INFORMATION TRANSMISSION BETWEEN CRUDE OIL MARKETS *

Transcription

1 INFORMATION TRANSMISSION BETWEEN CRUDE OIL MARKETS * Sang Hoon Kang, Seong-Min Yoon Absrac Transmission mechanisms of volailiy beween crude oil markes have drawn he aenion of numerous academics and praciioners because hey boh play crucial roles in porfolio and risk managemen in crude oil markes. However, here has been no consensus on he evidence of volailiy spillover beween WTI and Bren oil prices in he lieraure. In his conex, we reexamined he volailiy linkages beween wo represenaive crude oil markes using a VECM and an asymmeric bivariae GARCH model. Firs, looking a he reurn ransmission hrough he VECM es, we found a long-run equilibrium and bidirecional relaionship beween wo crude oil markes. However, he esimaion resuls of he GARCH-BEKK model sugges ha here is unidirecional volailiy spillover from he WTI marke o he Bren marke, implying ha he WTI marke ends o exer influence over he Bren marke and no vice versa. Regarding asymmeric volailiy ransmission, we also found ha bad news volailiy in he WTI marke increases he volailiy of he Bren marke. Thus, WTI informaion is ransmied ino he Bren marke, indicaing ha he prices of he WTI marke seem o lead he prices of he Bren marke. JEL Classificaion: C32; C58; G14; G17 Keywords: Asymmeric volailiy ransmission; Causaliy; Coinegraion; GARCH-BEKK model; Volailiy spillover effec * The auhors would like o hank seminar paricipans a he Pusan Naional Universiy for helpful suggesions. This work was suppored by he Naional Research Foundaion of Korea Gran funded by he Korean Governmen (NRF B00008). Deparmen of Business Adminisraion, Pusan Naional Universiy, Jangjeon2-Dong, Geumjeong-Gu, Busan, , Korea. Tel.: ; fax: address: sanghoonkang@pusan.ac.kr (S. H. Kang). Corresponding auhor. Deparmen of Economics, Pusan Naional Universiy, Jangjeon2-Dong, Geumjeong-Gu, Busan, , Korea. Tel.: ; fax: address: smyoon@pusan.ac.kr (S.-M. Yoon). 147

2 1. Inroducion In recen years, crude oil prices have reached record highs, and rising oil prices have posed a new hrea o he global economy. Boh academicians and energy marke paricipans have focused on forecasing and modeling oil prices by quanifying and managing he risks inheren in heir frequen volailiies. In paricular, he informaion ransmission of crude oil prices has drawn he aenion of numerous academics and praciioners as crude oil prices play a prominen role in naional economies. Rising crude oil flucuaions affec he world economy in many differen and significan ways. For example, rising crude oil prices increase he producion coss of goods and services and he coss of ransporaion and heaing. Thus, consumers, governmens and praciioners are grealy concerned abou he volailiy of crude oil prices and is possible negaive economic effecs, such as hose on business cycles (Mork, 1994), macroeconomies (Lee, Lee and Rai, 2001; Rafiq, Salim and Bloch, 2009), and inflaion (Hooker, 2002; Hamilon and Herrera, 2004). In addiion, he informaion ransmission beween crude oil markes and oher financial markes is becoming of greaer ineres o porfolio managers and policy makers because of he increasing rend of globalizaion. Many empirical sudies have examined he ineracion beween he crude oil markes and oher financial markes such as energy markes (Ewing, Malik and Ozfidan, 2002), sock markes (Sadorsky, 1999, 2001, 2012; Malik and Hammoudeh, 2007; Aloui and Jammazi, 2009; Chiou and Lee, 2009), fuures markes (Bekiros and Diks, 2008; Kaufmann and Ullman, 2009), and exchange rae markes (Sadorsky, 2000; Zhang e al., 2008). These sudies have found evidence of significan reurn and volailiy spillovers beween crude oil markes and oher financial markes. In conras, his sudy invesigaed he informaion ransmission mechanism beween wo crude oil prices: WTI (Wes Texas Inermediae) and Bren (Bren Blend). 