This article was downloaded by: [Mehmet Balcilar]On: 06 February 2015, At: 14:12Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,37-41 Mortimer Street, London W1T 3JH, UK

Click for updates

Applied Economics LettersPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/rael20

Are there long-run diversification gains from the DowJones Islamic finance index?Mehmet Balcilarab, Charl Joosteb, Shawkat Hammoudehc, Rangan Guptab & VassiliosBabalosd

a Department of Economics, Eastern Mediterranean University, Famagusta, Turkeyb Department of Economics, University of Pretoria, Pretoria, 0002, South Africac Lebow College of Business, Drexel University, Philadelphia, PA 19104, USAd Department of Accounting & Finance, Technological Educational Institute of Peloponnese,Kalamata 241 00, GreecePublished online: 11 Dec 2014.

To cite this article: Mehmet Balcilar, Charl Jooste, Shawkat Hammoudeh, Rangan Gupta & Vassilios Babalos (2014):Are there long-run diversification gains from the Dow Jones Islamic finance index?, Applied Economics Letters, DOI:10.1080/13504851.2014.990613

To link to this article: http://dx.doi.org/10.1080/13504851.2014.990613

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) containedin the publications on our platform. However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of theContent. Any opinions and views expressed in this publication are the opinions and views of the authors, andare not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon andshould be independently verified with primary sources of information. Taylor and Francis shall not be liable forany losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoeveror howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use ofthe Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

Are there long-run diversification

gains from the Dow Jones Islamic

finance index?

Mehmet Balcilara,b, Charl Joosteb, Shawkat Hammoudehc,Rangan Guptab,* and Vassilios Babalosd

aDepartment of Economics, Eastern Mediterranean University,Famagusta, TurkeybDepartment of Economics, University of Pretoria, Pretoria, 0002,South AfricacLebow College of Business, Drexel University, Philadelphia, PA 19104,USAdDepartment of Accounting & Finance, Technological EducationalInstitute of Peloponnese, Kalamata 241 00, Greece

We compare a nonlinear (time-varying) cointegration test with thestandard cointegration test in studying the long-run relationship of theDow Jones Islamic finance index with three other conventional globalequity market indices. Our results show that there is a long-runnonlinear cointegrating relationship between the Dow Jones Islamicstock market index and other conventional stock market indices,which is not picked up by the linear cointegration test. Thus,Islamic markets seem to offer little, if any, long-run diversificationto international investors.

Keywords: Islamic and conventional finance; time-varying cointegration

JEL Classification: C5; C12; G1

I. Introduction

It has been argued that Islamic markets are oftendecoupled from their conventional counterpartsbecause of heavy restrictions on the former. Thedecoupling argument is empirically motivated bythe rejection of cointegration. At the core of thedecoupling lie the rules that dictate how investments

ought to take place.1 The main objective of thisarticle is to discern whether the constant coefficientcointegration tests are reliable in studying therelationship between seemingly different stockmarkets, which have become more interlinked byglobalization.2 We use the Park and Hahn (1999)test for single equation cointegration and also extend

*Corresponding author. E-mail: Rangan.Gupta@up.ac.za1 Speculation using derivative markets and buying government debt that issues fixed coupons are prohibited, and investingin certain industries is not allowed (Hammoudeh et al., 2013).2 Other methods might support diversification despite the presence of dynamic cointegration (see Guidi and Ugur, 2014).

