Contagion effects of the Global Financial Crisis on GCC stock markets By Zhang Hengchao* & Zarinah Hamid# Abstract The objectives of this study are twofold: first, it is to investigate whether contagion effects of current Global Financial Crisis present in Gulf Cooperation Council (GCC) stock markets; second, it is to evaluate the impact of such financial crisis on the long-run and short-run dynamic relationships between GCC stock markets and leading conventional and Islamic stock markets around the globe, over the period from July 7, 2004 to August 3, 2012. In order to achieve the two objectives, this study employs three major investigation techniques. First, both standard Pearson correlation and heteroscedasticity adjusted correlation tests are used to investigate the evidence of contagion effects in GCC stock markets. Then, Vector Autoregression (VAR) based Johansen-Juselius (JJ) Cointegration tests are employed to investigate the existence of long-run equilibrium relationships among the stock markets. Finally, Variance Decompositions (VDCs) function is utilized to evaluate the dynamic interactions and strength of causal relationships among the stock markets in the short-run. It is expected that the findings of this paper will not only shed some light on the benefits of international portfolios diversification into GCC stock markets, but also provide guidelines for policymakers to structure appropriate economic policies to stabilize the economy in the wake of the international financial shocks. Keywords: Global Financial Crisis, GCC stock markets, Correlation, Cointegration, VDCs, Diversification benefits * PhD Candidate, Institute of Islamic Banking and Finance, International Islamic University, 50728 Kuala Lumpur, Malaysia # Associate Professor, Faculty of Economics and Management Sciences, International Islamic University, 50728 Kuala Lumpur, Malaysia Email: zhanghengchao_919@hotmail.com ; Phone: +60-183954098 1 1. Introduction Ever since late 1980s, emerging economies have started liberalizing their financial markets to attract foreign investors with the hope of fueling their economic growth. However, for many of these countries their financial sectors are lack of sound financial supervision and management. With the deficiency of financial supervision, the vulnerability of these financial markets to the external shocks has increased significantly as they move aggressively towards more integrated and globalized financial markets (Khalid & Kawai, 2003). The incidents of such increased vulnerability can be evidenced from the variety of financial crises among emerging economies over the past twenty years, namely, the crises in Mexico in 1994, Asia in 1997, Russia in 1998, and Brazil in 1999. One of the most noticeable characteristics of these crises is that, these financial crises were all initiated by the local turbulences but eventually spilled over to other markets, despite the existence of little or no economic linkages between affected countries and the initiators. Such phenomenon has often been described as financial contagion (Khalid & Kawai 2003; RaΔickas & VasiliauskaitΔ 2011). Although effects of financial contagion are critical for both international portfolio investors and domestic financial policy makers, there have been precious few researches on this topic. Only after Asian financial crisis, researches on this subject have been intensified (Chung 2005; Fahami 2011; Khalid & Kawai 2003). However, most of these researches have been devoted to explore the interdependence and contagion effects among developed countries as well as Asian emerging economies. Despite the rapid growth and liberalization of GCC financial markets, trivial efforts have been made to explore the contagion effects of the financial crisis from global and regional stock markets to GCC stock markets (Sedik & Williams 2011). 2 During the last few years, the Global Financial Crisis triggered by U.S. Subprime Crisis has severely affected many countries all over the world. As GCC countries vigorously liberalize their financial markets and hence integrate into world economy, researches on the contagion effects of the Global Financial Crisis on the stock markets of the GCC member countries are greatly needed. It is because that the study of contagion effects may shed some light on the potential benefits of diversifying international portfolios into GCC stock markets, and hence help financial authorities in GCC countries to structure effective policies to curb the contagion effects of the financial crisis. The objectives of this paper are twofold: first, it is to investigate whether current Global Financial Crisis has contagion effects on the GCC stock markets. Second, it is to evaluate the impact of current Global Financial Crisis on the long-run and short-run dynamic relationships between GCC stock markets and leading conventional and Islamic stock markets around the globe. The structure of this paper is organized as follow: Section 2 briefs the past researches on contagion effect. Section 3 presents the data and methodology. Section 4 discusses the empirical results of correlation analysis, cointegration tests, and variance decomposition function tests. Some concluding remarks and policy implications have then been presented in Section 5. 2. Literature review To understand the contagion effects of the Global Financial Crisis comprehensively, this section of the paper provides a brief review of past researches on both concepts as well as empirical analysis of financial contagion. 3 As for the definition of contagion, there are various versions. In the broadest sense, some researches described contagion as the phenomenon where the country-specific events quickly spread to other countries around the globe (Caporale, Cipollini, & Spagnolo, 2005; Candelon 2010; RaΔickas & VasiliauskaitΔ 2011). Specifically, some researches took into consideration the role of the fundamental interdependence, such as the financial and trade linkages in the case of contagion. For instance, Fraztscher (2002) considered contagion as the transmission of a crisis to a particular country due to its real and financial interdependence with the crisis initiator. In contrast, Pericoli & Sbracia (2003) referred contagion as the phenomenon of excess comovements between markets which cannot be explained by macroeconomic fundamental factors. As extension, other researchers added that such excess comovements between markets may occur when a shock in one country is transmitted to another country through cross market balancing, which