The disribuion of he reurns in hese markes show ha i is also leokuric or has highes eak. A uanile-uanile (Q-Q) lo on he oher hand is a ool o check wheher wo disribuions are he same, i.e, normal disribuion agains he series disibuion. If boh disribuions are similar, he lo is assumed o be linear. In his Figure, boh disribuins aear o be differen. The reurns deviae from he sringh line and his confirms he heavy ails and high eakedness characerisic of he reurns. Table 1 dislays he descriive roeries of he reurns of DJIMI, FTSEGII, and KLSI from Aril 1999 o Ocober 007. Toal observaions in his sudy are 8 observaions. The mean reurns of he hree indices are osiive. The KLSI has he highes reurn of 0.035 (1.8% annually) while DJIMI (5.8% annually) and FTSEGII (5.1% annually) have lower reurns a 0.016 and 0.0143, resecively. In erm of volailiy, KLSI has he lowes volailiy followed by FTSEGII and finally he Fig.. Normalized reurns disribuion and Q-Q lo highes volailiy is DJIMI. Alhough he financial heory indicaes ha higher volailiy mus be comensaed by higher reurns his is no he case in hese hree indices. KLSI has he highes reurns bu he lowes volailiy. DJIMI seems o earn lower reurn han KLSI. However, he former reors higher volailiy. The reurns of all he hree indices are negaively skewed and leokuric. This indicaes ha heir reurns are asymmeric. In addiion, he hree indices are no normally disribued based 167
Invesmen Managemen and Financial Innovaions, Volume 8, Issue 3, 011 on J-B es of normaliy. Meanwhile, he Lung-Box auocorrelaion es on reurns and reurns suared a 10 lags. I indicaes ha linear and non-linear deendencies exis in he firs and second momen. Linear deendency migh be exlained as marke inefficiency (Koumos, 1996; Koumos and Booh, 1995; and Kovai, 008). On he oher hand, nonlinear deendency migh indicae he resence of GARCH effec (Kovai, 008). Table 1. Descriive saisics of DJIMI, FTSEGII and KLSI reurns DJIMI FTSEGII KLSI Mean 0.016 0.014 0.035 Sd. dev. 0.968 0.918 0.913 Skewness -0.116-0.105-0.590 Kurosis 5.015 4.931 10.40 Jarue-Bera 38* 350* 515* LB (10) 63.97* 57.19* 89.97* LB (10) 746.51* 65.18* 301.18* Noe: * Significan a 1 %. Table shows he correlaion coefficien or he uncondiional correlaion beween he hree indices reurns. The correlaion beween DJIMI and FTSE- GII is he highes reaching almos one which indicae erfec correlaion. However, he correlaion beween KLSI and each index is abou 0.13 ha indicae very weak bu osiive and significan relaionshi. This low correlaion beween DJIMI and FTSEGII can be an indicaion ha hese indices movemens do no affec KLSI. This is migh be because DJIMI, FTSEGII are in develoed markes, while KLSI is in a develoing marke. Anoher reason could be ha DJIMI and FTSEGII migh have many firms ha are cross-lised in boh indices while KLSI does no have his characerisic. This low correlaion beween KLSI and boh DJIMI and FTSEGII can be useful in erm of diversificaion by invesors. Table. Simle correlaion coefficien for he reurns of DJIMI, FTSEGII and KLSI Variable FTSEGII KLSI DJIMI 0.983* 0.133* FTSEGII 1 0.19* KLSI 0.19 1 Noe: * Significan a 1 %. Table 3 dislays he resuls of he difference in mean reurns -es. The resul in all cases indicaes ha here is no difference in mean reurns among he hree indices. Table 3. T-es for difference in mean reurns Reurns T-es value DJIMI and FTSEGII -0.0517 DJIMI and KLSI 0.673 KLSI and FTSEGII 0.745 Table 4 reors he resuls of Augmened Dickey fuller (ADF) es. The urose of his es is o find ou wheher hese series are saionary by esing he null hyohesis ha he series have uni roo. From he resuls, i is clear ha all he sock markes reurns are saionary in he mean bu no in he variance. Table 4. ADF uni roo es KLSI DJIMI FTSEGII None -39.57* -40.55* -41.07* Trend & inerce -39.63* -40.57* -41.1* Inerce -39.61* -40.55* -41.07* Noe: * Significan a 1%. Table 5 reors he resuls of hree esimaions, GARCH-M, EGARCH-M, and TARCH-M as secified in euaions (9), (10) and (11). These hree esimaions models were done for KLSI, DJIMI, and FTSEGII. Since DJIMI and FTSEGII have almos a erfec correlaion beween hem, he esimaions below were done in wo markes relaionshi (i.e., KLSI wih DJIMI wihou FTSEGII and KLSI wih FTSEGII wihou DJIMI) raher han hree markes o avoid biasness in he resuls. In he reurns euaion of KLSI wih DJIMI and KLSI wih FTSEGII, i is eviden ha KLSI is affeced osiively by is own one-day lag, one-day lag of DJIMI and one-day lag of FTSEGII. This resul indicaes ha here is a sillover in reurns from DJIMI and FTSEGII on KLSI. In addiion, he coefficien of he risk reurns rade off () is no significan in any of he hree models. In he variance euaion, he coefficien 1 and 1 are osiive and significan in all he hree esimaions indicaing ha KLSI curren volailiy is affeced by is as volailiy. The coefficien 1, which is suosed o es he asymmery in he marke, is no significan in any of he models indicaing ha here is no leverage effec. Moreover, he coefficien measuring he sillover from DJIMI o KLSI and from FTSEGII o KLSI are significan in he GARCH-M model oining o he fac ha here is sillover from DJIMI and FTSEGII owards KLSI. In oher words, here is informaion ransmission from DJIMI and FTSEGII volailiies o KLSI volailiy. The half-life 1, which measure he eriod i akes a shock o decay ino he fuure, for GARCH- M effec is 17.9 days for KLSI wih DJIMI and 18.4 days for KLSI wih FSEGII, resecively. I is clear ha i akes longer for he shock in volailiy o disaear in he KLSI wih FTSEGII esimaion han in KLSI wih DJIMI esimaion. To deermine he bes model among he hree models he log likelihood crieria is used. From he able i is clear ha 1 Half-life = In(0.5)/In( 1+ 1 ). 168
GARCH-M model is he bes fi where log likelihood is he minimum. For all he models, an ARCH es was done o es for heeroscedasiciy in he Invesmen Managemen and Financial Innovaions, Volume 8, Issue 3, 011 hree models. The resuls of ARCH in lag 1 and 10 sugges ha here is no roblem of heeroscedasiciy. Table 5. Parameer esimaes of fiing GARCH (1,1), EGARCH and TARCH for KLSI from 1999-007 Coefficien GARCH-M EGARCH TARCH 0.044 0.046 0.035 0.030 0.033 0.035 C 0.008 0.006 0.008 0.01 0.007 0.006 FTSEGII (-1) 0.04* 0.197* 0.04* DJIMI (-1) 0.194* 0.19* 0.195* KLSI (-1) 0.158* 0.157* 0.161* 0.161* 0.160* 0.159* 0.010* 0.011** -0.164* -0.160* 0.011* 0.011* 1 (ARCH) 0.095* 0.095* 0.06* 0.00* 0.078* 0.079* 1 (GARCH) 0.894* 0.893* 0.977* 0.978* 0.89* 0.891* 1-0.07-0.0 0.034 0.03 DJIMI o KLSI ( ) -0.07** -0.057*** -0.04*** FTSEGII o KLSI ( ) -0.08** -0.069-0.06*** Log likelihood -467-465 -465-46 -465-463 ARCH (1) 0.67 0.198 0.700 0.580 0.115 0.07 ARCH (10) 11.60 11.75 16.35 16.49 1.96 1.4 Noe: *, ** and *** significan a 1%, 5%, and 10% resecively. KLSI is he deendan variable. Table 6 reors he resuls for he esimaion of DJIMI on KLSI. In he reurns euaion, he coefficien is no significan indicaing ha here is no risk remium in DJIMI. On he oher hand, i is clear ha DJIMI is affeced osiively by is own lag and negaively by KLSI lagged reurns in GARCH-M model only. In he variance euaion, he coefficiens for ARCH are significan in he firs wo esimaions while GARCH coefficien is significan in all he models esimaed. The coefficien 1 in EGARCH and TARCH models is negaive and osiive resecively, and significan imlying ha here is a leverage effec and asymmery of news. This means ha bad news has a greaer effec on volailiy han good news. The sillover effec coefficien from KLSI o DJIMI is no significan in all he models indicaing ha