Applied Economics Letters, 2014http://dx.doi.org/10.1080/13504851.2014.990613

© 2014 Taylor & Francis 1

Dow

nloa

ded

by [

Meh

met

Bal

cila

r] a

t 14:

12 0

6 Fe

brua

ry 2

015

the analysis using the Bierens and Martins (2010)multivariate time-varying cointegration test.3

A number of studies show that there is no long-runstandard cointegrating relationship between theIslamic stock market and conventional stock marketindices, implying the possibility of significant diver-sification benefits (see Bakri et al., 2010; El Khalichiet al., 2014).Standard cointegration tests that rely on linearity

and normality assumptions might yield misleadingresults with the existence of multiple regimes or struc-tural breaks. Furthermore, Guidi and Ugur (2014)argue that a battery of tests is required to studywhether different stock markets are decoupled. Thisrequires using a combination of dynamic correlationsand time varying and constant cointegration proce-dures, as well as portfolio diversification strategies.There seems to be some evidence of risk sharing

and an element of contagion among Islamic and con-ventional markets (Ajmi et al., 2014b). However, themarket shocks seem to be systemic because commonfactors rather than idiosyncratic contributed to Shariarules. An important implication of our research is thatthe contagion or common forces effects link the twotypes of stock market indices, which implies no sig-nificant diversification strategy.Ajmi et al. (2014a) use linear and nonlinear caus-

ality tests to study the link between the Islamic stockmarket indices and major conventional equity mar-kets. They show that there is a strong causal relation-ship between them. They also show that the Islamicstock market indices are prone to the same shocksthat strike the conventional equity markets and, thus,are sensitive to changes in global financial factors.

II. Methodology

We use the Standard and Poor’s US, European andAsian stock market indices (SP500, LSPEU andLSPAS50, respectively) and test for possible cointe-gration with the Dow Jones Islamic Finance Index(LDJIM) using daily data from 1 April 1999 till 22July 2013, which gives a total of 3796 observations.All data are sourced from Bloomberg. Figure 1

shows the natural logarithm of the series and thekernel densities of the returns with their respectivemeans and SDs.4 These indices share common fea-tures. They are severely affected by the 2008 globalfinancial crisis, and their average returns and SDs aresimilar. On the other hand, although the SP500 andLSPEU moved in tandem, the LDJIM and LSPAS50have followed different trajectories post 2002.We use a combination of standard (linear with and

without breaks and nonlinear) tests to check for bothstationarity and cointegration.5 Our main test is thatof Bierens and Martins (2010), which tests for time-varying coefficient (TVC) cointegration in a multi-variate setup, unlike that of Park and Hahn (1999)that tests for cointegration in a single equation. TheTVCs are approximated by the Chebyshev polyno-mials. The implementation uses the Akaike informa-tion criterion (AIC), Bayesian information criterion(BIC) and Hannan-Quinn (HQ) information criterionto select the number of the Chebyshev polynomials.Bierens and Martins (2010) show that a time-

varying VECM(p) can be represented as follows:

ΔYt ¼ μþ�0tYt�1 þ

Xp�1

j¼1

Γ jΔYt�j þ εt (1)

where μ is a k� 1 vector of intercepts, Yt is a k� 1time series vector and the error εt ,Nk 0;Ωð Þ. Thereare fixed r < k linearly independent columns for thetime-varying cointegrated matrix. As in Bierens andMartins (2010), the objective is to test the null hypoth-esis of time-invariant cointegration �

0t ¼ �0 ¼ αβ0

against the time-varying cointegration�0t ¼ �0 ¼ αβ

0t.

The time-varying polynomials are defined as

P0;T tð Þ ¼ 1 and Pi;T tð Þ ¼ ffiffiffi2

pcos iπ t � 0:5ð Þ=Tð Þ.

Any function g tð Þ of discrete time can be (due toorthonormal property) represented as

gðtÞ ¼XT�1

i¼ 0

�i;T Pi;T tð Þ

where �i;T¼ 1T

PTt¼1

g tð ÞPi;T tð Þ.

3 Failure to detect parameter shifts in econometric specifications when they exist implies that the model is misspecified,which could lead to poor forecasting performance (Gabriel and Martins, 2011).4 The summary statistics show that for data in both log levels and log first difference, the null hypotheses of normality, noautocorrelation, and no ARCH effects are rejected.5 These tests fail to reject the null of unit roots and also the null of no cointegration (results available upon request) for thelog levels of the series.