can be commonly explained by investors’ herding behavior (Kose et al. 2007; Kaminsky & Reinhart 2000). Last but not least, Forbes & Rigobon (2002) defined contagion uniquely as the significant increase in the correlation between markets, after a shock hit a market or group of markets. According to this study, contagion occurs only if the cross-market movements increase significantly after the crisis. If there does not exist significant increase of the comovement, these markets are only believed to be interdependent. By defining the contagion effect as significantly increased comovements between markets, in comparison with other definitions, it offers two advantages for the empirical analysis: a) the phenomenon can be tested directly and simply through correlation coefficients, b) it allows one to distinguish between crisis-contingent and non-crisis contingent (Forbes & Rigobon 2002). Have noted the benefits of such definition, this paper have adopted the definition of contagion introduced by Forbes & Rigobon (2002). 4 A variety of empirical methodologies can be found in the past researches to test the effects of financial contagion. In sum, four methods have been commonly applied, namely, correlation test, time-varying volatility models, Cointegration techniques, and Probi-Logit test. The correlation analysis was believed to be first introduced by King & Wadhwani (1990). The study found that the correlations between U.K., U.S. and Japan stock markets have increased significantly after the stock market crash of 1987. Then, Calvo & C. Reinhart (1996) applied the same approach to test the contagion effects of the Mexico crisis, and found increased comovements across weekly equity and Brady bond returns among emerging Latin American markets after the Mexico crisis. However, this standard correlation test have been criticized by Forbes & Rigobon (2002) for the presence of heteroscedasticity bias. As the paper explained when stock market volatility increases, the market correlations will be biased upward. Such upward bias of correlations will ultimately lead to false conclusions on the existence of contagion effects between markets. To overcome the problem of such correlation bias, Forbes & Rigobon (2002) introduced heteroscedasticity adjusted correlation tests, where the heteroscedasticity bias has been corrected. After applying the heteroscedasticity adjusted correlation tests, the authors did not find significant contagion effects in the cases of the Asian, the Mexican and the 1987 US crises. Such findings were found to be contradicted with the results given by unadjusted correlation tests. The heteroscedasticity adjusted correlation tests have been applied extensively in the latter researches. For instance, Collins & Biekpe (2003) adopted the adjusted correlation test to investigate the whether African equity markets suffered from the contagion effects of Asian financial crisis. The study concluded that except Egypt and South Africa, all the other Africa equity markets did not suffer contagion effects from Asian financial crisis. In addition, Hon et al. 5 (2004) applied the same technique to examine whether the terrorist attack in the U.S. on September 11, 2011 resulted in a contagion effects. The study indicated that international stock markets, particularly in Europe, experienced contagion effects in the three to six months after the crisis. Furthermore, Chiang et al. (2007) used this method to examine whether contagion effects of Asian financial crisis present in the Asian markets, and the study confirmed the existence of the contagion effects in the region. Most recently, Lee (2012) applied this method to investigate whether international stock markets suffer from the contagion effects of the U.S. subprime crisis. The study concluded that out of the 20 international stock markets, only Hong Kong, Taiwan, Australia, and New Zealand suffered from the contagion effects. The Autoregressive Conditional Heteroscedasticity (ARCH) process proposed by Engle (1982) and the Generalized ARCH (GARCH) developed by Bollerslev (1986) have been well acknowledged for measuring the volatility spillover of the stock market returns. For instance, Hamao, Masulis, & Ng (1990) applied GARCH method and confirmed the existence of contagion effects across markets after the 1987 U.S. stock market crash. Similarity, in examining the volatility spillover effects of 1994 Mexico crisis on bond returns across markets, Edwards (2000) also applied GARCH method. He found that the volatility is transmitted from one country to another but it is uncertain about the propagation mechanisms during the crisis time. Recently, Engle (2002) introduced a multivariate GARCH model with time-varying conditional correlations—dynamic conditional correlation (DCC) model to examine the contagion effects of stock markets. In comparison with other estimation techniques, it is believed that DCC-GARCH model has three advantages. First, DCC-GARCH model accounts for heteroscedasticity. Second, additional explanatory variables can be included in the mean equation to measure the common factor. Third, it is the multivariate form of GARCH model which can be used to examine 6 multiple asset returns without adding too many parameters (Chiang et al. 2007). In Caporale et al. (2005), the study used DCC-GARCH to examine the contagion effects of Asian financial crisis on East Asian region, and then they found existence of contagion within the region. Similarly, applying the same method on a more extensive time duration, Chiang et al. (2007) echoed with the previous research and confirmed the existence of contagion effects during the Asia crisis period. However, since most of GARCH-modeled contagion studies measure contagion as the excess correlation in the residuals of the model after controlling for fundamentals, it is difficult to discriminate whether excess comovement captured by the residuals is caused by global shocks or the omitted variables in the model (Forbes 2012) Some studies adapt Cointegration tests to examine the long-run relationship between markets and investigate the contagion effect through analyzing the extent of increased comovement during the crisis period, in comparison with the tranquil period (Yang & Lim 2004). The change of comovement can be measured by the changes in the Cointegration vectors between markets over a long period of time. However, such approach does not specifically test for contagion, as the increased cross-market relationship over such long time can be attributed to number of reasons, such