International CAPM and Oil price: Evidence from selected OPEC countries

Transcription

1 Internatonal CAPM and Ol prce: Evdence from selected OPEC countres ANNA CRETI* LeDa-CGEMP, Unversté Pars Dauphne & Ecole Polytechnque, France KHALED GUESMI IPAG Busness School, IPAG Lab I. Introducton Ths study contrbutes to the exstng lterature on dynamc nternatonal captal asset prcng model (ICAPM) allowng for smooth transton between dfferent ntegraton regmes and takng nto account changes n ol prce, for the major OPEC countres. In our model, expected returns may move from a perfectly-segmented regme to a perfectly-ntegrated one, or vce versa, dependng on a certan number of natonal and nternatonal factors that are lkely to drve the process of fnancal ntegraton. The proposed model s developed n the sprt of that presented by Bekaert and Harvey (1995) and allows for dynamc condtonal correlatons between stock returns by usng the multvarate DCC-GJR-GARCH model of Tse and Tsu (2002). Ths method also enables to test the relevance of dynamc measures of fnancal ntegraton wth respect to condtonal correlatons, whch are frequently used n the lterature when referrng to the level of ntegraton. Our study dffers from past ones n that we nvestgate the ntegraton of ol-exportng countres nto the world market usng ol prce as a common source of rsk. In fact, ol-dependent economes are partcularly exposed to large and volatle shocks assocated wth ol prce fluctuatons. The mpact of these shocks s pervasve, encompassng the government s budget process and balance sheet, as well as prvate-sector producton and consumpton decsons (Burger et al. 2010). Volatlty of ol prce could undermne stablty of countres whose economy hghly depends on ol exports (Yang et al., 2002). However, the negatve mpact of ol prce nstablty can be counterbalanced by the (short-term) low elastcty of world ol demand, whch guarantees ncome and wealth to OPEC countres. Therefore, n addton to world and local sources of rsk whch are commonly used n prevous studes, as for nstance n Bekaert and Harvey (1995), Hardouvels et al. (2006), * Correspondng author : Anna Cret LEDA CGEMP, Unversté Pars Dauphne, Place du Maréchal Delattre de Tassgny, 75016, Pars. Emal: anna.cret@dauphne.fr. Carrer et al. (2007), Arour et al. (2012b), Guesm and Nguyen (2011, 2014), we analyze the mpact of ol as a potental drver of fnancal ntegraton. Investgatng market ntegraton and ol rsk n OPEC countres s an ntrgung ssue. On one sde, OPEC countres, gven ther role of major world ol exporters, are subject to nternatonal macroeconomc trends that n turn drve ol market demand and therefore ther supply decsons. However, besde these fundamentals drvers, ol prce s also mpacted by a context of progressve fnancalzaton of raw materals, whch mples a comovement between ol and stock prces. Hence, f fundamentals would tend to foster OPEC market ntegraton, ol fnancalzaton could bas ths trend. Moreover, both market ntegraton and ol fnancalzaton are very dynamc phenomena, changng over tme, therefore a pror each of these forces can counteract the other one dependng on ther relatve strength. The objectve of ths paper s therefore twofold: frst, to test OPEC countres market ntegraton, and secondly to analyse how the ol market renforces or weakens ths macroeconomc phenomenon n a tme horzon whch encompasses some mportant macro-economc events such as the fnancal and economc crss. To ths end, we brng together two streams of lterature: on one hand, works on captal asset prcng, and on the other, models on the relatonshp between ol prce and stock markets. We nvestgate the lnks between fnancal ntegraton and ol prce n the four major ol-exportng countres, that s Saud Araba, Kuwat, Venezuela and Unted Arab Emrates. These countres represent 64% of the total OPEC offer 1 (OPEC, 2012) and also provde easy access to ther stock market data (see Charts A n the Appendx). Venezuela s the most dense country n terms of populaton, but the less rch n terms of GDP per capta. All the four countres analyzed have a postve trade balance, due to ol export mportance, ths latter representng roughly 95% of the export value. The major producer and exporter s Saud Araba; followed by the Unted Arab Emrates, Kuwat and then Venezuela, whose ol proven reserves are slghtly bgger than those of Saud Araba. Over the perod we nvestgate, revenue of ol mnus producton costs represents between 40 and 60 percent of GDP n Kuwat. 64 Cret_Artcle.ndd Sec1:64 14/04/15 12:03

2 All the other countres reman below the 60% threshold, wth the Unted Arab Emrates stablzng around 20%. Ol revenue s rather cyclcal and has declned n all countres durng the 2009 fnancal crss, due to fallng demand. Ol consumpton s also relatvely hgh n Saud Araba compared to other countres, and n partcular Kuwat. All these four OPEC countres are also well endowed n terms of gas producton and reserves, but only the Unted Arab Emrates export t, therefore our analyss focuses on ol. The exchange rate wth the dollar s qute homogeneous except for the case of Kuwat, whose currency s very weak wth respect to the dollar. On the fnancal sde, stock market captalzaton n percentage of the GDP has peaked n all countres between 2004 and 2007, reachng an hstorcal maxmum of nearly 200% n 2005 n Saud Araba. Venezuela lags behnd the other countres, wth a percentage beng steadly below 10%. The number of companes lsted n the stock exchange has been growng over tme, up to a maxmum of 210 n Kuwat, n 2010, followed by a slow declne untl The Unted Arab Emrates has experenced an nflexon n the number of lsted companes n 2001 and then a constant rse. The number of lsted companes n Venezuela s around sxty on average, wth a declne n the latest years of the sample. Interestngly enough, n the tme frame we analyze, our sample dsplays smlar trends as for the dynamc pattern of ol revenues, stock market captalzaton, number of lsted companes, wth some heterogenety regardng the strength of ol dependence and the mportance of fnancal markets n each of the economes under nvestgaton. Ths allows us to study both common features and country-specfc dfferences among the largest OPEC ol exporters. Our analyss shows that, over