islamic equity market integration and volatility spillover between emerging and us stock markets

Please cite this article in press as: Majdoub, J., & Mansour, W. Islamic equity market integration andvolatility spillover between emerging and US stock markets. North American Journal of Economicsand Finance (2014), http://dx.doi.org/10.1016/j.najef.2014.06.011

ARTICLE IN PRESSG ModelECOFIN 453 1–19

North American Journal of Economics and Finance xxx (2014) xxx–xxx

Contents lists available at ScienceDirect

North American Journal ofEconomics and Finance

Islamic equity market integration and volatilityspillover between emerging and US stockmarkets

Jihed Majdouba, Walid Mansourb,∗Q1

a Higher Institute of Management, University of Tunis, Tunisiab Islamic Economics Institute, King Abdulaziz University, Saudi ArabiaQ2

a r t i c l e i n f o

JEL classification:G11G1F3C58

Keywords:Islamic financeVolatility spilloversMultivariate GARCHConditional correlationsBEKKDCCCCC

a b s t r a c t

The purpose of this paper is to study the conditional correlationsacross the US market and a sample of five Islamic emerging mar-kets, namely Turkey, Indonesia, Pakistan, Qatar, and Malaysia. Theempirical design uses MSCI (Morgan Stanley Capital International)Islamic equity index since it applies stringent restrictions to includecompanies. Indeed, two main restrictions must be met: (i) the busi-ness activity must be compliant with Shari’ah (i.e., Islamic law)guidelines and (ii) interest-bearing investments and leverage ratiosshould not exceed upper limits. Three models are used: multivari-ate GARCH BEKK, CCC, and DCC. The estimation results of the threemodels show that the US and Islamic emerging equity markets areweakly correlated over time. No sheer evidence supports that theUS market spills over into the Islamic emerging equity markets.Besides interpreting the results in terms of weak market integra-tion, the peculiar specificities of the Islamic finance industry and theadmittance conditions to the MSCI Islamic equity index contributeto explaining them. Indeed, Islamic finance bans interest-bearinginvestments and imposes some rules, such as asset-backing, whichhas sizeable impacts on volatility spillover and shocks transmis-sions, alongside with the close linkage between real and financial

∗ Corresponding author at: Islamic Economics Institute, King Abdulaziz University, PO Box 80214, Jeddah 21589, Saudi Arabia.E-mail addresses: [email protected] (J. Majdoub), [email protected] (W. Mansour).

http://dx.doi.org/10.1016/j.najef.2014.06.0111062-9408/© 2014 Elsevier Inc. All rights reserved.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

dx.doi.org/10.1016/j.najef.2014.06.011

dx.doi.org/10.1016/j.najef.2014.06.011

http://www.sciencedirect.com/science/journal/10629408

mailto:[email protected]

mailto:[email protected]

dx.doi.org/10.1016/j.najef.2014.06.011



2 J. Majdoub, W. Mansour / North American Journal of Economics and Finance xxx (2014) xxx–xxx

sectors. These findings suggest that investors should take cautionwhen investing in the Islamic emerging equity markets and diver-sifying their portfolios in order to minimize risk.

© 2014 Elsevier Inc. All rights reserved.

1. Introduction

This paper studies the volatility spillovers across regions. It provides empirical evidence fromsix equity stock markets using the conditional variances stemming from multivariate generalizedautoregressive conditional heteroskedasticity (MGARCH) estimations. The objective of this paper is toinvestigate the volatility spillovers between the US stock market and five Islamic emerging stock mar-kets. It explores the extent to which the volatility spillovers are significant and shocks are transmittedacross the pairs of stock markets under consideration. The empirical study of volatility spillovers isinteresting from the particular perspective of portfolio diversification and hedging strategies. Indeed,empirical studies (e.g., Bekaert, Harvey, & Ng, 2003) show that the international portfolio diversifi-cation is impaired by a high integration of international stock markets and correlated stock pricesvolatility.

King and Wadhwani (1990) explain the volatility spillover by the rational attempts of agents touse imperfect information about the events relevant to stock prices. They study the simultaneous col-lapse of financial markets, after the October 1987 crash, although they were operating under differenteconomic circumstances and markets mechanisms. They construct a model in which there is a conta-gion between stock markets because investors use information from price changes in other markets.Indeed, the authors argue that “this constitutes a channel through which a ‘mistake’ in one marketcan be transmitted to other markets.” (p. 5) The transmission of imperfect information regardingprice changes from one market to another means that a mistake is transmitted. Indeed, the empiricalevidence of King and Wadhwani (1990) shows that when the volatility in one market increases thisrenders the size of the contagion effects larger. The increase of the correlation between stock marketsis a consequence after the crash erupted. Duncan and Kabundi (2013) study the features of domesticand foreign sources of volatility transmission in South Africa. They show that the estimated spilloverlevels are dynamic and tend to increase during domestic and foreign extreme moments.

This paper attempts to study the volatility spillovers across the US stock equity market and a sampleof five Islamic emerging markets using MSCI Islamic equity indexes. The research question of the paperis interesting since it sheds some light on the transmission of shocks across international markets,especially during turmoil episodes and investigates volatility spillovers from a different prism, that ofIslamic finance. Indeed, we use Islamic indexes for equity markets that have distinguished specificities.Such indexes apply stringent conditions to include companies in terms of compliance of the businessactivity to Shari’ah and restrictions imposed on interest-bearing investments.

Although several reasons lay behind volatility spillovers and transmission of shocks across markets,the industry of Islamic finance1 provides different explanations. Indeed, since the companies includedin the Islamic equity indexes are supposed to have very low leverage ratios and very small interestinvolvement, this linkage is expected to be broken. A supplementary fact is that the asset-backed rulein Islamic finance ensures that real and financial sectors are closely linked, which does not exposeIslamic emerging markets to volatility spillovers from the US market.

The empirical design aims at studying the conditional correlations using three models, namelyBEKK-MGARCH, CCC and DCC. Our findings suggest that there exist low volatility spillovers betweenIslamic emerging stock markets and the US stock market. All pair countries exhibit weak conditionalcorrelations. Although the estimates are statistically significant in almost all cases, they are very low.

1 The industry of Islamic finance increased remarkedly over the recent past. Indeed, the assets of the industry have beenmultiplied by almost five times over the last five years (Hammoudeh et al., 2013). The market value of Islamic finance’s assetsamounts to 1.6 trillion US dollars.

22

23

2425

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

61

62

63

64

65

66

dx.doi.org/10.1016/j.najef.2014.06.011



J. Majdoub, W. Mansour / North American Journal of Economics and Finance xxx (2014) xxx–xxx 3

Therefore, the transmission of shocks across such equity markets is not significant. The peculiar speci-ficities of the Islamic finance industry contribute to explaining the originality of our empirical findings.The latter have interesting implications for institutional and individual investors to invest in thesemarkets and benefit from portfolio diversification in order to minimize risk.

The rest of the paper is organized as follows. Section 2 presents a brief review of existing empiricalliterature and points out the specificities of Islamic investing. The details of the model specification arepresented in Section 3. Section 4 gives a comprehensive statistical description of the dataset. Sections5 and 6 present the analysis of major empirical findings. Section 7 concludes.

2. Literature review

2.1. Background on volatility spillovers

The study of index return volatility is embedded in the context of modern portfolio theory. Indeed, itis tackled from various perspectives, e.g., asset allocation (Chan, Karceski, & Lakonishok, 1999; Elton &Gruber, 1973), risk management (e.g., Alexander & Leigh, 1997; Beder, 1995), and option pricing (e.g.,Byström, 2002; Gibson & Boyer, 1998). Alongside with these perspectives, the correlation betweenindexes volatilities can explain the observed risk trends among stock markets. Over the recent past, abooming literature either focused on the volatility spillovers between developed and emerging stockmarkets or between emerging or developed markets belonging to the same region.

Engle, Ito, and Lin (1990) maintain that domestic returns could be significantly influenced by foreignreturns, as in the case of Japanese and American stock markets. Ng (2000) examines the magnitude andchanging characteristics of volatility spillovers from the US and Japan. The author finds that regionaland international factors (e.g., cultural and religious) are important drivers of market volatility. Sim-ilarly, Calvo (1999) argues that developed stock markets can act as a conduit for volatility acrossemerging markets in different regions. Dungey, Fry, González-Hermosillo, and Martin (2007) provideempirical evidence on the role of developed markets in volatility transmission across emerging mar-kets. The results of the authors support the above facts on volatility spillovers across regions. Using asimilar methodology, Gebka and Serwa (2007) find mixed results on volatility spillovers among theemerging capital markets in Eastern Europe, East Asia and Latin America.