here is no ransmission of informaion from KLSI volailiy o DJIMI volailiy. The half-life in his case is 10.8 days for half of he shock o disaear ino he fuure. GARCH-M is he bes fi based on log likelihood crieria. ARCH diagnosic es for he heeroscedasiciy indicae ha in lag 1 and 10 here is no roblem of heeroscedasiciy. Table 6. Parameer esimaes of fiing GARCH (1,1), EGARCH and TARCH for DJIMI from 1999-007 DJIMI GARCH-M (1,1) EGARCH TARCH -0.006-0.004-0.014 C 0.048 0.03-0.03 KLSI (-1) -0.040*** -0.037-0.030 DJIMI(-1) 0.147* 0.147* 0.150* 0.005** -0.075* 0.006* 1 (ARCH) 0.053* 0.09* 0.00 1 (GARCH) 0.94* 0.991* 0.949* 1-0.054** 0.08* KLSI o DJIMI ( 1) 0.010 0.09 0.00 Log likelihood -789-77 -767 ARCH (1) 0.468 0.43 1.3 ARCH (10) 8.93 9.49 6.89 Noe: *, **, *** significan a 1%, 5%, and 10% resecively. DJIMI is he deendan variable Table 7 reors he resuls for he esimaion of FTSEGII on KLSI. In he reurn euaion for here is no risk remium in his marke. In addiion, FTSEGII curren reurn is affeced osiively and significanly by is own lagged reurns and negaively by one lag of KLSI in he firs model only. In he variance euaion, he coefficiens for ARCH are significan in he firs wo esimaions while GARCH coefficiens are significan in all he models esimaed. In addiion, he leverage effec coefficien in he EGARCH and TARCH models is significan. I is negaive in he EGARCH and osiive in he TARCH model. This indicaes ha here is a leverage effec and bad news has higher imac han good news on he index volailiy. The sillover effec from KLSI o FTSEGII is no significan in any of he models, which indicae ha here is no informaion ransformaion from KLSI volailiy owards FTSEGII volailiy. The half-life in his case is 11. days for half of he shock o disaear in he fuure. Based on he log likelihood crieria i is clear ha GARCH-M model is he bes model. ARCH diagnosic es for he heeroscedasiciy indicae ha in lag 1 and 10 here is no roblem of heeroscedasiciy. 169
Invesmen Managemen and Financial Innovaions, Volume 8, Issue 3, 011 Table 7. Parameer esimaes of fiing GARCH (1,1), EGARCH and TARCH for FTSEGII from 1999-007 FTSEGII GARCH-M (1,1) EGARCH TARCH -0.005-0.00-0.004 C 0.046 0.01 0.0 KLSI(-1) -0.039*** -0.035-0.07 FTSEGII(-1) 0.149* 0.150* 0.151* 0.005** -0.065* 0.005* 1 (ARCH) 0.054* 0.078* -0.006 1 (GARCH) 0.941* 0.989* 0.953* 1-0.065* 0.091* KLSI o FTSEGII ( 1) 0.008 0.07 0.016 Log likelihood -701-678 -674 ARCH (1) 1.6 1.6.14 ARCH (10) 11.9 14.43 10.58 Noe: *, ** and *** significan a 1%, 5%, and 10% resecively. FTSEGII is he deendan variable. To summarize, from he above models i is clear ha none of he markes has risk-reurns rade off. In oher words, here is no relaionshi beween he sock reurns of any of hese markes and heir volailiy. All he indices are affeced osiively by heir own lagged reurns. In addiion lagged reurns of DJIMI and FTSEGII are affecing KLSI reurns osiively indicaing informaion ransformaion from hese markes ino KLSI. On he oher hand, KLSI has a negaive one-lagged effec on boh DJIMI and FTSEGII in he GARCH-M model only. The variance euaions indicae ha he coefficien of 1 and 1 significan and osiive in mos of he cases indicaing ha as flucuaions has osiive influence on he fuure volailiy. In addiion, 1 is big and significan indicaing ha reurns has long-erm memory or he flucuaions are ersisen. Moreover, here is leverage effec in DJIMI and FTSEGII only bu no in KLSI. The leverage effec indicaes ha hese markes become volaile when here is a large decrease in he rices (i.e., bad news). When rices of a sock fall his causes deb o euiy raio o increase leading shareholder o erceive ha his sock is more risky. This is somehow erlexing. Boh DJIMI and FTSEGII have sric screening crieria regarding deb raio, which mus no exceed 