Kharoum Universiy Journal of Managemen Sudies Vol.3, No.- Volailiy Spillover Across GCC Sock Markes Ibrahim A.Onour Absrac The sudy of volailiy ransmission across markes commonly ermed volailiy spillover provides useful insighs ino how informaion disseminaes across markes. Research resuls in his area have useful implicaions for issues such as inernaional or regional diversificaion and marke efficiency. In his paper, mulivariae GARCH model was employed o invesigae volailiy and informaion ransmission across he Gulf Cooperaion Council (GCC) markes. The model separaes direc volailiy ransmission from indirec ransmission, which is mainly due o cross-regional diversificaion and hedging sraegies underaken by porfolio managers. Findings of he sudy show ha effecs of indirec volailiy ransmission are more prominen han direc ransmission effecs across he GCC markes. Inroducion Afer he crash of Ocober 987, he issue of volailiy iner-dependence among capial markes gained momenum and became he subjec maer of much research in financial economic lieraure. King and Wadhwani (99) invesigaed a number of US markes afer he crash and showed ha markes overreac o he evens of oher markes, irrespecive of he economic value of informaion ransmied. Eun and Shim (989) idenified ha abou 6% of inernaional sock markes variabiliy may be explained by variabiliy in reurn in oher sock markes. Cheung and Ng (996) showed ha variabiliy of sock reurns of Asian- Pacific markes is closely associaed wih he variabiliy of sock reurns in major US sock markes. The primary objecive of his paper is o invesigae volailiy inerdependence among six of Gulf Cooperaion Council (GCC) sock markes, namely Abu-Dhabi, Bahrain, Dubai, Kuwai, Musca, and Saudi sock markes. Professor of Quaniaive Mehods, School of Managemen Sudies, Universiy of Kharoum- Sudan Qaar sock marke is no included in his research due o missing daa gap during he sample period under invesigaion. 88
These markes exhibi some common characerisics ha idenify hem as a unique group. GCC counries have close and common economic, insiuional, and culural ies. Consequenly, hese markes share a number of common feaures beside dual sock lisings among hem. In recen years, hese markes have adoped srucural reforms aimed a rading sysems sophisicaion and ransparency improvemen by adoping new regulaory framework, rading rules, reporing, surveillance, selemens and clearance sysems. All hese effors came in conjuncion wih he newly adoped agreemen requiring GCC member saes equal reamen of all GCC naionals in all invesmen aciviies, including sock ownership and esablishmen of new business and allowing free mobiliy of capial and labor of GCC naionals in member counries. The new agreemen also calls for harmonizaion of all invesmen-relaed laws and regulaions among GCC counries. This paper is moivaed by he growing lieraure on he condiional variance analysis. In he lieraure, differen mehods are adoped for measuring volailiy spillover. Some of he mehods include he cross-marke correlaion approach (Cheung and Ng, 996). Ohers adop GARCH modeling approach (Bollerslev, 99, and Hamao e al., 99). In his paper, he laer approach is followed. Marke Growh Policy makers in GCC counries have realized ha in order o have diversified economies and be less dependen on oil resources, resricions on equiy invesmens should be removed so ha foreign invesmens can be channeled owards developmen needs. Since efficien and well-regulaed capial markes are crucial for achieving such a goal, all GCC counries during he pas five years, embarked on new regulaory reforms aimed a deepening heir sock markes. In his conex, laws have been enaced o improve prudenial regulaions of commercial banks. Ani-money laundering policies have been adoped o safeguard agains unwaned inflow of money o he region. Resricions have also been eased for capial mobiliy beween GCC counries. Following hese policy reforms, here has been a subsanial surge in he liquidiy of GCC sock markes as indicaed by he significan rise in urn-over raios and he expanding marke capializaion during he pas hree years. Table indicaes he size of GCC sock markes gaining average annual growh of 39%, and urn-over raio increase of 55 % annually. Despie sluggish progress of privaizaion in GCC counries in general, he number of lised companies increased from 33 o 6 companies. 