1

Volatility and risk spillovers between oil, gold, and Islamic and

conventional GCC banks

Walid Mensia,b

, Shawkat Hammoudehc,d

, Idries Mohammad Wanas Al-Jarrahe, Khamis

Hamed Al-Yahyaeeb*

, Sang Hoon Kang

a Department of Finance and Accounting, University of Tunis El Manar, Tunis, Tunisia

b Department of Economics and Finance, College of Economics and Political Science, Sultan Qaboos

University, Muscat, Oman

Email: walid.mensi@fsegt.rnu.tn

Email: yahyai@squ.edu.om

cLebow College of Business, Drexel University, Philadelphia, United States

dEnergy and Sustainable Development (ESD), Montpellier Business School, Montpellier, France

Email address: shawkat.hammoudeh@gmail.com

eCollege of Business and Economic, Qatar University, Qatar

Email: idries@qu.edu.qa

fDepartment of Business Administration, Pusan National University, Busan 609-735, Republic of

Korea

Email: sanghoonkang@pusan.ac.kr

Abstract

This paper examines time-varying risk spillovers and hedging effectiveness between two major

commodity markets (oil and gold) and both the Islamic and conventional bank stock indices for

five GCC countries (Bahrain, Kuwait, Qatar, Saudi Arabia and UAE), using the DECO-

FIGARCH model and the spillover index of Diebold and Yilmaz (2012). The results of the

DECO-FIGARCH model show evidence of a weak average conditional correlation between all

the GCC bank stock indices and the two commodity markets. Moreover, we find significant risk

spillovers between these Islamic and conventional GCC bank stock indices and the commodity

markets. The spillovers rise considerably during the 2008-2009 global financial crisis and the

2014-2015 oil price plunge periods Further, oil, gold, and the conventional bank stock index of

Saudi Arabia, Kuwait and Qatar are net sources of volatility spillovers into the other markets,

while all the Islamic banks and conventional banks of UAE and Bahrain are net volatility

recipients of volatility spillovers. Finally, we provide evidence asserting that including gold and

oil in a GCC portfolio offers better but different diversification benefits and hedging effectiveness

for the GCC banks.

Keywords: GCC, Islamic banking, Commodity markets, Risk spillovers, Hedging effectiveness

JEL classification codes: G14; G15

2

1. Introduction

The risk spillover analysis is an important tool for controlling and managing bank

risks. Significant efforts have been dedicated to the understanding of excessive risk in

financial systems and identification of those policies that may reduce such risk, particularly

following the 2007–2008 global financial crisis (e.g., the Lehman brother collapse on

September 15, 2008). The global financial crisis (GFC) has changed the global financial

landscape. Indeed, the GFC had amplified the shocks and transmitted them from one market

into other markets and then involved the whole global financial system. The information set

(including extreme movements or major events) for one bank may have a substantial

predictive power of other banks and other markets (e.g., foreign exchange markets, stock

markets and commodity markets). In addition to market fundamentals, many plausible causes

or events may justify the presence of risk spillovers between banks or markets including

contagion, investor sentiments and reaction to news among others. Thus, determining the

sources and recipients of risk spillovers is important for financial sectors and markets.

The presence of risk spillovers between international financial markets has motived

individual and institutional investors to find new hedges and new havens to protect their

investments. The Islamic finance industry is considered one of the new havens as reflected by

the phenomenal growth of this industry during the last decade, particularly in the wake of the

GFC. The industry has grown in depth and breaths and offers alternative interesting

instruments, which make it a modern and viable financial system. The Islamic banking is the

largest sector of the Islamic finance industry and has assets in most countries of the world.

The Islamic banking follows the Shariah-compliance principles which screen investments for

Islamic finance. In contrast to conventional finance, Islamic banking is strictly prohibited

from carrying out business activities such as those that involve gambling, firearms, alcohol,

3

speculation and interest-gaining. The banking Shariah compliance is also based on risk-

sharing and not debt accumulation.

Participation in Islamic banking assets has also increased despite the major

turbulences in the last few years. The core participating countries include Bahrain, Qatar,

Indonesia, Saudi Arabia, Malaysia, United Arab Emirates, Turkey, Kuwait and Pakistan

whose assets reached 93% of total banking industry assets, exceeding US$920 billion in 2015.

An important subset of those countries, known as QISMUT which includes Qatar, Indonesia,

Saudi Arabia, Malaysia, UAE and Turkey, has a market share standing at 80% of the

international banking assets. The Islamic banking industry in the Middle Eastern countries,

particularly those in the Gulf Cooperation Council (GCC), has grown more rapidly than their

conventional counterparts in those regions. Based on recent banking sector asset projections,

five countries including Saudi Arabia, Qatar, Pakistan, UAE and Turkey are identified to be

the key players by 2020 on the accounts of “average annual growth rate” and banking sector

assets. In terms of banking participation, Saudi Arabia, Kuwait, Bahrain and Qatar will be the

major players in terms of having the highest banking market shares by 2020.1

We note that the risk spillovers are more visible during financial turmoil periods

which increase asset volatility. In fact, during the two recent crisis periods, the cross-market

linkages among international financial markets and between financial and oil markets have

increased, particularly following the 2008-2009 GFC and the 2010-2012 European sovereign

debt crisis (ESDC). The “flight-to-quality” phenomenon has attracted the attention of

individual investors, institutional investors (e.g., banks) which are primarily interested in

financial safety, and policymakers alike. In the safe haven literature, numerous studies view

gold as a good refuge asset (i.e., a strong hedge and a good safe haven asset) in financial

1 http://www.ey.com/Publication/vwLUAssets/ey-world-islamic-banking-competitiveness-report-

2016/$FILE/ey-world-islamic-banking-competitiveness-report-2016.pdf

4

markets (Baur and Lucey, 2010; Baur and McDermott, 2010; Mensi et al., 2014a, 2015b,

2016).

Gold is an important asset for GCC banks for cultural, financial and geopolitical

reasons. Culturally, there is an obsession with gold in regions that populate the people and

businesses of the Orient including India, China and the GCC areas. Moreover, financial

reasons motivate us to include gold in our analysis because this shiny metal is used to hedge

against unanticipated fluctuations and excessive variations in oil prices, stock markets,

interest rates and foreign exchange markets (Allayannis and Ofek, 2001). Finally, the GCC

banks are located in a region plagued with high geopolitical risks and riven with regional

wars. Thus, the yellow metal should provide financial stability to both Islamic and

conventional GCC banks. The association between gold and the safe haven property is also

important for GCC banks to maintain liquidity, credit risk management and portfolio risk

assessment. For the GCC central banks, gold reserves are also a store of value, a way of

fostering domestic currencies which are pegged to the dollar, and a guarantee of payments to

depositors.

On the other hand, the oil market has experienced strong instability during the last ten

years. More explicitly, the oil price exceeded US$145/barrel in summer 2008 and then

plunged to less than US$30/ barrel in June 2014. This instability has affected the financial

sectors especially those of the oil-rich Gulf countries during the previous boom and bust oil

cycles. Oil price shocks influenced the GCC bank profitability through oil income (e.g.,

lending to private sector), increased business activity and enhanced excess liquidity in the

banking system.

Motivated by those considerations, it is important to analyze the risk spillovers

between the Islamic and conventional GCC bank stock indices in Bahrain, Kuwait, Qatar,

Saudi Arabia and UAE and two strategic commodity markets including the gold and oil

5

markets. We have constructed those Islamic and conventional indices from individual bank

stock prices for each GCC country to achieve this objective.

This study aims to examine the directional spillovers and net spillovers among the

West Texas Intermediate (WTI) crude oil and gold with both the Islamic and conventional

bank indices for the five GCC countries (Bahrain, Kuwait, Qatar, Saudi Arabia and UAE)

over the period June 1, 2006 to September 19, 2016. Further, we conduct portfolio risk

management by quantifying the Islamic and conventional GCC optimal portfolios’ weights,

hedge ratios and hedging effectiveness for those GCC markets. The choice of the Islamic

GCC banks is motivated by the large size of their assets and their important growth during the

last few years, in addition to their risk exposure and business models that are compliant with

the Shariah rules that govern the GCC economies. Regarding the commodity markets, gold is

widely known as a refuge asset and oil is selected because the GCC countries (except

Bahrain) are heavily oil dependent as indicated earlier.

Consequently, the GCC financial markets are sensitive to oil price shocks. Oil should

also command more importance in Islamic GCC markets because many of the major oil-

producing GCC countries strongly follow the Islamic faith, and this underlines a risk feedback

between these markets, particularly during periods of economic downturns and financial

turmoil. The increase in crude oil prices in the last few years has increased the growth rate of

the Islamic bank assets in the Middle East (Khan, 2010). We note that these two important

commodities (oil and gold) offer different volatility and return behaviors during the financial

turmoil periods. This is another reason that motivates us to select these two important

commodities with those GCC markets.

The study contributes to the existing literature in four aspects. First, it uses the

dynamic equicorrelation (DECO)-FIGARCH model of Engle and Kelly (2012) to investigate

the dynamic conditional correlations between the commodity markets with Islamic and

6

conventional bank indices. The (DECO)-FIGARCH model outperforms the standard GARCH

model because of its ability to detect the long-range memory process in the conditional

variance of the financial time series and assumes a time-varying correlation among financial

assets. The DECO model provides an efficient way to assess in depth the variability in

correlations during different market regimes (Hammoudeh et al., 2016). It is worth noting that

this model is a special case of the DCC model in which the correlations across all pairs of the

markets are equal, and the common equicorrelation is time-varying (Pan et al., 2016).

Moreover, the DECO model provides consistent estimates of the DCC parameters in large

systems, and in this study it allows one to quantify the linkages of the commodity and bank

markets as a common group for the purpose of portfolio diversification in assets issued by

these markets.

