How do oil prices affect stock returns in GCC markets? An asymmetric cointegration approach.

Transcription

1 How do oil prices affect stock returns in GCC markets? An asymmetric cointegration approach. Mohamed El Hedi AROURI (LEO-Université d Orléans & EDHEC, mohamed.arouri@univ-orleans.fr) Julien FOUQUAU (ESC Rouen & LEO, julien.fouquau@groupe-esc-rouen.fr) 2009 Abstract This paper investigates the existence of long-term relationships between oil prices and GCC stock markets. Since GCC countries are major world energy market players, their stock markets are likely to be susceptible to oil price shocks. To account for the fact that stock markets may respond asymmetrically to oil price shocks, we propose an approach based on asymmetric cointegration. The results of the empirical analysis show that oil price shocks indeed affect the stock returns in an asymmetric fashion in GCC countries. Key words: GCC stock markets, oil prices, asymmetric cointegration analysis JEL classifications: G12, F3, Q43. 1

2 1. Introduction There has been a large volume of studies on linkages between oil prices and macroeconomic variables. Most of these studies have established significant effects of oil price changes on economics activity for several developed and emerging countries [Cunado and Perez de Garcia (2005), Balaz and Londarev (2006), Gronwald (2008), Cologni and Manera (2008) and Kilian (2008)]. Furthermore, some papers have shown that the link between oil and economics activity is not entirely linear and that negative oil price shocks (price increases) tend to have larger impact on growth than positive shocks do [Hamilton (2003), Zhang (2008), Lardic and Mignon (2006, 2008) and Cologni and Manera (2009)]. In sharp contrast to the volume of work investigating the link between oil price shocks and economics activity, there has been less work done on the relationship between oil price variations and stock markets (For more details see Sadorsky, 2008). Furthermore, most of these works have focused on few industrial countries (US, Canada, Europe and Japan). Concerning emerging markets, very few studies have been carried out. These studies have mainly focused on the short term relationship between energy price shocks and stock markets. One rationale of using oil price fluctuations as a factor affecting stock prices is that value of stock in theory equals discounted sum of expected future cash-flows. These cash-flows are affected by macroeconomic events that possibly can be influenced by oil shocks. Thus, oil price changes may influence stock prices. Most of previous studies have investigated this relationship within the framework of a macroeconomic model using data from net oil importing countries. Using new asymmetric cointegration tests, this article investigates the long-term relationship between oil price shocks and stock markets in the Gulf Cooperation Council (GCC) countries. Studying the link between oil prices and stock markets in GCC countries is interesting for several reasons. First, GCC countries are major suppliers of oil in world energy markets, their stock markets may be susceptible to change in oil prices. Second, the GCC markets differ from those of developed and from other emerging countries in that they are largely segmented from the international markets and are overly sensitive to regional political events. Finally, GCC markets represent very promising areas for regional and world portfolio diversification. Studying the influence of oil price shocks on GCC stock market returns is important for investors to make necessary investment decisions and for policy-makers to regulate stock markets more effectively. All these reasons highlight the interest of a study centred on GCC countries. The pioneer paper by Jones and Kaul (1996) tests the reaction of international stock markets (Canada, UK, Japan and US) to oil price shocks on the base of a standard cash-flow dividend valuation model. They found that for the US and Canada this reaction can be accounted for entirely by the impact of the oil shocks on cash-flows. The results for Japan and the UK were inconclusive. Using an unrestricted vector autoregressive (VAR), Huang et al. (1996) show a significant link between some American oil company stock returns and oil price changes. However, they find no evidence of a relationship between oil prices and market indices such as the S&P500. In contrast, Sadorsky (1999) applies an unrestricted VAR with GARCH effects to American monthly data and shows a significant relationship between oil price changes and aggregate stock returns in US. In particular, he shows that oil price shocks have asymmetric effects: positive oil shocks have a greater impact on stock returns and economy activity than do negative oil price shocks. Relying on nonlinear causality tests, Ciner (2001) provides evidence that oil shocks affect in a non linear manner stock returns in US, which is consistent with the documented influence of oil on economics activity. Recently, some papers have tented to focus on major European, Asian and Latin American emerging markets. The results of these studies show a significant short term link between oil price changes and emerging stock markets. Using a VAR model, Papapetrou (2001) shows a 2