1 We began he research wih a quesion: Does one marke lead he oher in erms of informaion ransmission? The exisence of he spillover effec implies ha one large shock increases prices no only in is own asse or marke bu also in oher asses or markes. In paricular, price volailiy is ofen relaed o he rae of informaion flow (Ross, 1989). If informaion comes in clusers, prices may exhibi volailiy even if he marke perfecly and insananeously adjuss o he news. Thus, sudies on volailiy spillover can help us undersand how informaion is ransmied across crude oil markes. The main conribuions of his paper are wofold. Firs, his sudy focuses on he reurn causaliy relaionship beween WTI and Bren oil prices by employing he vecor error correcion model (VECM). This allows us o analyze boh he long- and shor-run lead-lag relaionship beween crude oil prices. Second, his sudy examines he casual direcion of volailiy spillover using boh symmeric and asymmeric bivariae GARCH models. In paricular, we explore wheher bad news in one marke leads o a larger volailiy in he oher marke han does good news. A good undersanding of he asymmeric volailiy response o news is an imporan ingredien for designing hedging sraegies and opimizing porfolios. The res of his paper is organized as follows. Secion 2 presens he economeric mehodology. Secion 3 provides descripive saisics of he sample daa. Secion 4 discusses he empirical resuls. Secion 5 presens our conclusions. 2. Mehodology 2.1. Coinegraion es Coinegraion is an economeric propery of ime series variables. If wo or more series are hemselves nonsaionary bu a linear combinaion of hem is saionary, hen he series are said o be coinegraed. In pracice, coinegraion is a means of correcly esing hose hypoheses concerning he relaionship beween wo variables having uni roos. In he lieraure, he Johansen (1991) coinegraion es is he mos popular approach for esing coinegraion. This coinegraion es is based on maximum likelihood esimaors of a vecor auo regressive (VAR) process, and he likelihood raio es saisic for he hypohesis of he a mos r coinegraed relaionship and a leas m n r common rend is given by 1 WTI is he base grade raded, as ligh swee crude, on he New York Mercanile Exchange (NYEMX), while Bren is raded on London s Inernaional Peroleum Exchange (IPE). 148

3 where race r n T ln 1 i, (1) ir1 max r, r 1 T ln 1 ri, (2) r is he race saisic, rr, 1 race max is he eigen-max saisics, i r denoes he esimaed eigenvalue, and T is he sample size. The null hypohesis esed in race is no coinegraion. In fac, for bivariae coinegraion ess, up o wo null hypoheses can be esed. If he null hypohesis ha r 0 is rejeced, a leas one coinegraing vecor may exis and he second hypohesis ha r 1 is subsequenly esed VECM The erm coinegraion implies ha causaliy exiss beween he wo series; however, i does no specify he direcion of he causal relaionship. If coinegraion exiss beween he variables, hen we can rule ou he possibiliy of a spurious correlaion. Thus, in his conex, we have employed he VECM o deec he direcion of he causaliy. The VECM disinguishes beween long- and shor-erm relaionships beween he variables and can idenify causaion sources ha canno be deeced by he usual Granger causaliy es (Belloumi, 2009). In his sudy, he VECM for he WTI ( x ) and Bren ( y ) series can be wrien in he following manner: where 1, (3) 2 2 x,0 x 1 j1 xx, j j j1 xy, j j x, x ec x y, (4) 2 2 y,0 y 1 j1 yx, j j j1 yy, j j y, y ec x y x and y represen he firs differences beween hese variables and capure heir shor-run disurbances; ec is he error correcion erm derived from he long-run coinegraion relaionship and measures he magniude of he pas disequilibrium (i.e., residuals). The error correcion coefficiens, he dependen variables from he long-run equilibrium, and x and/or x and y, represen he deviaion of y should be significanly differen from zero if he wo variables are coinegraed. By using he coefficiens of each explanaory variable, we can es heir shor-run causal relaionships. For example, if he coefficiens xy, j ( yx, j ) are significan, hen we can explain x ( y ) using y j ( x j) ; in oher words, we can explain he causaliy relaionship beween