the perod August 2000 to June 2012, the ntegraton degree of the four major OPEC ol-exportng countres vares wdely through tme. Ths phenomenon s explaned by the nterest rate spread, the level of market openness and the return of world market ndex. Although the general trend s towards ncreasng fnancal ntegraton, ol-exportng countres seem to be stll sgnfcantly segmented from the global market. A breakdown of the total rsk premum confrms ths fndng, n that t underlnes the domnant role of the local rsk factor n explanng varatons n the expected returns for the four countres studed. We fnd that the revenue effect of ol exports protects OPEC countres: ol rsk represents a small fracton of the global rsk for all the countres n our sample. The most exposed country s the Unted Arab Emrates, the bggest OPEC exporter, where ol rsk represents 11% of total rsk. Overall, we show that condtonal correlatons underestmate the level of ol-exportng markets ntegraton. At the same tme, we fnd an ncrease n the level of market ntegraton durng crss perods, and a rsng trend as from the end Ths result echoes the conclusons reached by Arour et al. (2012) and Awartan and Maghyereh (2013) for case of the Gulf Cooperaton Councl countres. Fnancal ntegraton n crss perods could be drven by the postve relatonshp between ol and stock markets n ol-exportng countres, as documented for nstance by Arour and Rault (2011) and Fls (2011). Among the four countres analyzed, Venezuela exhbts some pecular characterstcs. Local rsks play a more mportant role than the nternatonal ones, contrarly to the other ol-exportng countres. Venezuela economc ntegraton s manly drven by trade openness. Moreover, dynamc correlatons of ts stock market wth world and ol are qute weak or not sgnfcant. These results llustrate the mpact of ol n an emergng country where fnancal markets are stll underdeveloped. The paper s organzed as follows. Secton 2 presents the lterature revew related to our study. Secton 3 descrbes the condtonal verson of the Internatonal Captal Asset Prcng Model, where world market rsk, ol rsk, currency rsk and local rsk are prced. The data s descrbed n Secton 4. Results are reported and dscussed n Secton 5. Secton 6 brefly concludes. II. Lterature revew Our model reles on two dfferent strand of lterature. On one sde, we refer to models that test the ntegraton of emergng countres n the world market, usng several verson of the CAPM model. On the other, we add to the debate on the lnks between ol and stock markets. As for the frst strand of the lterature, partal ntegraton of emergng markets has frst been nvestgated and tested usng Stehle (1977) s methodology. 2 Claessens and Rhee (1994) use ths methodology to examne the rsk-return lnkages n 16 emergng markets over the perod from 1989 to The emprcal results obtaned contradct the hypothess of ntegraton n most of the markets. By combnng the two tests, the authors show that emergng countres under consderaton (Brazl, Greece, South Korea, Mexco, Pakstan, the Phlppnes, Tawan, and Thaland) were segmented from the world market. In a dfferent way, the emprcal evdence documented n studes such as Stulz (1981), Errunza and Losq (1985), and Wheatley (1988) supports the partal segmentaton hypothess n lght of sgnfcant effects of legal barrers on asset prcng rules n emergng markets. Bekaert and Harvey (1995) agree wth the dea of a partal ntegraton, but are aganst a statc measure of the degree of market ntegraton. Accordngly, they develop an alternatve model that combnes the two extreme cases of perfect segmentaton and ntegraton so that at each pont n tme expected return on an asset (or a market) depends smultaneously on a global rsk factor weghted by an ntegraton coeffcent, and a local rsk factor weghted by a segmentaton coeffcent. Ths model s reduced to a domestc CAPM for strctly segmented markets, and to an nternatonal CAPM for perfectly ntegrated markets. Bekaert and Harvey (1995) apply ther nested model to 12 emergng markets and show that ther level of ntegraton changes over tme. On the same lne, Guesm and Nguyen (2011) study the dynamcs of the global ntegraton process of four emergng market regons nto the world market wth DCC-GARCH process. They fnd that the level of market ntegraton vares wdely over tme and 65 Cret_Artcle.ndd Sec1:65 14/04/15 12:03

3 s satsfactorly explaned by the degree of trade openness and varaton n the US term premum. Even though market ntegraton reaches farly hgh values durng several perods, and exhbts an upward trend towards the end of the estmaton perod, the emergng market regons under consderaton stll reman segmented from the world market. None of the aforementoned models has studed the role of ol for specfc countres. Nevertheless, ol represents a pecular asset, n partcular for emergng economes that export much of ther natural resources. The lterature so far has analyzed ths phenomenon manly by lookng at the comovements between stock and ol markets. Most of ths lterature offers substantal evdence on the mpact of ol on stock prces, puttng forward a negatve relatonshp between ol prce and stock market returns. 3 For nstance, Jones and Kaul (1996), usng a standard cashflow dvdend valuaton model, fnd a sgnfcant negatve mpact of ol prce shocks on US and Canadan quarterly stock prces n the postwar perod. Several models, relyng on some varants of Vector Autoregressve Analyss, hghlght smlar fndngs (Park and Ratt 2008, Sadorsky 1999, Papapetrou, 2001). Shftng from the study of comovements to volatlty analyss, the most recent lterature focuses on volatlty spllovers between ol/ndustral commodty and stock markets. Hammoudeh et al. (2004) nvestgate the spllover effects, day effects, and dynamc relatonshps among fve daly S&P ol sector stock ndces and fve daly ol prces for the US ol markets usng contegraton technques as well as ARCH-type models. They fnd that there are two-way nteractons between the S&P Ol Composte ndex, and ol spot and futures prces. Chou and Lee (2009) examne the asymmetrc effects of WTI daly ol prces on S&P 500 stock returns. Usng the Autoregressve Condtonal Jump Intensty model wth expected, unexpected and negatve unexpected ol prce fluctuatons, they fnd