3 significant relationship between oil price changes and stock markets in Greece. Basher and Sadorsky (2006) reach the same conclusion for other emerging stock markets using an international multifactor model. However, less attention has been given to smaller emerging markets, especially in the GCC countries where share dealing is a relatively recent phenomenon. Using VAR models and cointegration tests, Hammoudeh and Eleisa (2004) show that there is a bidirectional relationship between Saudi stock returns and oil prices changes. The findings also suggest that the other GCC markets are not directly linked to oil prices and are less dependent on oil exports and more influenced by their own domestic factors. Bashar (2006) employs VAR analyses to study the effect of oil price changes on GCC stock markets, and shows that only Saudi and Oman markets have predictive power of oil price increase. More recently, Hammoudeh and Choi (2006) have examined the long-run relationship among the GCC stock markets in the presence of the US oil market, the S&P 500 index and the US treasury bill rate. They have found that the treasury bill rate has direct impact on these markets, while oil and S&P 500 have indirect effects. As we can see, the results of the few available works on GCC countries are too heterogeneous. These results are puzzling because the GCC countries are strongly oil exporters and are similar in their economic structure. Furthermore, the GCC economies are oil dependent and thus are sensitive to oil price changes. The conclusions of previous works could be due to the fact that the tests they rely on are not powerful enough to detect nonlinear linkages. As we have mentioned above, recent papers argue that there is an asymmetric relationship between oil prices and the economics activity. This suggests that asymmetric linkages between oil prices and the stock market could be uncovered. The current article extends the understanding of the relationship between oil prices and stock markets in GCC by testing for asymmetric long term linkages, in addition to linear linkages. The rest of the paper is organized as follows. Section 2 presents briefly the GCC markets and the role of oil. The methodology is introduced in section 3. Section 4 presents the data and discusses the empirical results. Summary conclusions and policy implications are provided in Section GCC stock markets and oil Table 1 presents some key financial indicators for the stock markets in GCC countries. The GCC was established in 1981 and it includes six countries: Bahrain, Oman, Kuwait, Qatar, Saudi Arabia and United Arab Emirates (UAE). GCC countries have many common patterns. Together, they produce about 20% of all world oil, control 36% of world oil exports and possess 47% of the world oil proven. Oil exports largely determine earnings, governments budget revenues and expenditures and aggregate demand. The contributions of oil to GDP range from 22% in Bahrain to 44% in Saudi Arabia. Moreover, Table 1 indicates that for the three largest economies in GCC countries, Saudi Arabia, UAE, and Kuwait, the stock market s liquidity indicator is positively associated with the oil importance indicator in these economies. Table 1: Stock markets in GCC countries in 2007 Market Number of companies* Market Capitalization ($ billion) Market Capitalization (% GDP) * Oil (% GDP)+ Bahrain Kuwait Oman Qatar UAE S. Arabia Sources: Arab Monetary Fund and Emerging Markets Database. * numbers in