3 3 The apparen increase in he number of lised companies in mos GCC markes is mainly due o dual-lising of companies from oher GCC markes, and change from privae and family-owned, o public companies. 89
Marke Capializaion (million US$) 7 Table. Growh Indicaors Turn-over raio* (%) No. of Lised Companies 7 7 Bahrain 6765 76.9 4. 4 5 Kuwai 696 3536.. 9 96 Musca 3559 386.3.6 95 5 Saudi 76364 58984 8.8 3.4 76 AbuDhabi 64 8.4 39.4 6 64 Dubai 8456 3879.3 74.8 55 *Defined as he raio of raded shares o he oal ousanding shares. Source: Arab Capial Markes Saisics/Arab Moneary Fund. Daa Analysis Daa employed in his sudy are daily closing price indices for GCC sock markes, and Bren crude oil price as repored in he Wall Sree Journal and recorded as daily series in he daa base of he Cener for Energy Sudies of Louisiana Sae Universiy.The sample period covers daa from May 4 o Sep, 6, including 363 observaions. 4 Summary saisics for sock reurns are presened in Table. Table. Summary Saisics Bahrain Kuwai Musca Saudi Abu Dubai Dhabi Mean.8.3.5.37.4.65 S.deviaion: 3.5.8.3 5.5 7.8. Skewness: -.3.6.6. 7.7.3 Kurosis: 93 6.6 65.5 58.7 34 66.4 JB es 769 57 68 55 77 74 Q(5) Q (5).6 3.7 (.58) 47.7 87.5 44.3 79.9 45.6 8..6 (.8).3 (.98) 5.4 43 LM ARCH().6 (.84) 86.4 79. 8.. (.98) 8 4 Due o differences in he weekend holidays among GCC sock markes viz a viz he Wall Sree, harmonizaion of rading days has reduced he sample size o 363 observaions. 9
LM ARCH(5) 3.7 (.47) 4. 9. 4.3 (.98) 3 While he six markes exhibi posiive mean reurns, Table shows varying uncondiional volailiy. The skewness and high values of kurosis coefficiens indicae he disribuions of reurns characerized by posiive skewness and peakness relaive o a normal disribuion. 5 The posiive skewness resuls imply a higher probabiliy for sock prices increase. The Jarque-Bera (JB) es saisic provides evidence o rejec he null hypohesis of normaliy for he uncondiional disribuion of he daily price changes. The sample auocorrelaion saisic indicaed by he Ljung-Box, Q saisic, rejecs he null hypohesis of uncorrelaed price changes up o five lags for five markes in he group, bu only he Abu Dhabi marke shows evidence of uncorrelaed price changes. Invesigaion of ARCH (Auoregressive Condiional Heroskedasiciy) behavior of sock reurns, conduced by Q (5) and LM (Lagrange Muliplier) es saisics show evidence of sock reurns persisence (ARCH effec) for all markes, excep for he Abu Dhabi and Bahrain sock markes. Since he sandard LM and Ljung-Box saisics canno deec nonlinear dependence in ime series, he persisence in sock reurns of hese wo markes, could be more complex han can be capured by he LM and Q saisics. To circumven he low power of he LM es in deecing condiional heroskedasiciy of price reurns, he Kocenda and Briaka (5) es known as KK is employed o accoun for hidden nonlinear dependence in sock reurns by esing for sric whie noise process ha reflec sequence of independen and idenically disribued (iid) random variable. 