The Journal of Applied Business Research January/February 2014 Volume 30, Number 1

Transcription

1 Dynamic Spillover Beween The Oil And Sock Markes Of Emerging Oil-Exporing Counries Frederic Teulon, IPAG Business School, France Khaled Guesmi, IPAG Business School & EconomiX Universiy of Paris Oues Nanerre La Défense, France ABSTRACT The paper invesigaes he ime-varying correlaions beween sock marke reurns and oil prices in oil-exporing counries. A mulivariae GARCH-DCC process is employed o evaluae his relaionship based on daa from Venezuela, he Unied Arab Emiraes, Saudi Arabia and Kuwai. The resuls show ha here are ime-varying correlaions beween he oil and sock markes in emerging, oil-producing counries, indicaing ha hey are affeced by condiions in world markes. In addiion, he relaionship beween oil prices and sock reurns is found o be influenced by he origin of shocks o oil prices, wih sock marke responses being sronger o demand-side shocks caused by poliical urmoil or flucuaions in he global business cycle han o supply-side shocks caused by cus in oil producion. The resuls also provide evidence of volailiy spillovers beween he oil and sock markes. Keywords: Time-Varying Inegraion; Oil Exporing Counries; Oil Marke. INTRODUCTION I n recen years, much empirical research has focused on he impac of oil prices on macroeconomic variables in developed counries (Kilian & Park, 2009). Among such sudies, aenion is increasingly being devoed o he correlaion beween oil prices and sock markes. Alhough here is no heoreical model ha describes he relaionship beween he oil and financial markes, academic research finds ha flucuaions in oil prices affec corporae performance, oupu, and earnings, which subsequenly influence sock reurns. Mos exising sudies focus on he relaionship beween he oil and financial markes in oil-imporing counries, such as he Unied Saes (US), where increases in oil prices have negaive economic effecs. When oil prices increase, oil imporers can experience srong negaive consequences and economic recession (Ferderer, 996). However, he effecs of changes in oil prices on oher ses of counries are worh invesigaing. For his reason, he paper addresses he relaionship beween oil prices and financial markes from he perspecive of emerging, oilexporing counries. The relaionship beween oil prices and financial markes is hough o have specific effecs on members of he Organizaion of he Peroleum Exporing Counries (OPEC). In conras o he case of oil imporers, increases in oil prices posiively influence he economies of oil-exporing counries by providing hem wih addiional income and wealh. If his surplus income is ransmied back o he economy, hen higher oil prices would be expeced o lead o boh higher levels of economic aciviy and higher sock prices. On he oher hand, decreases in oil prices have a negaive relaionship wih he economic growh of oil producers and can generae poliical and social insabiliy (Yang e al., 2002), puing downward pressure of financial reurns. Therefore, as underlined by Bashar (2006) and Arouri e al. (200), he relaionship beween sock markes and oil prices in oil-exporing counries can be ambiguous and needs o be invesigaed. In addiion o he sudy of co-movemens, anoher relevan research area is volailiy analysis. The mos recen lieraure on his subjec focuses on volailiy spillovers beween oil or indusrial commodiies and sock Copyrigh by auhor(s); CC-BY 5 The Clue Insiue