he wo crude oil markes. Finally, we can use he coefficiens of xx, j and yy, j o measure how he reurns of he wo markes reac o heir own lagged values, which implies he degree of mean-revering behavior from boh ime series Bivariae GARCH model Much aenion has focused on how news from one marke affecs he volailiy process of he oher. In his sudy, we analyze he volailiy spillovers effec beween he wo crude oil markes by using a bivariae framework of he BEKK parameerizaion (Engle and Kroner, 1995). In his model, he variance-covariance marix of equaions depends on he squares and cross producs of innovaion, which is derived from he following mean equaion: R, 1 ~ N(0, H ), (5) where R is he 2 1 vecor of reurns a ime for each marke. The 2 1 vecor of random errors,, represens he innovaion for each marke a ime wih is corresponding 2 2 condiional variance-covariance marix H. The marke informaion available a ime 1 is represened by 1. This bivariae srucure hus faciliaes he measuremen of he effecs of innovaions in he mean reurns of one marke on is own lagged reurns and hose of he lagged reurns of he oher marke. The sandard BEKK parameerizaion for he bivariae GARCH model is wrien as: 149

4 H CC A A BH B, (6) where H is a 2 2 marix of condiional variance-covariance a ime, and C is a 2 2 lower riangular marix wih hree parameers. A is a 2 2 square marix of coefficiens and measures he exen o which condiional variances are correlaed pas squared errors. B is a 2 2 squared marix of coefficiens and shows he exen o which curren levels of condiional variances are relaed o pas condiional variances. where h11, h12, c11 c11 h21, h, c21 c c21 c 2 a11 a12 1, 1 1, 1 2, 1 a11 a12 2 a21 a 2, 1 1, 1 2, 1 a21 a b11 b12 h11, 1 h12, 1 b11 b12, (7) b21 b h21, 1 h, 1 b21 b h 11, denoes he variance of he marke reurns, h 12, denoes he covariance of he WTI reurns and he Bren reurns, and h, denoes he variance of he Bren reurns. The significance of diagonal coefficiens a11 ( a ) suggess ha he curren condiional variance of 11, h, ( h ) is correlaed wih is own pas squared errors, while he significance of lagged variance b11 ( b ) indicaes ha he curren condiional variance of h11, ( h, ) is affeced by is own pas condiional variance. In addiion, he significance of he off-diagonal coefficiens a 12 and b 12 indicaes a volailiy spillover effec from he WTI marke o he Bren marke, whereas he significance of off-diagonal coefficiens a 21 and b 21 suggess a volailiy spillover effec from he Bren marke o he WTI marke. The sandard BEKK model implies ha only he magniude of pas reurn innovaions is imporan in deermining curren condiional variances and covariances. However, i has been well observed ha volailiy responds asymmerically o posiive and negaive innovaions of equal magniude: volailiy ends o rise higher in response o negaive shocks, such as bad news, han o posiive shocks, such as good news (Engle and Ng, 1993; Glosen, Jagannahan and Runkle, 1993; Kroner and Ng, 1998). To circumven his problem, Kroner and Ng (1998) exended he GJR-GARCH approach o a mulivariae seing ha can capure he asymmeric response o news on volailiy. The asymmeric BEKK model is wrien as: H CC A A BH B D D, (8) h11, h12, c11 c11 a11 a12 1, 1 1, 1 2, 1 a11 a12 h 2 21, h, c21 c c21 c a 21 a a 2, 1 1, 1 2, 1 21 a b11 b12 h11, 1 h12, 1 b11 b12 b21 b h21, 1 h, 1 b21 b 2 d11 d12 1, 1 1, 1 2, 1 d11 d12, (9) d 2 21 d d 2, 11, 1 2, 1 21 d max 0, 1, 1 where 1 max 0, 2, 1 variances and covariance hrough he definiion of 1. If he off-diagonal coefficien 12, D is a 2 2 squared marix of parameers and capures any asymmery in d 21 ( d ) is posiive and 150