6 Resuls in Table 3 confirm evidence of nonlinear dependence and rejec he null hypohesis of iid sock reurns for he six markes. Table 3. Nonlinear Dependence Tes (KK)* 5 The skewness (sk) and excess kurosis (k) saisics were calculaed using he formulas: m m sk =, and = 4 3 (m ( m ) 3 3 / ) k, where m sands for he jh momen around he j 4 mean. Under he null hypohesis of normaliy, he wo saisics are normally disribued wih sandard errors, σ sk = 6 N, and σ = k 4 N, where N is he sample size. 6 In fac, he KK es is a more general form of BDS es inroduced by Brock, Decher, and Scheinkman (987), which is used for esing he null hypohesis ha he daa are independenly and idenically disribued, agains unspecified alernaive. Kocenda and Briaka (5) developed a compuer program for calculaing KK saisics. KK compuer program is available a he websie: hp://home.cerge-ei.cz/kocenda/papers/kk. 9
Dimension Bahrain Kuwai Musca Saudi Abu Dhabi Dubai.98.57.587.48.38.45 3.46.744.84.64.55.59 4.86.98.7.78.69.77 5.7.6.9.9.8.845 6.68..48.3.93.95 7.3.33.64.5.4.4 8.35.45.8.6.6.3 9.39.55.96.34.5..44.64.9.43.34.8 *Values in enries are KK saisics. All values of KK rejec he null hypohesis of iid a he % significance level criical values included in Kocenda and Briaka (5). Volailiy Transmission To idenify condiional volailiy of sock reurns, common facors such as oil price changes ha simulaneously influence GCC markes, need o be conrolled. Thus, in he following sock reurns saed as a funcion of oil price change, E, and own lagged values beside condiional sandard deviaion of reurns, as a measure of risk facor 7. One imporan moive for aking he condiional sandard deviaion as explanaory variable is o ensure consisency propery of a quasi-maximum likelihood esimaor. 8 As a resul, condiional volailiy of sock reurns, R, is deermined as: R where, and, h = a = β k e / q a R i i i i k j= ~N(,h ) β e p δ j α h i E i j λ h e Equaion Equaion,, 7 Sock reurns in inernaional major markes also seem o be a relevan variable explaining changes in GCC sock markes, bu according o recen research findings, e.g Shawka and Choi (6); and Abraham and Fazal (6), GCC sock markes are no coinegraed wih US equiy markes. 8 Newey and Douglas (6) showed ha when esimaing he parameers in a ime-varying condiional variance using a QMLE if he densiy from which he likelihood is consruced is non- Gaussian (or asymmeric) for a QMLE o be consisen, a condiional sandard deviaion needs o be included as an addiional regressor. 9
The significance and sign of he coefficien λ, reflec risk aiude of invesors. A significan and negaive sign of he coefficien, λ, indicaes risk aversion behavior, and insignificance implies risk neural behavior. Condiional volailiy of sock reurns is depiced in Equaion as GARCH (Generalized Auoregressive Condiional Heroskedasiciy) process, where h sand for condiional variance; p, and q are lag parameers for AR(p) and MA(q) componens. Mulivariae GARCH model ha accommodaes volailiy spillover among sock markes, as well as volailiy persisence wihin each marke, is he VECH model which was inroduced by Bollerslev, Engle, and Wooldridge (988), saed as: Vech(H p q ) = B Bivech(H i ) Ai vech(e ie i ) Equaion3, The noaion vech(.) is he vecor half operaor which ransforms asymmeric (dxd) marix ino a vecor of lengh d=(d)d/ by sacking he elemens of he upper riangular half of he marix. H denoes he condiional variance marix. One major problem relaed o vech specificaion of mulivariae GARCH models is he large number of parameers included in he esimaion process. An alernaive approach developed by Engle and Kroner (995) and hey ermed he Baba, Engle, Kraf and Kroner or BEKK represenaion specifies he condiional variance in GARCH (,) as: H = α α β H β A e e A Equaion 4, where α in his case is a (6x6) lower riangular marix and β is a (6x6) square marix of parameers. The marix β reflecs he exen o which curren levels of condiional variances are relaed o pas condiional variances. Parameers in marix A esimaes he exen o which condiional variances are linked wih pas squared errors. The elemens in A capures he impac of news on condiional volailiy. Despie he BEKK model including smaller number of parameers (N(5N)/ ) compared o he number of parameers in vech model ((N(N)(N(N))/ ), he number of parameers in he case of six markes sill seems large (93 parameers). Anoher problem relaed wih he general specificaion of BEKK model in Equaion 4, as noed by Bauwens (5), is ha inerpreaion of he basic parameers is no obvious since Equaion 4 has nonlinear parameers. 93