as greater trade integration or higher capital mobility (Forbes & Rigobon 2002; Lee 2012). Thus, this approach could miss periods of contagion as cross market relationships only increase briefly after a crisis. The probability approach predicts the probability that a country will be affected by a crisis when the crisis has occurred in another country. In Eichengreen, Rose, & Wyplosz (1996), the study applied Probit model to test the probability of currency crisis contagion, and found that the currency crisis occurred in elsewhere will significantly increase the probability of the currency crisis domestically. Although such approach is relative easy to perform quantitative 7 analysis on the existence of contagion, if heteroscedasticity is not addressed, the estimate will be biased (Xu & Liu 2010). Empirical researches on contagion effect in Gulf Cooperation Countries have gained attention recently. In (Moosa 2010), the study applied Structural Time Series Model of Harvey (1989, 1997) to examine the stock market contagion from the U.S. to the GCC countries during the period 2007-2008. The paper has shown limited evidence for the contagion effects of U.S. stock market shocks on GCC stock prices. In addition, Suliman (2011) applied GARCH method to investigate the evidence of contagion in GCC economies from 1960 to 2002. He found that Saudi Arabic first suffered from the contagion effect of 1987 US stock market crash and the 1997 Thai currency crisis, and then it was propagated to other GCC nations. 3. Data and methodology Data This study has chosen Saudi Tadawul All Share Index (KSA), Kuwait Stock Exchange Index (KWT), Abu Dhabi Securities Exchange General Index (UAE), Qatar Exchange Index (QAT), Bahrain Bourse All share index (BHN), and Oman Muscat Securities 30 Index (OMN) as the proxies of six GCC stock markets. In addition, we have chosen SP 500 (US), FTSE 100 (UK), Nikki 225 (JAP); as well as Dow Jones US Islamic Market Total Return Index (IUS), Dow Jones Islamic Market UK Index (IUK), and MSCI Japan Islamic Index (IJP) to represent the conventional and Islamic stock markets of world’s three largest stock exchanges in U.S., U.K., and Japan (Majid et al. 2007; Kassim 2010). The data used in this study are two-day rolling average of daily closing indices denominated in the local currency, for the period from July 7, 2004 to August 3, 2012. Since 8 stock markets in GCC countries and U.S., U.K., and Japan are located in different time-zones, the analysis is expected to encounter the problem of non-synchronous trading. In order to overcome such problem, this study has adopted the method of two-day rolling average advocated by Forbes and Rigobon (2002) and Mun (2005). According to Forbes and Rigobon (2002) and Mun (2005), two-day rolling average data obtained from the average of two conservative daily series will help to avoid the problem of non-synchronous trading among the stock markets. As for the unavailable data because of weekends, national holidays, or other reasons, stock indices were assumed to stay as same as the adjacent indices prior to the non-trading date (Chiang et al. 2007; Karim et al. 2010). Besides, since the trading activities of GCC stock markets are believed to be dominated by GCC citizens (Bley & Chen 2006), the stock indices included in this study have chosen to be dominated in their respective local currencies (Hon et al. 2004). In addition, both natural logarithm and the first differenced natural logarithm value (stock market return) of each series are calculated for the convenience of latter use. To examine the impact of Global Financial Crisis on GCC stock markets before and after the financial turmoil, we have divided the whole sample into two sub-periods: a) Tranquil Period, from July 7, 2004 to July 25, 2007; and b) Crisis Period, from July 26, 2007 to August 3, 2012. Since it is believed that the US subprime crisis debuted on July 26, 2007 (Dungey et al. 2002; Karim et al. 2010; Majid & Kassim 2009), we use this date to be the breaking point of the entire sample. 9 Methodology Correlation analysis: Since correlation analysis has been widely used to investigate the existence of financial contagion (Chiang et al. 2007), this study has started with simple pair-wise correlation between stock returns. The conditional correlation coefficients are measured as following: ππ₯π¦ ρ=π (1) π₯ ππ¦ where ππ₯π¦ is the covariance of stock returns in markets ‘x’ and ‘y’, ππ₯ and ππ¦ are the standard error of stock returns in market ‘x’ and ‘y’, respectively. Due to the presence of heteroscedasticity bias in returns, Forbes & Rigobon (2002) proposed to adjust the bias in stock index returns by manipulating the equation (1) to obtain the unconditional correlation coefficient: π π∗ = (2) √1+πΏ[1−π2 ] where πβ δ = ππ₯π₯ π −1 (3) π₯π₯ It measures the relative changes of variance in stock market ‘x’ between crisis and tranquil periods. To calculate the adjusted correlation coefficients, the crisis period is often used as the high volatility period while the tranquil period as the low volatility period. The following hypothesis is then tested: 10 π»0 :ππ ≤ ππ‘ π»1 :ππ > ππ‘ ππ is the adjusted correlation coefficient during the crisis period, and ππ‘ is the adjusted correlation coefficient during the tranquil period. Based on the definition of contagion in Forbes & Rigobon (2002), contagion presents when there is significant increase of markets comovements. In order to make statistical inference on the significance of the increased comovements, the standard Z-test is applied (Chiang et al. 2007). Fisher’s Z transformations convert standard coefficients to normally distributed Z-scores. Thus, the initial hypothesis is then transformed into: π»0 : ππ ≤ ππ‘ → π»0 : ππ ≤ ππ‘ π»1 : ππ > ππ‘ → π»1 : ππ ≤ ππ‘ where 1 1+π ππ = 2 ln (1−ππ) π 1 1+π ππ‘ = 2 ln (1−ππ‘) π‘ Z= ππ −ππ‘ 1 1 √π −3+π −3 π π‘ where, ππ and ππ‘ are Fisher transformations of correlation coefficients during crisis and tranquil periods, respectivelyοΌππ‘ = 1114 and ππ = 1836 are the number of observations during crisis and tranquil periods, respectively. The critical value of right-tailed Z-test at 1%, 5%, and 10% are 2.33, 1.65, and 1.28, respectively. 