Beirne, Guglielmo, Caporale, Schulze-Ghattas, and Spagnolo (2009) model volatility spillovers frommature to emerging stock markets by examining the changes in the transmission mechanisms duringturbulences in mature markets. The authors use trivariate returns GARCH BEKK model for 41 (localand regional) emerging market economies and show that many emerging markets are affected bythe volatility of the mature market. The conditional correlations between local emerging and maturemarkets tend to increase in turbulent episodes. Mukherjee and Mishra (2008) investigate the returnand volatility spillovers between Indian stock market and 12 developed and emerging Asian countriesto examine the stock market integration. The authors show that four Asian markets (Hong Kong, Korea,Singapore, and Thailand) have a significant flow of information in India. Zheng and Huo (2013) examinethe volatility spillover effect among a sample of developed markets including US, UK, Germany, Japanand Hong Kong. They introduce a Markov switching causality method in order to model the instabilityof volatility spillover relationships over market tranquil and turmoil periods. They show sheer evidenceof spillover effects among the markets under consideration. More specifically, they show that thebilateral volatility is striking during crisis periods, especially during the last subprime mortgage crisis.

Martinez and Ramirez (2011) use data for a group of five Latin American countries to study thevolatilities of asset returns. The authors show that the dynamic multivariate models are more pow-erful than the constant conditional correlation models. The major findings suggest that, although theco-volatility of the region is still moderate, the domestic volatilities within the markets have beenincreasing. Indeed, the authors document that the DCC model shows a long memory of asset marketssince the results maintain a moderate reaction to shocks with high persistency. As a consequence of2008 Subprime crisis, some clusters of volatility are identified in the Latin American equity markets.

Duncan and Kabundi (2013) investigate the specificities of domestic and foreign sources of volatil-ity transmission in South Africa using a sample of daily observations ranging from October 1996 toJune 2010. The results show that the estimated spillover levels are time-varying and increase during

67

68

69

70

71

72

73

74

75

76

77

78

79

80

81

82

83

84

85

86

87

88

89

90

91

92

93

94

95

96

97

98

99

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

dx.doi.org/10.1016/j.najef.2014.06.011




domestic and foreign extreme episodes. Indeed, the domestic spillovers of 38% exceed average foreignspillovers of 4.7%, as well as the highest domestic spillovers estimate in the US for a similar sampleperiod. The authors claim: “a stylized fact that, when a financial system is connected (via fundamen-tals, sentiment, geographical location, or some other common factor), a volatility shock to one assetcreates added uncertainty regarding future values of other assets in the system. The externalities tomarket volatility strengthen during crisis periods; that is, spillovers are at a maximum exactly whenprices are declining most rapidly. Hence, cross-market volatility linkages pose a threat to financialstability” (p. 572).

Jiang, Konstantinid, and Skiadopoulos (2012) study the extent to which the US and European stockmarkets spill over as a consequence of news announcements. The authors find significant spilloversof volatility across the US and European stock markets as well as within European ones. The resultsof the authors depend on the nature of news release, scheduled or unscheduled. Indeed, “scheduled(unscheduled) news releases resolve (create) information uncertainty, leading to a decrease (increase)in implied volatility. Nevertheless, news announcements do not fully explain the volatility spillovers,although they do affect the magnitude of volatility spillovers. Our results are robust to extreme marketevents such as the recent financial crisis and provide evidence of volatility contagion across markets”(p. 2260). Singh, Kumar, and Pandey (2010) examine return and volatility spillovers across NorthAmerican, European, and Asian stock markets. Fifteen indexes from these three regions are considered.The volatility spillover is modeled using an AR-GARCH model. The empirical results show that volatilityspills over across regions, particularly between Asian and European stock markets.

Ben Nasr, Ajmi, and Gupta (2013) study the conditional volatility of the Dow Jones Islamic MarketWorld Index (DJIM2) by taking account of both the long memory and structural changes in the char-acterization of the structure of financial returns volatility. Following Ben Nasr, Boutahar, and Trabelsi(2010), the authors allow the parameters of the conditional variance to be time-varying and allow thechange of the parameters to evolve smoothly over time. Indeed, they use two models, namely the Frac-tionally Integrated Generalized Autoregressive Conditional Heteroskedasticity (FIGARCH) model andthe Fractionally Integrated Time Varying Generalized Autoregressive Conditional Heteroskedasticity(FITVGARCH) model. Based on the model diagnostics, the results of the authors show that the FITV-GARCH model dominates the FIGARCH model in terms of performance, which stresses the importanceof including both the long-memory and structural changes in modeling the volatility of returns.

Al-Khazali, Lean, and Samet (2013) compare the performance of 18 Islamic and conventional DowJones indexes using their daily returns. They apply the stochastic dominance3 and show that all con-ventional indexes dominate the Islamic indexes. However, the domination of the conventional indexesdisappears during the periods of financial crises. Indeed, the US and European indexes dominate theirconventional counterparts during the recent financial crisis. Similar to Ben Nasr et al. (2013), Al-Khazaliet al. (2013) contribute to the literature of Islamic finance by documenting the ethical dimension ofIslamic investing.

Yusof and Majid (2007) investigate whether the conditional volatilities of Malaysian conventionaland Islamic stock markets are related to that of monetary policy variables such as the narrow moneysupply, the broad money supply, interest rates, and exchange rates. Kuala Lumpur Composite Index(KLCI) is used to measure the conventional stock market return, while Rashid Hussain Berhad IslamicIndex (RHBII) is used to measure the Islamic stock market return. The findings are interesting sincethe volatility of interest rate does not affect that of RHBII, whilst it does for KLCI. This particularresult confirms the non-interference of interest rate with the stock market volatility. Furthermore,the stabilization of interest rate as a monetary policy variable would be neutral to the volatility of

2 DJIM is used by other empirical studies. For instance, Hammoudeh et al. (2013) justify its use in terms of comprehensiverepresentativeness. Further, they claim that it is the most suitable for time-series studies of Islamic indices.

3 The stochastic dominance method is more robust than the widely used parametric methods, namely mean-variance andCAPM methods. Indeed, the latter methods may yield conflicting results in analyzing portfolio performance since they useparametric statistics that rely on the normality hypothesis of stock returns and quadratic utility functions. If the returns arenot normal and the investors’ utility functions are not quadratic, the results are not suitable. The stochastic dominance methodranks investments but does not depend on these two requirements and is not limited to the first two moments of returns.

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

143

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

dx.doi.org/10.1016/j.najef.2014.06.011




Islamic stock markets. In contrast, the interest rate volatility is an important driver of macroeconomicinstability, which increases consequently the return of stock market volatility during turmoil periods.

Although the study of Yusof and Majid (2007) was conducted prior to the last global financial crisis,their findings were corroborated by empirical studies (e.g., Al-Khazali et al., 2013; Hammoudeh et al., Q32013) exploring the role of ethical dimension of Islamic investing in alleviating the impairing impactsof stock market volatility. Hammoudeh et al. (2013) show that the global financial crisis has a lowerimpairing impact on the Islamic market, defined based on the Dow Jones stock universe, by comparisonto three conventional markets, namely the United States, Europe, and Asia. The restrictions imposed bythe principles of Shari’ah seem to have their positive side since they alleviate the impact of the globalfinancial crisis’ negative shocks. In addition, Hammoudeh et al. (2013) document interesting empiricalfindings particularly by showing that the dichotomy hypothesis does not hold true. Indeed, the Islamicmarket turns out to have a positive and stimulating effect on the three conventional markets.4

2.2. Islamic investing

Islamic investing differs from the conventional one on many counts. Indeed, the institutionaland individual investment decisions are restricted by a range of rules determined by scholars whoare members of Shari’ah5 Boards. Those scholars use Shari’ah in order to determine the generalrules according to which the investment decisions are permissible. The main prohibitions are riba,gharar/maysir, and illicit and unethical business activities. Firstly, the prohibition of riba (usury) is notexclusive in Islam. It is also prohibited in Judaism and Christianity (Schoon & Nuri, 2012). Its prohibi-tion is based on the fact that it is considered as a profit gained without running risks. Ariss (2010, p.102) claims that “the prohibition of interest is not exclusive to Islam, but common to all three Abra-hamic faiths. Although the Koran does not explicitly justify the prohibition of dealings based on apre-determined rate of interest, it is believed that the primary reason for doing so is to remove anyform of injustice in business transactions.” Islam prohibits both payment and receipt on the basis ofa predetermined rate. In fact, money itself has no intrinsic value and no reward for time preference.Bonds and other interest-bearing securities are avoided, while long-term equity investments couldprotect against inflation.