To resolve he over-parameerizaion problem, Bollerslev (99) proposed consan condiional correlaions among he elemens of covariances in Equaion 4 hij ( ) = ρij hii ( h jj( so ha and resric he elemens of marices A and B o only diagonal erms. However, since he off-diagonal erms of marix B represen indirec volailiy ransmission across markes in his paper, all elemens of marix B are mainained and only he diagonal erms of marix A are reserved. The cross produc erms of residuals (he off-diagonal erms of marix A) does no have meaningful inerpreaion of volailiy ransmission effecs. When including hese changes, he se of Equaion 4 may be saed as: h jj, for = c i 6 6 a ii β h e i ii( j =,... 6 i( 6 ε β h j i i( 6 3 β h 3i 3 i( 6 4 β h 4i 4i( 6 5 β h 5i 5i( Equaion 5, Where c i are consans, β ki ( i =,...6) are condiional variance-covariance parameers, and ε are residuals erms. The coefficiens in he variance erms in Equaion 5 reflec direc volailiy ransmission, and he coefficiens of covariance erms represen indirec volailiy ransmission, whereas coefficiens of squared residuals reflec ransmission of news across sock markes. Esimaion of parameers in Equaion 5 is performed maximizing he loglikelihood funcion: L( Ω) = N ln(π ) (/ ) N = (ln H e H e ) Equaion 6, where N is he number of observaions and Ω, represens he parameer vecor o be esimaed. 9 Esimaion Esimaion resuls of Equaion, reveal significan shor-erm effec of oil price change on sock reurns of he Musca and Bahrain markes, albei hey are smaller in erms of marke capializaion (Table A, Appendix), and relaively less oil-dependen economies among GCC counries. Significan and negaive coefficien values of (λ), show risk aversion aiude characerizing he Saudi, Kuwai, and Bahrain sock markes. However, insignifican (λ) coefficiens 9 Maximizaion of he log likelihood in Equaion 6, has QMLE feaures. Using ADF and PP uni roo ess, i has been verified ha sock reurns for he six GCC markes are I(). This could be due o speculaive facors ha characerize sock price changes in hose markes. 94
for he Abu-Dhabi and Dubai markes, reflec risk neural behavior of invesors. Table A, Appendix, signify saionariy condiions sipulaed by GARCH-ype volailiy of Equaion. Esimaion of Equaion 5 presened in Table 4 indicaes here is direc volailiy ransmission from he Saudi and Dubai markes o he Kuwai sock marke, and from he Musca o he Abu Dhabi marke. Esimaion resuls show ha he effecs of indirec volailiy ransmission are more prominen han he direc volailiy shocks. This is revealed by significan indirec volailiy ransmission across all GCC markes, which is indicaed by significan covariance ( β i j) coefficiens ij for he six GCC markes. Significan indirec volailiy ransmission across GCC markes is probably due o cross-regional porfolio managemen and hedging sraegies underaken primarily by invesmen funds managers. News ransmission effec indicaes ha he Kuwai and Bahrain markes are he only GCC markes ha respond significanly o ouside news. Volailiy in he Kuwai sock marke reacs o is inernal news, and o news originaing from he Saudi and Dubai sock markes. Table 4. Esimaes of Volailiy Transmission Parameers* Bahrain () Kuwai () Musca (3) Saudi (4) Abu Dhabi (5) Dubai (6) β - -.6(.55).(.74).7(.85).3(.36) -.78(.4) ( β 5.(.) -.65(.4).(.46).39(.86) 4.5(.5) ( β 7.(.48).4(.45) - -.9(.36) 5.4(.3) -.3(.87) 33( β -.(.7) 6.9(.5) 5.6(.) - -.89(.85) 8.5(.53) 44( β.99(.6) -.3(.63).3(.5) -.(.66) -.69(.5) 55( β -.4(.34).5(.5).(.94).4(.77) -.5(.6) - 66( β 3.9(.) -.(.85).4(.5).39(.39).(.98).(.4) ( β.6(.96) -.(.).8(.6).8(.64).4(.7).(.) 3( β -.9(.).4(.7).(.58) -.6(.49) -.49 -.9(.84) 4( β -.36(.5) -.5(.) -.(.55).(.49).3(.).(.75) 5( β.4(.59).(.54) -.(.5).7(.5) -.9(.4) -.5(.) 6( β.4(.79).(.) -.6(.5).5(.76) -.4(.53).8(.9) 3( β.68(.37).3(.44).(.66) -.53(.).5.9(.85) 4( β -.3(.96) -.7(.4) -.3(.3) -.3(.3).6(.) -.75(.7) 5( β.83(.).(.93).(.47) -.36(.3).48 -.44 6( β.98(.) -.8(.5).(.7).3(.3) -.4(.83) -.56(.9) 34( 95