11 Unit root tests: In order to avoid the problem of spurious results with regressions of non-stationary time series variables, it is necessary to determine the order of integration for each variable in the model. Thus, Unit root tests should be applied to establish whether the variables are non-stationary, and how many times the variables needed to be differenced so that the series will be stationary. As for the order of integration, a variable is said to be integrated of order d or I (d) if it is stationary after differencing d times. In this study, both Augmented Dickey-Fuller (ADF) and PhillipsPerron (PP) tests are applied to investigate the non-stationarity of the series used. The two tests differ from one another in terms of autocorrelation patterns used in the model. As for PP test, it allows for more general autocorrelation patterns in comparison with ADF test. Cointegration tests: The next step of analysis is to investigate the existence of long-run equilibrium relationships between variables. If there is a long-run equilibrium relationship between studied variables, although the variables are individually non-stationary, they cannot drift arbitrarily far away from each other. For a system of variables sharing the long-run equilibrium relationship, they are said to be cointegrated. The requirements for cointegration between these variables are that they have to be integrated of the same order and stationary in their linear combination. Besides giving information on the long-run equilibrium relationships among variables, the cointegration tests also provide guidance on the proper specification of the VAR models (Ibrahim 2003). Specifically, if the variables are found to be non-stationary and not cointegrated the first differenced VAR model should be used, while if non-stationary variables are found to be cointegrated a Vector Error Correlation Model (VECM) should be used ( Engle & Granger 1987). 12 However, since past researches are still unclear on whether VECM outperforms the VAR at all forecasting horizons (Ibrahim 2003), for our study we employ VAR model using variables in levels even with the findings of cointegration among variables. In order to test for cointegration, the Vector Autocorrelation (VAR) based JohansenJuselius (JJ) Cointegration approach is applied. By implementing JJ Cointegration technique on VAR model, it is believed that such combination is a realistic representation of market linkages (Dekker et al. 2001). The advantage of VAR model is that it needs not to distinguish between exogenous and endogenous variables in the model. Thus, employing such model will be less complicated (Gujarati 2003). Given that VAR approach models every endogenous variable in the system as the function of the lagged values of all of the endogenous variables in the system (Khalid & Kawai 2003); the VAR model can be specified as following: ππ‘ = δ + Ππ ππ‘−1 + β― + Ππ βππ‘−π + ππ‘ (4) where ππ‘ is an n×1 vector of non-stationary variables integrated of order d, δ is an n×1 vector of constant terms, Ππ and Ππ are the n×n matrices of coefficients, Δ denotes the first difference, k is the lag length, and ππ‘ is an n×1 vector of white noise error terms. Since the appropriate lag structure is important for the VAR model, this study has chosen the lag structure based on the criteria that the residuals from each VAR model are white noise process. To examine the existence of white noise process in the residuals of VAR model, the Ljung-Box Q-statistics has been applied. In order to determine the number of cointegration vectors, Johansen (1988) and Johansen and Juselius (1990) developed two test statistics, which are the Trace and the Maximal Eigenvalue statistics: 13 ππ‘ππππ = −π ∑ππ=π+1 ln(1 − ππ ) (5) ππππ₯ = −π ln(i − λr+1 ) (6) The Trace statistics tests the null hypothesis that there are at most r cointegration vectors. On the other hand, the Maximal Eigenvalue test is based on the hypothesis that the number of cointegration vectors is r against the alternative hypothesis that it is r+1. In addition, since Trace statistics are more robust than the Max-Eigenvalue statistics, if Trace and Max-Eigenvalue statistics give contradictory results, the Trace statistics shall prevail (Johansen & Juselius 1990; Cheung & Lai 1993). VDCs: Lastly, the Variance Decompositions (VDCs) analysis based on the VAR model in levels will be used for the cointegrated stock markets. In particular, the VDC analysis investigates how much of the variations in a certain market return can be explained by innovations from other markets in the system. Besides, it also provides us insights on the relative importance of other markets in driving the variations of particular stock market (Ravichandran & Maloain 2010). The variance decomposition analysis in this study provides decompositions of 2-day, 5-day, and 20-day horizons, for both tranquil and crisis periods, respectively. 4. Empirical results Correlation analysis: Table 1 presents the results of both unadjusted (conditional) and adjusted correlation (unconditional) coefficient analysis for GCC stock market indices in respect of the current Global Financial Crisis. The cross-market correlations of stock market index returns (1st differenced natural logarithmic indices) are compared between tranquil period and crisis period. 14 Table 1: GCC stock markets contagion to the current Global Financial Crisis--unadjusted and adjusted correlation coefficient tests Tranquil Period Crisis Period Contagion unadjusted ρ σ unadjusted ρ σ KSA 0.0021 1.2907 0.2295 0.9730 5.9808 *** Y KWT 0.0022 0.5688 0.0762 0.5051 1.9463 ** Y UAE -0.0370 0.9324 0.1580 0.7940 5.1287 *** Y QAT -0.0361 1.0411 0.1640 1.0264 5.2628 *** Y BHN 0.0416 0.5074 0.0799 0.8635 1.0073 N OMN 0.0307 adjusted ρ 0.4066 σ 0.0162 adjusted ρ 0.3947 σ -0.3814 Z-score N KSA 0.0028 1.2907 0.3016 0.9730 7.8574 *** Y KWT 0.0025 0.5688 0.0867 0.5051 2.2138 ** Y UAE -0.0436 0.9324 0.1854 0.7940 6.0228 *** Y QAT -0.0370 1.0411 0.1681 1.0264 5.3962 *** Y BHN 0.0246 0.5074 0.0473 0.8635 0.5967 Y/N Z-score N N OMN 0.0307 0.4066 0.0162 0.3947 -0.3819 Note: a) This table shows the Pearson cross-market correlation coefficients (ρ) and variance (σ) of the correlation between the U.S. and 6 GCC countries. b) Z-score are Fisher z-transformed values testing for the equality of tranquil period cross-market correlations with the corresponding crisis period correlations. The null hypothesis is no increase in correlation (or no contagion). c) "Y" and "N" indicates the presence and absence of contagion effects, respectively, d) Unadjusted correlation coefficients are conditional correlation coefficients, using Eq. (1). e) Adjusted correlation coefficients are adjusted for changes in variance, using Eq. (2). e) The 1%, 5%, and 10% right-tail critical Z values are 2.33, 1.65, and 1.28, respectively. ***, **, and * indicates statistical significance at the 1%, 5%, and 