Second, the paper analyzes the directional spillovers and net spillovers across the oil

and gold commodities and GCC bank markets, using the spillover index developed by

Diebold and Yilmaz (2012). Within this framework, we use a rolling sample approach to

detect the time-varying dynamics of the volatility spillover index, since the recent financial

crises directly affected the volatility structures between these markets. Third, we assess the

net volatility spillover index and the sensitivity of this index as a robustness check. Finally,

we quantify the optimal portfolio weights, hedge ratios, and hedging effectiveness for the

major commodity and bank stock asset portfolios. In fact, investors attempting to offset their

risk exposures and risks against downturn market movements should adjust their hedge ratios

and rebalance strategies in accordance with the movement of the bank market (bear or bull

markets) conditions.

Our empirical results show that the average conditional correlation between these

markets is close to zero. They also indicate that oil, gold, and the conventional banks of Saudi

Arabia, Kuwait and Qatar are a net source of volatility spillovers, while all the Islamic banks

7

and the conventional banks of UAE and Bahrain which house foreign banks are a net

recipient of volatility spillovers. Regarding the optimal weights for the commodities, we show

that it is optimal for the GCC banks to hold more oil than gold in a diversified portfolio. The

results for the average optimal hedge ratios show that all the ratios are weak for gold and

close to zero for oil. Finally, gold offers the best hedging effectiveness for UAE, Qatar and

Saudi Arabia banks, while oil provides the highest hedging effectiveness for the banks of

Bahrain which is a minor oil producer.

The remainder of this paper is organized as follows. Section 2 develops the

methodology used in this study. Section 3 describes the data and conducts preliminary

analyses. Section 4 provides and discusses the empirical results. Section 5 draws implications

for risk management and portfolio diversification. Section 6 provides concluding remarks.

2. Empirical method

In this section, we describe the empirical methods which begin with a description of the

bivariate DECO-FIGARCH model that assesses the dynamic correlations between the

commodities and banking stock indexes under consideration. We also employ the volatility

spillover index of Diebold and Yilmaz (2012), which in this study identifies the dynamics of

directional volatility spillovers across the Islamic and conventional GCC banking stock

indexes.

2.1. The DECO-FIGARCH model

To explore the time variation in conditional correlations between the commodity and

GCC bank markets, this study presumes that the return-generating process tr can be

described by an autoregressive moving average (ARMA (1,1)) model as follows:

8

1111 tttt rr , t with t t tz h , (1)

where [0, ) , 1 , 1 and the innovations { }tz follow the Student-t

distribution ,1,0~ STzt .2 The conditional variance th is positive with probability one and

is a measurable function of the variance-covariance matrix 1t .

The FIGARCH qdp ,, model of Baillie et al., (1996) is expressed as follows:

1 1 2[1 ( )] [1 [1 ( )] ( )(1 ) ]d

t th L L L L , (2)

where is the mean of the process, is the GARCH parameter, is the finite order lag

polynomials and d is the fractional differencing parameter capturing long memory to be

estimated where 0 1d , and L denotes the lag operator. The fractional differencing operator

(1 )dL is defined as:

0

( )(1 )

( ) ( 1)

kd

k

k d LL

d k

(3)

The FIGARCH model provides a greater flexibility for modeling the conditional variance and

can distinguish between the covariance stationary GARCH model for 0d and the non-

stationary IGARCH model when 1d , while for 0 1d the FIGARCH model is

sufficiently flexible to allow for an intermediate range of persistence.

In order to obtain dynamic correlations between the variables under consideration, we

review the DCC model of Engle (2002). Assume that the conditional expectation 1 0t tE

and the conditional variance-covariance matrix 1t t t tE H , where tE is the conditional

expectation for using the information set available at time t . The conditional variance-

covariance matrix tH can be written as:

1/2 1/2

t t t tH D R D , (4)

2 The Student-t distribution is estimated with the parameter , which represents the number of degrees of

freedom (df) and measures the degree of leptokurtosis displayed by the density (Fiorentini et al., 2003).

9

where ,t i j tR is the conditional correlation matrix, while the diagonal matrix of the

conditional variances is given by , ,, ,t i t n tD diag h h . Engle (2002) models the right-hand

side of Eq. (4), rather than tH , directly by proposing the following dynamic correlation

structure:

1/2 1/2

* *

t t t tR Q Q Q

, (5)

*

t tQ diag Q , (6)

, 1 1 11t i j t t t tQ q a b S au u bQ , (7)

where , ,, ,t i t n tu u u is the standardized residuals (i.e. , , ,/i t i t i tu h ),

i j t tS s E u u is the n n unconditional covariance matrix of tu , and a and b are non-

negative scalars satisfying 1a b ). The resulting model is called the DCC model.

In this context, Aielli (2013) proves that the estimation of the covariance matrix tQ in

this way is inconsistent because t tE R E Q , and suggests the following the consistent

DCC model (cDCC model) for the correlation-driving process:

* *1/2 *1/2

1 1 1 1 11t t t t t tQ a b S a Q u u Q bQ , (8)

where *S is the unconditional covariance matrix of *1/2

t tQ u .

Engle and Kelly (2012) suggest that t be modelled by using the cDCC process to obtain

the conditional correlation matrix tQ and then taking the mean of its off-diagonal elements.

This approach which reduces the estimation time is called the dynamic equicorrelation

(DECO) model. The scalar equicorrelation is defined as:

1 ,

1 1

, ,

1 2,

1 1

n n i j tDECO cDCC

t n t n i j i

ii t jj t

qJ R J n

n n n n q q

(9)

10

where , , 1 , 1 ,

DECO DECO DECO

i j t t DECO i t j t t DECO i j t tq a u u b q , which is the ),( jith

element of the matrix tQ from the cDCC model. We then use this scalar equicorrelation to

estimate the conditional correlation matrix:

1t t n t nR I J , (10)

where nJ is the n n matrix of ones and nI is the n -dimensional identity matrix. This process

allows one to represent the comovement degree of a group of commodity futures with a single

time-varying correlation coefficient.

Note that the estimation of the DECO model is carried out using a two-step maximum

likelihood of the probability density function of a bivariate Student-t distribution expressed

as:

2/1log2/122

/2

2log

tt Hv

vvv

vl

2/1log2 1 vHv ttt , (11)

where is the Gama function, v is the degree of freedom for the Student t distribution,

tH is a conditional variance-covariance matrix. is a parameter vector with all of the

coefficients of the DECO-FIGARCH model.

2.2. Spillover Index framework

We apply the generalized VAR methodology, variance decomposition, and the

generalized volatility spillover index of Diebold and Yilmaz (2012) to examine the directional

spillovers and net spillovers across the Islamic and conventional GCC banking stock indexes.

Following Diebold and Yilmaz (2012), we assume a covariance stationary n -variable

VAR( p ):

11

p

t i t tiy y , (12)

11

where ty is the 1n vector of endogenous variables, i are n n autoregressive coefficient

matrices, and t is a vector of error terms that are assumed to be serially uncorrelated. If the

VAR system above is a covariance stationary, then a moving average representation is written

as 0t j tj

y A

, where the n n coefficient matrix jA obeys a recursion of the form

1 1 2 2j j j p j pA A A A , with 0A being the n n identity matrix and 0jA for

0j . The total, directional and net spillovers are generated by the generalized forecast-error

variance decompositions of the moving average representation of the VAR model. The

framework of the generalized variance decompositions eliminates any dependence of the

results on the ordering of the variables.

Koop et al. (1996) and Pesaran and Shin (1998) propose the following H -step-ahead

generalized forecast-error variance decomposition:

211

0

1

0

,

H

jj i h jh

ij H

i h h ih

e A eH

e A A e

(13)

where is the variance matrix of the vector of errors , and jj is the standard deviation of

the error term of the j th

equation. Finally, ie is a selection vector with a value of one for the

i th element, and zero otherwise. The spillover index yields a n n matrix ijH H ,

where each entry gives the contribution of variable j to the forecast error variance of variable

i . The own-variable and cross-variable contributions are contained in the main diagonal and

the off-diagonal elements of the H matrix, respectively.

Because the own- and cross-variable variance contribution shares do not sum to one

under the generalized decomposition (i.e., 1

1n

ijjH

), each entry of the variance

decomposition matrix is normalized by its row sum as follows:

12

1

ij

ij n

ijj

HH

H

, (14)

with 1

1n

ijjH

and

1

n

ijjH n

by construction.

This allows one to define a total spillover index as:

, 1, , 1,

, 1

100 100.

n n

ij iji j i j i j i j

n

i j

H HTS H

nH

(15)

This index measures the average contribution of the spillovers of a volatility shock from

one market to all (other) markets to the total forecast error variance. Additionally, this index is

flexible and enables an identification of the directional spillovers among all markets.

Specifically, the directional spillovers received by market i from all other markets j are

defined as:

1, 1,

, 1

100 100.

n n

ij ijj j i j j i

i j n

iji j

H HDS H

nH

(16)

Similarly, the directional spillovers transmitted by market i to all other markets j are defined

as:

1, 1,

, 1

100 100.

n n

ji jij j i j j i

i j n

jii j

H HDS H

nH

(17)

The set of directional spillovers provides a decomposition of total spillovers into those

coming from (or to) a particular market. In the present application, this means that this

spillover matrix H consists of the main diagonal elements reflecting own-market spillovers,

and the off-diagonal elements reflecting cross-market spillovers.

Finally, by subtracting Eq. (17) from Eq. (16), we compute the net volatility spillovers

from each market to all other markets as:

.HDSHDSHNS jijii (18)

13

The net spillovers demonstrate whether a market is a receiver or a transmitter of

spillovers in net terms. It is also our interest to examine the net pairwise spillovers (NPS) as

following:

, 1 , 1

100.ji ij

ij n n

ik jk

i k j k

H HNPS H

H H

(19)

The net pairwise spillovers between markets i and j is simply the difference between

the gross shocks transmitted from market i to market j and those transmitted from j to i .