β -.98(.).(.76).6(.4).(.9).(.9).6(.6) 35( β.(.65) -.(.45) -.(.3) -.3(.85).8(.49) -.68(.3) 36( β -.8(.4).3(.4).3(.).(.4) -.(.) -.7(.53) 45( β -.7(.) -.(.) -.4(.78) -.3(.64).(.68) -.5(.7) 46( β.34(.9).(.6).(.).7(.6) -.(.59).45 56( a.(.63).6(.38) -.(.77) -.(.95) -.7(.39).74(.4) ( a -6.5(.7).(.5) -.65(.4) -.8(.53) -.7(.75) -4.(.5) ( a -7.(.48) -.4(.45).(.).95(.36) -5.4(.3).3(.87) 33( a.3(.73) -6.9(.5) -5.6(.).5(.34).9(.85) -8.6(.5) 44( a -.83(.34) -.(.73) -.(.59).(.39) -.(.7) -.8(.63) 55( a.44(.3) -.4(.4) -.(.79) -.8(.5).8(.43).7(.4) 66( *Lagged coefficien subscrips refer o he sock markes: = Bahrain; = Kuwai; 3 = Musca; 4 = Saudi; 5 = Abu Dhabi; 6 = Dubai. Values of consans are no repored in he able. Values in parenhesis are p-values. Bold numbers are significan up o 5% significan levels. All values are up o wo decimal numbers. Conclusion This paper invesigaes volailiy and informaion ransmission across GCC sock markes, using mulivariae GARCH specificaion of condiional volailiy. The GARCH model employed in he paper separaes he effec of direc volailiy ransmission from he indirec ransmission effec. This laer ype of volailiy ransmission is aribued o cross-regional porfolio diversificaion and hedging sraegies underaken mainly by managers of invesmen funds. The mulivariae GARCH approach employed in his sudy also capures he effec of news and informaion ransmission on volailiy of sock markes. Resuls of he paper reveal ha he Kuwai sock marke is he mos vulnerable o direc volailiy shocks in GCC markes, as volailiy shocks a Saudi and Dubai markes ransmi o Kuwai sock marke. The findings of he paper also reveal evidence of significan indirec volailiy ransmission across all GCC markes. Wih regard o sock markes reacion o news and informaion spillover, he Kuwai and Bahrain markes are he only GCC markes responding significanly o ouside news and informaion. Volailiy in he Kuwai sock marke reacs o is own inernal news, and o news originaing from he Saudi and Dubai sock markes. I should be noed ha invesmen funds are he only equiy invesmens accessible o foreigners in GCC counries over he sample period of his research. 96
Evidences of indirec volailiy ransmission across all GCC markes enhance he currency unificaion policy planned for he year. This is because as correlaion of shocks becomes sronger among GCC capial markes, adjusmen o such shocks becomes faser. This, in urn, reduces he cos of adjusmen using moneary insrumens. 3 More specifically, when he effec of an adverse emporary shock on a cerain GCC marke is ransmied o oher GCC markes, is impac will be realized on varying degrees by oher GCC markes. As a resul, he adverse effec of markes downurn, such as capial ransfer from one GCC marke o anoher marke in he region, becomes relaively smaller since he impac is no longer specific o a cerain marke in he region. On he oher hand, when shocks are uncorrelaed, he impac of any shock o any specific marke will be limied o ha marke. Consequenly, his may induce capial ouflow from he affeced marke. This may require he use of moneary insrumens o miigae he impac of capial ransfer in he affeced counry. While volailiy ransmission provides some advanage in erms of gains in marke efficiency, i also offers poenial pifalls. Greaer spillover effecs among GCC markes imply sronger co-movemens beween markes, herefore reducing he opporuniies for regional diversificaion. Furhermore, marke co-movemens may also lead o marke conagion as invesors incorporae ino heir rading decisions informaion abou price changes in oher markes. References Abraham A. and S. Fazal. 