2 markes. Hammoudeh e al. (2004) invesigae he spillover effecs, day effecs, and dynamic relaionships (among he daily prices of five S&P 500 oil secor socks and he prices of five ypes of oil in he US oil marke) using coinegraion echniques as well as ARCH-ype models. The auhors find evidence of volailiy spillovers from he oil fuures marke o some oil socks. Chiou and Lee (2009) examine he daily asymmeric effecs of WTI oil prices on S&P 500 sock reurns. Using he Auoregressive Condiional Jump Inensiy model wih expeced, unexpeced, and negaive unexpeced oil price flucuaions, hey find ha high flucuaions in oil prices have unexpeced, asymmeric effecs on sock reurns. Malik and Ewing (2009) rely on bivariae GARCH models o esimae he ransmission of volailiy beween weekly WTI oil prices and equiy reurns, finding evidence of spillover mechanisms. Choi and Hammoudeh (200) exend he ime-varying correlaions analysis by examining prices of commodiies including Bren oil, WTI oil, copper, gold, and silver in relaion o he S&P 500 index. They show ha correlaions beween commodiy prices have increased since 2003, limiing porfolio hedging subsiuabiliy. More recenly, Arouri e al. (200) examine he relaion beween oil prices and 2 sock markes in European counries. Their resuls show ha he reacions of reurns o changes in oil prices vary considerably across indusrial secors; furhermore, he inclusion of oil asses ino a porfolio of socks allows for improvemens in he porfolio s riskreurn characerisics. Of paricular ineres for our sudy is he work of Filis e al. (20) who analyze ime-varying correlaions beween Bren oil prices and sock markes by differeniaing beween oil-imporing and oil-exporing counries. They find ha ime-varying correlaions depend on he origins of oil shocks; responses are much greaer o aggregae-demand-side shocks han o supply-side shocks originaing from cus in OPEC producion. In order o beer undersand hese issues, his sudy evaluaes oil and sock marke inerdependence in oilexporing counries by measuring ineracions beween oil prices and sock marke indices as per he correlaion analysis defined by Engle (2002). The remainder of his paper is organized as follows: Secion 2 presens he empirical mehodology, he daa are described in Secion 3, he resuls are repored and discussed in Secion 4, and Secion 5 concludes. 2. METHODOLOGY In his research, we apply ime-varying correlaion coefficiens esimaed from a muli-variae GARCH- DCC model inroduced by Engle (2002). By allowing condiional correlaions o vary over ime, his specificaion is viewed as a generalizaion of he Consan Condiional Correlaion (CCC) model (Bollerslev, 990). To illusrae he dynamic condiional correlaion model for our purposes, le x be a 6 vecor (four sock markes, world marke and oil prices) conaining reurn, volume, and implied volailiy series in he following condiional mean equaion: x = µ + ε () where = E[ x Π ] µ is he condiional expecaion of x given he pas informaion Π, and ε is a vecor of errors in auoregression AR(). Assuming ha he reurn, volume and implied volailiy series x are deermined by he informaion se available a ime -, he model may be esimaed using maximum likelihood mehods, provided he condiional covariance marix, H. We also assume ha µ akes he following form: µ = Φ + Φ x, i, 0 i, i (2) Φ measures he ARCH effec in daa series. In he radiional mulivariae GARCH framework, he condiional variance-covariance marix can be wrien as: H = G R G where G = diag{ } (3) h i Copyrigh by auhor(s); CC-BY 52 The Clue Insiue

3 h i is he esimaed condiional variance from he individual sandard univariae GARCH(,) models in he following manner: h i 2 = i + αiε i, + βihi, ω i (4) R is he ime-varying condiional correlaion coefficien marix. As specified in Equaion (4), he variance of each marke is modeled as a funcion of he consan, he square of he las period s own residuals h ε 2 i,, and is lagged condiional variance i,. Afer he basic consrucion delineaed above, he dynamic correlaion coefficien marix of he DCC model can be denoed by: R [ ( )] diag Q 2Q [ diag( Q )] 2 = ( ) Q = q ij,!" diag( Q )# & $ 2 = diag ( ' q,, q 22,,... q 2,2, ) + * (5) In order o sandardize he residual error erm, Engle ses of condiional sandard deviaions and w = ε where G G is a 6 6 diagonal marix z is he sandardized residuals vecor wih a mean of zero and variance of one. Engle also suggess he esimaion of he following ime-varying correlaion process: ρ ij, = q q ij, ii, q jj, where q ij, = ρ ij + a( z i, z j, ρ ij ) + b( q ij, ρ ij ) = ( a b ) ρ ij + a z i, z j, + b q ij, (6) The ime-varying correlaion coefficiens in GARCH-DCC model can be divided ino wo componens. The firs componen, ρ, on he righ-hand side of Equaion (6) represens he uncondiional expecaion of he cross ij produc z z i j ; i.e., he uncondiional correlaion coefficien. The second componen, a zi, z j, + b qij,, also on he righ-hand side of Equaion (6), indicaes he condiional ime-varying covariance. 3. DATA AND STOCHASTIC PROPERTIES 3. Descripion of he Daa The daa sources for our sudy are Daasream Inernaional, MSCI, and he Inernaional Moneary Fund s (IMF) Inernaional Financial Saisics daabases. We used monhly daa colleced from he sock marke indices of oil-exporing counries (Venezuela, he Unied Arab Emiraes, Saudi Arabia, and Kuwai), a world sock marke index (WORLD), and Bren crude oil index (OIL) over he period Augus 3, 2000 o June 3, 200. All reurns are expressed in US dollars (USD). Copyrigh by auhor(s); CC-BY 53 The Clue Insiue