3. Data and preliminary analysis

3.1. Data

Our data set consists of daily closing spot prices for the Islamic and conventional bank

stock price indexes for the five GCC countries (Bahrain, Kuwait, Qatar, Saudi Arabia and

UAE). We exclude Oman because of a lack of sufficient data. We use the daily closing spot

price data for the three-month futures contract of the West Texas Intermediate (WTI) crude

oil, which is the benchmark or the reference price for the U.S. crude oil. These futures

contacts are traded on the NYMEX. The data for the closing oil price, which is expressed in

USD/barrel, is extracted from the US Energy Information Administration (EIA) website

(www.eia.gov). The gold futures price, expressed in USD/troy ounce, is traded on COMEX in

New York. The data for gold and bank stock price series are extracted respectively from the

World Gold Council and Bloomberg database.

The sample period runs from June 1, 2006 through September 19, 2016 (a total of

2221 observations) for the two commodity markets and all Islamic and conventional GCC

countries. The beginning of the sample period is dedicated by the data availability for the Gulf

banks. This period covers several episodes of wide instabilities and crises including the

14

dramatic increases in oil prices throughout 2007 and early 2008, the 2008-2009 GFC, the

2010-2012 ESDC, the 2011-2013 Arab spring , the drastic oil price plunge since June 2004

and the gradual recovery of the global stock markets.

To analyze the risk spillovers between the two commodity and five bank stock

markets, we construct two banking stock price indices for the Islamic banks and conventional

banks for each of the five GCC countries. Following Fakhfekh et al. (2016), each constructed

banking stock index is defined as a weighted average of the bank stock prices for each GCC

country. Formally, each Islamic or conventional GCC banking stock price index is defined

as

n

iT

i

CB

CB1

Price, with iCB denoting the stock market capitalization of the bank i,

n

i iT CBCB1

, “Price” denotes the daily observed ith bank stock price and n is the number

of banks in a specific GCC market.

We calculate the continuously compounded daily returns by taking the difference in

the logarithms of two consecutive GCC bank stock indices or commodity prices. Figure 1

displays the dynamic returns of the oil, gold, Islamic and conventional GCC bank stock index

returns. The figure shows some periods of significant fluctuations and that all commodities

and GCC banks exhibit a volatility clustering. We apply the Markov-switching-dynamic

regression (MS-DR) to detect two tranquil and volatile regimes in the return series that allows

one to identify the beginning and end of each phase of the financial crises, which is of great

importance when one deals with the cross-market spillovers issue.3

The shaded regions highlight the regimes of excess volatility according to the MS-DR

and show the effects of GFC on these banks. In fact, the oil and gold prices and all of the

Islamic and conventional GCC bank stock index returns exhibit significant abrupt variations

3 The MS-DR model identifies the existence of two regimes: regime zero (the “stable” regime) and regime one

(the “volatile/crisis” regime). For further details of the MS-DR model, see Hamilton (1988) and Hamilton and

Susmel (1994).

15

during the crisis period 2008-2009 but with different magnitudes. Concerning the two

commodity markets, the figure reveals that oil is more volatile than gold. This graphical

analysis will be confirmed by the descriptive statistics below. Among the Islamic GCC bank

stock indices, we note that the Islamic bank index for Bahrain (Saudi Arabia) is less (more)

affected by the GFC due to the concurrent collapse in oil prices which is the main revenue

source for the Saudi economy. Bahrain is a minor oil producer. Further, the Saudi banks are

more segmented than the Bahrain banks and do not receive much help from international

banks but are still affected by global oil shocks. As for the conventional GCC bank stock

price indices, the Kuwait bank index is more affected by the GFC.

Another volatility clustering can be observed during the period 2014-2016 which

corresponds to the recent oil price plunge. In fact, the economies of these Gulf countries,

particularly that of Saudi Arabia, are highly oil dependent as indicated.

16

Fig. 1. Dynamics of the commodity and GCC index returns

Note: The shaded areas highlight the regimes of excess volatility according to the Markov switching-dynamic

regression (MS-DR).

17

3.2. Preliminary analysis

Table 1 provides the descriptive statistics of the daily oil, gold and Islamic and

conventional bank stock index returns for the GCC banks. This table shows that the average

daily (price) index returns are negative for oil and all GCC bank index return series except for

the Bahrain conventional bank index and the gold price. For the case of the conventional and

Islamic GCC bank indices, the (unconditional) volatility as measured by the standard

deviations is the highest for the Islamic GCC banks than for their conventional counterparts.

Among the Islamic (conventional) GCC bank stock indexes, that of UAE (Saudi Arabia) is

the most volatile, while that for Bahrain (UAE) is the least volatile. The UAE banks may be

affected by the heavy borrowing of Dubai and the aftermath of the 2006 Dubai debt crisis. On

the other hand, the Bahrain banks have a life line support with major American and European

banks since Bahrain is an international financial center in that region.

Fig. 1 plots the time variations of the commodity (oil and gold) price returns and the

GCC bank share index returns. This figure shows evidence of stylized facts for all of the

return series, such as volatility clustering and asymmetric volatility. This graph also shows

evidence of a presence of nonlinearity and structural breaks. Thus, a nonlinear model (such

our GARCH model) fits the data. The shaded regions underline the regimes of excess

volatility according to the Markov-switching dynamic regression (MS-DR) and provide

evidence of strong volatility clustering during the GFC period for all the considered markets

but with different magnitudes. Despite the similarity of the Islamic GCC banks, those banks

present different magnitudes of stylized facts. Further, the effect of the oil price plunge on the

bank returns is observed between years 2014-2015.

Among the commodity markets, the oil market is more volatile than the gold market

and is also more volatile than the GCC bank markets. These descriptive statistics are in

concordance with the dynamics of the commodity and stock index returns displayed in Fig. 1.

18

The skewness coefficients are negative for the majority of the GCC bank stock index return

series (except for the Qatari Islamic bank stock index). The kurtosis coefficients are above

three for all the index return series, which indicates that the probability distributions of the

commodity price returns, Islamic and conventional GCC banking index returns are skewed

and leptokurtic, thus rejecting normality. This finding is also confirmed by the Jarque-Bera

statistics.

To examine the stationarity process, we use two popular conventional unit root tests,

the augmented Dickey and Fuller (1979) and Phillips and Perron (1988) tests, and the

stationarity test of the Kwiatkowski et al. (1992) test to check stationary of the series. The

results indicate that all the index return series are stationary at the conventional levels.

Further, we examine the existence of the ARCH effect using the LaGrange multiplier

test of Engle (1982), and the results show that all the return series exhibit an ARCH behavior.

Therefore, the estimation of a GARCH model is appropriate for modeling stylized facts such

as fat-tails, clustering volatility and persistence for the commodity price and bank index

returns. According to the Ljung-Box test results, we provide evidence for serial correlations

for the residual and squared residual at the conventional level of 1%.

The presence of a long memory process in the financial markets is essential for

analyzing the time-varying risk spillovers. For this purpose, we justify the use of the

FIGARCH model. However, we use four popular conventional long memory (LM) tests

namely the Hurst-Mandelbrot R/S test, Lo’s modified R/S test, the Gaussian semi-parametric

(GSP) test of Robinson and Henry (1999), and the GPH test of Geweke and Porter-Hudak

(1983).4 Tables 2 and 3 present the LM tests’ results for the return and the squared return (as a

proxy of volatility) series for the commodity (oil and gold) prices and the Islamic and

conventional bank indexes for the GCC countries. The results do not reject the null hypothesis

4 For further information on the LM tests, the reader can read the article of Mensi et al. (2014b, 2014c).

19

of no LM property when examining the return series for stocks and commodities, with the

exception of the UAE conventional bank stock index when we consider the GSP test. This

result is consistent with the LM literature. However, the LM results for the stock index returns

change radically when we consider volatility. In fact, we strongly reject the null of no LM

property for all commodity and GCC bank squared returns at the 1% significance level,

regardless of the applied LM tests. In sum, the squared returns may be governed by a

fractionally integrated model for all cases. These results also justify the use of the FIGARCH

model.

20

Table 1: Descriptive statistics of the pairs of the daily oil price, gold price, and the Islamic and conventional GCC bank stock index returns

Gold WTI SAI SAC QAI QAC UAI UAC BAI BAC KUI KUC

Mean 0.0328 -0.0233 -0.0522 -0.0347 -0.0051 -0.0002 -0.0121 -0.0188 -0.0639 0.0007 -0.0114 -0.0187

Max. 8.6249 16.414 11.208 21.284 13.644 10.011 14.353 11.645 12.548 5.9954 13.351 9.1586

Min. -9.8104 -19.164 -20.749 -22.564 -14.252 -15.131 -11.372 -15.348 -15.934 -7.7241 -14.734 -6.5142

Std. dev. 1.3681 2.6971 2.0145 1.9846 1.9233 1.6406 2.0367 1.3952 1.5634 1.0076 2.0204 1.0811

Skewness -0.3736 -0.1939 -0.8969 -0.3656 0.0281 -0.7132 -0.0282 -0.3574 -0.5238 -0.4423 -0.2489 -0.0605

Kurtosis 8.9056 9.9403 15.406 30.084 13.815 15.740 11.602 25.250 19.799 9.6833 10.650 10.161

J-B 3277.***

3278.***

14535.***

67904.***

10819.***

15202.***

6844.***

45839.***

26206.***

4204.***

5436.***

4745.***

Q(30) 41.102 58.246**

56.743***

61.969***

90.842***

109.79***

92.137***

65.814***

63.585***

74.581***

71.125***

69.236***

Q2(30)

395.08

*** 1952.

*** 371.05

*** 467.44

*** 1294.9

*** 1449.8

*** 1708.7

*** 177.88

*** 126.86

*** 462.07

*** 3650.

*** 1912.

***

ADF -48.26***

-31.74***

-44.27***

-50.10***

-40.67***

-40.87***

-42.85***

-30.02***

-43.86***

-46.82***

-48.31***

-45.83***

PP -48.27***

-49.33***

-44.33***

-50.08***

-40.58***

-40.83***

-42.81***

-44.16***

-43.90***

-47.38***

-48.40***

-45.82***

KPSS 0.2458 0.1244 0.1987 0.1195 0.0787 0.0963 0.3256 0.1762 0.0695 0.1508 0.1606 0.1851

LM-

ARCH(10) 10.876

*** 35.259

*** 13.777

*** 68.843

*** 31.540

*** 34.564

*** 54.238

*** 14.602

*** 5.4886

*** 16.158

*** 56.025

*** 44.023

***

Notes: SAI, SAC, QAI, QAC, UAI, UAC, BAI, BAC, KUI and KUC are respectively the Saudi Arabia Islamic bank index, Saudi Arabia conventional bank index, UAE Islamic bank index,

UAE conventional bank index, Bahrain Islamic bank index, Bahrain conventional bank index, Kuwait Islamic bank index and Kuwait conventional bank index. J-B denotes the empirical

statistics of the Jarque-Bera test for normality. The Ljung-Box Q(30) and Q2(30)

tests for no autocorrelation of residuals and square residuals, respectively. ADF, PP and KPSS are the

empirical statistics of the Augmented Dickey and Fuller (1979), and the Phillips and Perron (1988) unit root tests and the Kwiatkowski et al. (1992) stationarity test, respectively. The ARCH-

LM(10) test of Engle (1982) checks the presence of ARCH effects. The asterisk *** denotes the rejection of the null hypotheses of normality, no autocorrelation, unit root, stationarity, and

conditional homoscedasticity at the 1% significance level.