that hgh fluctuatons n ol prces have asymmetrc unexpected effects on stock returns. Malk and Ewng (2009) rely on bvarate GARCH models to estmate the volatlty transmsson between weekly WTI ol prces and equty sector returns and fnd evdence of spllover mechansms. Cho and Hammoudeh (2010) and Cret et al. (2013) extend the tme-varyng correlatons analyss between the prces of ol and several other commodtes wth stock market ndexes. Both the studes show that commodty correlatons ncrease snce 2003, lmtng hedgng substtutablty n portfolos, and become stronger after the 2008 fnancal crss. Usng Wavelet analyss, Roboredo and Rvera-Castro (2014) analyse the daly connecton between ol prce, the aggregate S&P 500, Dow Jones Stoxx Europe ndexes and European ndustral sectors. They conclude that ol prce changes do not dsplay a szable effect on stock market returns n the perod of pre-crss. Of partcular nterest for our study, Fls et al. (2011) analyze tme-varyng correlatons between Brent ol prces and stock markets by dfferentatng ol-mportng (USA, Germany, and the Netherlands) and ol-exportng (Canada, Mexco, and Brazl) countres. Usng the multvarate DCC-GARCH approach, they fnd that the condtonal varances of ol and stock prces do not dffer for olmportng and ol-exportng economes. Tme-varyng correlatons depend on the orgn of the ol shocks: the response from aggregate demand-sde shocks s much greater than supply-sde shocks orgnated by OPEC s producton cuts. Two other papers document how peculer s the relatonshp between ol and stock markets n ol-mportng and ol-exportng countres. Cret et al. (2014) apply the frequency approach of the evolutonary co-spectral analyss allowng a tme-varyng dynamc correlaton measure between the stock market ndex and the ol prce seres. They fnd that nterdependence between the ol prce and the stock market s stronger n exporters markets than n mporters markets. Besdes, they conclude that ol does not play a role n counteractng the changng returns of a portfolo of stocks n any of the countres studed. To measure volatlty spllover between ol and stock markets, Guesm and Fattoum (2014) apply the multvarate GJR- DCC-GARCH models. They show that ol prce shocks n perods of global turmol or durng global busness cycle fluctuatons (downturn or expanson) appear to have a sgnfcant mpact on the relatonshp between ol and stock market prces, both n ol- mportng and ol-exportng countres. In exportng countres, ther analyss unvels hgher and multple peaks, whch concde wth major events (lke the 2008 ol prce crss). In the case of mportng countres, the pattern of nteracton s far smoother compared to exportng countres. The relatonshp between stock markets and ol prces has also been analyzed for the specfc case of Gulf Cooperaton Councl (or GCC) countres. Results seem qute controversal and vary wth the sample selecton and the tme span under nvestgaton. Some studes document postve co-movements, 4 others do not fnd sgnfcant relatonshps, 5 or asymmetrc ones. 6 To our knowledge, the present study s the frst to analyze partal fnancal ntegraton and the mpact of ol as rsk factor n a dynamc settng for the 4 major OPEC ol exporters, whch nclude an emergng country as Venezuela, and Mddle East countres such as Unted Arab Emrates, Saud Araba and Kuwat. Moreover, on the methodologcal vewpont, we add to the prevous lterature as we estmate the tme-varyng condtonal correlaton relatonshps among dfferent varables by employng the multvarate DCC-GJR-GARCH framework of of Tse and Tsu (2002). The DCC-GJR-GARCH framework has the followng advantages: () t nests other GARCH process that exsts n the lterature; () t s relatvely parsmonous compared wth other multvarate models found n the lterature. Addtonally, ths technque s sutable to account for asymmetres, whch are typcally observed n stock markets and ol prces. III. Emprcal strategy Our emprcal strategy conssts of consderng global and local factors (Bekaert and Harvey, 1995, Hardouvels et al., 2006, Carrer et al., 2007, Guesm and Nguyen, 66 Cret_Artcle.ndd Sec1:66 14/04/15 12:03

4 2011) for market ntegraton, ol rsk Fls et al. (2011), and exchange rate rsk (Adler and Dumas, 1988; Carrer et al., 2007; Ta, 2007). The excess return E t 1 ) ssued n country, condtonally on a set of nformaton avalable to nvestors up to tme t-1 s gven by: E t 1 ) = Ω t 1 δ m,t 1 Cov t 1, R mt ) L + δ p,t 1 Cov t 1, R pt )+ δ k,t 1 Cov t 1, R kt ) + k=1 + (1 Ω t 1 )δ,t 1 Var t 1 ) (1) where E t 1 ) s the excess return ssued n country, condtonally on a set of nformaton avalable to nvestors up to tme t-1. Ω t 1 s the condtonal probablty of transton between segmentaton and ntegraton states, whch falls wthn the nterval [0,1] and can be thus nterpreted as a condtonal measure of ntegraton of market nto the world market. Cov t 1 s the condtonal covarance between the securty r s return and the returns R mt on the world market portfolo. The varable δ mt 1 refers to the condtonally expected world prce of covarance rsk. The second term δ p,t 1 Cov t 1, R pt ) has a structure smlar the frst one n the RHS of equaton (1): t takes nto account the condtonal covarance between the securty r return and the rewards for ol prces rsk on the world market portfolo, that s R pt. In the thrd term, the sub ndex k denotes the currences of four countres that we consder: Emrates Unted Arab, Saud Araba, Kuwat, and Venezuela. Ths term ncludes the covarance wth the return on the exchange rate of the currency of country k, denoted by the varable R kt, aganst the currency of the reference country; δ k,t 1 expresses the expected prce of the exchange rate rsk for currency k, condtonally on the nformaton avalable up to t-1. L s the number of markets ncluded n the sample, that s 4. We consder the U.S. as the reference country and use blateral real exchange rates to measure the exchange value of local currences aganst the U.S. dollar. Fnally, Var t 1 ) s the condtonal varance of natonal market return. Equaton (1) can be wrtten by decomposng the rsk premum. More specfcally, the total rsk market premum (TRM) can be broken down nto two components. The frst component, called a global rsk market premum (GRM), conssts of world market rsk