2 M. Balcilar et al.

Dow

nloa

ded

by [

Meh

met

Bal

cila

r] a

t 14:

12 0

6 Fe

brua

ry 2

015

Bierens and Martins (2010) then substitute

�0t ¼ αβ

0t ¼ α

Pmi¼0

�i;TPi;T tð Þ� �0

into (1). Here the

Chebyshev polynomial is a smooth function, whichallows βt to change gradually over time. This yieldsthe VECM(p):

ΔYt ¼ αXmi¼0

�i;TPi;T tð Þ !0

Yt�1

þXp�1

j¼1

Γ jΔYt�j þ εt (2)

Ymt ¼ Y

0t�1P1;T tð Þ;Y 0

t�1P2;T tð Þ; ::;Y 0t�1Pm;T tð Þ� �0

and�0i ¼ ð�0

0; �01; ::; �

0mÞ is a r � mþ 1ð Þk matrix of rank r.

The null hypothesis on the time-invariant

cointegration is �0i ¼ ðβ0

;Or;k;mÞ. This test is then con-ducted using a likelihood ratio (LR) test (Bierens andMartins, 2010).

III. Results

The constant parameter cointegration tests show thatthere is no cointegration among the Islamic and con-ventional stock variables. However, cointegrationexists when we allow for the possibility that eachpoint in time represents a different regime.6

We initially use the standard Johansen cointegra-tion test for multiple cointegrating relationships. Weuse a VAR(3) as given by the BIC criterion. Table 1shows the results using the maximal eigenvalue λmax

and trace λtrace cointegration order tests of Johansen.

6.4

6.8

7.2

7.6

8.0

8.4

8.8

2000 2002 2004 2006 2008 2010 2012

LDJIMLSP500LSPAS50LSPEU

.0

.1

.2

.3

.4

.5

.6

–12 –10 –8 –6 –4 –2 0 2 4 6 8 10 12 14

RDJIM - Mean = 0.01, SD = 1.1RSP500 - Mean = 0.01, SD = 1.3RSPAS50 - Mean = 0.04, SD = 1.4RSPEU - Mean = –0.00, SD = 1.4

Den

sity

Fig. 1. Stock market indices in logs and daily returns

6 This is supported by the Park and Hahn (1999) tests. The null hypothesis of fixed coefficient cointegration is rejected atthe 1% level, favouring the alternative that the fixed coefficients model is not cointegrated.Wewere unable to reject the nullhypothesis of cointegration in the TVC model.

Are there long-run diversification gains 3

Dow

nloa

ded

by [

Meh

met

Bal

cila

r] a

t 14:

12 0

6 Fe

brua

ry 2

015

A nonrejection of r = 0 for the Johansen (1991) testsimplies no cointegration. These standard tests showthat there is at least one cointegrating relationship,implying a long run relationship between the Islamicstock index and their counterparts.Table 2 reports the Bierens and Martins (2010)

multivariate TVC cointegration tests based on theChebyshev time polynomials. These tests are con-structed as LR tests under the null of time-invariantcointegration tested against the TVC cointegration.The AIC selected an extremely large number ofpolynomials and therefore is not used. m denotesthe number of the Chebyshev time polynomials andr denotes the number of cointegration relationships.The distribution of the LR test is a chi-square withm� k � r degrees of freedom, where k is the numberof variables. The p values of the LR tests are given inbrackets and ‘<’ means ‘less than’.The BIC selects 1 polynomial, the HQ selects 4 and

the AIC selects 376 polynomials for all the cases of 1to 3 cointegration vectors. The results are robust toVAR orders between 1 and 9. In all cases, the null ofthe time-invariant cointegration against the TVC coin-tegration is rejected for 1 to 3 cointegration vectors.Figure 2 plots the normalized parameter estimates

from the TVC cointegration model with 1 cointegra-tion relationship imposed on the estimation. Thenormalized time-varying cointegration relationshipis specified as

