The Quarterly Review of Economics and Finance 49 (2009) 829842

Contents lists available at ScienceDirect

The Quarterly Review of

Economics and Finance
journal homepage: www.elsevier.com/locate/qref

Shock and volatility spillovers among equity sectors of the

Gulf Arab stock markets
Shawkat M. Hammoudeh a,, Yuan Yuan a,1, Michael McAleer b,c,d,2
a
Lebow College of Business, Drexel University, 3141 Chestnut Street, Philadelphia, PA 19104, United States
b
School of Economics and Commerce, University of Western Australia, Australia
c
Econometric Institute Erasmus University Rotterdam, Tinbergen Institute, The Netherlands
d
Department of Applied Economics, National Chung Hsing University, Taiwan

a r t i c l e i n f o a b s t r a c t

Article history: The major objectives of this study are twofold. The rst objective is
Received 7 November 2008 to examine the dynamic volatility and volatility transmission in a
Received in revised form 31 March 2009
multivariate setting using the VAR(1)GARCH(1,1) model for three
Accepted 15 April 2009
major sectors, namely, Service, Banking and Industrial/or Insur-
Available online 24 April 2009
ance, in four Gulf Cooperation Council (GCC)s economies (Kuwait,
Qatar, Saudi Arabia and UAE). The second is to use the models
JEL classication:
results to compute and analyze the optimal weights and hedge
C33
G12 ratios for two-sector portfolio holdings, comprised of the three sec-
tors for each country. The results suggest that past own volatilities
Keywords: matter more than past shocks and there are moderate volatility
VAR(1)GARCH spillovers between the sectors within the individual countries, with
Volatility
the exception of Qatar. Moreover, the values for ratios of hedging
Shocks
long positions with short positions in the GCC sectors are smaller
Spillovers
Portfolio designs than those for the US equity sectors. The optimal portfolio weights
favor the Banking/nancial sector for Qatar, Saudi Arabia and UAE
and the Industrial sector for Kuwait.
2009 Published by Elsevier B.V. on behalf of the Board of
Trustees of the University of Illinois.

Corresponding author. Fax: +1 215 895 6673.

E-mail addresses: hammousm@drexel.edu (S.M. Hammoudeh), yy23@drexel.edu (Y. Yuan), michael.mcaleer@gmail.com
(M. McAleer).
1
Fax: +1 215 895 6673.
2
Fax: +31 0 10 408 1263.

1062-9769/$ see front matter 2009 Published by Elsevier B.V. on behalf of the Board of Trustees of the University of Illinois.
doi:10.1016/j.qref.2009.04.004
830 S.M. Hammoudeh et al. / The Quarterly Review of Economics and Finance 49 (2009) 829842