33%, References while KLSI does no have any screen agains deb raio. In addiion, here is asymmeric effec of news in hese DJIMI and FTSEGII since 1 0. Therefore, bad news has sronger imac han good news in DJIMI and FTSEGII bu no KLSI. Lasly, in erms of sillover or informaion ransmission, i is clear ha here is eviden sillover from KLSI o boh DJIMI and FTSEGII bu no vice versa. This means ha here a ransmission of informaion from KLSI o DJIMI and FTSEGII markes. Therefore, volailiy in KLSI affecs DJIMI and FTSEGII bu no vice versa. Conclusion Our resuls sugges ha here is no significan difference in sock marke reurns beween he hree Islamic sock marke indices, KLSI, DJIMI, and FTSEGII. Therefore invesing in any of hem will yield he same reurns. In addiion, i was found ha here is no risk remium in any of he hree markes. Moreover, our resuls show ha here is leverage effec risk in he case of DJIMI and FTSEGII bu no KLSI. These wo Islamic sock marke indices seem o be affeced more by bad news han good news, which could be due o heir larger marke caializaion han KLSI. Moreover, DJIMI and FTSEGII are inernaional indices while KLSI is a local index. In addiion, here is asymmeric imac of news on volailiy, which means ha bad news has a greaer effec on volailiies han good news. Based on he half-life values he marke ha revers o mean faser is DJIMI followed by FTSEGII and lasly KLSI. I means ha KLSI ake longer ime o rever o i mean or for any shock in volailiy o decay. This could be because boh DJIMI and FTSEGII includes securiies from differen counries and have a larger number of socks hen KLSI which includes local socks and is smaller comared o DJIMI and FTSEGII. Lasly, here is informaion ransmission DJIMI and FTSE- GII from oward KLSI bu no vice versa. This migh be a resul of cross lising of some securiies in KLSI a DJIMI and FTSEGII bu no vice versa. 1. Abdul Rahim, F., Ahmad, N. & Ahmad, I. (009). Informaion ransmission beween Islamic sock indices in Souh Eas Asia, Inernaional Journal of Islamic and Middle Easern finance and Managemen, (1),. 7-19.. Anoniou, A., Pesceo, G. & Violaris, A. (003). Modeling inernaional rice relaionshis and inerdeendencies beween he sock index and sock index fuures of hree EU counries: a mulivariae analysis, Journal of Business Finance & Accouning, 30 (5/6),. 645-667. 3. Baur, D. & Jung, R.C. (006). Reurn and volailiy linkages beween he US and he German sock marke, Journal of Inernaional Money and Finance, 5 (4),. 598-613. 4. Black, F. (1976). Sudies in sock rice volailiy changes, Paer resened a he 1976 Meeing of he Business and Economic Saisics Secion. 5. Bollerslev, T. (1986). Generalized auoregressive condiional heeroskedasiciy, Journal of Economeric, 31 (4),. 307-37. 170
Invesmen Managemen and Financial Innovaions, Volume 8, Issue 3, 011 6. Booh, G., Marikainen, T. & Tse, Y. (1997). Price and volailiy sillover in Scandinavian sock markes, Journal of Banking and Finance, 1,. 811-83. 7. Brooks, C. (008). Inroducory Economerics for Finance (nd Ed.), New York: Cambridge Universiy Press. 8. Caorale, G.M., Piis N. & Sagnolo, N. (006). Volailiy ransmission and financial crises, Journal of Economics and Finance, 30 (3),. 376-390. 9. Cheung, Y. & Ng, L. (199). Sock rice dynamics and firm size: an emirical invesigaion, Journal of Finance, 47,. 1985-1997. 10. Chrisie, A. (198). The Sochasic Behavior of Common Sock Variances, Journal of Financial Economics, 10,. 407-43. 11. Daly, K.J. (003). Souheas Asian sock marke linkages evidence from re- and os Ocober 1997, Asean Econmics Bullein, 0 (1),. 73-85. 1. Engle, R. (198). A general aroach o Lagrange mulilier model diagnosics, Journal of Economerics, 0 (1),. 83-104. 