The second prohibition is related to gharar/maysir. Gharar is related to uncertainty. Schoon (2009)maintains that there are two ways to interpret gharar. First, it applies exclusively to situations ofdoubtfulness or uncertainty as in the case of knowing whether something will take a place or not.Second, it applies to the unknown. The combination of these two cases is always taken as gharar.Maysir is associated with games of gambling. It occurs when one party looses a fraction or all hiswealth. Maysir is one example of gharar. However, the latter is more general. Shari’ah formally forbidscontracts and transactions that contain or are based on maysir and/or gharar.

The last prohibition is realted to the illicit and unethical acivities that are also a restriction toinvestment from the perspective of Shari’ah. Such acitivities are embedded in a sort of a black listincluding investing in activities violating human rights, child labor, armament, pork, and pornography.The challenge to any Shari’ah-compliant portfolio manager is to pick up winning stocks according tothe aforementioned restrictions. Indeed, the portfolio manager is supposed to manage stocks withthe specificities that are like socially responsible investment with the peculiar feature of avoidingparticularly interest-bearing stocks.6

4 The innovation of the contribution of Hammoudeh et al. (2013) consists in using appropriate econometric specifications.Indeed, the previous studies are alternating since they did not reach a consensus regarding the sign of the spillover effects ofIslamic markets on conventional markets. Hammoudeh et al. (2013) use the threshold and Markov-switching models to explorethe spillover effects.

5 Shari’ah can be translated as the Islamic law. It has two primary sources, namely Qur’an (Muslims Holy Book) and Hadith(the Prophet’s sayings, deeds and his endorsement of his Companions’ practices).

6 Nienhaus (2011) calls interest-bearing investment “interest pollution”. Shari’ah scholars set a permissible threshold toinvest in firms having activities with interest pollution. In effect, if they forbid investing in firms paying financial fees on debtsand/or earning risk-free financial returns, then the set of permissible investments will almost be empty. Shari’ah scholars allowfor such a permissible threshold on approval that a fraction of generated dividends will be donated for social purposes.

163

164

165

166

167

168

169

170

171

172

173

174

175

176

177

178

179

180

181

182

183

184

185

186

187

188

189

190

191

192

193

194

195

196

197

198

199

200

201

202

203

dx.doi.org/10.1016/j.najef.2014.06.011




Alongside with the aformentioned prohibitions, Islamic investing is based on major principles.The most important principles are profit-and-loss sharing and asset-backing. Profit-and-loss sharingmeans that the contracting parties should share the generated streams of cash flows, either hit by goodor bad stochastic shocks. Asset-backing means that financing is based on illiquid assets that createreal assets, since Islamic finance does not recognize money as the sole subject-matter of financialtransactions.

Investors in Muslim countries such as Malaysia, Turkey, and Indonesia, among others, can invest inthe stock market via Islamic equity funds7 (IEFs). The latter are the Islamic counterpart to conventionalmutual funds, which are investment vehicles that allow investors with lack of knowledge, skills, ortime to manage their own wealth to benefit from the returns of international equity markets. As amatter of fact, the IEFs are not allowed to invest in the sectors that are not compliant with Shari’ah.As reported by Failaka Advisors,8 the IEFs had grown to around US$20 billion in assets by February2008, reaching 5 times the value of 2003. Failaka Advisors maintains that the IEFs had grown rapidly,tripling over the previous five years. Although most of IEFs are located in the Gulf Cooperation Council(GCC) countries (especially Saudi Arabia and Bahrain), many investment banks in the world offer thesefunds, such as Citibank, Merrill Lynch, HSBC, and Deutsche Bank. Hayat and Kraeussl (2011) study theregional performance of IEFs and compare them to the conventional benchmarks. They report that thenumber of IEFs in Malaysia and Asia-Pacific amount to 51 and 14, respectively, whilst the number ofIEFs in the Middle-East region is only three.

A variety of equity benchmarks have been launched as a consequence of the rapid growth of theIFEs. The most important ones are FTSE Global Islamic Index Series, Dow Jones Islamic Market Index,and MSCI (Morgan Stanley Capital International) Islamic Index. Most of the Islamic equity bench-marks have a common device in their screening. Indeed, all stocks that fail to meet the guidelinesof Shari’ah are deleted. Since the empirical design of the paper will focus on MSCI Islamic Index, wewill particularly deal with its methodology. Severe conditions are applied to screen stocks. After thefirst screen related to the business industry, the second one consists in some financial filters, namelyratios, to guarantee that the company does not excessively earn from interest-bearing instrumentsand its financial leverage is not high. For instance, none of the financial ratios9 is allowed to exceed33.33%. Accordingly, stocks are deleted if any of the financial ratios exceeds 33.33%. There are somepurification factors applied to securities. For example, if a company earns a fraction of its incomefrom interest-bearing activities and prohibited business activities, a percentage of dividends must behanded over by shareholders for charity purposes.

3. Model specification

A multivariate MGARCH model is utilized to capture the dynamic relationship between indexes.Firstly, we employ the BEKK model of Engle and Kroner (1995). The authors propose this model becauseit has a good property according to which the conditional covariance matrices are positive definiteby construction. Secondly, we use the multivariate GARCH with constant conditional correlation ofBollerslev (1990) and the multivariate GARCH model with the dynamic conditional correlation of Engle(2002).

7 The first IEF, Amana Income Fund, was launched in North America in June 1986 by members of the North American IslamicTrust (Unified Management Corporation, Indianapolis, IN). This fund is the counterpart of an American trust. It still servesMuslims and non-Muslims in the United States.

8 For a complete analysis of the growth of IEFs, refer to: www.failaka.com.9 MSCI uses three financial ratios to screen companies, namely total debt over total assets, the sum of a company’s cash and

interest bearing securities over total assets, and the sum of a company’s accounts receivables and cash over total assets.

204

205

206

207

208

209

210

211

212

213

214

215

216

217

218

219

220

221

222

223

224

225

226

227

228

229

230

231

232

233

234

235

236

237

238

239

240

241

242

dx.doi.org/10.1016/j.najef.2014.06.011

http://www.failaka.com/




3.1. Conditional volatility: MGARCH (1,1) model

Bollerslev, Engle, and Wooldridge (1988) propose a VECH-GARCH model where the conditionalvariance and covariance are a function of all lagged conditional variance and covariance. The model isgiven by:

vech(Ht) = A0 +q∑

i=1

Aivech(εt−1ε′t−1) +

p∑

i=1

Givech(Ht−i) (1)

where ‘vech’ is the operator that stacks the lower triangular portion of a symmetric matrix into avector. A0 is a k(k + 1)/2 ×1 vector, and Ai and Gi are k(k + 1)/2 × k(k + 1)/2 matrices of parameters. Theproblem in this formulation of multivariate GARCH model is that the number of parameters is verylarge. The BEKK (1,1) model has the following form:

Ht = C ′C + A′εt−1ε′t−1A + G′Ht−1G (2)

where C is a k × k lower triangular matrix of constants, A and G are k × k matrices. The diagonal param-eters in A and G measure the effects of past shocks and past volatility of ith market on its conditionalvolatility, respectively. The off-diagonal elements in matrix A(aij) and G(gij) measure respectively thecross-market effect of shock spillover and the cross-effect of volatility spillover. The parameters ofthe BEKK model can be obtained by using the maximum likelihood estimation assuming a normaldistribution of errors. The following likelihood function is maximized

L(�) = −T log(2�) − 12

T∑

t=1

(log∣∣Ht

∣∣ + ε′tH

−1t εt) (3)

where T is the number of observations and � is the vector of parameters to be estimated. We usenumerical maximization techniques to maximize the non-linear likelihood function. We use both thesimplex and BFGS (Broyden–Fletcher–Goldfarb–Shanno) algorithms to obtain the initial condition andthe final parameter estimates of the variance–covariance matrix.