4 3.2 Main Saisics and Sochasic Properies Table repors he main saisics of he reurn series for he sock marke and Bren crude oil indices. All he series depar from condiions of normaliy and condiional heeroskedasiciy. The marke of he Unied Arab Emiraes was he mos volaile during he sudied period in erms of sandard deviaion (2.67%), while ha of Kuwai was he leas volaile (5.96%). The skewness coefficiens are posiive for he Unied Arab Emiraes and Venezuela. They are significanly differen from zero for almos all markes, indicaing he presence of asymmery in he disribuion of reurns. In addiion, all he reurn series are characerized by a kurosis coefficien ha is saisically significan and greaer han hree, meaning ha hey have faer ails han hose of a normal disribuion. Engle s (982) es for he firs order of condiional heeroskedasiciy is also performed and he hypohesis of no ARCH effecs for all considered reurn series canno be rejeced. This resul moivaes our choice of he GARCH modeling approach for condiional variance processes. Table : Descripive Saisics of Reurn Series Panel B: Excess Reurns On Sock Marke Indices Mean Sd. Dev. Skewness Kurosis ARCH() Unied Arab Emiraes Kuwai Saudi Arabia Venezuela World Oil Prices Noes: ARCH() is he empirical saisics of he Engle (982) s es for he 6 h order of ARCH effecs. +, ++, and +++ indicae ha he null hypohesis of no ARCH effecs is rejeced a he 0%, 5% and % levels respecively. The highes correlaion is observed for he pair, including he Unied Arab Emiraes and Kuwai (0.28) (see Table 2). The sock marke of he Unied Arab Emiraes has he lowes uncondiional correlaion wih he world marke (-0.3). The correlaions of he Kuwaii and Saudi markes wih he world marke are and -0.06, respecively. We also noice a negaive correlaion beween sock reurns in he Unied Arab Emiraes and he real exchange rae reurns in Kuwai (-0.03), which is consisen wih he view ha he foreign exchange marke in he Unied Arab Emiraes may serve as a hedge agains risks in he Kuwaii sock marke during bear marke periods. Table 2: Correlaions RE_EU RE_KU RE_SA RE_VE RE_W RE_OI RE_EU,00 0,29 0,42-0,8-0,3-0,3 RE_KU 0,29,00 0,28-0,05-0,04-0,09 RE_SA 0,42 0,28,00 0,08-0,06-0, RE_VE -0,8-0,05 0,08,00 0,03-0,3 RE_W -0,3-0,04-0,06 0,03,00 0,2 RE_OI -0,3-0,09-0, -0,3 0,2,00 Noes: DU, SA, KU, VE, W, US and OI idenify he sock marke of U. A. Emiraes, Saudi Arabia, Kuwai, Venezuela, World and Oil prices. ARCH(6) is he empirical saisics of he Engle (982) s es for he 6 h order of ARCH effecs. +, ++, and +++ indicae ha he null hypohesis of no ARCH effecs is rejeced a he 0%, 5% and % levels respecively. Table 3 shows he auocorrelaions and parial auocorrelaions of excess reurns on sock markes and Bren crude oil indices, as well as reurns on real exchange raes. Paricularly noeworhy is ha only firs-order auocorrelaions are significan a he % level for sock reurns, oil prices, and exchange raes. This finding suggess ha he GARCH (,) parameerizaion for second momens is appropriae. Copyrigh by auhor(s); CC-BY 54 The Clue Insiue