10% levels, respectively. As table 1 reveals, the cross-market correlations between U.S. and six GCC countries during the crisis period are larger than those during the tranquil periods for both unadjusted and adjusted correlation coefficients. The exceptions have found to be Bahrain (BHN) and Oman (OMN) in both scenarios. When unadjusted correlation analysis is applied, contagion effects can be observed from Saudi Arabia (KSA), Kuwait (KWT), Abu Dhabi (UAE), and Qatar (QAT). In addition, similar findings can be evidenced from adjusted correlation analysis. Since adjusted correlation analysis has taken into account the heteroscedasticity bias, it provides more reliable results in comparison with unadjusted correlation analysis. Thus, it is safe to conclude that among six GCC stock markets only Saudi Arabia, Kuwait, Abu Dhabi, and Qatar have suffered 15 contagion effects from the current Global Financial Crisis. However, our findings are found to be contradicted with Moosa (2010), which has rejected the presence of stock market contagion from the U.S. to the GCC markets in wake of 2007 U.S. Subprime Crisis. Two factors can be attributed for such differences. First, different methodologies have been used in testing the contagion hypothesis. Opposing to our correlation analysis, structural time series model has been applied in Moosa (2010). Second, different time intervals have been applied in the study. Moosa (2010) used a turmoil period from January 2, 2007 to December 31, 2010. As for this study, it covers the time intervals between July 7, 2004 and August 3, 2012, which sets July 26, 2007 as the breaking point for the crisis period. Due to these two differences, it is expected that our investigation on the contagion effects will be different from the study of Moosa (2010). Unit root tests: Table 2: Unit Root Tests Tranquil Period KSA KWT UAE QAT BHN OMN US UK JAP IUS IUK Level (Natural Log) ADF PP -1.5679 -1.3901 -1.6369 -1.4729 -1.6557 -1.8672 -1.7479 -1.9144 -1.7249 -1.5147 -2.6827 -2.1503 -2.6625 -2.9830 -3.5239** -3.4430** First Difference (Log difference) ADF PP -6.6698*** -19.4903*** -6.7303*** -18.1323*** -7.8766*** -18.5103*** -8.3517*** -16.2133*** -8.9127*** -17.5444*** -6.9397*** -16.9439*** -9.6120*** -17.3563*** -12.0111*** Crisis Period Level (Natural Log) ADF PP -1.5127 -1.5879 -1.9895 -1.2376 -1.4960 -1.5673 -1.8407 -1.6914 -1.8066 -1.7828 -1.5769 -1.0754 -1.7698 -1.8446 First Difference (Log difference) ADF PP -12.4450*** -21.2454*** -21.832*** -21.8411*** -13.2334*** -19.1765*** -10.0039*** -21.0635*** -7.9506*** -20.7549*** -6.9619*** -21.7009*** -11.6484*** -24.9369*** -18.3922*** -2.1713 -2.1765 -10.2573*** -22.5428*** -2.1119 -2.3983 -2.1132 -8.7842*** -18.7243*** -2.5651 -2.5004 -9.6963*** -23.0495*** -2.7863 -9.6178*** -17.2764*** -1.8978 -1.9081 -10.0528*** -25.2004*** -3.8135** -3.7759** -10.4382*** -17.7488*** -1.7420 -1.9974 -13.4841*** 22.3476*** IJAP -2.6953 -2.8750 -9.6408*** -18.1612*** -2.1014 -2.1468 -10.4694*** -25.7605*** Note: *,**, *** denote significance at the 10%, 5%, and 1% level, respectively. The above tests of ADF and PP are based on model with constant and trend. 16 As for the stationary nature of the stock market indexes (natural logarithmic indexes), and indexes returns (1st difference of natural logarithmic indices), Table 2 illustrates the unit root test results for both ADF and PP methods during the two sub-periods. For the two sub-periods, both ADF and PP statistics indicate that at level all the series are non-stationary, except for UK’s precrisis Conventional and Islamic indices. In general, this implies that the null hypothesis of the presence of unit root cannot be rejected at all levels of significance before and during the Global Financial Crisis, when market indices are at level. On the other hand, when the series are at first difference, the null hypothesis of the presence of unit root can be rejected at all levels of significance. In other words, all of these stock market returns are said to be integrated at order one, or I (1). Cointegration tests: Having noted that all stock market indices are integrated at order one, or I (1), next we applies VAR-based JJ Cointegration analysis to investigate the evidence of long-run equilibrium relationships among markets. The cointegration tests have been employed on two models, for both tranquil period and crisis period. For the first model, six GCC stock markets together with world’s three largest conventional stock markets (i.e. U.S., U.K., and Japan) are included. As for the second model, six GCC stock markets together with world’s three largest Islamic stock markets (i.e. U.S. Islamic, U.K. Islamic, and Japan Islamic) are included. The appropriate lag lengths for both models are determined on the criteria that the residuals from each VAR model are white noise process. In particular, based on the Ljung-Box Q-statistics, the inclusion of 24 lags for the two models during the tranquil period are sufficient to render all the residuals of the VAR models to be white noise. As for the crisis period, the appropriate lag length for the model 17 with conventional stock markets is 25, whereas it is 23 lags for the model with Islamic stock markets. Subsequently, Trace and Max-Eigenvalue statistics are applied to determine the number of the cointegration vectors at 5% significance level. In addition, since Trace statistics are more robust than the Max-Eigenvalue statistics, if Trace and Max-Eigenvalue statistics give contradictory results, the Trace statistics shall prevail (Johansen & Juselius 1990; Cheung & Lai 1993). The results of the cointegration tests are presented in Table 3. When GCC stock markets and conventional stock markets of U.S., U.K. and Japan are included in the model, before the crisis Trace statistics indicates that there are three cointegration relationships among them. In contrast, Max-Eigenvalue statistics tend to reject the existence of long-run equilibrium relationship among these markets. As for the crisis period, both Trace and Max-Eigen statistics confirm the existence of one cointegration relationship among these markets. Since Trace statistics prevail over Max-Eigen statistics in case of conflicting results, it is safe to conclude that there are long-run equilibrium relationships between GCC stock markets and conventional stock markets of US, UK, and Japan, for both of the sub-periods. This result suggests that GCC stock markets and selected three world’s largest conventional stock markets share long-run stable relationship regardless the