LDJIMt ¼ β0 þ β1tLSP500þ β2tLSPEU

þ β3tLSPAS50

where these variables are defined as above.The parameters are quite unstable for the whole

period; however, they are relatively stable after 2002

Table 1. Multivariate linear (standard) cointegration tests

Lag (p) 1 2 3 4 6 8 10

Panel A: VAR order selection criteriaAIC −37.54 −38.04 −38.07 −38.09 −38.10 −38.10 −38.10HQ −37.53 −38.02 −38.04 −38.05 −38.04 −38.01 −38.00BIC −37.51 −37.98 −37.99 −37.98 −37.93 −37.86 −37.83

Panel B: Johansen cointegration tests

Eigenvalues 0.0021 0.0019 0.0011 0.0003

Critical values

H0 λmax 10% 5% 1%

r = 3 1.17 6.50 8.18 11.65r = 2 4.15 12.91 14.90 19.19r = 1 7.39 18.90 21.07 25.75r = 0 8.01 24.78 27.14 32.14

H0 λtrace 10% 5% 1%r ≤ 3 1.17 6.5 8.18 11.65r ≤ 2 5.32 15.66 17.95 23.52r ≤ 1 12.72 28.71 31.52 37.22r = 0 20.73 45.23 48.28 55.43

Table 2. Multivariate time-varying cointegration test

m r = 1 r = 2 r = 3

Likelihood Ratio (LR) test for time-varying cointegration1 42.86 (<0.01) 53.71 (<0.01) 62.12 (<0.01)2 63.65 (<0.01) 105.57 (<0.01) 121.47 (<0.01)3 89.42 (<0.01) 139.22 (<0.01) 175.66 (<0.01)4 118.49 (<0.01) 184.07 (<0.01) 245.91 (<0.01)

Log likelihood of TVC cointegration model1 50767.18 52090.39 52865.622 50777.58 52116.32 52895.303 50790.47 52133.14 52922.394 50805.00 52155.57 52957.52

HQ for TVC cointegration model1 −26.70 −27.39 −27.782 −26.71 −27.39 −27.793 −26.71 −27.39 −27.794 −26.71 −27.40 −27.79

BIC for TVC cointegration model1 −26.64 −27.31 −27.692 −26.63 −27.30 −27.683 −26.63 −27.30 −27.674 −26.63 −27.29 −27.66

4 M. Balcilar et al.

Dow

nloa

ded

by [

Meh

met

Bal

cila

r] a

t 14:

12 0

6 Fe

brua

ry 2

015

compared to 1990s. The parameters during the late1990s were markedly different, which could be dueto a number of reasons, such as the dot-com bubble.We also observe that large stock market swings inducemore parameter volatility. The 2008 financial crisisseems to have affected the cointegrating relationships.7

Contrary to other findings in the literature, ourresults suggest that the benefits of diversification interms of using the DJIM are slightly overstated. TheDJIM, the SP500 and SPAS50 enjoy a positive long-run relationship, despite controlling for a parametershift. But the relationship with SPEU is primarilynegative in the majority of the sample. It does, how-ever, seem as though the stock market crises slightlyinvert these relationships.

IV. Conclusion

We use various cointegrating tests to analysethe cointegrating relationship between the DowJones Islamic Market (DJIM) index with other con-ventional counterparts because of its implications forportfolio diversification. Our results show that theconstant parameter cointegration tests tend to reject

cointegration in the presence of regime shifts, imply-ing possible benefits from diversification. However,we are able to identify cointegrating relationships ina multiple equation setup with parameter shifts.The basic result is that there is little benefit in using

the DJIM to diversify and hedge against movementsin conventional stockmarket indices. There is a strongand positive cointegrating relationship betweenSP500 and DJIM, which share many same stocksaccessed from the global Dow Jones stock universethat meet the Sharia-compliance qualifications,whereas the remaining unshared stocks are disposedto the same fluctuating global common forces. Alsothe Sharia principles’ restrictions seem to make aslight difference during major crises like the 2008global financial crisis, which is probably why we seethe time-varying cointegrating parameters of the threeconventional stock markets in relation to the DJIMtends to converge post the financial crisis.