3.2. Conditional correlation: the CCC and DCC models

Bollerslev (1990) introduced a class of multivariate GARCH models, specifically the ConstantConditional Correlation (CCC-MGARCH) in which the conditional correlations are time-invariant.Accordingly, the conditional variances are proportional to produce the corresponding conditionalstandard deviations. This restriction greatly reduces the number of unknown parameters and thussimplifies the estimation. The conditional covariance matrix can be expressed as follows:

Ht = DtRDt = (�ij√hii,thjj,t) (4)

where R = (�ij = E(�t�′t)) is a symmetric k × k definite positive matrix containing the conditional cor-

relation �ij, εt = Dt�t, �t is an i.i.d. random vector and Dt = diag(h1/211t ......h

1/2kkt

). This model assumesthat the conditional variance hiit follows a univariate GARCH model. Indeed, this is shown through thefollowing equation:

hiit = wi +q∑

j=1

˛ijε2i,t−j +

p∑

j=1

ˇijhii,t−j, i = 1, ..., k (5)

However, the assumption that the random shocks have a time-invariant conditional correlationmay not be supported empirically. In order to let the conditional correlation matrix be time-variant,Christodoulakis and Satchell (2002), Tse and Tsui (2002), Engle (2002), and Engle and Sheppard (2001)proposed a generalization of the CCC model.10 Tse and Tsui (2002) introduced a varying correlation

10 For more details, see Bauwens et al. (2006).

243

244

245

246

247

248

249

250

251

252

253

254

255

256

257

258

259

260

261

262

263

264

265

266

267

268

269

270

271

272

273

274

275

276

277

278

279

dx.doi.org/10.1016/j.najef.2014.06.011




GARCH model in which the conditional correlations are a function of the conditional correlations of theprevious period. The dynamic conditional correlation (DCC) of Tse and Tsui (2002) has the followingform:

Ht = DtRDt (6)

Rt = (1 − �1 − �2)R + �1 t−1 + �2Rt−1 (7)

where Dt is defined as in Eq. (4), R is a symmetric k × k positive definite parameters matrix with unitdiagonal elements, t−1 is the k × k correlation matrix of the past P standardized residuals (ε̂t−1...ε̂t−p).A necessary condition to ensure the positivity of t−1 is P ≥ k and �1 and �2 are non-negative scalarparameters satisfying �1 + �2 < 1 ensuring that R is mean reverting. Moreover, Engle (2002) proposed adifferent dynamic conditional correlation model. The DCC model of Engle assumes that the covariancematrix is decomposed as follows:

Ht = DtRtDt

Rt = diag(q1/211t ......q

1/2kkt

)Qtdiag(q1/211t ......q

1/2kkt

) (8)

where Qt is a symmetric k × k positive definite matrix containing the conditional covariance of stan-dardized residuals given by:

Qt = (1 − �1 − �2)Q0 + �1�t−1�′t−1 + �2Qt−1 (9)

where Q0 is the unconditional covariance matrix of �t, �t is defined as in Eq. (5), �1 and �2 are non-negative scalar parameters satisfying �1 + �2 < 1, �1 represents the impact of last shocks on a currentconditional correlation and �2 captures the impact of the past correlation. If �1 and �2 are statisti-cally significant, the conditional correlations are not constant. Engle (2002) shows that the likelihoodfunction can be written as:

L(�) = − 12

T∑

t=1

(log 2� + 2 log∣∣Dt

∣∣ + log∣∣Rt

∣∣�tR−1t �t) (10)

The DCC model can be estimated consistently in two stages. First, Qt is used to calculate the dynamicconditional correlation:

�ij,t = qij

(qii,tqjj,t)1/2

(11)

Second, �ij,t is used to estimate conditional covariance:

hij,t = �ij(hii,thjj,t)1/2 (12)

where hii,t(hjj,t) and hij,t are the conditional variance and conditional covariance generated by usingunivariate GARCH models.

4. Data description

We use daily data over January 2008 through January 2013 of five Islamic indexes,11 namelyIndonesia, Malaysia, Pakistan, Qatar, Turkey, alongside with the US index, which gives 1306 obser-vations. The indexes are provided by Morgan Stanley Capital International (MSCI) and expressed inUS Dollar so as to get a homogenous dataset and take into account the currency risk. The MSCI Islamicindex reflects Shari’ah investment guidelines and is designed to measure the performance of the large,mid and small cap segments across markets that are relevant for Islamic investors. Table 1 presents

11 It is common in the previous studies to study the volatility spillover on a regional basis through the study of economies inone region, e.g., Middle-East North Africa (MENA), Gulf Countries Council (GCC). In our context, the choice of the five countries isbecause they are emerging economies and they have an MSCI Islamic index since they are among the first countries to developIslamic finance.

280

281

282

283

284

285

286

287

288

289290

291

292

293

294

295

296

297

298

299

300

301

302

303

304

305

306

307

308

309

310

311

312

313

314

315

dx.doi.org/10.1016/j.najef.2014.06.011




Table 1Stock markets and Indexes.

Stock Market Benchmark

USA USA MSCI Islamic IndexPakistan Pakistan MSCI Islamic IndexIndonesia Indonesia MSCI Islamic IndexTurkey Turkey MSCI Islamic IndexQatar Qatar MSCI Islamic IndexMalaysia Malaysia MSCI Islamic Index

Table 2aSummary statistics for Islamic daily stock returns (log differences).

Stock Market Min. Max. Mean Median Std. Dev Skewness Kurtosis Jarque–Bera

USA −0.09 0.11 0.000081 −0.004912 0.015032 −0.102585 8.825341 4237.375277Pakistan −0.10 0.08 −0.000078 −0.000701 0.018142 −0.377816 3.253267 606.5383030Indonesia −0.16 0.16 −0.000079 −0.004867 0.022145 −0.298003 7.780420 3310.902808Turkey −0.15 0.17 −0.000112 −0.002791 0.024713 −0.145173 4.891100 1305.389491Qatar −0.15 0.10 −0.000169 0.000375 0.017586 −0.821378 15.68229 13,519.41331Malaysia −0.10 0.05 0.000042 0.002089 0.011601 −0.811149 9.111220 4657.010807

Table 2bCorrelation matrix of market return.Q4

Indonesia Malaysia Pakistan Qatar Turkey USA

Indonesia 1.0000Malaysia 0.5018 1.0000Pakistan 0.0619 0.0916 1.0000Qatar 0.3693 0.2739 0.0198 1.0000Turkey 0.3156 0.3430 0.0158 0.1690 1.0000USA 0.1471 0.1556 0.0037 0.0093 0.4063 1.0000

the main indexes we use. Table 2a gives some descriptive statistics about these indexes with logdifferences. Turkey has the highest spread ranging between − 0.15 and 0.17. Indonesia is ranked sec-ond with a spread ranging from −0.16 and 0.16. This means that the Turkish market undergoes largefluctuations compared to the other markets, seeing it has the most extreme values.

The return volatility of the Turkish market is the highest. Its standard deviation amounts to0.024713 followed by the Indonesian market. The fact that Turkey exhibits the highest standard devi-ation means that it encourages the inflow of short-term capital, which is highly volatile. Abu-Zaid(2011) studies the volatility spillover among the US and UK markets and three economies from theMiddle East (Egypt, Turkey, and Israel). He finds that Turkey has the highest volatility, which is inter-preted in terms of smallness and openness of the stock market and vulnerability to global shocks. Thethird-order moments of the distributions (the skewness coefficients) indicate that the tail of the leftside of the probability density function is fatter for all indexes than the right side.

Qatar has the fattest left-skewed Index distribution. The fourth-order moments (the kurtosiscoefficients) are positive and higher than 3 for all countries, reflecting Leptokurtic distributions thatare sharper than the normal distribution. Table 2b reports the correlation matrix across stock marketsreturns.

Fig. 1 shows two plots, namely the daily series of indexes and the daily indexes returns for the 6countries during the sample period spanning over 2008–2012. Although the index series have almostthe same trend over time, they clearly cross each others. Indeed, the Qatari index crosses simulta-neously the Malaysian and Turkish ones. Similarly, the Pakistani index crosses the US, Malaysian, andTurkish series. The fact that the indices cross each others one or two times means that they exhibit anegative correlation. The second plot show that the index returns in log-differences

Fig. 2 shows 5 panels of the country pairs’ daily moving correlations of returns. The Turkey–USApair exhibits the highest correlation since both indices are highly bonded over the sample period.

316

317

318

319

320

321

322

323

324

325

326

327

328

329

330

331

332

333

334

335

336

337

338

339

dx.doi.org/10.1016/j.najef.2014.06.011




Fig. 1. Time series plots of Islamic MSCI Index – levels and returns.