4 Furthermore, GCC countries are importers of manufactories products from developed and emerging countries. Therefore, oil price fluctuations can indirectly impact GCC markets through their influence on prices of imported products and then increases in oil prices are often indicative of inflationary pressure in the GCC economies which in turn could indicate the future of interest rates and investment of all types. Thus, oil price fluctuations should affect corporate output and earnings, domestic price levels and stock market share prices in GCC countries. However, unlike net oil importing countries where the expected link between oil prices and stock markets is negative, the transmission mechanism of oil price shocks to stock market returns are ambiguous and the total impact of oil price shocks on stock returns depends on which of the positive and negative effects counterweigh the other. Figure 1 : GCC countries oil dependency Source: International Monetary Fund. Saudi Arabia leads the region in terms of market capitalization. However, in comparison to each country s GDP, Qatar is the leader. Stock market capitalization exceeded GDP for all counties except Oman. In terms of the number of listed companies, Kuwait is the leading market followed by Oman. Overall, GCC stock markets are limited by many structural and regulatory weaknesses: relatively small numbers of listed firms, large institutional holdings, low sectoral diversification, and several other deficiencies remain. However, in recent years a broad range of legal, regulatory, and supervisory changes has increased market transparency. More interestingly, GCC markets begin improving their liquidity and opening their operations to foreign investors. Recently, in March 2006 the Saudi authorities lifted the restriction that limited foreign residents to dealing only in mutual investment funds and the other markets have progressively followed. 1 Finally, even if GCC countries have several economic and political characteristics in common, they have different oil dependence degrees and efforts to diversify and liberalize the economy. In particular, each country in GCC depends on oil to a different degree. For example, UAE and Bahrain are less oil-dependent than Saudi Arabia and Qatar (Figure 1). Thus, comparison between GCC stock markets forms an interesting subject. 1 For interested readers, further information and discussions about market characteristics and financial sector development of these countries can be found in Creane et al. (2004), Neaime (2005), and Naceur and Ghazouani (2007). 4

5 3- Methodology The aim of this article is to test for long term relationships between oil prices and stock markets in GCC countries. We test for both linear and asymmetric cointegration. For linear cointegration, we use the traditional approach. To prove the presence of asymmetric cointegration, we need to follow a three-step process [for more details about this methodology see Schorderet (2004) and Lardic and Mignon (2006, 2008)]. In a first time, it is necessary to decompose a time series in positive and negative parts. According to Schorderet (2004) 2, by distinguishing its positive and negative increments, any time series yt T t 0 can be broken down into its initial value and its negative and positive cumulative sums: yt y0 yt yt (1) where y 0 is the initial value and t yt 1 1 y t i 0 yt i (2) i0 and t yt 1 1 y t i 0 yt i (3) i0 1 is the indicator function taking 1 if the event in brackets occurs and zero otherwise.. In a second step, we consider now two nonstationary time series y 1 t and y 2 t. We suppose that there is no traditional cointegration but a linear combination z t of their components such that: zt 1 y1 t 2 y1t 3 y2t 4 y2t (4) If there is a vector ( 1, 2, 3, 4) with 0, 1 2 and 3 4 such as z t is a stationary process, the time series y 1 and y 2 are said to be asymmetrically cointegrated. t t Other alternative, suppose that only one component of each series appears in the cointegrating relationship (4). This may be seen as a cointegration relation which operates only in one direction: or y1t y2t z1t t 1,, T (5) y1t y2t z2t t 1,, T (6) Because of nonlinear characteristics of z t, OLS estimation of (5) and (6) is likely to be biased in finite sample and the usual techniques of statistical inference are misspecified. Schorderet (2004) suggests estimating using OLS the auxiliary models (7) and (8). As proven 2 The asymmetric cointegration approach of Schorderet (2004) we used in this paper have been applied in several recent studies, see for instance Lardic and Mignon (2006, 2008) and Shen et al. (2007). 5

6 by West (1988), under fairly general conditions the OLS estimation of (7) and (8) is asymptotically normal and statistical inference can then proceed in the usual way: or y1t y1t y2t 1t y1t y1t y2t 2t (7) (8) To test for no cointegration against asymmetric cointegration, the traditional Engle and Granger procedure can be applied to (7) and (8) instead of (5) and (6). 3 Thus, in order to test for asymmetric cointegration between oil prices and stock prices in GCC countries, we can estimate the following long-term relation: LStockt LStockt 0 1 LOilt 1 t (9) LStockt LStockt 0 1 LOilt 2t (10) where LStock t, LStock t, LOilt and LOil t are respectively the positive and the negative component of logarithms of stock price indices and the oil prices as defined in (2) and (3). Traditional cointegration tests can be applied to (9) and (10). 4- Data and Empirical Results 4.1- Data Our goal is to investigate the existence of a long-term relationship between oil prices and stock markets in GCC countries. We use monthly data obtained from Arab Monetary Fund (AMF) over the period January 1996 December Note that stock exchanges in UAE and Qatar are newly established and did not participate in the AMF database when it began in Thus, we study only four of the six GCC stock markets: Bahrain, Kuwait, Oman and Saudi Arabia. 4 As a robustness test, we equally study the six GCC stock markets using data from Morgan Stanley Capital International (MSCI). Data are available from June 2005 for the six GCC markets at different frequencies. We use weekly data which may adequately capture relationship between oil and stock prices in the region. We do not use daily data in order to avoid time difference problems with international markets. In fact, the equity markets are generally closed on Thursdays and Fridays in GCC countries, while the developed and international oil markets close for trading on Saturdays and Sundays. Furthermore, for the common open days, the GCC markets close just before US stocks and commodity markets open. Accordingly, we opt to use weekly data and choose Tuesday as the weekday for all variables because this day lies in the middle of the three common trading days for all markets. For oil, we use the OPEC countries spot price weighted by estimated export volume obtained from the Energy Information Administration (EIA). OPEC prices are often used as 3 According to Schorderet (2004), various extensions of models (7) and (8) are possible. For instance, an intercept can be added. 4 Data for 2008 are not available in AMF database. Furthermore, weekly data are not available. 6