premum (WRM), ol rsk premum (ORM) and exchange rate rsk premum (ERM). It s weghted by the level of ntegraton Ω t 1. The second one, referred to as the local rsk premum (LRM), s weghted by the level of market segmentaton (1 Ω t 1 ). TRM = GRM + LRM =WRM + ORM + ERM + LRM δ m,t 1 Cov t 1, R mt ) GRM = Ω t 1 +δ p,t 1 Cov t 1, R pt ) L + δ k,t 1 Cov t 1, R kt ) k=1 LRM = (1 Ω t 1 )δ,t 1 Var t 1 ) (2) Let X m,t 1, X,t 1 and F,t 1 denote respectvely the vector of nternatonal nformaton varables, the vector of local nformaton varables and the vector of ntegraton varables, avalable at tme (t-1). The expected prces of rsk and the dynamcs of market ntegraton can then be modelled as δ k,t 1 = δʹ k X m,t 1 δ,t 1 = e ( γ ʹX,t 1 ) δ m,t 1 = e ( ʹ = e g ʹF,t 1 Ω t 1 δ m X m,t 1 ) (3) Under the hypothess of ratonal expectatons, the econometrc specfcaton of equaton (1) s characterzed by the followng system of equatons: c r r,t c r k,t c r p,t r c,t r = δ t 1 h rr,t + ε r,t = δ k t 1 h kk,t + ε k,t = δ k t 1 h pp,t + ε p,t = Ω t 1 r δ t 1 + (1 Ω 1 h r,t + δ p k t 1 h p,t + δ t 1 l k=1 h k,t d )δ,t 1 h,t + ε,t ; ε t ψ t 1 ~N (0, H t ) (4) = E (Unted Arab Emrates), S (Saud Araba), K (Kuwet), and V (Venezuela). where r c r,t, r c k,t, r c p,t and r c,t refer to the (10 1) vector of excess returns on regonal market, exchange rate, ol prce and local market respectvely. The varables r t are assumed to be normally dstrbuted and expressed n dollars. h r,t, h k,t, h p,t and h,t are, respectvely, the condtonal covarance between market s return and regonal market return, the condtonal covarance between market s return and exchange rate k s return, the condtonal covarance between market s return and ol prce and the condtonal varance of market, all beng ssued from the (10 10) varance-covarance matrx H t (5 equatons for excess returns on world market, 4 real exchange rate ndces and the Brent crude ol ndex). We model H t by usng a multvarate DCC-GARCH model 67 Cret_Artcle.ndd Sec1:67 14/04/15 12:03

5 as descrbed n Engle (2002) wth H t = D t R t D t ʹ. R t s the (10 10) symmetrc matrx of dynamc condtonal correlatons. D t s a dagonal matrx of condtonal standard devatons for each of the return seres, obtaned from estmatng a unvarate GJR-GARCH process of Glosten et al. (1993) n the equaton of varance expressed as 2 h,t = w + α ε,t β h,t 1 + γ I,t ε,t 1 (5) where persstence s measured by the coeffcents β and the ndcator varables I,t captures asymmetry n the estmate of coeffcents γ. A negatve value of γ mples that negatve resduals ncrease the varance more than postve resduals. Followng Bollerslev and Wooldrdge (1992), the parameters are estmated by quas-maxmum lkelhood (QML), assumng condtonal normally dstrbuted errors. Gven the hghly nonlnear structure of the model and the large number of parameters nvolved n estmaton, we estmate the model n two steps. We frst estmate a subsystem of fve equatons for excess returns on world market, four real exchange rate ndces and the Brent crude ol ndex. Ths stage allows us to obtan the condtonal varance of world market, real exchange rate ndces and Brent crude ol ndex, ther condtonal covarance s as well as the prces of world market, Brent crude ol ndex and exchange rate rsks. In the second stage, we estmate the prce of local market rsk and the tme-varyng level of ntegraton for each emergng market n the system (3). Ths strategy s also used by Bekaert and Harvey (1995), Hardouvels et al. (2006) and Guesm and Nguyen (2011; 2014). IV. Data Our dataset conssts of monthly tme seres related to stock market ndces, exchange rates, as well as global, regonal and local economc and fnancal varables. The use of monthly frequency s a common feature among the studes focusng on fnancal market ntegraton whch s a relatvely long-run phenomenon. Usng monthly data thus allows us not only to have suffcent data ponts to make relable statstcal nferences, but also to compare our results wth those of prevous studes. Also, as noted by Harvey (1991), monthly data help reduce potental bases that may arse from emergng market mperfectons such as the bd-ask effect and non-synchronous tradng days. Data are extracted from MSCI DataStream Internatonal. We can summarze the varables we use as follows: Global nstrumental varables: they are used to explan changes n the prces of nternatonal markets, Brent crude ol ndex and foregn exchange rsk. We choose the followng varables: the world market portfolo (Morgan Stanley Captal Internatonal-MSCI World ndex) n excess of the 30-day Eurodollar nterest rate whch s denoted by (IRENT), t he varaton n the US term premum (USTP) and the return on the S&P s 500 stock market ndex (RSP). Local nstr umental varables: They are used to nfer the changes n the local prce of rsk, nclude the return on the local stock market ndex n excess of the 30-day Eurodollar nterest rate (LRENT), and the varaton n the trade-weghted average local nflaton rate (INFRT). 7 Fnancal market ntegraton factors: A set of canddate factors that cause the movements n the degree of regonal fnancal ntegraton s chosen based on the fndngs of prevous studes (e.g., Bekaert and Harvey, 1997, 2000; Bhattacharya and Daouk, 2002). 8 Snce there s a numercal convergence problem at the estmaton stage when we have more than three unknown parameters, only three nformaton varables are used to explan the changes n fnancal ntegraton measure: the nterest rate spread (SWAY), the return of world market ndex (RENT) and the level of market openness (OPEN) of the market under consderaton. The specfcaton wth these three factors provdes best ft to the data as wtnessed by the AIC, BIC and Log-lkelhood crtera. Accordngly, the degree of market openness can be a potental factor n promotng fnancal ntegraton. The tme-varyng degree of market ntegraton s thus modeled as follows: 9 Ω t 1 = e α 0 +α 1 SWAY,t 1 +α 2 OPEN,t 1 +α 3 RENT,t 1. IV.2. MAIN STATISTICS AND STOCHASTIC PROPERTIES Table 2 reports the man statstcs of return seres for the stock market, real exchange rate and Brent crude ol ndces for all the countres n the sample. All the seres depart from normalty condtons and condtonal heteroscedastcty. The Unted Arab Emrates stock Market s