21

Table 2. Results of the long memory tests for the returns of oil price, gold price, and the Islamic and conventional GCC banking stock indexes

Gold WTI SAI SAC QAI QAC UAI UAC BAI BAC KUI KUC

Panel A: Hurst-Mandelbrot R/S test

R/S statistic 1.413 1.217 1.639 1.395 1.240 1.755 1.777 1.689 1.135 1.794 1.331 1.407

Panel B: Lo’s modified R/S test ( 1)q 1.431 1.224 1.591 1.441 1.159 1.643 1.695 1.632 1.101 1.789 1.349 1.389 ( 5)q 1.473 1.226 1.528 1.444 1.099 1.619 1.657 1.509 1.063 1.703 1.375 1.414

Panel C: GSP test

4/Tmd -0.029

(0.021)

0.0015

(0.021)

0.029

(0.021)

-0.028

(0.021)

-0.008

(0.021)

-0.004

(0.021)

0.041

(0.021)

0.062***

(0.021)

-0.002

(0.021)

0.069***

(0.021)

-0.031

(0.021)

-0.0002

(0.021)

16/Tmd -0.070

(0.042)

0.106

(0.042)

-0.018

(0.042)

-0.044

(0.042)

-0.074

(0.042)

0.0387

(0.042)

0.241***

(0.042)

0.183***

(0.042)

-0.089

(0.042)

0.229

(0.042)

0.026

(0.042)

0.030

(0.042)

32/Tmd 0.0612

(0.063)

0.195

(0.063)

-0.011

(0.063)

-0.012

(0.063)

0.038

(0.063)

-0.041

(0.063)

0.122

(0.063)

0.345***

(0.063)

-0.001

(0.063)

0.124

(0.063)

0.184***

(0.063)

0.238***

(0.063)

64/Tmd 0.113

(0.084)

-0.166

(0.084)

-0.019

(0.084)

0.033

(0.084)

0.133

(0.084)

0.050

(0.084)

0.074

(0.084)

0.270***

(0.084)

0.012

(0.084)

0.124

(0.084)

0.169

(0.084)

0.265***

(0.084)

Panel D: GPH test

45.0Tmd 0.192

(0.134)

-0.141

(0.134)

0.007

(0.134)

0.019

(0.134)

0.044

(0.134)

0.026

(0.134)

0.151

(0.134)

0.209

(0.134)

0.243

(0.134)

0.054

(0.134)

0.078

(0.134)

0.181

(0.134)

5.0Tmd 0.034

(0.106)

0.095

(0.106)

-0.004

(0.106)

0.076

(0.106)

0.031

(0.106)

0.102

(0.106)

0.150

(0.106)

0.244

(0.106)

0.198

(0.106)

0.037

(0.106)

0.144

(0.106)

0.196

(0.106)

55.0Tmd 0.129

(0.085)

0.220

(0.085)

-0.026

(0.085)

0.037

(0.085)

0.021

(0.085)

0.037

(0.085)

0.193

(0.085)

0.281***

(0.085)

0.051

(0.085)

0.170

(0.085)

0.225

(0.085)

0.188

(0.085)

6.0Tmd 0.024

(0.069)

0.170

(0.069)

-0.003

(0.069)

0.022

(0.069)

0.090

(0.069)

0.083

(0.069)

0.289***

(0.069)

0.190***

(0.069)

-0.040

(0.069)

0.231***

(0.069)

0.118

(0.069)

0.165

(0.069) Notes: The critical values of the Hurst-Mandelbrot R/S test and Lo’s modified R/S analysis are 1.862 and 2.098 at the 5% and 1% significance levels, respectively. The numbers in the

parentheses are the standard deviations of the estimates. “ q ” in Lo’s modified R/S test is the number of lags of autocorrelation. m denotes the bandwidth for the GSP and GPH tests. ** and ***

indicate significance at the 5% and 1% levels, respectively.

22

Table 3. Results of the long memory tests for the squared returns of oil price, gold price, and the Islamic and conventional GCC banking stock indexes

Gold WTI SAI SAC QAI QAC UAI UAC BAI BAC KUI KUC

Panel A: Hurst-Mandelbrot R/S test

R/S statistic 3.190***

4.459***

4.566***

2.433***

4.938***

5.019***

4.466***

2.799***

3.422***

4.343***

5.894***

5.189***

Panel B: Lo’s modified R/S test

( 1)q 3.093***

4.038***

4.247***

2.232***

4.542***

4.690***

3.871***

2.515***

3.300***

4.018***

5.250***

4.753***

( 5)q 3.190***

3.011***

3.654***

2.120***

3.617***

3.665***

2.818***

2.265***

2.933***

3.478***

3.937***

3.626***

Panel C: GSP test

4/Tmd 0.234***

(0.021)

0.374***

(0.021)

0.206***

(0.021)

0.100***

(0.021)

0.266***

(0.021)

0.296***

(0.021)

0.373***

(0.021)

0.115***

(0.021)

0.181***

(0.021)

0.201***

(0.021)

0.344***

(0.021)

0.308***

(0.021)

16/Tmd 0.415***

(0.042)

0.417***

(0.042)

0.353***

(0.042)

0.175***

(0.042)

0.438***

(0.042)

0.481***

(0.042)

0.378***

(0.042)

0.196***

(0.042)

0.238***

(0.042)

0.355***

(0.042)

0.625***

(0.042)

0.431***

(0.042)

32/Tmd 0.541***

(0.063)

0.830***

(0.063)

0.268***

(0.063)

0.257***

(0.063)

0.402***

(0.063)

0.390***

(0.063)

0.399***

(0.063)

0.327***

(0.063)

0.282***

(0.063)

0.454***

(0.063)

0.768***

(0.063)

0.558***

(0.063)

64/Tmd 0.555***

(0.084)

0.574***

(0.084)

0.587***

(0.084)

0.275***

(0.084)

0.494***

(0.084)

0.474***

(0.084)

0.074***

(0.084)

0.431***

(0.084)

0.448***

(0.084)

0.413***

(0.084)

0.527***

(0.084)

0.566***

(0.084)

Panel D: GPH test

45.0Tmd 0.583

***

(0.134)

0.576***

(0.134)

0.727***

(0.134)

0.307***

(0.134)

0.496***

(0.134)

0.605***

(0.134)

0.358***

(0.134)

0.501***

(0.134)

0.557***

(0.134)

0.371***

(0.134)

0.493***

(0.134)

0.585***

(0.134)

5.0Tmd 0.607

***

(0.106)

0.828***

(0.106)

0.592***

(0.106)

0.276***

(0.106)

0.425***

(0.106)

0.475***

(0.106)

0.435***

(0.106)

0.417***

(0.106)

0.475***

(0.106)

0.508***

(0.106)

0.689***

(0.106)

0.579***

(0.106)

55.0Tmd 0.603

***

(0.085)

0.871***

(0.085)

0.339***

(0.085)

0.279***

(0.085)

0.423***

(0.085)

0.489***

(0.085)

0.461***

(0.085)

0.374***

(0.085)

0.375***

(0.085)

0.448***

(0.085)

0.845***

(0.085)

0.642***

(0.085)

6.0Tmd 0.549

***

(0.069)

0.529***

(0.069)

0.413***

(0.069)

0.230***

(0.069)

0.506***

(0.069)

0.442***

(0.069)

0.470***

(0.069)

0.256***

(0.069)

0.340***

(0.069)

0.384***

(0.069)

0.828***

(0.069)

0.573***

(0.069)

Notes: See the notes of Table 2.

23

4. Empirical results

4.1. Estimation of marginal model

To avoid the non-synchronous trading effect in the world’s financial markets, we have

followed Forbes and Rigobon (2002) to employ two-day rolling returns based on each

aggregated market index. Furthermore, we have applied the two-day rolling returns to the

multivariate DECO-FIGARCH model for modelling all stylized facts for stock returns such as

volatility clustering, volatility persistence (long memory), and time-variations in conditional

volatility and correlation.

The results of the estimation of the multivariate DECO-FIGARCH (1, d, 1) model

between the commodity (oil and gold) and GCC bank stock (Islamic and conventional)

markets are summarized in Table 4. Panels A, B and C summarize the mean and variance

equations, and the average correlation for each pair and the diagnostic tests, respectively.5

Looking at Panel A, we find that the autoregressive (AR(1)) parameter of the mean equation

is positive and statistically significant at the 1% level for all cases (except for the

conventional bank stock index of UAE). This result indicates that the past returns are

instantaneously and rapidly included in the current returns for these banks. Moreover, the

fractional integrated coefficient (d) is highly significant for all the return series, suggesting a

high level of persistence. Among all series, the Islamic bank stock index for Saudi Arabia

exhibits the highest parameter, while the conventional bank stock index of UAE presents the

lowest long memory parameter. We also note that the Islamic bank stock indices for Saudi

Arabia, Qatar and Kuwait are more persistent than their conventional counterparts.