With the exception of this pair, the remaining pairs show low correlations. Fig. 3 shows the movingdaily correlation coefficients of the returns in country pairs with windows of 2008. Indeed, when theSubprime crisis erupted, the coefficients of correlations of classical (i.e., non-Islamic) indices werepositive and extremely high. Nonetheless, the MSCI Islamic indices show low correlations for thesame period. The highest correlation is exhibited by the pair Turkey–USA, which reached 0.4 in lateApril 2008. Similarly, the country pairs Indonesia–USA and Malaysia–USA exhibit correlations almostequal to 0.4, but did not hit it though. The remaining pairs have very low coefficients that barely go

340

341

342

343

344

345

346

dx.doi.org/10.1016/j.najef.2014.06.011




USA/PAKISTAN I SLAMIC INDE X RETUR N

20122011201020092008-0.15

-0.10

-0.05

0.00

0.05

0.10

0.15RDUSA

RDPA KISTAN

USA/TUR KEY ISLAMIC I NDEX RE TURN

20122011201020092008-0.20-0.15-0.10-0.05-0.000.050.100.150.20

RDTURKE Y

USA/MALAYSIA ISLAMIC IND EX RETUR N

20122011201020092008-0.15

-0.10

-0.05

0.00

0.05

0.10

0.15

RDUSA

RDMALAYS IA

USA/INDONES IA ISL AMIC I NDEX RETU RN

20122011201020092008-0.20-0.15-0.10-0.05-0.000.050.100.150.20

USA/QATAR ISLAMIC INDEX RE TURN

20122011201020092008-0.20

-0.15

-0.10

-0.05

-0.00

0.05

0.10

0.15RDUSA

RDQATAR

Fig. 2. Country pairs’ daily moving correlation coefficients.

beyond 0.1. Further, in some cases the coefficients are negative, which could be interpreted in termsof negative correlations between Islamic indices in emerging economies and the US.

The latter fact is at odds with the studies on correlations documented during the Subprime crisis.For instance, while Beirne et al. (2009) study the volatility spillovers from mature to emerging stockmarkets in order to examine the changes in the transmissions mechanisms, they show that the condi-tional correlations between local emerging and mature markets tend to increase in turbulent periods.However, the Islamic indices show different correlation coefficients, which sheds some light on thetransmission of shocks and volatility spillovers among the US and emerging markets in our sample.The composition of Islamic indices and the characteristics of stocks play a crucial role in lowering thecorrelation coefficients, especially during crisis periods. The investors have a higher confidence in thestocks included in MSCI equity Islamic indexes, which explains the extent to which their returns aremoderated relative to the ones included in non-Islamic indexes.

5. Volatility spillover

In this section, we examine the estimates results of time-varying variance–covariance by the BEKK(1,1) model. The existence of any causal relation among variance and covariance included in Ht impliesthat the off-diagonal coefficients of the matrices A(aij) and G(gij) are statistically significant. The mostimportant feature of the BEKK model is that it can explain the causality relation among both varianceand covariance. The results of estimated BEKK model are shown in Table 3. Throughout the empiricalinvestigation, we denote the countries USA, Turkey, Qatar, Pakistan, Malaysia, and Indonesia by 1, 2,3, 4, 5, and 6, respectively.

The forecast of the volatility is very complicated with the imperfections observed in the market,especially during crises and extreme episodes. Indeed, the estimation of volatility turns out to becomplicated since it depends on several factors of the same market and other ones. This is shownthrough the transmission of shocks to several markets in the world. From the same point of view and

347

348

349

350

351

352

353

354

355

356

357

358

359

360

361

362

363

364

365

366

367

368

369

370

dx.doi.org/10.1016/j.najef.2014.06.011

Please cite

this

article in

press

as: M

ajdou

b, J.,

& M

ansou

r, W

. Islam

ic equ

ity m

arket in

tegration an

dvolatility

spillover

between

emergin

g an

d U

S stock

markets.

North

Am

erican Journal

of Econom

icsand

Finance (2014),

http

://dx.d

oi.org/10.1016/j.najef.2014.06.011

AR

TIC

LE

IN P

RE

SS

G M

odelEC

OFIN

453 1–19

12

J. M

ajdoub, W

. M

ansour /

North

Am

erican Journal

of Econom

ics and

Finance xxx

(2014) xxx–xxx

Table 3Estimates results of multivariate GARCH BEKK(1,1) Model.

Variable Coefficient T-statistic Prob. Variable Coefficient T-statistic Prob.

Mean(1) 0.000787552 2.88425 0.00392342* a(4,4) −0.136791259 −6.96752 0.00000000*Mean(2) 0.000924495 1.53299 0.12527828** a(4,5) 0.006065590 0.52158 0.60196120Mean(3) 0.000060477 0.24497 0.80648113 a(4,6) −0.017017848 −0.82643 0.40855832Mean(4) 0.000766187 1.84340 0.06526992 a(5,1) 0.236962582 8.34689 0.00000000*Mean(5) 0.000768214 3.07922 0.00207543* a(5,2) 0.164271732 3.37586 0.00073584*Mean(6) 0.000919869 1.97027 0.04880697** a(5,3) 0.002930430 0.13048 0.89618758c(1,1) 0.001465152 6.58737 0.00000000* a(5,4) −0.014081355 −0.47339 0.63593594c(2,1) −0.003351720 −7.85731 0.00000000* a(5,5) −0.139321340 −5.88187 0.00000000*c(2,2) 0.004515669 8.21834 0.00000000* a(5,6) −0.108966414 −2.32550 0.02004498c(3,1) 0.000184016 1.16459 0.24418699 a(6,1) −0.017203683 −1.22127 0.22198306c(3,2) 0.000493554 3.10726 0.00188833* a(6,2) −0.102266702 −4.12734 0.00003670*c(3,3) −0.000031043 −0.16215 0.87118866 a(6,3) 0.001215362 0.09133 0.92723043c(4,1) −0.000185450 −0.75567 0.44984732 a(6,4) −0.064636521 −3.64594 0.00026642*c(4,2) −0.000535718 −2.14154 0.03223018* a(6,5) 0.007641866 0.67041 0.50259454c(4,3) 0.000269150 0.89074 0.37306852 a(6,6) −0.058236484 −2.53084 0.01137897**c(4,4) 0.001300223 3.39946 0.00067518* g(1,1) 0.830734019 96.31732 0.00000000*c(5,1) −0.000381706 −1.48933 0.13640007 g(1,2) −0.219699504 −9.82345 0.00000000*c(5,2) 0.000808505 4.16624 0.00003097* g(1,3) 0.009200513 1.04268 0.29709811c(5,3) 0.000110827 0.46008 0.64545787 g(1,4) −0.096708044 −7.75928 0.00000000*c(5,4) 0.000381701 1.83457 0.06656898 g(1,5) −0.083218881 −6.95869 0.00000000*c(5,5) 0.000096335 0.45755 0.64727259 g(1,6) −0.256293368 −11.83181 0.00000000*c(6,1) −0.000315194 −0.66972 0.50303898 g(2,1) 0.087804152 12.97478 0.00000000*c(6,2) 0.001626235 4.68361 0.00000282* g(2,2) 0.954939489 78.59231 0.00000000*c(6,3) 0.000300183 0.70531 0.48061951 g(2,3) −0.014689354 −3.37707 0.00073263*c(6,4) 0.001061694 2.36808 0.01788067** g(2,4) 0.028675825 4.49379 0.00000700*c(6,5) 0.000205224 0.55704 0.57749702 g(2,5) −0.010304753 −1.50272 0.13291117c(6,6) 0.000002083 0.00657 0.99475703 g(2,6) 0.022667618 1.66574 0.09576561a(1,1) −0.082407673 −3.29080 0.00099904* g(3,1) −0.068432108 −9.26542 0.00000000*a(1,2) −0.582409640 −13.98583 0.00000000* g(3,2) 0.023533599 2.20652 0.02734790**

dx.doi.org/10.1016/j.najef.2014.06.011

Please cite

this

article in

press

as: M

ajdou

b, J.,

& M

ansou

r, W

. Islam

ic equ

ity m

arket in

tegration an

dvolatility

spillover

between

emergin

g an

d U

S stock

markets.