the most volatle durng the studed perod n terms of standard devaton (12.67%), whle Kuwat s the least volatle (5.96%). The skewness coeffcents are postve for Unted Arab Emrates and Venezuela. They are sgnfcantly dfferent from zero for almost all markets, ndcatng the presence of asymmetry n the return dstrbuton. In addton, all the return seres are characterzed by a kurtoss coeffcent statstcally sgnfcant and greater than 3, and thus have fatter tals than those of a normal dstrbuton. Engle (1982) s test for the 1 st order of condtonal heteroscedastcty s also performed and we cannot reject the hypothess of no ARCH effects for all return seres consdered. Ths result motvates our choce of GARCH modelng approach for condtonal varance processes. V. Results V.1. PRICES OF WORLD MARKET, OIL PRICE AND FOREIGN EXCHANGE RISKS As dscussed above, we frst estmate the system (4) for excess returns on world market, Brent crude ol and 68 Cret_Artcle.ndd Sec1:68 14/04/15 12:03

6 returns on real exchange rates. The estmaton results and resdual dagnoss are reported n Table 3. Panel A and B present the estmated parameters for the prce of ol and world rsk respectvely. The prce of ol rsk s negatvely correlated wth IRENT and postvely wth RSP (both coeffcent beng sgnfcant at 1% level), confrmng the co-movement between ol and stock markets, on one sde, and dversfcaton opportuntes, on the other. Results of the Wald tests of nullty and constancy restrctons on the prce of ol prce rsk, reported n Panel D, clearly rejects the null hypotheses that the latter s equal to zero and constant, consstently wth the lterature (Fls and al., 2011). The prce of world market rsk also presents a negatve relatonshp both wth IRENT and RSP (n the two cases, the coeffcent are sgnfcant at 1% level). The prce of world market therefore ncreases as captal and stock markets are less effcent. Results of the Wald tests of nullty and constancy restrctons on the prce of world market rsk, reported n Panel D, clearly reject the null hypotheses that the latter s equal to zero and constant, whch confrms fndngs of prevous studes, such as Bekaert and Harvey (1995), and Carrer et al. (2007). Turnng out to the analyss of ol-exportng countres (Panel C), the coeffcents assocated wth the S&P s 500 ndex, and the coeffcents assocated wth the US term premum have nsgnfcant effect on the evoluton of the prce of exchange rate rsk. Ths latter s negatvely affected by the Eurodollar nterest rate IRENT (at 1% level). Venezuela presents a specfc pattern: ts exchange rate rsk s explaned (at 1%level) by the US term premum, as the Venezuelan economy s strongly dollarzed, and postvely wth the stock market, ndcatng a lkely relatonshp between the exchange rate and the stock markets, perhaps nduced by the mportance of ol exports n ths country wth respect to all other economc actvtes. We calculate the Wald test to nvestgate the null hypotheses that the prce of exchange rsk s zero and constant respectvely. The obtaned results, reported n Panel D and E of Table 3, ndcate rejecton of these null hypotheses at the 1% level for all markets consdered. These fndngs are coherent wth those of prevous studes, ncludng Carrer et al. (2007) and Ta (2007), n that the exchange rate rsk s a relevant factor of rsk for asset prcng n emergng markets, and that they change over tme. We fnally examne the hypotheses of jont nullty and constancy of all the four prces of exchange rate rsk and fnd evdence aganst ther valdty. Panel F of Table 2 presents a detaled analyss of the modelstandardzed resduals. Normalty s rejected at the 1% level for four currency returns. The departure from normalty decreases substantally for world returns, but t remans sgnfcant at the 1% level. The Ljung-Box test reveals that the frst-order autocorrelatons of the standardzed resduals are no longer sgnfcant, and ther values decrease substantally. The Engle (1982) s test for condtonal heteroscedastcty of the standardzed resduals ndcates that ARCH effects no longer exst n all cases, thus revealn g the sutablty of the GARCH approach. Although all the coeffcents n the multvarate DCC-GJR-GARCH process for condtonal varances and covarance are not reported, most of them are sgnfcant at the 1% and 5% levels. Ths result confrms the tme-varaton n both prces and quanttes of rsk as we have found based on Wald tests. Fgure 1 to 3 represent dynamc evoluton of exchange rsk prces for the countres under nvestgaton, world market and ol rsk, both for the estmates and the Hodrck Prescott fltered values. 9 These latter are substantally constant, wth varatons amplfyng n the last three years for all countres but Venezuela, whch has recently acheved a strong Real Effectve Exchange Rate (REER) stablty (World Bank, 2012). Fltered values of world and ol rsk prces reman stable over the sample, even though ol rsk prce estmates show skyrocketng values after V.2. TIME-VARYING WORLD MARKET INTEGRATION OF STOCK MARKET Table 4 reports the descrptve statstcs of our tmevaryng measure of market ntegraton, whch s obtaned by estmatng the whole system (4), whle mposng the estmates from the subsystem for world, ol prces and exchange rate returns. 10 Fgure 1 depcts the tme-paths of fnancal ntegraton measure (estmates and Hodrck Prescott fltered values), showng a cycle wth an upward trend startng n 2007 and reachng ts maxmum n The degree of nternatonal ntegraton dffers from one market to another, reflectng the heterogenety of the economc and monetary polces. Stock markets of the countres n the sample are well ntegrated nto world markets snce the ntegraton measures average between (Venezuela) and (Kuwat). Kuwat, Saud Araba and Venezuela ntegraton levels depend on openness and the nterest rate spread (at 1% level), ths latter varable affectng also market ntegraton of Unted Arab Emrates (at the same sgnfcance level), together wth the return of world market ndex (sgnfcant at 5% level). Fnally, the statstcal sgnfcance of the coeffcent assocated wth the degree of trade openness suggests that ths s the most mportant determnant of Venezuela degree of market ntegraton. The observed changes n the level of regonal ntegraton are crucal for the markets under consderaton, as ncreased fnancal ntegraton may also nduce adverse effects, beyond ts assocated benefts (.e. greater rsk dversfcaton, better captal allocaton and hgher economc growth potental). 11 Furthermore, the level of nternatonal ntegraton serves as the bass of all fnancal ssues relatng to asset prcng, determnaton of market rsk premum, and ol prce as well as polces for economc cooperaton between ol-exportng and ol-mportng countres. V.3. FORMATION OF TOTAL RISK PREMIUM AND MARKET INTEGRATION Table 5 reports average values of the total, the global, ol and local rsk premums. The two-sded Student-t test ndcates that global, ol and local rsk premums are sgnfcantly dfferent from zero at the 1% level for all the markets consdered. The Unted Arab Emrates has the hghest total rsk premum (9.277%), followed by Venezuela (8.362%), Saud Araba (7.339%), and Kuwat 69 Cret_Artcle.ndd Sec1:69 14/04/15 12:03