1. Introduction

In developed countries, equity sector investing has been popular for many years. Investors invest
in defensive stocks, such as those of the non-cyclical consumer goods sector, when the economy is
teetering into recession. They invest in high tech sectors stocks when the economy is booming. In
international investing, portfolio managers who follow the top down approach usually pick countries
and then sectors. Even informed investors choose sectors without paying much attention to interac-
tions and volatility transmission among sectors. In frontier markets such as the markets of the rich
oil-producing countries, sector investing has not yet reached similar popularity and their markets
lack organized sector indices. While there have been studies that examine the transmission of returns
among individual sectors within a system, information is still needed on how volatility spillovers occur
among sectors in multivariate settings. This knowledge is particularly useful because of the increase in
globalization and contagion among world nancial markets. The current transmission of high volatility
among sectors of individual countries and among countries is a vivid and topical example.
More recent literature on Middle East and North Africa (MENA) market volatility uses univariate
GARCH models and examines volatility behavior at the market index level. Hammoudeh and Li (2008)
examine sudden changes in volatility for ve GCC stock markets at the market index level, using the
iterated cumulative sums of squares (ICSS) algorithm, and analyze their impacts on the estimated
persistence of volatility. They nd that most of these stock markets are more sensitive to major global
events than to local and regional factors.
Zarour and Siriopoulos (2008) use the univariate CGARCH model developed by Engle and Lee (1993)
to investigate the existence of volatility decomposition into short run and long run components. They
apply this model to daily market index returns data for nine emerging markets in the Middle East
region, including three of our GCC countries. Hammoudeh and Choi (2007) employ the univariate
GARCH approach with Markov-switching to study the volatility behavior for the transitory and per-
manent components of the individual GCC market indices, allowing for two volatility regimes to exist.
While Malik and Hammoudeh (2007) use trivariate GARCH models, their systems include one individ-
ual GCC market index, the WTI oil price and S&P 500 index to analyze return volatility transmission
for three GCC markets. The volatility transmission does not involve more than one GCC market within
one system.
This paper uses a more recent multivariate technique that examines shock and volatility transmis-
sion among three sectors, namely Banking, Industrial and Service, for Kuwait, Qatar and Saudi Arabia,
and Financial, Insurance and Service, for UAE which does not have data for the Industrial sector.3 The
technique is the vector autoregressive moving average GARCH (VARGARCH) model developed by Ling
and McAleer (2003) (see Chan, Lim, & McAleer, 2005 for an early application of the model). This method
enables us to examine the conditional volatility and conditional correlation cross effects with mean-
ingful estimated parameters and less computational complications, as compared with other methods
such as the BEKK model of Engle and Kroner (1995). BEKK is a multivariate GARCH(1,1) model with
dynamic covariances and dynamic correlations, but typically is not attached to a VAR(1) model. The
VAR(1) version of BEKK has not yet been analyzed theoretically (see McAleer, Chan, Hoti, & Lieberman,
2008 for further theoretical details). For more than four or ve assets or commodities, BEKK typically
does not converge because it has far too many parameters. In short, there is little argument in favor of
BEKK, other than that it leads to a positive denite dynamic covariance matrix (see McAleer, 2005 for
further elaboration).

3
The paper focuses on the three largest equity sectors in the GCC countries, which are Service, Industrial and Banking sectors,
taking into account market capitalization, data availability and convergence reasons. We included the Insurance sector for the
UAE only because there is no data available on the Industrial sector for that country. Moreover, this small sector is relatively
more important for UAE based on the number and capitalization of the insurance companies that are traded on the Abu Dubai
and Dubais stock markets. Dubai is the third largest re-export hub in the world. We cannot include the Insurance sector for
Saudi Arabia as a fourth sector for three reasons: First, the sector is dominated by Sharia-compliant companies; second there is
no adequate time series data for it, and third most of the 21 insurance companies that exist in the country were established in
2007 and 2008, and four IPOS will be issued in 2009. The same reasons apply to Qatar which has only few insurance companies.
The ratio of the Insurance sectors market capitalization to total market capitalization for each of Kuwait, Qatar and Saudi Arabia
on December 15, 2008 is less than 1%, while it is 2.7% in UAE (see Table 1). This information was provided to us privately.
 S.M. Hammoudeh et al. / The Quarterly Review of Economics and Finance 49 (2009) 829842 831

The paper has two main objectives. (1) To examine own conditional volatility for each sector and
conditional cross sector volatility transmission of the three major equity sectors in the four individual
GCC stock markets, using the popular version of the vector autoregressive moving average GARCH
model, which is the VAR(1)GARCH model. This method enables an examination of the conditional
volatility and conditional correlation cross effects with meaningful estimated parameters and less
computational complication compared with several other methods. (2) To use the estimated results to
compute the weights of the sectors in an optimal portfolio of each GCC country, and the optimal hedge
ratios that minimize overall risk for holding the sectors in portfolios without affecting the expected
returns in the individual country.
The empirical results for the rst objective suggest that past own volatility and not past shocks is
the stronger driver in determining future volatility for the GCC frontier stock markets. This implies that
fundamentals matter more than news. Those results also show moderate volatility spillovers between
the sectors within the individual countries, with the exception of Qatar which demonstrates strong
spillovers. The results on the second objective imply that optimal portfolio weights of investors should
own much more Banking stocks than Service or Industrial stocks in Saudi Arabia and Qatar, and more
nancial stocks than Service or Insurance stocks in UAE in order to minimize risk without lowering
expected returns. Investors in Kuwait hold more Industrial stocks. The values for the hedge ratios
for the GCC sectors are smaller than those for US equity sectors, reecting the possibility of greater
hedging effectiveness in GCC markets than in the USA, thereby leading to more sophisticated hedging
techniques and strategies. These empirical results are important for the GCC countries which have
recently embarked on establishing equity funds for both individual and institutional investors.
The remainder of the paper is organized as follows. Section 2 provides a description of the data
and summary statistics. Section 3 presents the empirical VARGARCH model. Section 4 discusses the
empirical results and Section 5 provides the economic implications for designing optimal portfolios
and formulating optimal hedging strategies. Section 6 gives some concluding comments.