5 significan, he bad news volailiy of he WTI marke (or he Bren marke) causes a higher volailiy of he Bren (or he WTI marke) han he good news volailiy of he WTI marke (or he Bren marke). The parameers of he bivariae GARCH model can be esimaed by he maximum likelihood esimaion mehod opimized wih he Bernd, Hall, Hall, and Hausman (BHHH) algorihm. The condiional log likelihood funcion L( ) is expressed as: T T 1 1 1, (10) L( ) T log log H ( ) 0.5 ( ) H ( ) where T is he number of observaions and denoes he vecor of all he unknown parameers. 3. Daa and descripive saisics This sudy analyzes he informaion ransmission beween he crude markes. To do his, we consider wo represenaive crude oil prices, he WTI and Bren prices. The daa ses consis of he weekly Friday closing prices spanning January 5, 1990 o Augus 26, 2011 (a oal of 1,130 observaions) provided by he US Energy Informaion Adminisraion (EIA). Figure 1 shows he dynamics of he wo ses of oil price daa. The spo prices of crude oil have been influenced by evens wih economic and geo-poliical effecs. For example, (1) he Firs Gulf war in collapsed oil producion, resuling in he doubling of he price of crude oil; (2) oil prices wen down in owing o he Asian currency crisis; (3) Organizaion of Peroleum Expor Counries (OPEC) curailed he producion of crude oil by 4.2 million barrels per day beween 2000 and 2001, resuling in increased crude oil prices; (4) he uncerainies associaed wih he 9/11 error aack in 2001 and he subsequen US miliary acion in Iraq beginning in March 2003 reversed he rend in oil prices; (5) crude oil prices rose owing o global economic growh, growing demand, and sagnan supply in ; (6) crude oil prices experienced heir bigges decrease owing o he US recession in ; (7) curren oil prices have rebounded up o $120 owing o concerns abou he poliical unres in he Norhern Africa region and he European deb crisis. Figure 1. Dynamics of weekly crude oil prices The reurn series of he wo prices are compued by Ri, pi, pi, 1 ln 100 for 1,2,, T, where R, denoes he coninuously compounded reurns for indices i a ime, and, i p denoes he closing price of indices i a ime. The wo reurn series clearly show volailiy clusering in Figure 2. i 151

6 Figure 2. Weekly reurns for crude oil prices Table 1 shows he descripive saisics for he wo reurn series. The sample mean of reurns is very small, and he corresponding sandard deviaions of reurns are much higher. The disribuion of reurns is no normally disribued, as is indicaed by he skewness, kurosis, and Jarque-Bera es. In addiion, he null hypohesis of no serial correlaion is saisically rejeced a he 1% significance level by he Ljung-Box es saisic, LB 2 (32), wih a lag of 32 for he squared reurn series, implying ha he squared reurns exhibi significan signs of serial correlaion. These resuls are in favor of a model ha incorporaes ARCH/GARCH feaures. Table 1. Descripive saisics of sample reurns Saisics WTI Bren Mean Sd. dev Skewness Kurosis Jarque-Bera (J-B) *** *** LB 2 (32) *** *** Noes: The J-B corresponds o he es saisic for he null hypohesis of normaliy in sample reurns disribuion. 2 The Ljung-Box saisic, LB (32), checks for he serial correlaion of he squared reurns up o he 32 nd order. *** indicaes he rejecion of he null hypohesis a he 1% significance level. Table 2 provides he resuls of augmened Dickey-Fuller (ADF) and Phillips-Perron (PP) uni roo ess for he log price series and he reurn series. The null hypohesis of he ADF and PP ess is ha a ime series conains a uni roo. As shown in Table 2, he calculaed values of boh he ADF and PP es saisic indicae ha he log price series conain a single uni roo a he l% significance level, implying ha he log prices series are non-saionary. However, in he case of reurn series, boh hese saisics rejec he null hypohesis of a uni roo a he l% significance level, implying ha he reurn series are saionary in all samples. 152