7 (6.132%). World rsk premums are on average greater than local premums for all markets. The share of local rsk premum over the total rsk premum ranges from 40% for Unted Arab Emrates to 88% for Saud Araba. Ol rsk represents a small part of the global rsk for all the countres. The most exposed country s the largest exporter, Unted Arab Emrates (0.988%). Ol rsk represents however only 10% of ts total rsk. Lookng at the dynamc evoluton of market ntegraton, Table 6 presents the results of the DCC-GJR-GARCH estmaton between stock markets of the 4 countres and the world market, whose evoluton s depcted n Fgure 2. The mean correlaton s negatve, meanng that n these countres stock market can be countercyclcal. These results suggest that, n general, condtonal correlatons overestmate the degree of global ntegraton of stock markets wth the world one durng certan perods, and underestmate t durng others. Condtonal correlatons also appear to be less stable than the ntegraton measure. The average values of dynamc ntegraton exceed those of market condtonal correlaton ndces for the Unted Arab Emrates (0.774 versus 0.052), and Venezuela (0.683 versus ). Overall, our results show that on average condtonal correlatons underestmate the level of ol-exportng countres market ntegraton. At the same tme, we fnd an ncrease n the level of market ntegraton durng crss perods, and a rsng trend at the end These results confrm the fndngs of several recent papers on ntegraton and suggest that ol exports do not counteract general trends observed for other emergng economes. 12 VI. Concluson The purpose of ths paper s to study the dynamcs of the global ntegraton process of stock markets of the bggest OPEC exporters nto the world market, whle takng nto account the role of ol prce as a potental source of rsk. We develop an nternatonal captal asset prcng model sutable for partally ntegrated markets, wth departs from purchasng power party, n lne wth Bekaert and Harvey (1995) s regme-swtchng approach. We thus explan tme-varatons n expected returns on stock market ndces. In ts fully functonal form, the model allows the market ntegraton measure as well as global, local and ol rsk premums to vary through tme. Future research however should take nto account the mpact of nsttutonal reforms amed at removng trade barrers whch affect captal movements. We fnd that the level of market ntegraton of the four major OPEC ol-exportng countres (Unted Arab Emrates, Saud Araba, Kuwat and Venezuela) vares wdely over tme and depends by the degree of trade openness and varaton n the US term premum. Even though ntegraton reaches farly hgh values durng several perods, and exhbts an upward trend towards the end of the estmaton perod, the stock markets under consderaton stll reman segmented from the world market. These results suggest that dversfcaton nto OPEC market assets contnue to produce substantal profts and that the asset prcng rules reflects a state of partal ntegraton. Venezuela s the less ntegrated country, gven that the development of the ol sector s deeper than the growth of ts fnancal actvtes and stock markets. Ol prce rsk does not affect to a hgh extent OPEC countres, whch reman protected by the earnngs due to ol export, stll representng the major source of ther wealth. 1 Accordng to OPEC (2012), for the latest avalable statstcs, n 2006, producton allocated to these countres s respectvely (n 1,000 barrel/day): Saud Araba (9,099), Venezuela (2,223), Unted Arab Emrates (2,444) and Kuwet (2,247). Table 1 n the Appendx llustrates the man 2011 facts and fgures relatve to these countres. 2 Stehle (1977) derves both a prcng model for an ntegrated state and a model for a segmented state. The frst model requres that an asset s expected return s a functon of the global systematc rsk, and the adjusted local systematc rsk, whch corresponds to the uncorrelated porton between the natonal and world market portfolos. Under the null hypothess of perfect ntegraton, the local beta should be zero. The prcng model n case of segmented markets s constructed n a smlar fashon, except that the roles of the local and global systematc rsks are reversed. 3 For a detaled revew of the lterature on ths topc, see Fls et al. (2011). 4 Arour and Rault (2011) study the mpact of ol prces shocks on GCC countres, wth a boostrap panel contegraton model, and provde evdence that the stock market performance of the Gulf markets s affected by postve ol prce shocks. Smlar results were also documented by Bashar (2006). Hammoudeh and Alesa (2004) nd spllovers from ol markets to the stock ndces of ol-exportng countres, ncludng Bahran, Indonesa, Mexco and Venezuela. 5 Al Janab et al. (2010) use bootstrap test for causalty to study non-normal fnancal data wth tme-varyng volatlty. They conclude that ol prces do not tend to affect these stock markets and thus ol prces cannot be used as predctors for the GCC stock markets. Hammoudeh et al. (2004) examnes the long-run nteracton between fve GCC stock markets (Bahran, Kuwat, Oman, Saud Araba, and UAE) and three global factors (ol spot prce ndces, US 3-month Treasury bll rate, and S&P ndex). They apply contegraton tests and VEC model to weekly data from February 1994 to December The authors fnd that ol prce movements do not have drect effects on any GCC stock markets. 