7 reference prices for crude oil including oil produced by GCC countries. 5 All prices are denominated in American dollar. Next we discuss results obtained using monthly data from AMF database. Results obtained using MSCI weekly data are reported in Appendix Unit root tests In the first step of our analysis, we have to test for the presence of unit roots in the oil price and stock market series in logarithm. To this aim, we apply alternatively three standard tests: the Augmented Dickey Fuller (ADF), the Phillips Perron (PP) and the Kwiatkowski et al. (1992) (KPSS) test. Contrary to the two first tests 6, the KPSS test has the advantage to be based on the null hypothesis of series stationnarity. In performing an ADF test, we will face two practical issues. First, we have the choice 7 to include a constant, a constant and a linear time trend, or neither in the test regression. Indeed, including irrelevant regressors in the regression will reduce the power of the test to reject the null of a unit root. Then, we apply a fisher test strategy to select the optimal model. Under the null hypothesis, the critical values of this test are not standard but they have been computed by Dickey Fuller (1981, p.1063). Second, we will have to specify the number of lagged difference terms of the dependent variable to be added to the test regression. The usual advice is to include a number of lags sufficient to remove serial correlation in the residuals. In respect of it, we retain the number of lags that minimize information criteria. Table 2: Unit root tests Levels First difference ADF PP KPSS ADF PP KPSS LOil 1.49a -2.12c 1.19***b -11.4***a -11.4***a 0.14b LStock Bahrain 0.69a -8.88***c 1.01***b -9.41***a -41.9***a 0.30b LStock Kuwait 2.02a -1.56c 1.12***b -9.38***b -9.61***b 0.28b LStock Oman 0.29 a -1.72b 0.28b -11.6***a -11.6***a 0.09b LStock Saudi 1.18a -1.94c 1.09***b -12.8***a -12.8***a 0.12b Notes : All variables are in natural logs. ADF is the Augmented Dickey-Fuller test, PP the Phillips-Perron test and KPSS the Kwiatkowski- Phillips-Schmidt-Shin test. (a) model without constant nor deterministic trend, (b) model with constant without deterministic trend, (c) model with constant and deterministic trend. *, ** and *** denote rejection of the null hypothesis respectively at 10%, 5% and 1%. To specify the PP or KPSS tests, we have to select the regression form to test as in the ADF test 8. Then, we must choose the kernel and the bandwidth parameter needed for estimating the residual spectrum at frequency zero. The usual solution is to use Bartlett kernel and Newey West (1994) data-based automatic bandwidth parameter method. In Table 2, we report the results of the three unit root tests. The majority of the tests give the same conclusion that the series are integrated of order (1) and in all the cases at the 1% significance level. The two exceptions are the PP test for Bahrain and KPSS test for Oman in level. Since the other tests give the same conclusion of unit root, we can however conclude that all series are integrated of order (1). 5 Very similar results are obtained with West Texas Intermediate and Brent spot prices. Oil prices are in US dollar per barrel. Note also that GCC currencies are pegged at fixed rates to the U.S. dollar officially since The ADF and PP test are based on the null hypothesis of a unit root. 7 The specification choice in ADF or PP tests doesn t modify the result except for the Oman stock markets series in level in which we find unit root presence for other specifications. These results are available upon request. 8 Two alternative test regressions are only possible in KPSS test: inclusion of a constant or a constant and a trend. We choose the same or the close form that in ADF test regression. 7