North

Am

erican Journal

of Econom

icsand

Finance (2014),

http

://dx.d

oi.org/10.1016/j.najef.2014.06.011

AR

TIC

LE

IN P

RE

SS

G M

odelEC

OFIN

453 1–19

J. M

ajdoub, W

. M

ansour /

North

Am

erican Journal

of Econom

ics and

Finance xxx

(2014) xxx–xxx

13

Table 3(Continued)

Variable Coefficient T-statistic Prob. Variable Coefficient T-statistic Prob.

a(1,3) −0.070914985 −3.73715 0.00018612* g(3,3) 0.961302756 212.84229 0.00000000*a(1,4) −0.071075750 −3.59958 0.00031874* g(3,4) −0.006444306 −1.07543 0.28218271a(1,5) −0.206662030 −12.16537 0.00000000* g(3,5) 0.002572361 0.53552 0.59228841a(1,6) −0.377610810 −10.84861 0.00000000* g(3,6) −0.004239723 −0.45427 0.64963538a(2,1) 0.027773229 2.07840 0.03767281** g(4,1) 0.043203303 7.07305 0.00000000*a(2,2) −0.026585161 −1.11002 0.26698917 g(4,2) −0.030543369 −3.12722 0.00176466*a(2,3) −0.019324230 −1.97677 0.04806724** g(4,3) 0.005645245 1.85176 0.06406083a(2,4) 0.018034232 1.35124 0.17661996 g(4,4) 0.978350891 182.41136 0.00000000*a(2,5) 0.001401006 0.13899 0.88945997 g(4,5) −0.007776802 −2.01640 0.04375828**a(2,6) 0.040424761 1.99093 0.04648885** g(4,6) 0.001908922 0.23543 0.81387861a(3,1) 0.048833852 3.03250 0.00242534 g(5,1) 0.067169739 5.71208 0.00000001*a(3,2) 0.035584352 1.14040 0.25412175 g(5,2) 0.024709041 1.39825 0.16203801a(3,3) 0.313316472 14.38163 0.00000000* g(5,3) −0.003289891 −0.72669 0.46741652a(3,4) 0.050224307 2.72262 0.00647669* g(5,4) 0.023627815 3.01881 0.00253771*a(3,5) 0.051238431 3.81328 0.00013714* g(5,5) 0.983435401 195.20223 0.00000000*a(3,6) 0.132495867 4.79964 0.00000159* g(5,6) −0.015634054 −1.10837 0.26770198a(4,1) 0.114257738 8.83053 0.00000000* g(6,1) 0.071001577 10.82705 0.00000000*a(4,2) 0.004107348 0.17223 0.86325733 g(6,2) −0.050406511 −5.13026 0.00000029*a(4,3) −0.012857502 −1.10333 0.26988386 g(6,3) −0.003973344 −1.34568 0.17840572

g(6,4) −0.033676248 −6.57866 0.00000000*g(6,5) −0.011251250 −2.55773 0.01053579**g(6,6) 0.971940234 148.70537 0.00000000*

Notes: This table shows the estimates of the multivariate GARCH BEKK (1,1) model.‘*’ and ‘**’ denote respectively the 1% and 5% significance levels. 1, 2, 3, 4, 5, 6 denote respectively: USA,Turkey, Qatar, Pakistan, Malaysia, and Indonesia. The parameters cij , aij , and gij are the off-diagonal elements of the matrices C, A and G, as presented in Section 3.

dx.doi.org/10.1016/j.najef.2014.06.011




Fig. 3. Moving daily correlation coefficients of returns in country pairs with windows of 2008.

after the Subprime crisis, the international banks lost enormously, whereas Islamic banks had bornelimited losses. This fact pushed academicians and practitioners to particularly take into account thespecificities of the financial instruments of Islamic banks as well as the peculiar characteristics of theIslamic finance industry. Analogically, the Islamic financial markets have a moderate volatility relativeto those of the non-Islamic emerging markets. The financial markets of Islamic countries exhibit a weakcorrelation of volatility with the US market as well as with the other emerging markets.

The estimation results of BEKK show that the majority of pairs are statistically significant. Theoff-diagonal elements of the matrix A reflect the cross past shocks. For instance, the coefficient a(1,4)is equal to −0.071 and is statistically significant at 1%. It shows that the cross past shocks are nottransmitted from the US stock market to the Pakistani stock market. This means that, when shocks hitthe US market, the Pakistani market does not capture them. The coefficient a(4,1) reflects the sameeffect but in the opposite direction. It exhibits a low value (0.11) that is statistically significant.

The off-diagonal coefficients of the matrix G reflect the cross past conditional volatility of theother markets. For instance, the coefficient of g(1,4) is equal to −0.09 and is significant at the 1%significance level. This means that the US market does not spill over the past conditional volatility tothe Pakistani market. In other terms, the volatility of the MSCI Islamic equity index of Pakistan does notdepend on the volatility of the US. The coefficient g(4,1) is equal to 0.04 and is statistically significant.

371

372

373

374

375

376

377

378

379

380

381

382

383

384

385

386

387

dx.doi.org/10.1016/j.najef.2014.06.011




Consistently with previous literature, the volatility spillover effects are not symmetric, which meansthat the markets do not transmit shocks uniformly. Evidence demonstrates that Malaysia is the maintransmitter among the US and Islamic emerging markets. Indeed, g(5,1) is equal to 0.067. The volatilitytransmission from Malaysia to the US amounts to 6.7%, which implies that a 1% increase in returnsof MSCI Islamic index for Malaysia transmits 6.7% volatility to the US index. By comparison to non-Islamic indices, the conditional correlations are very low. For example, Xioa and Dhesi (2010) showthat the S&P500 index transmits 29.39% and 21.43% volatility spillovers to the DAX and CAC indices,respectively.

The findings suggest a weak correlation of volatility across markets. Such results justify the weaklinkages between the stock markets, which constitutes a reason for international diversification seeingthe weak conditional correlations and shocks transmissions between countries. We can conclude thatthe volatility of Islamic markets indexes is not correlated with that of the US in both directions. Themajor reason for the weak correlation and avoidance of spillovers is that the main conduit for suchspillovers is interest rate. Since MSCI Islamic equity indexes are based on investments having lowleverage ratios and very small interest rate involvement is permitted, this linkage is broken. Accord-ingly, institutional and individual investors could grasp the opportunity to invest in these markets andbenefit from portfolio diversification in order to minimize risk.

6. Constant and dynamic conditional correlations

The extent of market integration is given by the conditional correlations estimated by the CCC andDCC models. The CCC estimates across markets are very low and some are not statistically significant.Thus, the shocks are correlated only in the same market, and not across markets. For example, thelowest CCC estimate between Pakistan and Turkey, r(4,2), is equal to 0.01 which is the lowest value;however, it is not statistically significant. The implications of the CCC estimation are consistent withvery low conditional correlations between the volatilities. These phenomena can be explained basicallyby the fact that, as the stock market index is calculated according to Shari’ah guidelines, the volatilitydoes not spill over across markets.

However, our results do not support the hypothesis of constant conditional correlations but arein favor of dynamic ones.12 Indeed, the DCC estimation shows that the coefficient of the parametera captures the effect of previous shocks on the conditional correlation, whilst the coefficient of theparameter b captures the effect of the previous period’s conditional correlations. The figures in Table 4show that the estimates for all countries are statistically significant and exhibit low values. For instance,the US equity market has the following statistically significant estimates: a1 = 0.24 and b1 = 0.72. Thesums of these parameters are fairly close to one for all countries, except for Qatar, which means thatthe conditional volatility is persistent. The conditional correlation of the US market with the otherIslamic emerging markets is low. The volatilities in the Islamic emerging markets are accordinglynot affected by that of the US market. Fig. 4 shows the dynamic conditional correlations plotted. Bycomparison with previous studies (e.g., Xioa & Dhesi, 2010), there are clearly low correlations for allemerging Islamic markets with the US market.

Table 4 shows that the parameters �1 and �2 are statistically significant at the 1% level. The firstparameter reflects the impact of the past shocks on current conditional correlation, whilst the secondreflects the impact of past correlation. Since both parameters are statistically significant, then theconditional correlations are not constant, which supports the hypothesis of the dynamic correlations.Accordingly, the DCC is favorable to the CCC. The sum of the parameters �1 and �2 is higher than theunity. This means that the process described by the model is not mean reverting. In other terms, afterthe shocks occurred in the equity market, the conditional correlations will not return to the long-rununconditional level (i.e., changes in conditional correlation cannot last for long time). Further, shocks

12 In a different context, Chang et al. (2013) study the conditional correlations and volatility spillovers between the crude oiland financial markets. They find similar results to ours inasmuch as the estimation of the CCC model shows that the conditionalshocks are correlated only in the same market and not across markets. In contrast, the estimation of the DCC model showsthat the conditional correlations are always significant. This result corroborates the fact that the assumption of conditionalcorrelations is supported empirically by the data.

388

389

390

391

392

393

394

395

396

397

398

399

400

401

402

403

404

405

406

407

408

409

410

411

412

413

414

415

416

417

418

419

420

421

422

423

424

425

426

427

428

429

430

431

432

433

dx.doi.org/10.1016/j.najef.2014.06.011




Table 4Estimates results of CCC and DCC Models.

CCC DCC

Coefficient T-statistic Prob. Coefficient T-statistic Prob.