6 Arour et al. (2012a) study sx GCC and, by usng a wde range of contegraton technques, they fnd that the relatonshp between ol and stock-prces s postve and evdent n the short-term, but not n the long-term. When causalty exsts, t runs from ol prces to stock markets n most cases. The effects of ol prce changes on stock returns n the GCC countres are asymmetrc: negatve ol prce changes have larger mpact on stock returns than postve ol prce changes. Asymmetrc effects are also found by Awartan and Maghyereh (2013), who nvestgate the dynamc spllover of return and volatlty between ol and equtes n the GCC countres durng the perod 2004 to See prevous studes of Bekaert and Harvey (1997, 2000), Bhattacharya and Daouk (2002) and Guesm and Nguyen (2011, 2014). 8 It encompasses the degree of market openness, the development of the local stock market, ndustral producton, nflaton rate, the short-term nterest rate, the nterest rate spread, the long-term nterest rate, the exchange rate volatlty, the economc growth rate, the current account defct, the local and regonal market returns, the local market dvdend yeld, the regonal market dvdend yeld, the world nterest rate, the world market return and the world market dvdend yeld. It s expected that these factors, beng mportant determnants of cross-border nvestment flows and nternatonal market convergence, have an explanatory power for the non-monotonous process of fnancal ntegraton. Notce that the currency exchange rate s a proxy for market openness, that s one of the drver of market ntegraton, ths latter beng one part of our nvestgaton purpose. Ol prce s quoted n dollar s used n our analyss as a source of rsk as t can be correlated to the stock market. 9 The Hodrck Prescott flter s a mathematcal tool to separate the cyclcal component of a tme seres from raw data. It s used to obtan a smoothed-curve representaton of a tme seres. 10 Most of the estmates of the ndvdual coeffcents on the local nformaton varables are sgnfcant, whch suggests tme-varaton n the local prces of rsk. They are avalable under request to the correspondng author. 11 For example, Levne and Zervos (1996), Stgltz (2002), and Bekaert et al. (2002), among others, document the ncreased fnancal nstablty as an mportant threat for emergng markets, due to external shocks and dspartes n trade wth developed countres. 12 Longn and Solnk (1995) show that correlatons of nternatonal stock markets vary over tme, whle Ang and Bekaert (1999) and Guesm and Nguyen (2011) detect an ncrease n correlatons durng perods of fallng markets and a reducton n the correlaton n perods of rsng markets. Other studes document that correlatons between nternatonal stock markets are hgher durng crss perods than durng normal perods (Kng and Wadhwan, 1990; Calvo and Renhart, 1995). Pukthuanthong and Roll (2009) demonstrate formally that the correlaton among ndex returns s an mperfect measure of fnancal ntegraton. Carrer et al. (2007) and Guesm and Nguyen (2011) also conclude that the correlaton of an emergng market s ndex returns wth the world market sgnfcantly underestmates the ntegraton ndex, whose estmaton s condtonal on real economc actvtes. 70 Cret_Artcle.ndd Sec1:70 14/04/15 12:03

8 References Adler M. and Dumas B. (1983), Internatonal Portfolo Selecton and Corporaton Fnance: A Synthess, Journal of Fnance, 38, Adler M. and Q R. (2003), Mexco s Integraton nto the North Amercan Captal Market, Emergng Markets Revew, 4, Al-Janab H.J.A. and Irandoust H. (2010), Modelng Tme-Varyng Volatlty and Expected Returns: Evdence from GCC and MENA Regons, Emergng Markets Trade and Fnance, 46(5), Ang A. and Bekaert G. (1999), Internatonal Asset Allocaton wth Tme-varyng Correlatons, NBER Workng Paper Avramov D. (2002), Stock Return Predctablty and Model Uncertanty, Journal of Fnancal Economcs, 64, Arour M. E. H., Joun J., Tuyen N. L and Nguyen D.C. (2012a), On the Relatonshp between World Ol Prces and GCC Stock Markets, Journal of Quanttatve Economs, Vol 10 (1), Arour M., Nguyen D.K. and Pukthuanthong K. (2012b), An Internatonal CAPM for Partally Integrated Markets: Theory and Emprcal Evdence, Journal of Bankng and Fnance 36, ), Arour M. E. H. and Rault C. (2011), Ol Prces and Stock Markets n GCC Countres: Emprcal Evdence from Panel Analyss, Internatonal Journal of Fnance and Economcs, do: /jfe.443 Awartan B. and Maghyereh A.I. (2013), Dynamc Spllovers between Ol and Stock Markets n the Gulf Cooperaton Councl Countres, Energy Economcs 36, Bashar A. Z. (2006), Wld Ol Prces, but Brave Stock Markets! The Case of GCC Stock Markets, Operatonal Research, 6(2), Awokuse T.O., Chopra A. and Bessler D.A. (2009), Structural Change and Internatonal Stock Market Interdependence: Evdence from Asan Emergng Markets, Economc Modellng, 26, Bekaert G. and Harvey C.R. (1995), Tme-varyng World Market Integraton, Journal of Fnance, 50, Bekaert G. and Harvey C.R. (2000), Foregn Speculators and Emergng Equty Markets, Journal of Fnance, 55, Bekaert G., Harvey C.R. and Lumsdane R. (2002), The Dynamcs of Emergng Market Equty Flows, Journal of Internatonal Money and Fnance, 21, Bekaert G., Harvey C.R. and Lumsdane R. (2005), Does Fnancal Lberalzaton Spur Growth?, Journal of Fnancal Ec onomcs, 77, Bekaert G., Hodrck R.J.R. and Zhang X. (2009), Internatonal Stock Return Comovements. Journal of Fnance, 64, Bhattacharya U. and Daouk H. (2002), The World Prce of Insder Tradng, Journal of Fnance, 57, Bollerslev T. (1990), Modelng the Coherence n Short-run Nomnal Exchange Rates: A Multvarate Generalzed ARCH Approach, Revew of Economcs and Statstcs, 72,Bollerslev T. and Wooldrdge J.M. (1992), Quas-maxmum Lkelhood Estmaton and Inference n Dynamc Models wth Tme Varyng Covarances, Econometrc Revews, 11, Burger J.D., Rebucc A., Warnock F.E. and Warnock V. (2010), External Captal Structures and Ol Prce Volatlty, IDB Workng Paper Seres, No. IDB- WP-107 Calvo S. and Renhart C. (1995), Captal flows to Latn Amerca: Is the re Evdence of Contagon Effects?, World Bank Polcy Research Workng Paper Seres. Carrer F. (2001), The Effects of Lberalzaton on Market and Currency Rsk n the European Unon, European Fnancal Management, 7, Carrer F., Errunza