8 4.3- Traditional cointegration tests In the second step, we proceed to standard cointegration tests between the oil prices and stock markets series. In this respect, we begin by applying the Engle and Granger (1987) testing. This leads to estimate the following relation (11) by OLS and to test the unit root hypothesis in the residuals: LStockt LOil t 0 1 t (11) By definition, there is cointegration if the residual sequence t is stationary. To this end, we use the same strategy that we have applied in the unit root tests: ADF test and PP test. To confirm our results, we have equally applied cointegration tests using the methodology developed in Johansen (1991). The null hypothesis in Johansen test is the absence of a cointegration relationship. Results are reported in Table 3. Table 3: Traditional cointegration tests ADF PP Johansen LStock Bahrain -4.88*** -8.41*** LStock Kuwait LStock Oman LStock Saudi Notes: ADF is the Augmented Dickey-Fuller test, PP the Phillips-Perron test and Johansen the trace statistics. *, ** and *** denote rejection of the null hypothesis respectively at 10%, 5% and 1%. The long-term relation is (11). We apply unit root tests on residual series t. For Bahrain, the cointegration hypothesis is not rejected according to ADF and PP tests. For all the other countries, the findings indicate that residual series are nonstationary and thus that oil prices and GCC stock markets are not cointegrated. Results reported in appendix 1 indicate that we reach the same conclusions for the six GCC countries when we use weekly data from MSCI. The absence of long term relationships between oil prices and stock markets in too oil dependent economies such as GCC countries seems to be counterintuitive. However, as we have previously mentioned, some recent papers have shown that the link between oil and economics activity is not entirely linear and that there is strong evidence of asymmetric relationships between the two variables [Hamilton (2003), Zhang (2008) and Lardic and Mignon (2006, 2008)]. Therefore, one possible explanation for our findings is that the traditional cointegration tests are too restrictive and cannot reproduce such asymmetric long term relationships. In the rest of the paper, we test for asymmetric cointegration between oil prices and stock markets in GCC countries Asymmetric cointegration tests In the last stage, we are looking at whether the decomposition of a time series into its positive or negative partial sum alters the conclusions of the traditional cointegration test. In other words, is there an asymmetric cointegration relationship between the oil prices and stock markets in GCC countries? As shown by Shorderet (2004), standard unit root and cointegration tests can be applied to check for asymmetric cointegration. Then, we have to estimate the equations (9, 10) by OLS and to test for the presence of unit roots in the residual sequences 1t and 2t. As before, we follow the same methodology and equally the ADF, PP and Johansen tests. The results of 8