Acknowledgements

We thank an anonymous referee for many helpfulcomments. However, any remaining errors are solelyours.

Fig. 2. Time-varying cointegrating parameters

7 Following the suggestions of an anonymous referee, we also followed the approach of Guidi and Ugur (2014) in usingbivariate (which included the DJIM and one conventional stock market in turn) symmetric and asymmetric dynamicconditional correlations on returns data, to check for the robustness of our findings. Our basic result obtained from the time-varying cointegration approach still continued to hold in the sense that the correlation increased over time, especially since2008. Details of these results are available upon request from the authors. It must, however, be pointed out, that since weare trying to analyse long-run diversification, the time-varying cointegration applied on the log-levels of data is the moreappropriate approach rather than the DCC-GARCH approach of Guidi and Ugur (2014) used on the returns of the series,which in turn, tends to capture short-run comovements instead.

Are there long-run diversification gains 5

Dow

nloa

ded

by [

Meh

met

Bal

cila

r] a

t 14:

12 0

6 Fe

brua

ry 2

015

ReferencesAjmi, A. N., Ben Nasr, A., Gupta, R. et al. (2014a)

Forecasting the volatility of the Dow Jones Islamicstock market index. Long memory vs. regimeswitching, Working Paper No. 201412, Departmentof Economics, University of Pretoria

Ajmi, A. N., Hammoudeh, S., Nguyen, D. K. et al.(2014b) How strong are the causal relationshipsbetween Islamic stock markets and conventionalfinancial systems? Evidence from linear and non-linear tests, Journal of International FinancialMarkets, Institutions and Money, 28, 213–27.doi:10.1016/j.intfin.2013.11.004

Bakri, A. K., Nor, A. M. and Kassim, M. A. A. (2010) Thesubprime crisis and Islamic stock markets integra-tion, International Journal of Islamic and MiddleEastern Finance and Management, 3, 363–71.doi:10.1108/17538391011093298

Bierens, H. J. and Martins, L. F. (2010) Time-varyingcointegration, Econometric Theory, 26, 1453–90.doi:10.1017/S0266466609990648

El Khalichi, A., Humayun, K. S., Arouri, M. et al. (2014)Are Islamic equity indices more efficient than their

conventional counterparts? Evidence from majorglobal index families, Working Paper No. 2014–091, IPAG Business School, IPAG, Paris.

Gabriel, V. J. and Martins, L. F. (2011) Cointegration testsundermultiple regime shifts: an application to the stockprice-dividend relationship, Empirical Economics, 41,639–62. doi:10.1007/s00181-010-0401-8

Guidi, F. and Ugur, M. (2014) An analysis of south-easternEuropean stock markets: evidence on cointegrationand portfolio diversification benefits, Journal ofInternational Financial Markets, Institutions andMoney, 30, 119–36. doi:10.1016/j.intfin.2014.01.007

Hammoudeh, S., Jawadi, F. and Sarafrazi, S. (2013)Interactions between conventional and Islamicstock markets: a hybrid threshold analysis,Mimeo, Drexel University, Philadelphia, PA.

Johansen, S. (1991) Estimation and hypothesis testing ofcointegration vectors in Gaussian vector autoregres-sive models, Econometrica, 59, 1551–80.doi:10.2307/2938278

Park, J. Y. and Hahn, S. B. (1999) Cointegrating regressionswith time varying coefficients, Econometric Theory,15, 664–703. doi:10.1017/S0266466699155026

6 M. Balcilar et al.

Dow

nloa

ded

by [

Meh

met

Bal

cila

r] a

t 14:

12 0

6 Fe

brua

ry 2

015