V. and Hogan K. (2007), Characterzng World Market Integraton through Tme, Journal of Fnancal and Quanttatve Analyss, 42, Chambet A. and Gbson R. (2008), Fnancal Integraton, Economc Instablty and Trade Structure n Emergng Markets, Journal of Internatonal Money and Fnance, 27, Chnn M. and Forbes K. (2003), A Decomposton of Global Lnkages n Fnancal Markets over Tme, NBER Workng Paper n Chou J.S. and Lee Y.H. (2009), Jump Dynamcs and Volatlty: Ol and the Stock Markets, Energy 34(6), Cho K. and Hammoudeh S. (2010), Volatlty Behavor of Ol, Industral Commodty and Stock Markets n a Regme-swtchng Envronment, Energy Polcy 38(8), Claessens S. and Rhee M. (1994), The Effects of Equty Barrers on Foregn Investment n Developng Countres, In: Frankel J.A. (Ed.), Internatonalzaton of Equty Markets, Unversty of Chcago Press, Chcago and London, Cret A., Joets M. and Mgnon V. (2013), On the Lnks between Stock and Commodty Markets Volatlty, Energy Economcs, Vol. 37, Cret A., Ftt Z. and Guesm K. (2014), Ol Prce and Fnancal Markets: Multvarate Dynamc Frequency Analyss, Energy Polcy, Vol. 73, Engle R. (1982), Autoregressve Condtonal Heteroskedastcty wth Estmates of the Varance of U.K Inflaton, Econometrca, 50, Engle R. (2002), Dynamc Condtonal Correlaton: A Smple Class of Multvarate GARCH models, Journal of Busness and Economc Statstcs, 20, Engle R. and Kroner K. (1995), Multvarate Smultaneous Generalzed ARCH, Econometrc Theory, 11, Errunza V. and Losq E. (1985), Internatonal Asset Prcng under Mld Segmentaton: Theory and Test, Journal of Fnance, 40, Fama E. and French K. (1992), The Cross-secton of Expected Stock Returns, Journal of Fnance, 47, Fls G., Degannaks S. and Floros C. (2011), Dynamc Correlaton between Stock Market and Ol Prces: The Case of Ol-mportng and Ol-exportng Countres, Internatonal Revew of Fnancal Analyss 20(3), Gerard B., Thanyalakpark K. and Batten J., (2003), Are the East Asan Markets Integrated? Evdence from the ICAPM, Journal of Economcs and Busness, 55, Cret_Artcle.ndd Sec1:71 14/04/15 12:03

9 References (contnued) Guesm K. and Nguyen, D N. (2011), How Strong s the Global Integraton of Emergng Market Regons? An emprcal assessment, Economc Modellng, 28, Guesm K. (2011), What Drve The Regonal Integraton of Emergng Stock Markets?, Economcs Bulletn, Vol. 31, n 3, Guesm K. and Nguyen D. N. (2014), Tme-varyng Regonal Integraton of Stock Markets n Southeast Europe, Appled Economcs, 46 (11), Guesm K. and Fattoum S. (2014), Return and Volatlty Transmsson between Ol Prces and Olexportng and Ol-mportng Countres, Economc Modellng, Vol. 38, Glosten L., Jagannathan R., and Runkle D. (1993), On the Relatonshp between the Expected Value and the Volatlty of the Nomnal Excess Returns on Stocks, Journal of Fnance, 48, Hammoudeh S., Dbooglu S. and Alesa E. (2004), Relatonshps among U.S. Ol Prces and Ol Industry Equty Indces, Internatonal Revew of Economcs and Fnance 13(4), Hammoudeh S. and Alesa E. (2004), Dynamc Relatonshp among GCC Stock Markets and NYMEX Ol Futures, Contemporary Economc Polcy, 22, Hardouvels G.A., Mallaropulos D. and Prestley R. (2006), EMU and European Stock Market Integraton, Journal of Busness 79, Harvey C.R. (1991), The World Prce of Covarance Rsk, Journal of Fnance, 46, Harvey C.R (1995), The Rsk Exposure of Emergng Equty Markets, World Bank Economc Revew, 9, Harvey J. L. and Strauss W.A. (1987), The New Dollar Indexes are no Dfferent from the Old Ones, Federal Reserve Bank of Chcago Economc Perspectves, Hodrck R. and Prescott E. (1997), Post war US Busness Cycles: A Descrptve Emprcal Investgaton, Journal of Money, Credt and Bankng 29 (1), IMF (2012), Saud Araba Country Survey. Jones C.M. and Kaul G. (1996), Ol and the Stock Markets, Journal of Fnance 51, Jong F. and Roon F.A. (2005), Tme-varyng Market Integraton and Expected Returns n Emergng Markets. Journal of Fnancal Economcs 78, Kng M. and Wadhwan S. (1990), Transmsson of Volatlty between Natonal Stock Markets, Revew of Fnancal Studes, 3, Kose M., Otrook C. and Whteman C. (2008), Understandng the Evoluton of World Busness Cycles, Journal of Internatonal Economcs, 75, Levne R. and Zervos S. (1998), Captal Control Lberalzaton and Stock Market Development, World Development, 26, Longn F. and Solnk B. (1995), Is the Correlaton n Internatonal Equty Return Constant: ?, Journal of Internatonal Money and Fnance, 14, Malk F. and Ewng B.T. (2009), Volatlty Transmsson between Ol Prces and Equty Sector Returns, Internatonal Revew of Fnancal Analyss 18, Marashdeh Hazem A. and Shrestha Mn B (2010), Stock Market Integraton n the GCC Countres, Internatonal Research Journal of Fnance and Economcs, 37, OPEC (2012) Annual Statstcal Bulletn, OPEC Publcatons. Papapetrou E. (2001), Ol Prce Shocks, Stock Market, Economc Actvty and Employment n Greece, Energy Economcs 23(5), Park J. and Ratt R.A. (2008), Ol Prce Shocks and Stock Markets n the U.S. and 13 European Countres, Energy Economcs 30, Pukthuanthong K. and Roll R. (2009), Global Market Integraton: An Alternatve Measure and ts Applcaton, Journal of Fnancal Economcs, 94, Rajan R. and Zngales L. (2003), The Great Reversals: The Poltcs of Fnancal Development n the 20th Century, Journal of Fnancal Economcs, 69, Reboredo J. C. and Rvera-Castro M. A. (2014), Wavelet-based Evdence of the Impact of Ol Prces on Stock Returns, Internatonal Revew of Economcs & Fnance, 29, Sadorsky P. (1999), Ol Prce Shocks and Stock Market Actvty, Energy Economcs 21, Stehle R. (1977), An Emprcal Test of the Alternatve Hypotheses of Natonal and Internatonal Prcng of Rsky Asset, Journal of Fnance, 33, Stgltz J. E. (2002), Informaton and the Change n the Paradgm n Economcs, Amercan Economc Revew, 92, Ta C.-S. (2007), Market ntegraton and contagon: Evdence from Asan Emergng Stock and Foregn Exchange Markets, Emergng Markets Revew, 8, Tse Y. K. and Tsu Albert K.C. (2002), A Multvarate Generalzed Autoregressve Condtonal Heteroscedastcty Model wth Tme-Varyng Correlatons, Journal of Busness and Economc Statstcs, Amercan Statstcal Assocaton, 20(3): Yang C.W., Hwang M.J. and Huang B.N. (2002), An Analyss of Factors affectng Prce Volatlty of the US Ol Market, Energy Economcs, 24, Cret_Artcle.ndd Sec1:72 14/04/15 12:03