9 cointegration tests are reported in Table 4 for equation (9) and in Table 5 for the equation (10). Table 4: Tests for asymmetric cointegration (tests on ADF PP Johansen LStock Bahrain -2.72* -3.49*** 33.8*** LStock Kuwait * 28.4*** LStock Oman * 30.6*** LStock Saudi *** Notes: ADF is the Augmented Dickey-Fuller test, PP the Phillips-Perron test and Johansen the trace statistics. *, ** and *** denote rejection of the null hypothesis respectively at 10%, 5% and 1%. The long-term relation is (10). We apply unit root tests on residual series 1t Consider first tests on 1t. For Bahrain, the ADF, the PP and Johansen tests find cointegration between oil prices and stock markets. Little evidence of cointegration is also found for Kuwait, Oman based on PP test and Johansen tests based on the trace statistic test. For the other cases, the used tests find no cointegration between the two variables. Last point, only the Johansen tests find cointegration for Saudi. However, we should be careful with this result since the Johansen tests have always concluded in cointegration hypothesis. Next, turn to tests on 2t. There exists a stronger evidence of asymmetric cointegration than in the negative component except for the Kuwait. For Oman, the three tests used conclude to a cointegration relationship between oil prices and stock markets, and two tests for Bahrain and Saudi. In these two cases, the results are slightly different to those obtain previously with the negative component. Furthermore, note that appendix 1 indicates that we reach similar conclusions for the six GCC countries using weekly data from MSCI. Table 5: Tests for asymmetric cointegration (tests on 1t ) 2t ) ADF PP Johansen LStock Bahrain ** 36.8*** LStock Kuwait ** LStock Oman -2.66* -4.67*** 39.9*** LStock Saudi ** 28.9*** Notes: ADF is the Augmented Dickey-Fuller test, PP the Phillips-Perron test and Johansen the trace statistics. *, ** and *** denote rejection of the null hypothesis respectively at 10%, 5% and 1%. The long-term relation is (10). We apply unit root tests on residual series 2t As we expected, there is evidence of asymmetric cointegration for all GCC countries at least in the positive or in the negative components. Thus, there are asymmetric stable long term relationships between oil prices and stock markets in GCC countries, especially when oil prices increase Long term relationships Finally, we estimate the long term asymmetric relationships (equations 9 and 10). The results are summarized in Tables 6 and 7. T-statistics reported are computed based on corrected standard errors (due to the presence of serial correlation in the residuals, Schorderet (2004)). Following West (1988), consistent estimates of the standard deviations can be obtained by 1 scaling the OLS statistics by ˆ 2 2 s where 2 is the usual OLS standard error of regression and s is an estimator of the spectral density at frequency zero of the regression disturbance. 9

10 As we expected, the adjusted R squared are higher for the equation (10) than for equation (9) confirming the strong evidence of asymmetric cointegration obtained between the positive components of oil prices and stock markets in all GCC countries. Table 6: Long term relationships between oil prices and stock markets in GCC (equation (9)) 0 LStock Bahrain (13.5) LStock Kuwait (-10.6) LStock Oman (4.8) LStock Saudi (10.1) 1 Adjusted (44.3) (114.8) (95.1) (78.2) 2 R S.E of resid Notes: t-satistics corrected as in West (1988) are reported in parentheses. S.E of resid is the residual standard error and Adjusted 2 R is the adjusted R-squared. The coefficients 1 are significantly positive, suggesting that, in the long term, oil price and stock market changes move in the same direction in GCC countries. More interestingly, 1 and 1 are different. The elasticity of stock price changes to oil price decreases (Table 6) varies from (Bahrain) to (Kuwait). As for positive components, the coefficients 1 (Table 7) vary from (Bahrain) to (Kuwait). 1 are larger than 1, which implies that oil price decreases have more impact on stock markets in GCC countries than oil price increases. Taken together, our results confirm the asymmetric reaction of GCC stock markets to oil price changes. Table 7: Long term relationships between oil prices and stock markets in GCC (equation (10)) 0 LStock Bahrain (-11.8) LStock Kuwait (-6.49) LStock Oman (17.1) LStock Saudi (20.1) 2 S.E of resid 1 Adjusted R (58.2) (148.?) (14.?0) (149.?) Notes: t-satistics corrected as in West (1988) are reported in parentheses. S.E of resid is the residual standard error and Adjusted 2 R is the adjusted R-squared. 5- Conclusion and policy implications This paper extends the understanding of linkages between oil prices and stock markets in GCC countries. Since GCC countries are major world energy market players, their stock markets are likely to be susceptible to oil price shocks. We test for classic and asymmetric cointegration in order to focus on both linear and nonlinear dependencies. In fact, recent papers argue that there is an asymmetric relationship between oil prices and the economics activity. This suggests asymmetric linkages between oil prices and the stock markets. The 10