7 Table 2. Resuls of uni roo es for log price and reurns WTI Bren log price reurns log price reurns ADF [prob.] PP [prob.] [0.745] [0.764] [0.000] [0.000] [0.845] [0.843] Noe: MacKinnon s (1991) 1% criical value is for he ADF and PP ess [0.000] [0.000] 4. Empirical resuls 4.1. Resuls of coinegraion es and causaliy es Table 3 shows he resuls of he Johansen coinegraion es for he WTI and Bren reurn series. The race saisic is 35., which is above he 5% criical value of 15.49, rejecing he null hypohesis of no coinegraion beween WTI and Bren series variables, H0 : r 0 a he 5% significance level. Likewise, he max-eigen saisic is 34.60, which rejecs he null hypohesis of no coinegraion a he 5% significance level. However, under H0 : r 1, he race and max-eigen saisics are equal o 0.614, which are below he 5% criical value of Thus, in his case, we mus accep he null hypohesis of one coinegraion a he 5% significance level. As a resul, we have found a leas one coinegraion relaionship beween he WTI and Bren series. In oher words, we have found evidence of a long-run relaionship beween he wo series. Table 3. Resuls of Johansen coinegraion es Null hypohesis Trace Saisic 0.05 Criical value Max-eigen saisic 0.05 Criical value r 0 35.** ** r Noes: ** denoes rejecion of he hypohesis a he 5% significance level. The repored criical values are he Oserwald-Lenum (1992) criical values. Furhermore, we invesigaed he shor- and long-run causaliy using he VECM. Table 4 summarizes he esimaion resuls of he VECM obained from Equaions (3) and (4). According o he long-run relaionship using ec 1 (an error correcion erm), only he esimaed coefficien ec 1 for he WTI series is saisically significan a he 1% level, suggesing ha he Bren crude oil price is less likely o adjus han he WTI crude oil price when he wo crude oil markes deviae from heir long-erm coinegraion relaionship. Furhermore, we consider shor-run causaliy beween he WTI and Bren series. Mos of he ransmission coefficiens ( xy,1, xy,2, and yx,2 ) are saisically significan, implying a bi-direcional relaionship beween he WTI and Bren reurn series. This evidence indicaes ha he WTI reurns have an impac on he Bren reurns and vice versa. 153

8 Model Table 4. Esimaion resuls of he VECM 2 2 x,0 x 1 j1 xx, j j j1 xy, j j x, x ec x y 2 2 y,0 y 1 j1 yx, j j j1 yy, j j y, y ec x y x x,0 x xx,1 xx,2 xy,1 xy, (0.001) (0.033)** (0.056) (0.056)*** (0.056)*** (0.055)*** y y,0 y yx,1 yx,2 yy,1 yy, (0.001) (0.033) (0.057) Noe: ** and *** indicae significance a he 5% and 1% levels (0.055)*** (0.057) (0.055)*** Table 5. Esimaion resuls of he GARCH-BEKK model Symmeric Asymmeric Variable Coefficien Sd. error Coefficien Sd. error Panel A: Symmeric and asymmeric GARCH (1, 1)-BEKK esimaions c 1.092*** (0.143) 0.991*** (0.139) 11 c 0.383*** (0.160) 0.486*** (0.145) 21 c (0.260) (0.001) a 0.416*** (0.068) 0.302*** (0.070) 11 a *** (0.067) *** (0.064) a *** (0.065) (0.075) a 0.366*** (0.059) 0.385*** (0.060) b 0.862*** (0.051) 0.909*** (0.045) 11 b ** (0.052) 0.098*** (0.046) b (0.045) (0.045) b 0.876*** (0.046) 0.865*** (0.041) d 0.382*** (0.078) 11 d *** (0.075) d (0.066) d 0.242*** (0.093) Panel B: Diagnosic ess [0.633] [0.793] LB [0.594] [0.951] ARCH (10) [0.533] [0.747] ARCH (10) [0.933] [0.995] log-likelihood Noes: P-values are in brackes and sandard errors are in parenhesis. The ARCH i (10) es saisic checks he remaining ARCH effecs in sandardized residuals. The LB 2 i (32) es saisic checks for he serial correlaion of squared sandardized residuals. ** and *** indicae significance a he 5% and 1% levels, respecively. 154

9 4.2. Volailiy spillover beween he crude oil markes In order o examine he volailiy spillover effec, we use he symmeric and asymmeric GARCH (1,1) models based on he BEKK approach. The esimaion resuls of he BEKK model are repored in Table 5. To check he accuracy of he model specificaions, we employ wo diagnosic ess: he LM ARCH saisic, ARCH i (10), for sandardized residuals; and he Ljung-Box saisic, LB i2 (32), for squared sandardized residuals. Noe ha he ARCH i (10) es saisic checks he remaining ARCH effec in sandardized residuals and ha he LB i2 (32) es saisic checks for he serial correlaion of squared sandardized residuals. The insignificance of ARCH i (10) and LB i2 (32) saisics indicaes he appropriaeness of he symmeric and asymmeric GARCH-BEKK model. The imporan coefficiens in he bivariae GARCH model are a ii, and b ii, elemens of marices A and B, where i 1 sands for WTI, and i 2, for Bren. As menioned earlier, he diagonal