Panel B shows that the DECOa coefficient is positive and statistically significant at the

conventional level, underlying the importance of shocks between the commodity and both the

5 One should note that the lag order (1, d, 1) is selected by using the Akaike Information Criteria (AIC) and the

Schwarz Information Criteria (SIC). The results are not reported here to honor space but they are available upon

request.

24

Islamic and conventional bank GCC stock index return. The DECOb parameter is significant

and very close to one for all pairs. This result corroborates the results of higher persistence of

volatility across the considered markets as determined by the variance equation, particularly

the GARCH parameter and the d-FIGARCH parameter.

We show that the average correlation is close to zero (0.083) but is statistically

significant at the 1% level. This result exhibits the presence of diversification investment

opportunities between the markets. Additionally, the evidence on the degrees of freedom of

the Student-t distribution (df) indicates that fat-tailedness characterizes the distribution of all

return series. Taking together, the significance of the parameters DECOa , DECOb and df

demonstrates the appropriateness of using the DECO-FIGARCH model with Student-t

distributions. The results of the diagnostic tests using the Ljung–Box tests for serial

correlation in the standardized residuals and the squared standardized residuals results do not

reject the null hypothesis of no serial correlation in all pairs. This finding provides no evidence

of misspecification in our model.

Fig. 3 plots the dynamic equicorrelation for the group of the commodity and GCC

bank markets. From this figure, we can draw several interesting findings. First, we observe

time-varying correlations over the sample period, suggesting that institutional investors do or

should frequently change their portfolio structure. Second, the correlation is positive and

weak for all sample period, with a higher level of correlation during the GFC with a value

equals 0.35. This rise in correlation between the markets decreases the potential of

diversification benefits during crises. Finally, the correlation level increases during the more

recent period 2014–2015, which corresponds to the oil price collapse. This result supports the

hypothesis of financial contagion between the commodity and GCC stock returns under

consideration. Thus, the trajectories of the time-varying conditional correlations show that the

share prices of the Islamic and conventional GCC banks are not immune against international

25

factors and oil price shocks. Following the recent oil price decline, the conditional

correlations exhibit a gradual decrease, indicating a gradual recovery of these markets.

Fig. 2. Dynamic equicorrelation among the commodities and GCC banks

Note: The dynamic equicorrelation between Gold, WTI and the five GCC bank indices is estimated from the

multivariate ARMA-FIGARCH(1,d,1)-DECO model.

26

Table 4. Estimation of the multivariate ARMA-FIGARCH(1,d,1)-DECO model

Saudi Arabia Qatar UAE Bahrain Kuwait

Gold WTI SAI SAC QAI QAC UAI UAC BAI BAC KUI KUC

Panel A: Estimates of ARMA-FIGARCH(1,d,1) model

Const. 0.0304

(0.0264)

0.0365

(0.0426)

0.0243

(0.0290)

-0.0701

(0.0770)

0.0127

(0.0253)

0.0373

(0.0255)

0.0106

(0.0299)

-0.0361

(0.0235)

-0.0274***

(0.0081)

0.0163

(0.0169)

0.0008

(0.0255)

-0.0080

(0.0152)

AR(1) -0.0181

(0.0208)

-0.0175

(0.0244)

0.0849***

(0.0263)

0.1009**

(0.0503)

0.0858***

(0.0303)

0.0819***

(0.0268)

0.0475

(0.0254)

0.0069

(0.0548)

-0.0140***

(0.0293)

-0.0274

(0.0266)

-0.1242***

(0.0247)

-0.0547**

(0.0241)

MA(1) 0.9812***

(0.0032)

0.9773***

(0.0036)

0.9647***

(0.0133)

0.9636***

(0.0158)

0.9478***

(0.0088)

0.9543***

(0.0079)

0.9571***

(0.0055)

0.9685***

(0.0059)

0.9182***

(0.0094)

0.9862***

(0.0025)

0.9800***

(0.0030)

0.9806***

(0.0043)

Const. 1.1145

(0.8869)

1.7433**

(0.7276)

2.0059

(1.5310)

0.1318

(0.0854)

3.7552

(3.7946)

0.0070

(0.0036)

6.6453

(4.3303)

0.4249***

(0.1303)

4.0441

(3.2239)

0.2806**

(0.1188)

0.7906**

(0.3862)

0.3539**

(0.1627)

d-FIGARCH 0.4823***

(0.1324)

0.4971***

(0.0683)

0.7240***

(0.0752)

0.3189**

(0.1623)

0.6628***

(0.0908)

0.5266***

(0.0700)

0.6126***

(0.0609)

0.1859***

(0.0706)

0.7097***

(0.0924)

0.3615***

(0.0894)

0.4542***

(0.0662)

0.4339***

(0.0652)

ARCH 0.2442***

(0.0990)

0.2956***

(0.0771)

0.0072

(0.0894)

0.2331***

(0.2497)

0.2956***

(0.0982)

0.3354***

(0.0683)

0.0546

(0.0915)

0.8563***

(0.0981)

0.3686**

(0.1676)

0.4126***

0.0801)

0.4296***

(0.0761)

0.1818

(0.1076)

GARCH 0.6718***

(0.1265)

0.6844***

(0.0902)

0.6116***

(0.0847)

0.1291***

(0.1818)

0.7155***

(0.0981)

0.7197***

(0.0596)

0.5268***

(0.0836)

0.8851***

(0.0831)

0.7183***

(0.1279)

0.6749***

(0.0868)

0.6735***

(0.0759)

0.4409***

(0.1223)

Panel B: Estimates of the DCC model DECO

t 0.10059***

(0.0098)

DECOa 0.0584***

(0.0155)

DECOb 0.8929***

(0.0327)

df 6.1278***

(0.1884)

AIC 27.86028

SIC

28.02476

Panel C: Diagnostic tests

Q(30) 25.804

[0.6850]

19.099

[0.9378]

38.392

[0.1399]

37.093

[0.1744]

31.792

[0.3772]

20.926

[0.8901]

40.661

[0.0926]

32.991

[0.3229]

18.941

[0.9412]

36.962

[0.1782]

35.309

[0.2315]

28.963

[0.5195]

Q2(30)

19.458

[0.9298]

19.963

[0.9174]

7.0039

[0.9999]

8.4525

[0.9999]

27.581

[0.5925]

11.093

[0.9993]

23.626

[0.6235]

12.126

[0.9984]

0.2375

[1.0000]

20.394

[0.9058]

35.922

[0.2106]

29.802

[0.4758]

Notes: Q(30) and Q2(30) are the Ljung-Box test statistic applied to the standard residuals and the squared standardized residuals, respectively. The P-values are reported in brackets [.]. The

standard error values are reported in parentheses (.). The asterisks ** and *** indicate significance at the 5% and 1% levels, respectively.

27

4.2. Total volatility spillover index and rolling-sample spillover analysis

Table 5 summarizes the estimated results of the total volatility spillover matrix. We

note that the (i, j)th entry in each panel is the estimated contribution to the forecast-error

variance of variable i coming from innovations to market j. The row sums excluding the main

diagonal elements (termed ‘From others’) and the column sums (termed ‘To others’) report

the total spillovers to (received by) and from (transmitted by) each volatility.

The total volatility spillovers value is 29%, indicating diversification gains. Let us first

focus on the directional spillovers transmitted ‘To others’. Gold has a lower impact on the all

Islamic and conventional GCC bank stock indices except for the bank stock index of Saudi

Arabia than oil. In contrast, the WTI crude oil price contributes more significantly to the

conventional banks than the Islamic counterparts for Bahrain, Saudi Arabia and Qatar. The

crude oil acts as the price discovery tool for the conventional bank of these three countries.

Further, the risk spillovers between the GCC bank markets themselves are low because those

banks are isolated according to country lines and heavily involved with the national

governments. For instance, the Islamic Saudi banks contribute 3.8%, 2.3%, 0% and 0.2% to

the forecasting variance of the Islamic banks of Qatar, UAE, Bahrain and Kuwait,

respectively. Moreover, the volatility transmission from the Islamic and conventional GCC

bank to oil and the yellow metal is weak. Taking for example the gold market, the risk

spillover coefficient is close to zero from the Bahrain Islamic and UAE conventional banks to

the shiny metal, and is less than 1% for the spillovers from Saudi Arabia and Qatar. The

Kuwait Islamic (conventional) bank stock index contributes to 0.98% (2.4%) to the

forecasting variance of gold.

28

Table 5. Total volatility spillovers for the commodity and GCC bank.