11 results of our empirical analysis show that oil price shocks indeed affect the stock index returns in an asymmetric fashion. Our findings should be of interest to researchers, regulators, and market participants. In particular, GCC countries as policy makers in OPEC should keep an eye on the effects of oil price fluctuations on their own economies and stock markets. For investors, the significant relationship between oil prices and stock markets imply some degree of predictability in the GCC stock markets. In particular, portfolio diversification, speculation, arbitrage and hedging strategies have to be built differently when one expects increase or decrease in oil prices. The findings of this article offer several avenues for future research. First, the link between oil and stock markets in GCC countries can be expected to vary across different economic sectors. A sectoral analysis of this link would be informative. Second, evidence from international equity markets should be produced to examine the robustness of the findings. Third, the methodology applied in this article can be used to examine the effects of other energy products such as gaz. Finally, further research could compare causality between oil and stock markets in GCC countries and other oil exporting countries. 11

12 References: Balaz, P., Londarev, A., (2006) Oil and its position in the process of globalization of the world economy. Politicka Ekonomie, 54 (4), Bashar, Z. (2006), Wild oil prices, but brave stock markets! The case of Gulf Cooperation Council (GCC) stock markets, Middle East Economic Association Conference, Dubai. Basher, S. A. and P. Sadorsky (2006) Oil Price Risk and Emerging Stock Markets, Global Finance Journal, 17, Ciner, C. (2001). Energy Shocks and Financial Markets: Nonlinear Linkages. Studies in Non-Linear Dynamics and Econometrics, 5, Cologni, A. and M. Manera (2008). Oil prices, inflation and interest rates in a structural cointegrated VAR model for the G-7 countries. Energy Economics, 30, Cologni, A. and M. Manera (2009). The Asymmetric Effects of Oil Shocks on Output Growth: A Markov-Switching Analysis for the G-7 Countries, Economic Modelling, 26, Creane, S., Goyal, R., Mobarak, A.M. and Sab, R. (2004), Financial sector development in the Middle East and North Africa, Working Paper n 04/102, IMF, Washington, DC. Cunado, J., Perez de Garcia, F., Oil prices, economic activity and inflation: evidence for some Asian countries. The Quarterly Review of Economics and Finance 45 (1), Dickey, D.A., and W.A. Fuller (1981), Likelihood Ratio Statistics for Autoregressive Time Series with a Unit Root, Econometrica, 49(4): Engle, R.F., and C.W.J. Granger (1987), Co-integration and Error Correction: Representation, Estimation, and Testing, Econometrica 55(2): Gronwald, M. (2008). Large oil shocks and the US economy: Infrequent incidents with large effects, Energy Journal, 29, Hamilton, J. D. (2003). What is an Oil Shock? Journal of Econometrics, 113, pp Hammoudeh, S. and Aleisa, E. (2004) Dynamic relationship among GCC stock markets and NYMEX oil futures, Contemporary Economic Policy, Vol. 22, pp Hammoudeh, S. and K. Choi, "Behavior of GCC Stock Markets and Impacts of US Oil and Financial Markets," Research in International Business and Finance, Vol. 20 no. 1,pp Huang, R. D., Masulis, R. W., and H. R. Stoll (1996). Energy shocks and financial markets, Journal of Futures Markets, 16, Johansen S. (1991), Estimation and Hypothesis Testing of Cointegration Vectors in Gaussian Vector Autoregressive Models, Econometrica, 59, pp Jones, C.M., Kaul, G. (1996), Oil and the Stock Markets, Journal of Finance, vol. 51, n 2, pp Kilian, L., (2008). Exogenous Oil Supply Shocks: How Big Are They and How Much Do They Matter for the US Economy? Review of Economics and Statistics, 90, Kwiatkowski, D., Phillips, P. and Y. Shin, (1992), Testing for the Null Hypothesis of Stationarity against the Alternative of a Unit Root, Journal of Econometrics 54, Lardic S. and Mignon V. (2006), "The impact of oil prices on GDP in European countries: An empirical investigation based on asymmetric cointegration", Energy Policy, vol. 34(18), pp Lardic S. and Mignon V. (2008), "Oil prices and economic activity: An asymmetric cointegration approach", Energy Economics, vol. 30(3), pp Naceur, S.B. and Ghazouani, S. (2007), Stock markets, banks, and economic growth: empirical evidence from MENA region, Research in International Business and Finance, Vol. 21, pp Neaime, S. (2005), Financial market integration and macroeconomic volatility in the MENA region: an empirical investigation, Review of Middle East Economics and Finance, Vol. 3, pp Newey, W. K., and K. D. West (1994). Automatic Lag Selection in Covariance Matrix Estimation, Review of Economic Studies, vol. 61, no 4, pp