5 Table 3: Auocorrelaions of Reurn Series Panel A: Auocorrelaions of Excess Reurns on Sock Marke Indices and Oil Prices Lag Oil Prices U. A. Emiraies Kuwai Saudia Arabia Venezuela AC PAC AC PAC AC PAC AC PAC AC PAC *** *** 0.85 *** 0.85 *** 0.34 *** 0.34 *** 0.74 *** 0.74 *** 0.47 *** 0.47 *** Noes: This able repors he serial correlaion and parial auocorrelaion funcions for excess sock marke reurns. ** and *** indicae he significance a he 5% and % levels respecively. 4. EMPIRICAL RESULTS Table 4 shows ha he average dynamic condiional correlaion beween sock markes and oil prices is posiive and significan a he % level, hus providing evidence of volailiy spillovers beween oil and sock markes. However, his is no he case for Venezuela, whose dynamic correlaions are no significan, probably because of he very limied developmen of is sock marke compared o he imporance of oil in ha economy (World Bank, 202). Table 4: Dynamic Condiional Correlaions beween Sock Markes and he World Marke Unied Arab Emiraes Kuwai Saudi Arabia Venezuela ρ max ρ min ρ mean -0.5*** -0.07* -0.0*** * Sd. dev Noes: This able repors some saisics of dynamic condiional correlaions, esimaed from he DCC-GARCH model. ρ max, ρ min, and ρ mean indicae he maximum, minimum and average values of dynamic condiional correlaions. *, **, and *** indicae ha he average degree of inegraion is significanly differen from zero a he 0%, 5%, and % levels wih respec o he wo-sided Suden- es, respecively. As Figure illusraes, correlaions increase saring from he end of 2007 in he cases of he Unied Arab Emiraes, Kuwai, and Saudi Arabia. These counries display posiive correlaions during all he invesigaed periods and exhibi considerable decreases in correlaion coefficiens beween 2000 and This period is characerized by changes in he precauionary demand for crude oil as a resul of he Iraq war (see Filis e al., 20). Our resuls are consisen wih hose of Forbes and Rigobon (2002) who sressed ha increased correlaions during imes of crisis are caused by increased volailiy in global sock markes Emiraes Arab Unis 0.3 Kuwai Saudi Arabia 0.04 Venezuela Figure : Dynamic Condiional Correlaions wih Oil Noe: The uni is he correlaion coefficien Copyrigh by auhor(s); CC-BY 55 The Clue Insiue