Mean(1) 7.78705e−004 3.14132 0.00168186* 5.7957e−004 1.36794 0.17133179Mean(2) 0.00169 3.01077 0.00260584* 1.3965e−003 1.07505 0.28235435Mean(3) 1.86646e−004 0.97964 0.32726411 −4.4799e−004 −2.03050 0.04230538**Mean(4) 7.55131e−004 1.79635 0.07243872 2.1396e−003 5.13123 0.00000029*Mean(5) 6.69395e−004 2.54735 0.01085449** 1.4692e−004 0.59538 0.55159019Mean(6) 7.43933e−004 1.85218 0.06400046 4.6301e−004 1.90117 0.05727968c(1) 2.82000e−006 3.71745 0.00020124* 6.7093e−006 4.27828 0.00001883*c(2) 7.47084e−006 2.16244 0.03058390** 1.8061e−004 15.66637 0.00000000*c(3) 3.41683e−007 1.88371 0.05960437 8.0084e−006 14.30770 0.00000000*c(4) 1.41003e−005 6.10799 0.00000000* 7.4246e−005 19.65198 0.00000000*c(5) 9.73853e−007 2.12459 0.03362125** 1.3568e−005 6.24709 0.00000000*c(6) 1.52208e−006 2.07539 0.03795071** 9.7538e−005 15.53114 0.00000000*a(1) 0.11018 8.47766 0.00000000* 0.2410 30.35137 0.00000000*a(2) 0.06266 5.06905 0.00000040* 0.2186 19.01563 0.00000000*a(3) 0.12138 10.78340 0.00000000* 0.3480 36.82893 0.00000000*a(4) 0.12213 8.22853 0.00000000* 0.2841 15.46384 0.00000000*a(5) 0.04765 3.86647 0.00011042* 0.3060 20.74479 0.00000000*a(6) 0.05196 5.86416 0.00000000* 0.1607 18.90848 0.00000000*b(1) 0.87306 67.12729 0.00000000* 0.7255 72.72139 0.00000000*b(2) 0.92262 55.81773 0.00000000* 0.5473 47.44151 0.00000000*b(3) 0.90111 116.89296 0.00000000* 0.7433 110.90180 0.00000000*b(4) 0.82810 44.03751 0.00000000* 0.4749 35.91158 0.00000000*b(5) 0.94323 63.39315 0.00000000* 0.5722 53.74828 0.00000000*b(6) 0.94446 107.50495 0.00000000* 0.6876 81.17404 0.00000000*r(2,1) 0.38766 17.25327 0.00000000* – – –r(3,1) 0.05216 1.89940 0.05751143 – – –r(3,2) 0.11432 4.01765 0.00005878* – – –r(4,1) 0.00371 0.13831 0.88999492 – – –r(4,2) 0.01006 0.36179 0.71750720 – – –r(4,3) 0.05291 1.97310 0.04848437** – – –r(5,1) 0.17496 6.36857 0.00000000* – – –r(5,2) 0.30593 11.62443 0.00000000* – – –r(5,3) 0.24328 9.06565 0.00000000* – – –r(5,4) 0.12246 4.31495 0.00001596 – – –r(6,1) 0.16392 5.73909 0.00000001 – – –r(6,2) 0.28697 10.94891 0.00000000 – – –r(6,3) 0.24626 9.42113 0.00000000 – – –r(6,4) 0.06733 2.33992 0.01928764 – – –�1 – – – 0.1465 33.18274 0.00000000�2 – – – 0.9604 25.17192 0.00000000

Notes: This table shows the estimates of the multivariate GARCH CCC AND DCC models.‘*’ and ‘**’ denote respectively the 1% and5% significance levels. 1, 2, 3, 4, 5, 6 denote respectively: USA, Turkey, Qatar, Pakistan, Malaysia, and Indonesia. The parametersai , bi , and rij are the off-diagonal elements of the matrices A, B and R, as presented in Section 3.

hitting Islamic emerging markets from the US market are transitory. The latter result corroborates thefact that the US market does not strikingly spill over the volatility to the Islamic emerging markets.

This stylized fact does not confirm previous studies. For instance, Christodoulakis (2007) showsthat there is persistence of the memory in the correlations. Yu, Fung, and Tam (2010) maintain thatmarkets become more integrated when the conditional correlation increases over time. On anotherside, Duncan and Kabundi (2013) find close results to ours. Indeed, the authors argue that “Althoughthere is some evidence that global shocks impact significantly on domestic volatility, particularlyduring global crises, foreign spillovers are small on average. Thus, idiosyncratic risk is identified asthe key determinant of volatility in South Africa. Weak synchronization between domestic and worldvolatilities provides a compelling motive for international diversification.” (p. 573)

The results can be valuable in terms of minimization of risk and portfolio choice. The DCC modelcould further be interpreted in terms of its forecasting ability relative to the unconditional correlations.

434

435

436

437

438

439

440

441

442

443

444

445

dx.doi.org/10.1016/j.najef.2014.06.011




Fig. 4. Time-varying conditional correlations.

A better forecasting ability suggests an improvement in risk management techniques. Indeed, whena portfolio manager could predict better the correlations of assets in a dynamic manner, this plays aninteresting role in portfolio theory since it is useful to compute the risk of diversified portfolios. Asa consequence, a high forecasting ability of the dynamic correlation between indexes among Islamicemerging markets helps to optimize the portfolio diversification.

As it may be noticed, the results of the CCC and DCC models corroborate with those of the multi-variate GARCH BEKK (1,1) model. Indeed, the conditional correlations are very low, which means thatthe US market has not been a channel for transmission of shocks and volatility spillover to Islamicmarkets. These findings are certainly related to some features of Islamic finance industry. Firstly, thescreening of the MSCI Islamic equity index excludes all companies working in Islamically prohibitedsectors including gambling which is a cause of volatility. Secondly, it also imposes stringent restric-tions on leverage ratios and interest-related dealings. Due to these restrictions these investments areless exposed to the volatility of interest rates and to the impairing impacts of bubbles that may occurin financial markets. Thirdly, the asset-backed principle, which is peculiar to Islamic finance, ensuresa close linkage between the real and financial sectors and hence prevents purely speculative invest-ments. In Western financial markets such speculative investments, e.g., derivatives, cause a lot ofvolatility in stock markets. Islamic restrictions that do not allow short selling raise a strong protectivewall against such volatility. Indeed, Islamic finance offers financial innovations that could engender agreater social responsibility, ethical and moral values and sustainable finance, and hence, can possiblyendure financial crises better than the conventional system (Ben Nasr et al., 2013).

7. Conclusion

The purpose of this paper is to analyze the correlation of volatility between indexes of a sampleof Islamic emerging equity markets and the US Islamic market through the study of the dynamicconditional correlation based on a multivariate GARCH model. Our empirical design used the MSCIIslamic equity index that includes companies on the basis of a severe screening. Indeed, the MSCI

446

447

448

449

450

451

452

453

454

455

456

457

458

459

460

461

462

463

464

465

466

467

468

469

470

dx.doi.org/10.1016/j.najef.2014.06.011




Islamic equity index has two main requirements. On the one hand, the business activity must becompliant with Islamic Shari’ah. On the other hand, some financial ratios related to financial fees andleverage must not exceed upper limits.

In order to study the volatility spillovers among American and emerging Islamic markets, we usedthe multivariate GARCH BEKK (1,1) model alongside with the CCC and DCC models. The estimatesstemming from the estimation of the multivariate GARCH BEKK (1,1) model indicate that all paircountries exhibit a weak correlation of volatility, which suggests a weak correlation among the USand Islamic emerging markets. Although the estimates of conditional correlations are statisticallysignificant in almost all cases, they are very low. Therefore, the transmission of shocks among thesemarkets is not significant.

A plausible interpretation can be argued in terms of diversification of risk seeing the weak cor-relation and the absence of transmission of shocks among Islamic financial markets. Indeed, thesemarkets do have a long memory and are weakly integrated, which can be a reason for internationaldiversification. The results of the CCC and DCC models corroborate with those of the multivariateGARCH BEKK (1,1) model. Indeed, the constant and dynamic conditional correlations are statisticallysignificant for all countries. However, they have very low estimates for all countries. Our empiricalregularities do not confirm previous empirical studies (e.g., Christodoulakis, 2007; Yu et al., 2010).In this regard, the weak conditional correlations over time suggest that the Islamic emerging and USmarkets are weakly integrated and the volatility spillovers among them are weak as well. The char-acteristics of stocks included in the MSCI Islamic equity indexes and the peculiar specificities of theIslamic finance industry play an important role in explaining our results. Indeed, the prohibition ofinvestment in interest-bearing activities and the stringent restrictions on leverage coupled with the‘asset-backed’ principle contribute to explaining why the Islamic equity indexes have very low condi-tional correlations. The investment in Islamic financial markets is less affected by volatility spilloversand shocks transmission. However, a better forecasting of conditional correlations in these marketshelps managers to optimize portfolio diversification.