13 Papapetrou, E. (2001). Oil Price Shocks, Stock Market, Economic Activity and Employment in Greece, Energy Economics, 23, Sadorsky, P. (1999), Oil Price Shocks and Stock Market Activity, Energy Economics, vol. 2, pp Sadorsky, P. (2008). Assessing the impact of oil prices on firms of different sizes: Its tough being in the middle, Energy Policy, 36, 10, Schorderet, Y. (2004), Asymmetric Cointegration, Working Paper, Department of Econometrics, University of Geneva. Shen, C.H., Chen C.F. and Chen, L.H. (2007), An empirical study of the asymmetric cointegration relationships among the Chinese stock markets, Applied Economics, vol. 39, n 11,, pp (13); West K.D. (1988), Asymptotic normality when regressors have a unit root, Econometrica, 56, Zhang D. (2008), Oil shock and economic growth in Japan: A nonlinear approach, Energy Economics, vol. 30, issue 5, pages pp

14 Appendix 1: Results obtained for the six GCC countries using MSCI weekly data from the first week of June 2005 to the third week of October 2008 Table A1: Unit root tests Levels First difference ADF PP KPSS ADF PP KPSS LOil 0.24a 0.49a 1.30***b -8.35***a -8.26***a 0.17b LStock Bahrain -0.33a -0.42a 0.94***b ***a ***a 0.18b LStock Kuwait 1.06a -1.93b 1.39***b ***a ***a 0.30b LStock Oman -0.55a -0.04a 0.84***b -2.72**a ***a 0.20b LStock Qatar -0.23a -0.14a 0.38*b ***a ***a 0.15b LStock Saudi -0.72a -0.72a 0.90***b ***a ***a 0.12b LStock UAE -1.04a -0.95a 0.53**b ***a ***a 0.18b Notes : All variables are in natural logs. ADF is the Augmented Dickey-Fuller test, PP the Phillips-Perron test and KPSS the Kwiatkowski- Phillips-Schmidt-Shin test. (a) model without constant nor deterministic trend, (b) model with constant without deterministic trend, (c) model with constant and deterministic trend. *, ** and *** denote rejection of the null hypothesis respectively at 10%, 5% and 1%. Table A2: Traditional cointegration tests ADF PP Johansen LStock Bahrain -3.59** -3.31** LStock Kuwait LStock Oman LStock Qatar LStock Saudi LStock UAE Notes: ADF is the Augmented Dickey-Fuller test, PP the Phillips-Perron test and Johansen the trace statistics. *, ** and *** denote rejection of the null hypothesis respectively at 10%, 5% and 1%. The long-term relation is (11). We apply unit root tests on residual series t. Table A3: Tests for asymmetric cointegration (tests on ADF PP Johansen LStock Bahrain ** 20.96** LStock Kuwait ** LStock Oman ** LStock Qatar LStock Saudi * LStock UAE Notes: ADF is the Augmented Dickey-Fuller test, PP the Phillips-Perron test and Johansen the trace statistics. *, ** and *** denote rejection of the null hypothesis respectively at 10%, 5% and 1%. The long-term relation is (10). We apply unit root tests on residual series 1t Table A4: Tests for asymmetric cointegration (tests on 2t ) 1t ) ADF PP Johansen LStock Bahrain ** 33.88*** LStock Kuwait **** LStock Oman -3.03* -3.58** 39.9*** LStock Qatar -3.36** -5.41*** 38.78*** LStock Saudi ** 35.47*** LStock UAE ** 37.75*** Notes: ADF is the Augmented Dickey-Fuller test, PP the Phillips-Perron test and Johansen the trace statistics. *, ** and *** denote rejection of the null hypothesis respectively at 10%, 5% and 1%. The long-term relation is (10). We apply unit root tests on residual series 2t 14