6. Informaion ransmission beween he Gulf equiy markes of Saudi Arabia and Bahrain. Resrearch in Inernaional Business and Finance Vol., Issue 3: 76-85. Arab Moneary Fund. Various saisical repors. Emiraes. Abu Dhabi, Unied Arab Bauwens, L., S. Lauren and J. Rombous. 3. Mulivariae GARCH models: A Survey. CORE discussion paper (3/3), Deparmen of Economics, Universie Caholique de Louvian, Belgium. Bayoumi, T. and B. Eichengreen. 993. Shocking aspecs of European moneary unificaion. In The Transiion o Economic and Moneary Union in Europe. Edied by F. Giavazzi and F. Torres. Cambridge, UK: Cambridge Universiy Press. 3 Bayoumi and Eichengreenn (993) showed ha while demand and supply shocks across US regions are higher han across European Union counries, he adjusmen o shocks is faser in he US han in Europe. 97
Bollerslev, T. 99. Modelling he cohernce in shorrun nominal exchange raes: A mulivariae generalized ARCH approach. Review of Economics and Saisics 7:498-55., R. Engle. and J. Wooldridge. 988. A capial asse pricing model wih ime-varying covariances. Journal of Poliical Economy 96:6-3. Brock W., W. Decher and J. Scheinkman. 996. A es for independence based on correlaion dimension. Economeric Reviews Vol. 5, No.3: 97-35. Cener for Energy Sudies, Louisiana Sae Universiy, Inerne websie: hp://www.enrg.lsu.edu/les/spo. Cheung, Y. and L. Ng. 996. A causaliy-in-variance es and is applicaion o financial marke prices. Journal of Economerics 7:33-48. Engle, R. and K. Kroner. 995. Mulivariae simulaneous generalized ARCH. Economeric Review :-5. Eun C. and S. Shim. 989. Inernaional ransmission of sock marke movemens. Journal of Financial and Quaniaive Analysis 4: 4-56. Hamao, Y., R. Masulis, and V. Ng. 99. Correlaion in price changes and volailiy: A cross inernaional sock markes. Review of Financial Sudies Vol. 3: 8-37. King M. and S. Wadhwani. 99. Transmission of volailiy beween sock markes. Review of Financial Sudies 3:5-33. Kocenda, E.and L. Briaka. 5. Advancing he iid es based on inegraion acroos he correlaion inegral, ranges, and power. Economeric Reviews Vol.4, No. 3:.65-95. Newey W. and D. Seigerwald. 8. Consisency of quasi-maximum likelihood esimaors for models wih condiional heeroskedasiciy. Social Science Research Nework Elecronic Library. Working Paper Series. Inerne websie: hp://ssrn.com/absrac=659. Shawka H. and K. Choi. 6. Behavior of GCC sock markes and impacs of US oil and financial markes. Research in Inernaional Business and Finance. Vol., No.: -44. Appendix Table A. Sock Reurns' Volailiy 98
Parameers* Bahrain Kuwai Musca Saudi Abu Dhabi a a a a 3 δ δ δ δ 3 δ 4 λ (pvalue) -8.4 -.39 -. -.4 (.8) -7.3.6 (.3) -.5 -.83 (.35) 8.9 -.78 -.5 (.95).99 83.8 (.).5 - -.59-5. (.83).99 35.4 (.88) -. - - -.6.6 (.3) - -6. -.4 (.6) -.4 (.5) (.7).4 (.) - 5.8 - -8.6 - - - - - - - - - - -.77 (.).75-5.9. (.59) Dubai 8..59. (.7).7 (.6).5 (.75).65 (.7) -.8 (.7) -.9 (.33) *Lags in Equaion have been deermined by AIC crieria. Saionariy condiions of parameers of equaion, impose he resricion ha lagged variables corresponding o dashed "-" cells in he able be excluded. Table A. Parameer Esimaes* Parameers** α α φ φ Bahrain GARCH(,) 73.5(.9).(.7) - Kuwai 9.6(.4).(.35) - 99
GARCH(,) Musca 336.3.(.87).58 GARCH(,) Saudi 449.5.66 - GARCH(,) Abu Dhabi 375.89(.).(.5) - GARCH(,) Dubai.79(.34).6.7 - GARCH(,) ** Values in paranhesis are p-values. *Saionariy condiions of he equaion h = α q p α ie i φi sipulae ha: α >, α i, φi, for all i and ( α i φi ) < be saisfied for all markes. q p h i