This study can be extended in various ways since several questions remain unanswered. Futureresearch could focus on the asymmetric characteristics of volatility spillovers. Our findings could befurther studied by taking into account the effect of scheduled and unscheduled news release. Thecontagion effect on the stability of volatility spillovers in Islamic financial markets constitutes alsoan interesting research route. At the empirical level, the results could be strengthened by the use ofintra-daily data. The high frequency data may reveal interesting empirical implications.

Acknowledgements

We would like to thank the Guest Editors, Shawkat Hammoudeh and Duc Khuong Nguyen, theEditor, Hamid Beladi, and two anonymous referees for thoughtful comments and suggestions.

References

Abu-Zaid, A. (2011). Volatility spillover effects in emerging MENA stock markets. Review of Applied Economics, 7(1–2), 107–127.Alexander, C., & Leigh, C. (1997). On the covariance matrices used in value-at-risk models. Journal of Derivatives, 4(3), 50–62.Al-Khazali, O., Lean, H. H., & Samet, A. (2014). Do Islamic stock indexes outperform conventional stock indexes? A stochastic

dominance approach. Pacific-Basin Finance Journal, 29–46.Ariss, R. (2010). Competitive conditions in Islamic and conventional banking: A global perspective. Review of Financial Economics,

19(3), 101–108.Bauwens, L., Laurent, S., & Rombouts, J. (2006). Multivariate GARCH models: A survey. Journal of Applied Econometrics, 21(1),

79–109.Beder, T. S. (1995). VaR: Seductive but dangerous. Financial Analyst Journal, 51(5), 12–24.Beirne, J., Guglielmo, M., Caporale, M., Schulze-Ghattas, M., & Spagnolo, N. (2009). Volatility spillovers and contagion from mature

to emerging stock markets European Central Bank WP#1113.Bekaert, G., Harvey, C. R., & Ng, A. (2003). Market integration and contagion. Journal of Business, 78(1), 39–70.Ben Nasr, A., Ajmi, A. N., & Gupta, R. (2013). Modeling the volatility of the Dow Jones Islamic Market World Index using a fractionally

integrated time varying GARCH (FITVGARCH) model.Ben Nasr, A., Boutahar, M., & Trabelsi, A. (2010). Fractionally integrated time varying GARCH model. Statistical Methods and

Applications, 19(3), 399–430.Bollerslev, T., Engle, R. F., & Wooldridge, J. M. (1988). A capital asset pricing model with time-varying covariances. Journal of

Political Economy, 96(1), 116–131.

471

472

473

474

475

476

477

478

479

480

481

482

483

484

485

486

487

488

489

490

491

492

493

494

495

496

497

498

499

500

501

502

503

504

505

506

507

508

509

510

511

512

513

514

515

516

517

518

519

520

521

522

523

524

dx.doi.org/10.1016/j.najef.2014.06.011




Bollerslev, T. (1990). Modeling the coherence in short-run nominal exchange rates: A multivariate generalized ARCH approach.Review of Economics and Statistics, 72(3), 498–505.

Byström, H. N. (2002). Using simulated currency rainbow options to evaluate covariance matrix forecasts. Journal of InternationalFinancial Markets, Institutions and Money, 12(3), 216–230.

Calvo, G. A. (1999). Contagion in Emerging Markets: When Wall Street is a Carrier. University of Maryland Working Paper.Chan, L. K., Karceski, J., & Lakonishok, J. (1999). On portfolio optimization: forecasting covariances and choosing the risk model.

Review of Financial Studies, 12(5), 937–974.Chang, C., McAleer, M., & Tansuchat, A. (2013). Conditional correlations and volatility spillovers between crude oil and stock

index returns. North American Journal of Economics and Finance, 25, 116–138.Christodoulakis, G. A. (2007). Markovian credit transition probabilities under inequality constraints: The US portfolio

1984–2004. Journal of Credit Risk, 3(3), 25–39.Christodoulakis, G. A., & Satchell, S. E. (2002). Correlated ARCH: Modeling the time-varying correlation between financial asset

returns. European Journal of Operations Research, 139(2), 351–370.Dungey, M., Fry, R., González-Hermosillo, B., & Martin, V. L. (2007). Contagion in global equity markets in 1998: The effects of

the Russian and LTCM crises. North American Journal of Economics and Finance, 18(2), 155–174.Duncan, A., & Kabundi, A. (2013). Domestic and foreign sources of volatility spillover to South African asset classes. Economic

Modeling, 31, 566–573.Elton, E. J., & Gruber, M. J. (1973). Estimating the dependence structure of share prices – Implications for portfolio selection.

Journal of Finance, 28(5), 1203–1232.Engle, R. F. (2002). Dynamic conditional correlation: a simple class of multivariate generalized autoregressive conditional

heteroskedasticity models. Journal of Business and Economic Statistics, 20(3), 339–350.Engle, R. F., & Sheppard, K. (2001). Theoretical and empirical properties of dynamic conditional correlation multivariate GARCH,

NBER Working Paper Series No. 8554.Engle, R. F., Ito, T., & Lin, W. L. (1990). Meteor showers or heat waves? Heteroscedastic intra daily volatility in the foreign

exchange market. Econometrica, 58(3), 525–542.Engle, R. F., & Kroner, K. (1995). Multivariate simultaneous GARCH. Econometric Theory, 11(1), 122–150.Gebka, B., & Serwa, D. (2007). Intra- and inter-regional spillovers between emerging capital markets around the world. Research

in International Business and Finance, 21(2), 203–221.Gibson, M. S., & Boyer, B. H. (1998). Evaluating forecasts of correlation using option pricing. Journal of Derivatives, 6, 13–38.Hayat, S., & Kraeussl, R. (2011). Risk and return characteristics of Islamic equity funds. Emerging Markets Review, 12(2), 189–203.Jiang, G., Konstantinid, E., & Skiadopoulos, G. (2012). Volatility spillovers and the effect of news announcements. Journal of

Banking and Finance, 33, 2260–2273.King, M., & Wadhwani, S. (1990). Transmission of volatility between stock markets. Review of Financial Studies, 3(1), 5–33.Martinez, C., & Ramirez, M. (2011). Dynamic conditional correlation in Latin American asset markets University of Rosario WP#107.Mukherjee, K., & Mishra, R. (2008). Stock market integration and volatility spillovers: India and its major Asian counterparts Working

Paper. India: National Institute of Bank Management.Ng, A. (2000). Volatility spillover effects from Japan and the US to the Pacific Basin. Journal of International Money and Finance,

19(2), 207–233.Nienhaus, V. (2011). Islamic finance ethics and Shari’ah law in the aftermath of the crisis: Concept and cractice of Shari’ah

compliant finance. Ethical Perspectives, 18(4), 591–623.Schoon, N., & Nuri, J. (2012). Comparative financial systems in Judaism, Christianity and Islam: The case of interest Unpublished

Working Paper. UK: University of Surrey.Schoon, N. (2009). Islamic banking and Finance. London: Spiramus.Singh, P., Kumar, B., & Pandey, A. (2010). Price and volatility spillovers across North American, European and Asian stock markets.

International Review of Financial Analysis, 19(1), 55–64.Tse, Y. K., & Tsui, A. (2002). A multivariate GARCH model with time-varying correlations. Journal of Business and Economic

Statistics, 20(3), 351–362.Xioa, L., & Dhesi, G. (2010). Volatility spillover and time-varying conditional correlation between the European and US stock

markets. Global Economy and Finance Journal, 3(2), 148–164.Yu, I. W., Fung, K. P., & Tam, C. S. (2010). Assessing financial market integration in Asia-equity markets. Journal of Banking &

Finance, 34(12), 2874–2885.Yusof, R. M., & Majid, S. A. (2007). Stock market volatility transmission in Malaysia: Islamic versus conventional stock market.

Journal of King Abdulaziz University: Islamic Economics, 20(2), 17–35.Zheng, T., & Huo, H. (2013). Reexamining the time-varying volatility spillover effects: A Markov switching causality approach.

North American Journal of Economics and Finance, 26, 643–662.

525

526

527

528

529

530

531

532

533

534

535

536

537

538

539

540

541

542

543

544

545

546

547

548

549

550

551

552

553

554

555

556

557

558

559

560

561

562

563

564

565

566

567

568

569

570

571

572

573

574

575

576

577

578

579

dx.doi.org/10.1016/j.najef.2014.06.011

islamic equity market integration and volatility spillover between emerging and us stock markets

Documents