6 The nex sub-period of ineres is from 2006 o 2008, when oil prices increased significanly (excep in Venezuela) due o rising demand, mainly from China. The correlaion coefficien showed an increasing and posiive paern for all counries. This aggregae-demand-side oil price shock is expeced o have a posiive effec on sock markes because i signals an increase in world rade. This finding is in line wih Hamilon (2009) and Kilian and Park (2009), who suggesed ha aggregae-demand-side oil price shocks resuling from global economic growh have a posiive impac on sock prices. Two main conclusions can be drawn from our invesigaion. Oil price shocks during periods of global urmoil and poliical insabiliy have an imporan impac on he relaionship beween oil prices and sock marke reurns. Regarding he sign of his correlaion, we find wo rends - a negaive one, in he case of Venezuela beween 2007 and 2009, and a posiive one, when aggregae-demand-side oil price shocks (Asian crisis, Chinese economic growh, and he global financial crisis) resul in a significan posiive correlaion beween sock marke and oil prices. The repercussions of hese phenomena are no symmeric in all he sudied counries; herefore, we do no see a proper "conagion effec". However, some specific rends appear among counries whose sock markes exhibi posiive correlaions. 5. CONCLUSION The curren sudy invesigaes ime-varying correlaions beween sock marke reurns and oil prices in oilexporing counries, considering he origin of shocks o oil prices (i.e., aggregae-demand-side, precauionary demand, or supply-side). The daase consiss of monhly sock and oil prices from he major oil-exporing counries - Venezuela, he Unied Arab Emiraes, Saudi Arabia, and Kuwai - from Augus 3, 2000 o Jun 3, 200. Condiional correlaions are esimaed following he GARCH-DCC mehod of Engle (2002). The sudy provides evidence for ime-varying correlaions beween oil and sock prices. The resuls show a significan increase in he mean of he dynamic condiional correlaion coefficien beween world and he examined emerging, oilexporing counries. Furhermore, he correlaion is found o change as a resul of he origin of shocks o oil prices semming from periods of global urmoil or changes in he global business cycle. In paricular, precauionary demand shocks, caused by wars or erroris aacks, and aggregae-demand-side shocks, caused by flucuaions in he global business cycle (i.e., housing marke boom, Chinese growh, global financial crisis), end o influence he correlaion beween oil and sock marke prices o a much greaer exen han supply-side shocks originaing from OPEC producion cus. AUTHOR INFORMATION Frédéric Teulon is Head of he Research cener (IPAG research Lab) and made a career as a professor a various universiies in Paris. He now eaches a he IPAG Business School. His main research areas are macroeconomics and inernaional finance. f.eulon@ipag.fr Khaled Guesmi is Professor a he IPAG Business School and is a specialis in quaniaive sudies. His main research areas are quaniaive finance and economerics. khaled.guesmi@ipag.fr (Corresponding auhor) REFERENCES. Arouri, M. E. H., & Nguyen, D. K. (200). Oil prices, sock markes and porfolio invesmen: Evidence from secor analysis in Europe over he las decade. Energy Policy, 38, Bashar, A. Z. (2006). Wild oil prices, bu brave sock markes! The case of GCC sock markes. Operaional Research, 6(2), Bollerslev, T. (990). Modelling he coherence in shor-run nominal exchange raes: A mulivariae generalized ARCH model. Review of Economics and Saisics, 72, Chiou, J. S., & Lee, Y. H. (2009). Jump dynamics and volailiy: Oil and he sock markes. Energy, 34(6), Copyrigh by auhor(s); CC-BY 56 The Clue Insiue

7 5. Choi, K., & Hammoudeh, S. (200). Volailiy behavior of oil, indusrial commodiy and sock markes in a regime-swiching environmen. Energy Policy, 38(8), Engle, R. (982). Auoregressive condiional heeroskedasiciy wih esimaes of he variance of U.K Inflaion. Economerica, 50, Engle, R. F. (2002). Dynamic condiional correlaion: A simple class of mulivariae generalized auoregressive condiional heeroskedasiciy models. Journal of Business and Economic Saisics, 20, Federer, J. P. (996). Oil price volailiy and he macroeconomy. Journal of Macroeconomics, Winer 996, 8(), pp Filis, G., Degiannakis, S., & Floros, C. (20). Dynamic correlaion beween sock marke and oil prices: The case of oil-imporing and oil-exporing counries. Inernaional Review of Financial Analysis, 20(3), Forbes, K. & Rigobon R. (200). No conagion, only inerdependence: measuring sock marke comovemens. NBER Working Paper Hamilon, D. J. (2009). Causes and consequences of he oil shock of Brooking papers on economic aciviy, spring, Hammoudeh, S., Dibooglu, S., & Aleisa, E. (2004). Relaionships among U.S. oil prices and oil indusry equiy indices. Inernaional Review of Economics and Finance, 3(4), Kilian, L., & Park, C. (2009). The impac of oil price shocks on he U.S marke. Inernaional Economic Review, 50, Malik, F., & Ewing, B. T. (2009). Volailiy ransmission beween oil prices and equiy secor reurns. Inernaional Review of Financial Analysis, 8, Yang, C. W., Hwang, M. J., & Huang, B.N. (2002), An analysis of facors affecing price volailiy of he US oil marke. Energy Economics, 24, Copyrigh by auhor(s); CC-BY 57 The Clue Insiue

8 NOTES Copyrigh by auhor(s); CC-BY 58 The Clue Insiue