2. Data description

The data cover the three major sector daily indices for four of the six GCC countries, namely Saudi
Arabia, Kuwait, Qatar and UAE. The primary focus of the paper is on the three most important sectors
in each country. The sectors are the Service, Industrial and Banking sectors for the rst three countries,
and Service, Insurance and Financial for UAE, which does not have an index for the Industrial sector. Size
of market capitalization and data availability on the sector indices have been factors for the selection
of the UAE sectors. Bahrain was excluded because this kingdom changed its index series in 2003, so
that there are currently no adequate series for its sectors. Furthermore, reasonable sectoral data do
not exist for Oman.4
The sample covers the daily period from December 31, 2001 until December 31, 2007 for Saudi
Arabia, Kuwait and Qatar. The sample period for UAE starts with the same date but ends on December
10, 2007. It should be noted that the GCC countries do not share the same week-end, and their week-
ends are different from week-ends in western markets. Therefore, we cannot pool variables across
countries on a daily basis. The data set also includes a dummy variable for the 2003 Iraq war, which
is intended to capture the impact of geopolitics on sector return and volatility. We represent this
geopolitical event by a dummy variable that takes the value of zero before March 4, 2003 and one
after that. This dummy captures a geopolitical regional event of historical proportion that in reality
will continue to impact the region for many years. Statistically speaking, such events can cause breaks
in the data.
It is useful to note that the aggregate GCC stock market is dominated by the Saudi Arabian stocks,
accounting for more than 40% of the total GCC market capitalization. Moreover, across the aggregate
GCC sectors the Banking/Financial sector dominates the other sectors. In UAE the Financial sector
accounts for about 58% of the total market capitalization, followed by the Service sector based on

4
Aksel Kibar of Abu Dhabi Investment Company indicated in a private communication that It is a fact that these two markets
[Bahrain and Oman] are very illiquid. Sometimes stocks do not trade for weeks.
832 S.M. Hammoudeh et al. / The Quarterly Review of Economics and Finance 49 (2009) 829842

Table 1
Descriptive statistics for GCC sector returns.

Sector Service Industrial Banking

Saudi Arabia
Mean 0.0007 0.0014 0.0007
S.D. 0.0265 0.0229 0.0163
Skewness 0.8801 0.6010 0.5906
Kurtosis 5.2299 5.0387 7.9321
Capitalization (%) n.a. 8 31.1

Kuwait
Mean 0.0015 0.0010 0.0011
S.D. 0.0100 0.0093 0.0097
Skewness 0.1055 0.3479 0.4375
Kurtosis 2.4901 2.8554 3.2788
Capitalization (%) 26.3 8.3 34.0

Qatar
Mean 0.0009 0.0011 0.0014
S.D. 0.0143 0.0168 0.0155
Skewness 0.3276 0.0230 0.1107
Kurtosis 4.7814 3.0792 2.8739
Capitalization (%) 15.9 28.9 54.9

Sector Service Insurance Financial

UAE
Mean 0.0009 0.0004 0.0009
S.D. 0.0143 0.0101 0.0093
Skewness 0.2640 0.1611 0.1254
Kurtosis 6.9276 6.3911 5.9536
Capitalization (%) 38.9 2.7 58.3

Notes:
(1) The numbers are log differences, or returns.
(2) UAE does not have an index for the Industrial sector and Banking is called Financial.
(3) The market capitalizations for Kuwait, Qatar and Saudi Arabia are shares of sector capitalization in total market capitaliza-
tion based on trading on December 15, 2008. They should be considered just indicators of importance of sectors. These were
calculated by Abu Dhabi Investment Company for this study only. The capitalization shares for the UAE sectors are taken from
NBAD Indices Daily report issued on December 16, 2008 by National Bank of Abu Dhabi based on the NBAD market indexs
capitalization.