occurrence of Global Financial Crisis. Thus, the opportunity of benefiting from long-term international portfolio diversification into GCC stock markets has diminished regardless the occurrence of the Global Financial Crisis. As for the second model, prior to the Global Financial Crisis both Trace and MaxEigenvalue statistics have evidenced the existence of cointegration relationship between GCC 18 stock markets and Islamic stock markets of U.S., U.K., and Japan. However, for the crisis period, only Trace statistics confirmed the presence of cointegration relationship among these markets. Against, since Trace statistics prevail over Max-Eigen statistics in case of conflicting results, it is rational to conclude that there is consistent cointegration relationship between GCC stock markets and the Islamic stock markets of U.S., U.K., and Japan, regardless the occurrence Global Financial Crisis. As implication, in the long-run, there are only marginal benefits for international investors holding stocks in the Islamic stock markets of U.S., U.K., and Japan to diversifying their portfolios into GCC stock markets. Table 3. Cointegration Tests Model KSA, KWT , UAE, QAT , BHN, OMN, US, UK, JAP KSA, KWT , UAE, QAT , BHN, OMN, IUS, IUK, IJAP Tranquil Period Crisis Period MaxMaxH0 Trace Trace Max 5% Trace Trace 5% Max 5% Eigen. Eigen. S tatistic 5% CV CV S tatistic CV CV S tatistic S tatistic r = 0 220.0237** 51.6528 197.3709 58.43354 209.8721** 60.0213** 197.3709 58.43354 r ≤ 1 168.3709** 38.0509 159.5297 52.36261 149.8507 42.88828 159.5297 52.36261 r ≤ 2 130.3200** 35.1866 125.6154 46.23142 106.9625 25.61891 125.6154 46.23142 r≤3 26.4947 95.75366 40.07757 81.34355 24.79998 95.75366 40.07757 r = 0 221.5914** 61.2804** 197.3709 58.43354 207.8249** 54.06407 95.13341 197.3709 58.43354 r ≤ 1 160.3110** 39.9273 159.5297 52.36261 153.7608 37.64956 159.5297 52.36261 r≤2 120.3837 37.5624 125.6154 46.23142 116.1112 35.53098 125.6154 46.23142 r≤3 82.82134 23.7097 95.75366 40.07757 27.22911 95.75366 40.07757 80.5803 Note: a) the variables included in the two models are the natural logarithm values of the stock market indices, b) ** denotes statistical significance at the 5% level. VDCs: In the previous section, the cointegration analysis has indicated the existence of long-run relationships among studied stock markets. In order to gauge the relative strength of market shock in one market on other markets in the system, VAR-based VDCs analysis has been employed. The VDCs analysis also helps to infer on the dynamic interactions between stock markets in the system. In particular, the results of the VDCs analysis document the extent to 19 which the variance of the stock market indices in one GCC member country can be explained by the variance of other stock markets in the system. Table 4 and Table 5 in Appendix 1 provides information on the VDCs for the two previously mentioned VAR models in both tranquil and crisis periods. In addition, the reports of the results for VDCs are at 2-day, 10-day, and 20-day horizons. As for VDCs analysis in Table 4, it follows the Cholesky ordering of Saudi Arabia, Kuwait, U.A.E., Qatar, Bahrain, Oman, U.S., U.K., and Japan. On the other hand, the VDCs analysis in Table 5 follows the Cholesky ordering of Saudi Arabia, Kuwait, U.A.E., Qatar, Bahrain, Oman, U.S. Islamic, U.K Islamic, and Japan Islamic. As indicated by Table 4, four interesting observations can be found by comparing the after-crisis VDCs results with the pre-crisis ones. First, after the debut of the Global Financial Crisis there is deceased importance of own market shocks in all GCC markets. This may indicate the increased vulnerability of GCC market to external shocks from other markets either within or outside the region. In particular, the stock markets of Qatar, Bahrain, and Oman have severely subject to the external shocks after the debut of the Global Financial Crisis. Statistically, the selfexplanatory powers of their stock markets variations have dropped extensively from over 90 percent prior to the crisis to nearly 50 percent after the crisis, within the time horizon of 20 days. Similar findings can be evidenced from (Ravichandran & Maloain 2010). Similarly, as can be observed from Table 5, when Islamic stock markets are included in the model GCC stock markets still tend to depend more on the variations of other stock markets both within and outside of the region, after the debut of the financial crisis. 20 As for the second observation, the conventional U.S. stock market plays increased role in explaining the variations of GCC stock markets after the debut of the crisis. Specifically, the explanation power of conventional stock markets have increased handsomely from less one 1 percent prior to the crisis to nearly 20 percent after the crisis, within the time interval of 20 days. Similar pattern of observation can be found from Table 5, where the extent of U.S. Islamic stock market in explaining the variations of the GCC stock market have increased significant after the debut of the financial crisis. In particular, the explanation power of U.S. Islamic stock market have surged from an average of less than 1 percent prior to the crisis to an average of 10 percent after the crisis. In terms of the third interesting observation, the Table 4 reveals that the regional stock markets play more important role than international conventional stock markets in explaining the variations of the GCC stock markets, regardless the occurrence of the Global Financial Crisis. The results shown from Table 5 are found to be echoed with the findings of Table 4, where the regional stock markets also play more significant role than international Islamic stock markets in driving the movements of GCC stock markets. Lastly, regionally Saudi Arabia stock market has consistently acted as the major driver of the stock markets variations with in the region, whereas internationally U.S. conventional stock market plays a relatively more significant role than the other two leading conventional stock markets in variations of the GCC stock markets. In particular, the conventional stock markets of U.K., and Japan only have the average explanation power of less than 1 percent across the two studying periods, whereas the explanation power of U.S. stock markets have increased significantly form less 1 percent to nearly 10 percent after the debut of the crisis. Similarly, the U.S. Islamic stock market is also found to be more significant than the other two 21 international Islamic stock markets in influencing the variations of GCC stock markets, for both studying periods. 