To (i) From(j)

Gold WTI SAI SAC QAI QAC UAI UAC BAI BAC KUI KUC From others

Gold 87.7 2.18 0.9 0.7 0.7 0.5 0.3 0.07 0.05 0.59 0.98 2.4 9.3

WTI 5.44 67.9 0 0.21 2.1 0.5 0.9 0.03 0.23 1.04 7.81 14 32.1

SAI 0.16 0.02 76 4.48 1.8 3.5 5.2 0.03 0.6 0.05 1.13 6.9 23.8

SAC 0.69 0.58 1.4 90.8 0.7 0.6 0.4 0.06 0.56 0.25 0.58 2.4 8.2

QAI 1.57 3.29 3.8 0.73 48 21 4.8 0.09 0.08 2.67 4.69 9.9 52.1

QAC 3.35 3.68 4.9 0.78 18 46 5.6 0.03 0.01 4.9 3.52 8.9 54.1

UAI 0.31 3.88 2.3 1.29 2.4 5.7 71 0.05 1.29 0.2 1.82 9.6 28.9

UAC 0.03 0.13 0.4 0.17 0.1 0.1 0 96.9 0.53 0.75 0.09 0.6 3.1

BAI 0.01 0.1 0 0.11 0.6 0.4 0.8 0.43 91.6 1.96 3.54 0.4 8.4

BAC 0.63 2.59 0 0.06 2.7 6.9 1 0.14 0.55 74.5 5.91 5.1 25.5

KUI 2.13 15 0.2 0.2 5.3 3.7 2.7 0 0.9 6.32 46.9 17 53.1

KUC 1.47 13.2 0.3 0.95 4.9 4.8 2.9 0.14 0.04 3.61 17.6 52 49.9

To others 15.8 44.7 14 9.7 40 47 25 1.1 4.8 22.3 47.6 76 348

All 104 113 91 101 87 93 96 98 96.5 96.8 94.5 128 Total: 29%

Notes: The underlying variance decomposition is based on a daily VAR of order 4 (as determined by the Schwarz information criterion) using the generalized VAR spillover index

of Diebold and Yilmaz (2012). The (i,j)th element of the table shows the estimated contribution to the variance of the 10-step-ahead forecast error of i coming from innovation

shocks to variable j. The diagonal elements (i=j) are the own variance share estimates, which show the fraction of the forecast error variance of market i that is due to its own shocks.

The last column “From others” shows the total spillovers received by a particular market from all other markets, while the row “To others” shows the spillover effect directed by a

particular market to all other markets. The lower right corner “Total” indicates the level of total spillovers.

29

Similar results can be seen for the crude oil market. This result demonstrates the role

played by the precious metal market as a refuge asset during the market turmoil, and this

result is in line with previous studies including, among others, Baur and Lucey (2010), Baur

and McDermott (2010), Bredin et al. (2015), Mensi et al. (2015a).

For the graphical evidence, we plot the time-varying volatility spillover index in Fig.

3. A close inspection of this figure, we show that the volatility spillovers attain their

maximum level during 2008–2009 and 2015–2016, which as indicated earlier corresponds to

the GFC and the oil price plunge periods. Moreover, we see the commodity-bank linkages are

highly influenced by the political and economic events as illustrated in Figure 3. The 2007–

2008 commodity crisis, the 2011 Arab spring revolution and the changes of rates by the U.S.

Federal Reserve between 2009-2013 and the Chinese stock markets crash in summer 2015

increase the spillovers between these markets, which reduces investment diversification

opportunities for the considered markets. It is worth noting that in July 2015 the Shanghai

stock market had fallen 30% over three weeks, as more than half of the listed companies filed

for a trading halt in an attempt to prevent further losses. Again, the Shanghai index fell in

August by 8.48%, which is the largest fall since 2007.

30

Fig. 3. The dynamics of the total volatility spillover index

Notes: The dynamics of total volatility spillovers are calculated from the forecast error variance decompositions

of 10-step-ahead forecasts with 200-day rolling windows.

4.3. Net volatility spillover

We deepen our analysis by examining the time-varying behavior of the volatility

spillovers. More specifically, we study the net pairwise volatility spillovers, which provide a

fruitful information about the directional volatility spillovers among the bank-commodity

futures markets. We divide the total volatility spillover index into two directional spillovers: i)

the receivers of volatility spillovers, termed directionally as ‘from’, and ii) the transmitters of

volatility spillovers, termed directionally as ‘to’. The net dynamic volatility spillover index is

then computed by subtracting the directional ‘to’ spillovers from the directional ‘from’

spillovers. The positive (negative) values indicate a source (a recipient) of return and

volatility to (from) others.

Table 6 reports the net directional pairwise index to identify the main net recipients an

d contributors to the volatility spillovers. Regarding the directional spillovers received ‘From

others’, the results exhibit that the WTI oil receives more shocks from the rest of the markets t

31

han gold. Also, the Islamic GCC bank stock indices receive more risk than the conventional b

ank indices for all the GCC markets. In fact, the Qatar Islamic bank stock index receives more

risk from the remaining markets (commodities and the rest of banks) than any of the other Isl

amic bank stock indices, while the Islamic bank stock index of Bahrain receives less risk than

the other markets. The Qatar conventional bank index receives 6.73% and 47.3% of the risk s

pillovers from gold, oil and other Islamic and conventional GCC bank stock indexes.

From this table, we can conclude that both the gold and oil markets are net

contributors of risk. More precisely, gold receives risk spillovers from the Saudi Islamic and

conventional bank index, the UAE conventional bank index, the Bahrain Islamic bank index

and the Kuwait conventional bank index, while gold is a net contributor to the remaining

markets. In fact, gold contributes to the error forecast variance of the conventional bank index

of Qatar a meager of 2.88%, the conventional bank index of the Bahrain 0.04%, and the

Islamic bank index for Kuwait 1.15%. Again, this result is in line with previous works (see

Baur and Lucey, 2010; Mensi et al., 2015b; Mensi et al., 2016) on the ability of gold to be a

good hedge and/or a safe haven asset, not only for the stock markets but also for the bank

markets of the GCC economies.

The WTI crude oil receives risk spillovers from gold (3.26%), the Islamic banks of

Bahrain (0.1%) and the conventional banks of Kuwait (0.6%) On the other hand, oil

contributes to the other markets with risk spillovers ranging from 0.1% for the UAE

conventional banks to 7.21% for the Islamic banks of Kuwait. On the other hand, we find that

the risk spillovers between GCC bank stock markets is weak since these markets are well

capitalized, strongly supervised by their respective central banks and domestically isolated

from other GCC banks. The GCC governments also deposit their oil revenues in those banks

and borrow from them, and thus there is no room for contagion to take place among them.

32

To sum up, we conclude that the gold and oil markets as well as the conventional

banks of Saudi Arabia, Qatar and Kuwait are net-contributors of risk to the other markets. In

contrast, all Islamic banks of the five GCC economies and both the Bahrain and UAE

conventional banks are net-recipients of risk from the other markets. A large percentage of

investors in the Bahrain and UAE stock markets come from Saudi Arabia and the other GCC

markets. The UAE market may reflect the heavy borrowing by Dubai from international

markets.

Fig. 4 depicts the time-varying evolution of the net volatility spillover index for each

market. As shown in this figure, we can identify the source or the recipient of the net

volatility spillovers, despite the net volatility spillover oscillates in either the negative or the

positive direction and that their magnitudes have often changed over time. We observe that

gold and oil are net sources of risk spillovers. This result indicates that gold and oil exposure

poses a systemic risk to both Islamic and conventional banking sector in GCC countries. It is

worth noting that oil has become a net recipient of risk spillovers after 2015 in the light of

shocks from the U.S. shale oil shocks and Saudi Arab’s changing oil policy objectives. That

is, this result is due to the significant plunge in oil prices owing to the increase in the U.S.

shale oil production and the Saudi Arabian policy of maximizing oil market share. Further, we

see that the Islamic banks for all GCC banks as well as the conventional banks (stock indices)

of Bahrain and UAE are net recipients of risk spillovers. This result is due to market openness

to the foreign investors. Thus, these banks become integrated with international markets. The

rest of the conventional GCC banks are sources of risk spillovers. We note that the Islamic

banks of Saudi Arabia are a net recipient of risk spillovers between 2011-2012 which

corresponds to the ESDC. Also, the Islamic bank stock index of UAE is a net recipient of the

risk spillovers between 2015-2016 which again corresponds to the recent drop in oil prices.

33

On the whole, the huge exposure of GCC banks to the oil sector coupled with the low

price of crude oil in the international market continues to cause concerns for GCC central

banks. The Islamic banks are also more risky than conventional banks which can also be a

source of additional concern. This result is due to fact that Islamic banks have a smaller

universe than conventional banks. In addition, Islamic finance restricts hedging against risk

contrary to what conventional model does, thus it is difficult for Islamic banks to manage

stock markets and foreign exchange risks when foreign options and futures are not allowed.

Islamic banks face a legal risk as the legal systems in their countries do not have specific laws

or statutes that support the unique features of Islamic financial instruments. Islamic banks

face concertation risk as 60% of the financing provided by those banks is by the way of

Murabiha (cost plus). This graphical analysis is in line with the results reported in Table 5.

34

Table 6. Net directional pairwise indices spillovers

Gold WTI SAI SAC QAI QAC UAI UAC BAI BAC KUI KUC Net Conclusion

Gold 0 3.26 -0.8 -0.01 0.9 2.88 0 -0.04 -0 0.04 1.15 -0.9 6.48 Net-contributor

WTI -3.26 0 0 0.37 1.2 3.17 2.9 0.1 -0.1 1.55 7.21 -0.6 12.57 Net-contributor

SAI 0.76 0 0 -3.1 2.03 1.45 -2.9 0.41 -0.6 -0.04 -1 -6.6 -9.49 Net-recipient

SAC 0.01 -0.37 3.1 0 0.07 0.15 0.9 0.11 -0.5 -0.19 -0.4 -1.4 1.5 Net-contributor

QAI -0.9 -1.2 -2 -0.07 0 -2.19 -2.4 0.05 0.52 -0.01 0.59 -5 -12.56 Net-recipient

QAC -2.88 -3.17 -1.5 -0.15 2.19 0 0 0.11 0.41 2 0.19 -4 -6.73 Net-contributor

UAI -0.01 -2.94 2.88 -0.85 2.37 -0.05 0 -0.01 -0.5 0.78 0.91 -6.7 -4.11 Net-recipient

UAC 0.04 -0.1 -0.4 -0.11 -0.1 -0.11 0 0 -0.1 -0.61 -0.1 -0.5 -2 Net-recipient

BAI 0.04 0.13 0.57 0.45 -0.5 -0.41 0.5 0.1 0 -1.41 -2.6 -0.4 -3.56 Net-recipient

BAC -0.04 -1.55 0.04 0.19 0.01 -2 -0.8 0.61 1.41 0 0.41 -1.5 -3.16 Net-recipient

KUI -1.15 -7.21 0.96 0.38 -0.6 -0.19 -0.9 0.09 2.64 -0.41 0 0.96 -5.43 Net-recipient

KUC 0.91 0.58 6.59 1.4 4.95 4.03 6.7 0.47 0.39 1.46 -1 0 26.49 Net-contributor

Notes: The net directional pairwise spillovers, obtained as the difference between the contribution from variable i and the contribution from variable j in Table 5. The column “Net” indicates the

total sum of the net directional pairwise spillovers, expressed as a negative value (net-recipient) and a positive value (net-contributor), respectively.