elemens in marix A capure he own pas shock effec, while he diagonal elemens in marix B measure he own pas volailiy effec. From Table 5, he diagonal parameers ( b 11 and b ) in marix B are saisically significan, indicaing he presence of srong GARCH effecs; he own pas volailiy affecs he condiional variance of boh markes. Furhermore, he diagonal parameers ( a 11 and a ) are significan, implying an ARCH effec in boh markes. The off-diagonal elemens of marices A and B capure cross-marke effecs, such as shock spillover and volailiy spillover effecs beween he WTI and Bren markes. In he symmeric GARCH model, we find evidence of bidirecional shock spillover effec beween wo crude oil markes because he coefficiens a 12 and a 21 are negaively significan a he 1% level. However, in he asymmeric GARCH model, here is a unidirecional shock spillover effec from he WTI marke o he Bren marke. Furhermore, we idenify a unidirecional volailiy spillover from he WTI marke o he Bren marke in boh he symmeric and asymmeric GARCH models. For example, he pas volailiy of he WTI marke increases he presen volailiy of he Bren marke owing o he posiive value of he coefficien b 12, bu he reverse direcion is impossible. Thus, his evidence indicaes ha he WTI marke appears o play a more imporan role in influencing he volailiy of he Bren marke. As far as marix D is concerned, we find evidence of an asymmeric response o negaive shocks (bad news) of he own marke for boh reurns because of he significance of diagonal coefficiens d 11 and d. This evidence suggess ha he own negaive shocks have more effec han he own posiive shocks on he volailiy of each marke. In addiion, he cross-marke asymmeric response is eviden from he WTI marke o he Bren marke, as he coefficien d 12 is posiively significan a he 1% level. This means ha bad news in he WTI marke leads o a greaer volailiy change in he Bren marke han does good news in he Bren marke. In summary, our empirical resuls show a unidirecional volailiy spillover from he WTI marke o he Bren marke, indicaing ha WTI shocks heighen Bren marke volailiy bu no vice versa. More imporanly, bad shocks in he WTI marke cause greaer volailiy in he Bren marke han do good shocks in he WTI marke, owing o he cross-marke hedging demand. For example, a fall in WTI prices induces an acive hedger o shif funds from he WTI marke ino he Bren marke. Thus, bad news in he WTI marke will increase volailiy changes in he Bren marke. 5. Conclusions This paper invesigaed he reurn and volailiy spillover effecs beween crude oil markes using he VECM es and bivariae GARCH-BEKK model. In paricular, we considered he symmeric and asymmeric volailiy ransmissions beween he WTI and Bren markes. By employing he VECM es, we found bi-direcional reurns ransmissions beween he WTI and Bren markes. This evidence indicaes ha he WTI reurns have an impac on he Bren reurns and vice versa. Wih regard o symmeric and asymmeric volailiy, he empirical resuls show unidirecional volailiy spillover from he WTI marke o he Bren marke, indicaing ha informaion is ransmied from he WTI marke o he Bren marke bu ha a reverse direcion is impossible. More imporanly, bad news in he WTI marke seems o increase he 155

10 volailiy of he Bren marke more sharply han do good shocks in he WTI marke. Thus, hese findings indicae ha he WTI marke appears o play a more imporan role in influencing he volailiy of he Bren marke. These findings are of pracical imporance o crude oil marke paricipans and may be useful in making opimal porfolio allocaion decisions and developing hedging sraegies beween he wo crude oil markes. In he crossmarke hedging demand, falling WTI prices induce acive hedgers o shif funds from he WTI marke o he Bren marke. Thus, bad news in he WTI marke will increase volailiy changes in he Bren marke. References Aloui, C., Jammazi, R., The effecs of crude oil shocks on sock marke shifs behaviour: a regime swiching approach. Energy Economics 31, Bekiros, S.D., Diks, C.G.H., The relaionship beween crude oil spo and fuures prices: coinegraion, linear and nonlinear causaliy. Energy Economics 30, Belloumi, M., Energy consumpion and GDP in Tunisia: coinegraion and causaliy analysis. Energy Policy 37, Chiou, J.