5. Conclusion This article aims to achieve two major objectives: first, to investigate whether contagion effects of current Global Financial Crisis presents in Gulf Cooperation Council (GCC) stock markets; second, it is to evaluate the impact of the financial crisis on the long-run and short-run dynamic relationships between GCC stock markets and international leading conventional and Islamic stock markets around the globe. Noted that we have defined contagion as the circumstance where the cross-market movement increases significantly after the debut of financial crisis (Forbes & Rigobon 2002), the investigation on the presence of contagion effects has been conducted based on the correlation analysis with the adjustment of heteroscedasticity bias. As the correlation analysis indicates, contagion effects of Global Financial Crisis can be only evidenced in Saudi Arabia, Kuwait, Abu Dhabi, and Qatar stock markets. As for the effects of the Global Financial Crisis on the long-run relationships between studied stock markets, the VAR based JJ Cointegration tests reveal that there are stable long-run equilibrium relationships between GCC stock markets and the conventional as well as Islamic stock markets in U.S., U.K., and Japan, regardless the occurrence of the financial crisis. In other words, the debut of Global Financial Crisis does not have significant influence on the long-run relationship between GCC stock markets and the leading international stock markets around the globe. As implication, there are diminished diversification benefits for international investors to diversifying into GCC stock markets, both prior to and after the debut of financial crisis. 22 In addition, the VDCs analysis indicates four interesting observations pertaining to the impact of Global Financial Crisis on the short-run relationships between GCC stock markets and international stock markets. First, after the debut of the financial crisis, GCC stock markets have experienced increased vulnerability to external shocks from both with and outside of the region. Second, both conventional and Islamic stock markets of U.S. play increased role in explaining the variations of GCC stock markets after the debut of the financial crisis. Third, GCC stock markets seem to be more influenced by regional markets than international markets, regardless the occurrence of the Global Financial Crisis. Last but not least, Saudi Arabia has consistently acted as the major regional force for the stock markets variations within the region, whereas both conventional and Islamic stock markets of U.S. have constantly acted as the major international forces for the movements of the regional stock markets. In sum, GCC stock markets have been consistently integrated with global stock markets, both conventional and Islamic ones. In addition, GCC stock markets become more vulnerable to the external shocks. In order to mitigate the negative influence of the external shock, it is recommended for GCC countries to move towards more integrated regional markets. In addition, GCC countries are advised to increase their diversification in production and trade. With a larger and diversified economy, it is believed that GCC stock markets will be less vulnerable to the external shocks (Suliman 2011). 23 References: Bollerslev, T., 1986. Generalized Autoregressive Conditional Heteroskedasticity. Journal of Econometrics, Volume 31(3), pp. 307-327. Calvo, S. & Reinhart, C., 1996. Capital Flows to Latin America : is There Evidence of Contagion Effects?. World Bank Policy Research Series, No.1619. Candelon, B., 2010. Introduction to the Special Issue of Pacific. Pacific Economic Review, Volume 15(3), pp. 336-339. Caporale, G. M., Cipollini, A. & Spagnolo, N., 2005. Testing for Contagion: a Conditional Correlation Analysis. Journal of Empirical Finance, Volume 12(3), pp. 476-489. Chiang, T. C., Jeon, B. N. & Li, H., 2007. Dynamic correlation analysis of financial contagion: Evidence from Asian markets. Journal of International Money and Finance, Volume 26 (7), pp. 1206-1228. Chung, H., 2005. The contagious effects of the Asian financial crisis: some evidence from ADR and country funds. Journal of Multinational Financial Management, Volume 15(1), pp. 67-84. Edwards, S., 1998. Interest Rates, Contagion and Capital Controls. National Bureau of Economic Research, Volume Working Paper No. 6756. Eichengreen, B., Rose, A. & Wyplosz, C., 1996. Contagious Currency Crises: First Tests. Scandinavian Journal of Economics, Volume 4, pp. 463-484. Engel, R., 2002. Dynamic Conditional Correlation - A Simple Class Of Multivariate Garch Models. Journal of Business and Economic Statistics, Volume 20, pp. 339-350. Engle, R. F., 1982. Autoregressive Conditional Heteroscedasticity with Estimates of the Variance of United Kingdom Inflation. Econometrica, Volume 50, pp. 987-1008. Fahami, N. A., 2011. The structure of linkages and causal relationships between BRIC and developed equity markets. s.l., IACSIT Press, Singapore, pp. 72-77. Forbes, K. J. & Rigobon, R., 2002. No Contagion , Only Interdependence : Measuring Stock Market Comovements. The Journal of Finance, Volume LVII (5), pp. 2223-2261. Fratzcher, M., 2003. On Currency Crisis and Contagion. International Research Journal of Finance and Economics, Volume 8 (2), pp. 109-129. Gujarati, D. N., 2003. Basic Econometrics. 4 ed. s.l.:McGraw-Hill Companies. Hamao, Y., Masulis, R. W. & Ng, V., 1990. Correlations in Price Changes and Volatility across International Stock Markets. Review of Financial Studies, Volume 3 (2), pp. 281-307. 