35

36

Fig. 4. Time-varying net volatility spillover indices

Notes: The time-varying net volatility spillover indices are calculated by subtracting the directional ‘to’

spillovers from the directional ‘from’ spillovers. The positive (negative) values of the spillovers indicate that the

variable is a net contributor (recipient) of the spillovers.

4.4. Robustness tests

For robustness, we conduct two statistical tests to examine the sensitivity of our

spillover results. To start, we check the choice of the order of the VAR and compute the

spillover index for orders 2 to 6 and plot the minimum, maximum and the median values in

Fig. 5(a). Next, we plot the spillover index for the forecast horizons ranging from five to ten

days in Fig. 5(b). Both Figs. 5(a) and 5(b) reveal the spillover indices that appear to follow

similar patterns. This finding indicates that the total spillover plot is not sensitive to the choice

of the order of the VAR or the choice of the forecast horizon. Similar alternative values as

robustness tests are also adopted by previous studies in the literature (Diebold and Yilmaz,

2009, 2012, 2014; Chau and Deesomsak, 2014; Antonakais and Kizys, 2015 among others).

37

Fig. 5. Robustness tests

Note: (a) Sensitivity of the index to the VAR lag structure (max, min, and median values of the index for VAR

orders 2–6); (b) Sensitivity of the index to forecast horizon (max, min, and median values over 5- to 10-day

horizons).

5. Portfolio design and hedging strategy analysis

The above empirical results provide evidence of risk spillovers across the commodity

and GCC bank markets under consideration. These results have important implications for

having efficient diversified portfolios and conducting risk management. Practically, building

an optimal portfolio based on risk management and portfolio allocation decisions requires a

preliminary and accurate estimation of the temporal covariance matrix. To manage the

commodity-bank more efficiently, we use the estimated results of the DECO-FIGARCH

model, which allows investors to make optimal portfolio allocation decisions by constructing

dynamic risk-minimizing hedge ratios. Thus, we quantify the optimal portfolio weights and

the hedge ratios for designing optimal hedging strategies.

To minimize risk without reducing expected returns, we consider a portfolio

construction of commodity and bank assets. To do this, we assume an investor is holding a set

of bank assets and wishes to hedge her position against unfavorable effects with commodity

assets. Specifically, we follow Kroner and Ng (1998) to define the portfolio weight of the

holdings of commodity (gold or oil) assets by:

38

B

t

BC

t

C

t

BC

t

B

tC

thhh

hhw

,

,

2, with

11

10

00

C

t

C

t

C

t

C

t

C

t

w

ww

w

w , (20)

where C

th , B

th and BC

th , are the conditional volatility of the commodity markets, the

conditional volatility of the stock market and the conditional covariance between the

commodity and the bank asset at time t, respectively. From the budget constraint, the optimal

weight of the bank asset is equal to (1 )Ctw . For each commodity-stock pair, all information

needed to compute the weight Ctw is obtained from the DECO-FIGARCH model.

Following Kroner and Sultan (1993), we apply the beta hedge approach in order to

minimize the risk of this bank-commodity portfolio. We thus measure how much a long

position (buy) of one dollar in the commodity asset (gold or oil) should be hedged by a short

position (sell) of C

t dollar in the GCC bank shares, that is:

C

t

BC

tC

th

h ,

, (21)

The hedging effectiveness of the constructed portfolios can be assessed by comparing

the realized hedging errors (Ku et al., 2007), which are defined by Eq. (22)

unhedged

hedged

Var

VarHE 1 , (22)

where the variance of the hedge portfolio ( hedgedVar ) is the variance of the returns of the

weighted portfolio of a commodity and a bank stock (PF II), whereas the variance of the

unhedged portfolio (unhedgedVar ) is the variance of the returns of the benchmark portfolio (PF

I). A higher HE ratio implies a greater hedging effectiveness measured in terms of the

portfolio’s variance reduction, which thus implies that the associated investment policy can be

deemed a better hedging strategy.

Table 7 presents the values of the optimal portfolio weights, the hedge ratios and the

39

hedging effectiveness. A glance at the coefficients of the optimal weights for commodities, we

find that the GCC banks include holding more oil than gold. On the other hand, the other result

exhibits that institutional investors should hold on average larger weights of commodity assets

than bank assets. By taking Kuwait, for instance, we observe that the optimal weight is 67.27%

(38.55%) for gold while the rest of the wealth should be invested in the conventional (Islamic)

banks. For the oil, the optimal weights are 86.79% (65.16%) and the remainder of the wealth

is invested in the conventional (Islamic) banks of Kuwait. On the whole, the optimal allocation

for gold in a one-dollar conventional bank portfolio ranges from 22.06% to 67.27%, while the

Islamic bank portfolio varies from 38.55% to 68.97%. For oil, the optimal weights for all bank

markets are above 64% and less than 87% respectively to GCC banks. Among the bank

markets, the UAE Islamic (Kuwait conventional) banks should hold a higher proportion of the

gold (oil) commodity asset.

The average optimal hedge ratios show that all the ratios are weak for gold (less than

15%) or close to zero for oil. Looking first at the gold market, the largest ratio reaches 0.1506

for the SAI-gold pair, meaning that a one-dollar long position in the gold should be shorted by

15.06 cents of the Islamic banks of the Saudi Arabia. The lowest ratio is equal to 0.0401 for the

UAE Islamic bank index, indicating that a one-dollar long position in gold should be shorted

by 4.01%. The hedge ratio results for oil can be explained by the fact that these economies are

oil-dependent. More interestingly, the optimal hedge ratios vary slightly across the GCC

banking markets. This result means that investors should hold more of the yellow metal than

GCC bank stocks to minimize risk for investors with bank stock holdings in that region.

Finally, we can conclude that oil is the cheapest hedge for the Islamic bank index of

Saudi Arabia, whereas the most expensive hedge is for the Islamic bank index of UAE. As for

the yellow metal, the cheapest (most expensive) hedge is for the Islamic bank index of UAE

(the Islamic bank index of Saudi Arabia. Note that for oil (gold), the hedge ratio is higher for

40

Islamic banks than their conventional counterparts for all cases, except for Saudi Arabia (UAE).

The time-varying optimal hedge ratios from the estimates of the DECO-FIGARCH

model for the bank-gold pairs are plotted in Figure 6.6 The graphical evidence is in line with

the results reported in Table 7. This plot exhibits a higher variability of the estimated hedge

ratio for the bank stock markets of UAE and Saudi Arabia which demonstrate significant

changes during 2008 and 2014, which is due to the shocks of the onset of the global financial

crisis as well as the recent oil price plunge. The remaining pairs show similar patterns but with

different magnitudes, and for this reason, we will not interpret them. During oil price shocks

and GFC, investors tend to hold more long positions in god and short positions in GCC bank

shares.

Finally, we analyze the hedging effectiveness by actually running portfolio simulations

with our optimal portfolio designs and hedging ratios. More concretely, we build two

portfolios: a portfolio which is composed of only bank stock indexes (PF I) and a weighted

portfolio contains a precious metal (or oil) and a bank stock index with the optimal portfolio

weights calculated above (PF II). The results in Table 7 show that hedging strategies

involving the GCC bank stock and commodity markets make it possible to reduce portfolio

risks. It is worth noting that the hedging effectiveness value for all pairs is positive and high,

suggesting that a significant risk reduction can be realized and that the hedged portfolio is

able to decrease the risk exposure. For the UAE conventional (Islamic) banks, the variance

reduction ranges from 99.11% (99.6%) for gold to 79.7% (93.51%) for oil. More interestingly,

gold offers the best hedging effectiveness for UAE, Qatar, Bahrain and Saudi Arabia while oil

provides the highest hedging effectiveness for Bahrain followed by Qatar, Kuwait and Saudi

Arabia. In sum, gold (oil) provides the best hedging for the conventional (Islamic) banks of

Bahrain.

6 The plots of the dynamic hedge ratios for the bank-oil portfolio are available upon request.

41

Table 7. Optimal portfolios’ weights, hedge ratios and hedging effectiveness

Portfolio C

tw C

t HE (%)

SAI/ GOLD 0.3927 0.1506 97.38

SAC/ GOLD 0.3886 0.1487 99.36

QAI/ GOLD 0.4545 0.1377 98.87

QAC/ GOLD 0.2206 0.1194 96.34

UAI/ GOLD 0.6897 0.0401 99.11

UAC/ GOLD 0.5390 0.1008 99.60

BAI/ GOLD 0.4602 0.1173 97.44

BAC/ GOLD 0.6642 0.0769 99.62

KUI/ GOLD 0.3855 0.1471 90.43

KUC/ GOLD 0.6727 0.0802 97.79

SAI/ WTI 0.8554 0.0211 86.98

SAC/ WTI 0.6443 0.0822 98.98

QAI/ WTI 0.7007 0.0748 99.03

QAC/ WTI 0.7382 0.0659 99.41

UAI/ WTI 0.6418 0.0862 79.70

UAC/ WTI 0.7602 0.0578 93.51

BAI/ WTI 0.7016 0.0669 99.39

BAC/ WTI 0.8548 0.0430 99.42

KUI/ WTI 0.6516 0.0790 98.49

KUC/ WTI 0.8679 0.0431 96.75

Notes: The numbers in bold identify the hedged portfolio which has the highest variance reductions. PFI is a

portfolio of 100% GCC bank stocks, while PFII is the weighted stock and precious metal portfolio in which the

weights are given by the optimal weights.

42

Fig. 6. Time-varying hedging ratios between the GCC bank stock indices and gold.