13. Engle, R.F., Lilien, D.M. & Robins, R.P. (1987). Esimaing Time Varying Risk Premia in he Term Srucure: The ARCH-M Model, Economerica, 55 (),. 391-407. 14. Engle, R.F. & Ng, V.K. (1993). Measuring and Tesing he Imac of News on Volailiy, Journal of Finance, 48,. 1749-1778. 15. Glosen, L.R., R. Jagannahan & Runkle, D.E. (1993). On he Relaion beween he Execed Value and he Volailiy of he Nominal Excess Reurn on Socks, Journal of Finance, 48 (5),. 1779-1801. 16. Johnson, L. & Neave, E. (1996). Efficiency and effeciveness of Islamic financing: he cos of orhodoxy, Working Paer, No. 96-6, Queen s School of Business, Queen s Universiy, Canada. 17. Kasibhala, K.M., Sewar, D., Sen, S. & Malindreos, J. (006). Are daily sock rice indices in he maor Euroean euiy markes coinegraed? Tess and evidence, American Economis, 50 (1),. 47-57. 18. Koulakiois, A., Paasyriooulos, N. & Molyneux, P. (006). More evidence on he relaionshi beween sock rice reurns and voaliliy: a noe, Inernaional Research Journal of Finance and Economics, 1,. 1-8. 19. Koumos, G. & Booh, G. (1995). Asymmeric volailiy ransmission in Inernaional sock marke, Journal of Inernaional Money and Finance, 14,. 747-76. 0. Koumos, G. (1996). Modeling he dynamic inerdeendence of maor Euroean sock markes, Journal of Business Finance & Accouning, 3 (7),. 975-988. 1. Kovai, Z. (008). Forecasing volailiy: Evidence from he Macedonian sock exchange, Inernaional Research Journal of Finance and Economics, 18,. 18-1.. Lamba, A.S. & Ochere, I. (001). An analysis of he dynamic relaionshis beween he Souh African euiy marke and maor world euiy markes, Mulinaional Finance Journal, 5 (3),. 01-4. 3. Langbein, J.H. & Posner, R.A. (1980). Social invesing and he law of russ, Michigan Law Rewiev, 79 (1),. 7-11. 4. Liao, X. & Qi, G. (008). Analysis and Comarison of ARCH effecs for Shanghai comosie index and NYSE comosie index, Inernaional Journal of Business and Managemen, 3 (1),. 0-4. 5. Nelson, D.B. (1991). Condiional Heeroskedasiciy in Asse Reurns: A New Aroach, Economerica, 59 (),. 347-370. 6. Ozun, A. (007). Are he reacions of emerging euiy markes o he volailiy in advanced marke similar? Comaraive evidence from Brazil and Turkey, Inernaional Research Journal of Finance and Economics, 9,. 0-30. 7. Poon, S. & Taylor, S. (199). Sock reurns and volailiy: An emirical of UK sock marke, Journal of Banking and Finance, 16,. 37-59. 8. Rashid, A. & Ahmad, S. (008). Predicing sock reurns volailiy: an evaluaion of linear vs. nonlinear mehods, Inernaional Research Journal of Finance and Economics, 0,. 141-150. 9. Rosly, S.A. (005). Criical issues on Islamic banking and financial markes (1s Ed.). Kuala Lumur: Dinamas. 30. Rudd, A. (1981). Social resonsibiliy and orfolio erformance, California Managemen Review, 3 (4),. 55-61. 31. Soydemir, G. (000). Inernaional ransmission mechanism of sock marke movemens: evidence from emerging euiy markes, Journal of Forecasing, 19 (3),. 149-17. 3. Teer, J.A. (1991, May 13). The cos of social crieria, Pensions and Invesmen, 34. 33. Ulrich, D. & Marzban, S. (008). Review and analysis of curren Syariah-comlian euiy screening racices, Inernaional Journal of Islamic and Middle Easern Finance and Managemen, 1(4),. 85-303. 34. Shachmurove, Y. (005). Dynamic linkages among he emerging Middle Easern and he Unied Saes sock markes, Inernaional Journal of Business and Managemen, 10 (1),. 104-13. 35. Yalama, A. & Sevil, G. (008). Forecasing world sock markes volailiy, Inernaional Research Journal of Finance and Economics, 15,. 159-174. 36. Yeh, Y.-H. & Lee, T.-S. (000). The ineracion and volailiy asymmery of unexeced reurns in he greaer China sock markes, Global Finance Journal, 11,. 19-149. 37. Zakoian, J. (1994). Threshold Heeroskedasic models, Journal of Economic Dynamics and Conrol, 18,. 931-955. 171