24 Kaminsky, G. & Reinhart, C., 2000. On Crises, Contagion and Confusion. Journal of International Economics, Volume 51, pp. 145-168. Khalid, A. M. & Kawai, M., 2003. Was financial market contagion the source of economic crisis in Asia? Evidence using a multivariate VAR model. Journal of Asian Economics, Volume 14, pp. 131-156. King, M. A. & Wadhwani, S., 1990. Transmission of Volatility between Stock Markets. The Review of Financial Studies, Volume 3 (1), pp. 5-33. Kose, M., Eswar, P. & Marco, T., 2007. How Does Financial Globalization Affect Risk Sharing? Patterns and Channels. International Monetary Fund Working Paper, WP/07/238, pp. 1-41. Kouki, I., Harrathi, N. & Haque, M., 2011. A Volatility Spillover among Sector Index of International Stock Markets. Journal of Money, Investment and Banking, Volume 22, pp. 32-45. Lee, H. Y., 2012. Contagion in International Stock Markets during the Sub Prime Mortgage Crisis. International Journal of Economics and Financial Issues, Volume 2 (1), pp. 41-53. Moosa, I., 2010. Stock Market Contagion in the Early Stages of the Global Financial Crisis: the Experience of the GCC Countries. International Journal of Banking and Finance, Volume 7 (1), pp. 19-34. Naoui, K., Khemiri, S. & Liouane, N., 2010. Crises and Financial Contagion : The Subprime Crisis. Journal of Business Studies Quarterly, Volume 2(1), pp. 15-28. RaΔickas, E. & VasiliauskaitΔ, A., 2011. Channels of Financial Risk Contagion in the Global Financial Markets. Economic and Managment, Volume 16, pp. 1174-1185. Sedik, T. S. & Williams, O. H., 2011. Global and Regional Spillovers to GCC Equity Markets Global and Regional Spillovers to GCC Equity Markets. IMF Working Paper WP/11/138, pp. 127. Suliman, O., 2011. The Large Country Effect, Contagion and Spillover Effects in the GCC. Applied Economics Letters, Volume 18, pp. 285-294. Xu, Y. & Liu, D., 2010. Empirical Study on the Contagion Effect of Financial Crisis. International Conference on Computer Application and System Modeling, pp. 560-563. Yang, T. & Lim, J. J., 2004. Crisis , Contagion , and East Asian Stock Markets. Review of Pacific Basin Financial Markets and Policies, Volume 7 (1), pp. 119-151. 25 Appendix 1: Table 4: Variance Decomposition Analysis Cholesky Ordering: KSA KWT UAE QAT BHN OMN US UK JAP Variance Explained by innovations in: Days of KSA KWT UAE QAT OMN BHN US UK Tranquil Period KSA 2 99.7571 0.0000 0.0366 0.0840 0.0628 0.0136 0.0011 0.0013 5 97.6870 0.0670 0.0762 0.6060 1.1207 0.3294 0.0084 0.0271 20 93.3059 0.7058 0.2811 0.4065 3.9416 0.8310 0.0147 0.1498 KWT 2 3.3737 96.4014 0.0001 0.0031 0.0284 0.0571 0.0132 0.1091 5 9.7568 88.2016 0.6170 0.2967 0.0665 0.1840 0.0460 0.7551 20 25.9259 65.2194 4.2231 0.2886 0.1353 3.0292 0.3854 0.4231 UAE 2 3.7532 0.8545 95.2716 0.0037 0.0157 0.0940 0.0001 0.0031 5 9.3936 0.2913 89.0296 0.0798 0.5047 0.4850 0.0218 0.1165 20 13.5675 0.5475 77.3745 2.9301 1.3818 2.9927 0.2820 0.2162 QAT 2 1.3740 0.5832 1.5845 96.3843 0.0015 0.0053 0.0242 0.0274 5 4.1088 0.6851 1.4202 93.3454 0.0906 0.0044 0.2176 0.0627 20 8.6697 3.3931 0.9912 80.5101 0.1009 1.8933 0.3793 2.5094 BHN 2 0.5862 1.8593 0.0486 0.6471 96.7184 0.0652 0.0000 0.0000 5 3.1430 3.1749 0.0405 0.5299 91.6648 0.4860 0.1605 0.1312 20 11.6965 2.4377 1.6936 1.7369 72.7337 3.1943 2.5518 2.3699 OMN 2 0.1297 0.3976 0.3394 0.3558 0.1306 98.6251 0.0015 0.0191 5 0.8606 1.1256 1.5820 0.1409 0.6941 94.8446 0.1204 0.4580 20 8.7155 1.1090 13.6131 0.0535 1.1742 74.2767 0.1364 0.3276 Crisis period KSA 2 99.1825 0.0953 0.0351 0.0014 0.0301 0.0097 0.6017 0.0001 5 91.8816 0.5405 0.2746 0.0406 0.3664 0.2425 5.8884 0.5075 20 71.5129 2.1237 0.2162 0.8803 0.3598 0.6663 20.3600 1.1788 KWT 2 4.9610 94.8266 0.0039 0.0259 0.0012 0.0100 0.0428 0.0040 5 8.8640 89.4631 0.0122 0.2109 0.1124 0.0392 0.6787 0.1436 20 5.8748 84.7977 0.2424 0.1988 0.1352 1.5398 6.7282 0.0839 UAE 2 10.7485 6.5601 81.8357 0.0124 0.0099 0.0035 0.6070 0.0058 5 16.0192 9.2680 69.4431 0.0300 0.0513 0.1299 4.1996 0.1281 20 14.0970 12.9967 59.2752 0.3492 0.5182 0.4618 11.1580 0.3592 QAT 2 10.8004 9.1385 11.5318 67.3838 0.0121 0.0293 0.6979 0.0058 5 20.4319 11.9205 8.1415 53.6892 0.0206 0.0538 4.7038 0.0366 20 15.4023 20.9310 3.2284 44.0335 0.2142 0.3983 13.4326 0.2801 BHN 2 2.5510 12.0271 2.2152 0.2006 82.9200 0.0000 0.0167 0.0043 5 9.9001 13.2625 2.6255 0.6130 72.5904 0.0117 0.1977 0.1648 20 12.0532 21.3266 5.7595 0.4715 50.5874 0.0986 8.0332 0.3678 OMN 2 8.2803 4.1118 11.9883 5.1530 1.1806 68.5862 0.3753 0.0009 5 14.6399 6.5347 14.1842 4.8021 1.2455 55.2299 2.6098 0.1569 20 12.9717 10.1200 12.1052 5.2561 1.4420 43.4624 12.4707 0.4046 26 JAP 0.0436 0.0781 0.3636 0.0139 0.0764 0.3702 0.0041 0.0777 0.7077 0.0155 0.0651 1.5531 0.0751 0.6693 1.5856 0.0011 0.1738 0.5941 0.0440 0.2578 2.7021 0.1247 0.4759 0.3992 0.2171 0.7309 0.7846 0.4004 1.0022 2.0796 0.0649 0.6344 1.3021 0.3237 0.5971 1.7673 Table 5: Variance Decomposition Analysis Cholesky Ordering: KSA KWT UAE QAT BHN OMN IUS IUK IJAP Variance Explained by innovations in: Day of KSA KWT UAE QAT BHN OMN IUS IUK Tranquil Period KSA 2 99.7324 0.0003 0.0249 0.0577 0.0521 0.0121 0.0002 0.0149 5 97.3044 0.0490 0.0936 0.2861 1.3751 0.3274 0.0457 0.0585 20 88.9777 0.3861 0.3234 0.1519 5.6566 0.9167 0.3925 3.0309 KWT 2 3.2576 96.4298 0.0000 0.0047 0.0119 0.0555 0.0124 0.1969 5 9.3322 88.8851 0.5650 0.2744 0.0213 0.1285 0.0287 0.6861 20 23.5862 69.3685 3.5142 0.4862 0.1640 2.0712 0.4551 0.2226 UAE 2 4.1198 0.8446 94.9275 0.0000 0.0106 0.0914 0.0001 0.0016 5 9.6004 0.2831 88.7029 0.1414 0.6072 0.5756 0.0102 0.0738 20 11.9022 0.5780 76.4962 3.2357 2.4931 4.4130 0.3708 0.0867 QAT 2 2.0325 0.5458 1.5210 95.8719 0.0010 0.0061 0.0088 0.0037 5 5.8031 0.5637 1.0627 91.7975 0.0424 0.0041 0.0514 0.6513 20 10.4758 2.4903 0.6550 78.9637 0.0933 1.8094 0.1708 5.2960 BHN 2 0.6559 1.5324 0.0689 0.6208 96.9344 0.0565 0.0005 0.0002 5 3.4410 2.3495 0.0913 0.5101 92.2275 0.3553 0.0824 0.1018 20 11.2175 1.9392 0.4271 2.2741 76.8332 3.6266 2.5159 0.1233 OMN 2 0.1949 0.3302 0.5474 0.4197 0.1237 98.3742 0.0023 0.0031 5 1.1320 0.7444 1.9137 0.1581 0.6526 94.9685 0.0948 0.1261 20 7.5662 0.9243 13.6914 0.2304 1.0598 75.6771 0.2288 0.2911 Crisis period KSA 2 99.1762 0.1018 0.0166 0.0029 0.0200 0.0204 0.5771 0.0009 5 92.9361 0.4834 0.2290 0.0485 0.4740 0.0600 5.3106 0.1183 20 72.9831 1.5650 0.2821 1.5047 0.4986 0.1561 20.3728 1.5212 KWT 2 4.7978 94.9181 0.0042 0.0107 0.0027 0.0239 0.0335 0.0001 5 8.6967 89.8043 0.0235 0.1355 0.2008 0.0263 0.3790 0.0364 20 5.7050 85.5736 0.5455 0.1163 0.2203 1.5959 4.7637 0.9037 UAE 2 11.2472 6.3326 81.5177 0.0085 0.0096 0.0000 0.6274 0.0121 5 17.1172 8.8846 69.5224 0.0125 0.0727 0.0099 3.1943 0.0416 20 16.3935 11.0969 59.6386 0.2868 0.1866 0.1378 10.5307 0.1204 QAT 2 11.4164 9.2873 11.8646 66.3197 0.0138 0.0400 0.6769 0.0894 5 22.1978 11.6416 8.5614 52.5135 0.0221 0.0862 3.6897 0.5957 20 17.7889 19.4956 3.5432 44.9480 0.1919 0.1798 12.4198 1.0186 BHN 2 2.7939 11.9867 2.1198 0.2721 82.7020 0.0000 0.0183 0.0000 5 11.5559 12.9386 2.3928 0.8303 71.1477 0.0049 0.1146 0.0247 20 15.7796 20.4975 3.9970 0.8134 49.3652 0.2203 6.6274 0.0366 OMN 2 9.0205 3.7445 12.3598 5.4258 1.2094 67.6364 0.2962 0.0186 5 15.9070 6.0842 15.0916 5.3099 1.0789 53.2335 2.2443 0.2046 20 14.2899 8.7138 11.1883 6.1072 0.7580 43.4035 12.6690 0.3938 27 IJAP 0.1054 0.4602 0.1642 0.0312 0.0787 0.1320 0.0044 0.0056 0.4241 0.0092 0.0238 0.0455 0.1304 0.8411 1.0431 0.0046 0.2098 0.3309 0.0840 0.3402 1.1165 0.2091 0.6974 0.5760 0.2448 1.1448 1.6087 0.2919 0.6920 0.4143 0.1072 0.9905 2.6629 0.2887 0.8460 2.4766