-S., Lee, Y.-H., Jump dynamics and volailiy: oil and he sock markes. Energy 34, Engle, R.F., Kroner, K.F., Mulivariae simulaneous generalized ARCH. Economeric Theory 11, Engle, R.F., Ng, V.K., Measuring and esing he impac of news on volailiy. Journal of Finance 48, Ewing, B.T., Malik, F., Ozfidan, O., Volailiy ransmission in he oil and naural gas markes. Energy Economics 24, Glosen, L.R., Jagannahan, R., Runkle, D.E., On he relaion beween he expeced value and he volailiy of he nominal excess reurn on socks. Journal of Finance 48, Hamilon, J.D., Herrera, A.M., Commen: oil shocks and aggregae macroeconomic behavior: he role of moneary policy. Journal of Money, Credi and Banking 36, Hooker, M.A., Are oil shocks inflaionary? Asymmeric and nonlinear specificaion versus changes in regime. Journal of Money, Credi and Banking 34, Johansen, S., Esimaion and hypohesis esing of coinegraion vecors in Gaussian vecor auoregressive models. Economerica 59, Kaufmann, R.K., Ullman, B., Oil prices, speculaion, and fundamenals: inerpreing causal relaions among spo and fuures prices. Energy Economics 31, Kroner, K.F., Ng, V.K., Modeling asymmeric comovemens of asse reurns. Review of Financial Sudies 11, Lee, B.R., Lee, K., Rai, R.A., Moneary policy, oil price shocks, and he Japanese economy. Japan and he World Economy 13, MacKinnon, J.G., Criical values for coinegraion ess. In: Engle, R.F., Granger, C.W.J. (Eds.), Long-Run Economic Relaionships: Readings in Coinegraion, Oxford Universiy Press, New York, Malik, F., Hammoudeh, S., Shock and volailiy ransmission in he oil, US and Gulf equiy markes. Inernaional Review of Economics and Finance 16, Mork, K., Business cycles and he oil marke. Energy Journal 15, Oserwald-Lenum, M., A noe wih quaniles of he asympoic disribuion of he maximum likelihood coinegraion rank es saisics. Oxford Bullein of Economics and Saisics 54, Rafiq, S., Salim. R., Bloch, H., Impac of crude oil volailiy on economic aciviies: an empirical invesigaion in he Thai economy. Resources Policy 34, Ross, S.A., Informaion and volailiy: he no-arbirage maringale approach o iming and resoluion irrelevancy. Journal of Finance 44, Sadorsky, P., Oil price shocks and sock marke aciviy. Energy Economics 21, Sadorsky, P., The empirical relaionship beween energy fuures prices and exchange raes. Energy Economics, Sadorsky, P., Risk facors in sock reurns of Canadian oil and gas companies. Energy Economics 23, Sadorsky, P., Correlaions and volailiy spillovers beween oil prices and he sock prices of clean energy and echnology companies. Energy Economics 34, Zhang, Y.-J., Fan, Y., Tsai, H.-T., Wei, Y.-M., Spillover effec of US dollar exchange rae on oil prices. Journal of Policy Modeling 30,

11 Brief biography of auhors Sang Hoon Kang Sang Hoon Kang is a Professor of Business Adminisraion a Pusan Naional Universiy. He holds a PhD in Financial Economics from Universiy of Souh Ausralia. His main ineres is o model and forecas he volailiy of financial ime series. He has published several academic journal papers, in paricular Forecasing Volailiy of Crude Oil Markes (Energy Economics, 2009), Srucural Changes and Volailiy Transmission in Crude Oil Markes (Physica A, 2011). Seong Min Yoon Seong Min Yoon is a Professor of Economics a Pusan Naional Universiy. He holds a PhD in Economics from Korea Universiy. He has been a visiing scholar o he Universiy of Washingon in Seale and he Universiy of Colorado in Denver. His research fields include financial marke dynamics, corporae finance, regional economics, and applied microeconomics. He has published over 120 academic journal papers on hose fields including Forecasing Volailiy of Crude Oil Markes (Energy Economics, 2009) and various books including Vision and Issues of Busan Indusry (Asian Insiue for Regional Innovaion, 2009). 157