43

6. Conclusion

Risk spillovers are of great importance for institutional investors, particularly

conventional and Islamic banks in the GCC region, because of high geopolitical risks and

dependency on the highly volatile oil prices. However, counterparty credit risk management

and financial stability requires monitoring and quantifying the risk spillovers of these

financial institutions.

This paper’s objective is to explore the risk spillovers and examine hedging

effectiveness between the two major commodity markets (gold and crude oil), and the Islamic

and conventional banks for five GCC countries ((Bahrain, Kuwait, Qatar, Saudi Arabia and

UAE). For this purpose, we use the dynamic equicorrelation FIGARCH model (DECO-

FIGARCH) and the risk spillovers index developed by Diebold and Yilmaz (2012).

The results show that the average conditional correlations between the commodity and

GCC bank markets are weak due to the fact that most GCC banks are segmented from each

other and from the major international banks, suggesting an ample room for diversification

opportunities for international investors. They also show that gold has a lower impact on the

Islamic and conventional GCC bank than on the oil market. In fact, oil prices increase the risk

spillovers to the conventional GCC banks more than to their Islamic counterparts as measured

by the bank stock index for Bahrain, Saudi Arabia and Qatar. On the other hand, the volatility

spillovers from the Islamic and conventional GCC banks to the commodity markets are weak.

Oil and gold are strategic and global commodities, while most of the GCC banks are

segmented from major international financial markets. Furthermore, oil, gold, and the

conventional banks of Saudi Arabia, Kuwait and Qatar are a net source of volatility spillovers

to the remainder of the markets. In contrast, all the Islamic banks and the conventional banks

of UAE and Bahrain are a net receipt of volatility spillovers. A good portions of investors in

the stock markets of those countries come from Saudi Arabia.

44

By having a close inspection of the coefficients of the optimal weights for the two

commodities, the result reveals that the GCC bank stock indices should hold more oil than

gold. The average optimal hedge ratios show that all the ratios are weak for gold and close to

zero for oil. Finally, gold offers the best hedging effectiveness for UAE, Qatar and Saudi

Arabia which are major oil exporters, while oil provides the highest hedging effectiveness for

Bahrain which is a minor oil producer.

These results have several important insights for policy makers. Given that both

Islamic and conventional GCC banks are less exposed to the risk of gold, banks in oil-

exporting countries should seek diversification benefits in holding gold. For Bahrain, this

country is a minor oil producer and can find diversification benefits in holding oil. Having

more investors from Saudi Arabia and other GCC countries, buying stocks in Bahrain, Dubai

and Abu Dhabi makes the latter countries net receivers for volatility spillover from the former

countries. In this case investors in Bahrain, Dubai and Abu Dhabi should make sure they

include diversifiers from outside the GCC region. Investors dealing with Islamic GCC bank

stocks should be cognizant they need more hedges and safe havens than those who invest in

conventional GCC bank stocks. Islamic GCC banks should also hold collaterals as

conventional GCC banks do to safeguard depositors’ base. Islamic banks should have a higher

risk-weighted capital asset requirements than conventional banks because they get involved in

more risk activities. The Islamic GCC banks depend on more real estate investments than

conventional GCC banks. Moreover, the GCC governments deposit their oil revenues and

borrow money from the conventional banks than from the Islamic banks. Finally, GCC

central banks do not have an elaborate supervision of Islamic banks as they do for the

conventional banks.

45

References

Aielli, G. P. 2013. Dynamic conditional correlation: on properties and estimation. Journal of Business

and Economic Statistics 31, 282–299.

Antonakakis, N., Kizys, R., 2015. Dynamic spillovers between commodity and currency markets.

International Review of Financial Analysis 41, 303–319.

Baillie, R.T., Bollerslev, T., Mikkelsen, H.O., 1996. Fractionally integrated generalized autoregressive

conditional heteroskedasticity. Journal of Econometrics 74, 3–30.

Baur, D.G., Lucey, B.M., 2010. Is gold a hedge or a safe haven? An analysis of stocks, bonds and

gold. Financial Review 45, 217–229.

Baur, D.G., McDermott, T.K., 2010. Is gold a safe haven? International evidence. Journal of Banking

and Finance 34, 1886–1898.

Bredin, D., Conlon, T., Poti, V., 2015. Does gold glitter in the long-run? Gold as a hedge and safe

haven across time and investment horizon. International Review of Financial Analysis 41, 320–

328.

Chau, F., Deesomsak, R., 2014. Does linkage fuel the fire? The transmission of financial stress across

the markets. International Review of Financial Analysis 36, 57–70.

Dickey, D., Fuller, W., 1979. Distribution of the estimators for autoregressive time series with a unit

root. Journal of the American Statistical Association 74, 427–431.

Diebold, F.X., Yilmaz, K., 2009. Measuring financial asset return and volatility spillovers, with

application to global equity markets. Economic Journal 119, 158–171.

Diebold, F.X., Yilmaz, K., 2012. Better to give than to receive: predictive directional measurement of

volatility spillovers. International Journal of Forecasting 28, 57–66

Diebold, F.X., Yilmaz, K., 2014. On the network topology of variance decompositions: measuring the

connectedness of financial firms. Journal of Econometrics 182, 119–134.

Engle, R.F., 1982. Autoregressive conditional heteroskedasticity with estimates of the variance of

United Kingdom inflation. Econometrica 50, 987–1007.

Engle, R.F., 2002. Dynamic conditional correlation: a simple class of multivariate generalized

autoregressive conditional heteroskedasticity models. Journal of Business & Economic Statistics

20, 339–350.

Engle, R.F., Kelly, B., 2012. Dynamic equicorrelation. Journal of Business & Economic Statistics 30,

212–228.

Fakhfekh, M., Hachicha, N., Jawadi, F., Selmi, N., Cheffou, A.I., 2016. Measuring volatility

persistence for conventional and Islamic banks: an FI-EGARCH approach. Emerging Markets

Review 27, 84–99.

Fiorentini, G., Sentana, E., Calzolari, G., 2003. Maximum likelihood estimation and inference in

multivariate conditionally heteroscedastic dynamic regression models with Student-t innovations.

Journal of Business & Economic Statistics 21, 532–546.

Geweke, J., Porter-Hudak, S., 1983. The estimation and application of long memory time series

models. Journal of Time Series Analysis 4, 221–238.

Hamilton, J. D., 1988. Rational–Expectations Econometric Analysis of Changes in Regime. An

Investigation of the Term Structure of Interest Rates. Journal of Economic Dynamics and Control

12, 385–423.

Hamilton, J.D., Susmel, R., 1994. Autoregressive conditional heteroscedasticity and changes in

regime. Journal of Econometrics 64, 307–333.

46

Hammoudeh, S., Kang, S.H., Mensi, W., Nguyen, D.K., 2016. Dynamic global linkages of the BRICS

stock markets with the United States and Europe under external crisis shocks: implications for

portfolio risk forecasting. The World Economy 39, 1703–1727 .

Khan, F., 2010. How ‘Islamic’ is Islamic banking? Journal of Economic Behavior & Organization 76,

805–820.

Koop, G., Pesaran, H.M., Potter, S., 1996. Impulse response analysis in nonlinear multivariate models.

Journal of Econometrics 74, 119–147.

Kroner, K. F., Ng, V. K., 1998. Modeling asymmetric comovements of asset returns. Review of

Financial Studies 11, 817–844.

Kroner, K.F., Sultan, J., 1993. Time varying distribution and dynamic hedging with foreign currency

futures. Journal of Financial and Quantitative Analysis 28, 535–551.

Ku, Y. H., Chen, H., Chen, K., 2007. On the application of the dynamic conditional correlation model

in estimating optimal time-varying hedge ratios. Applied Economics Letters 14, 503–09.

Kwiatkowski, D., Phillips, P.C. B., Schmidt, P., Shim, Y., 1992. Testing the null hypothesis of

stationarity against the alternative of a unit root: how sure are we that economic time series are

non-stationary? Journal of Econometrics 54, 159–178.

Mensi W., Hammoudeh, S., Kang, S.H., 2015a. Precious metals, cereal, oil and stock market linkages

and portfolio risk management: evidence from Saudi Arabia. Economic Modelling 51, 340–358.

Mensi, W., Hammoudeh, Reboredo, J.C., Nguyen, D.K., 2014a. Do global factors impact BRICS

stock markets? A quantile regression approach. Emerging Markets Review 19, 1–17.

Mensi, W., Hammoudeh, Reboredo, J.C., Nguyen, D.K., 2015b. Are Sharia stocks, gold and U.S.

Treasury hedges and/or safe havens for the oil-based GCC markets? Emerging Markets Review

24, 101–121.

Mensi, W., Hammoudeh, S., Tiwair, A-K., 2016. New evidence on hedges and safe havens for Gulf

stock markets using the wavelet-based quantile. Emerging Markets Review 28 155–183.

Mensi, W., Hammoudeh, S., Yoon, S-M., 2014b. How do OPEC news and structural breaks impact

returns and volatility in crude oil markets? Further evidence from a long memory process. Energy

Economics 42, 343–354.

Mensi, W., Hammoudeh, S., Yoon, S.M., 2014c. Structural breaks and long memory in modeling and

forecasting volatility of foreign exchange markets of oil exporters: The importance of scheduled

and unscheduled news announcements. International Review of Economics and Finance 30, 101–

119.

Pan, Z., Wang, Y., Liu, L., 2016. The relationship between petroleum and stock returns: An

asymmetric dynamic equi-correlation approach. Energy Economics 56, 453-463.

Pesaran, H.M., Shin, Y., 1998. Generalized impulse response analysis in linear multivariate models.

Economics Letters 58, 17–29.

Phillips, P.C.B., Perron, P., 1988. Testing for a unit root in time series regression. Biometrica 75, 335–

346.

Robinson, P. M., Henry, M., 1999. Long and short memory conditional heteroscedasticity in

estimating the memory parameter of levels. Econometric Theory 15, 299–336.