the capitalization of the companies included in the NBAD market index. The Insurance sector repre-
sents 2.7% of the total capitalization. This is a good representation of equity sectors shares in market
capitalization for the individual GCC countries (Table 1).
This table also provides the descriptive statistics for the daily indices of the three sectors in each
of the four GCC countries over the sample periods. The average daily index return varies among the
sectors within the same country, and for the same sector across the four countries. In Saudi Arabia, the
Industrial sector gives the greatest average return relative to the other two sectors. This sector is close
to 10% of Saudi GDP which is dominated by the government owned-oil sector. The Saudi Industrial
sector is the rst in the Arab world in terms of quantity, diversity and exporting capacity. Its economy
can thus support a relatively large Industrial base. It is not surprising that the Industrial sector in Saudi
Arabia yields the highest average return because the country has the largest economy, dened in terms
of GDP, in the GCC, Middle East and North Africa (MENA) regions. This is consistent with the empirical
ndings reported in Hammoudeh and Al-Gudhea (2006).
In Kuwait and tiny Qatar, the Service sector and Banking sector, respectively, yield the greatest
returns. Qatar is competing with Dubai and Bahrain in having the best nancial center in the region,
but it does not have a solid industrial base. In UAE, both the Service and Financial sectors give the same
highest return, which is much higher than the return in the Insurance sector. Stocks of the Insurance
sector are not as liquid as those of the other sectors. Overall, among the three sectors in the four
countries, the Service sector in Kuwait returns the highest average yield, while the Insurance sector
in UAE gives the lowest. In terms of sector historical (unconditional) risk, as dened by the standard
 S.M. Hammoudeh et al. / The Quarterly Review of Economics and Finance 49 (2009) 829842 833

deviation, most of the risk is in the Service sector for Saudi Arabia, Kuwait and UAE, but in the Industrial
sector for Qatar, whose highly concentrated industries are based on the volatile oil and natural gas.
Thus, in Kuwait and UAE, sector historical risk is commensurate with return.
Most of the returns are skewed to the left, implying that there is a greater chance that the sectors
go down than up in a given period of time. This result suggests that investors invest in these sectors for
the long haul to override the intermittent declines. The kurtosis is mixed, with some indices having a
kurtosis that is higher than the normal distribution, while for others it is lower.

3. Empirical model

As indicated above, the univariate GARCH approach has been used in modeling volatility in the
general indices of the GCC stock markets. Our objective is to apply recent techniques in modeling
volatility to upgrade the use of the univariate GARCH approach to a multivariate system. While the
BEKK model provides a multivariate GARCH(1,1) framework with dynamic covariances and dynamic
correlations, this model does not have a VAR attached to it because the BEKK VAR distribution has
not theoretically been analyzed. Moreover, the BEKK model is too excessive in parameters, many of
which lack empirical explanations. Therefore, the VARGARCH has an analysis advantage over the
BEKK model. To our knowledge, this study is the rst to apply the VARGARCH model to frontier
stock markets. We will be able to use the results of the estimated VARGARCH results to compute the
weights of the sectors in an optimal portfolio of each GCC country and the optimal hedge ratios to
analyze hedging effectiveness. Also to our knowledge, this has not been done before.
This approach will enable us to examine the conditional volatility and correlation dependency, and
interdependency of equity sectors of the GCC markets. With this approach, we will be able to focus more
on the estimation of meaningful, interpretable parameters with minimal computational difculties
than for several other models. We use the VARGARCH model developed by Ling and McAleer (2003)
to focus on the interdependence of conditional variances and conditional correlations among these
sectors. We will then compute the optimal weights and hedge ratios.
The equity sectors for each GCC stock market in the VARGARCH system are indexed by i, and n is
the total number of sectors. Based on information criteria, the mean equation for the ith sector in this
system is AR(1), and is given by:
Ri,t = ai + bi Ri,t1 + di D03 + i,t ,
1/2
(1)
i,t = hi,t i,t ,

where Ri,t is the return of the ith sector of the nx1 vector Rt , dened as the log differences and D03 is the
geopolitical dummy for the 2003 Iraq war.5 The innovation i,t is an i.i.d. random shock, and hi,t is the
conditional variance of the ith sector at time t. Ling and McAleer (2003) specied the interdependent
conditional variance as:

n 
n
hi,t = ci + ij 2j,t1 + ij hj,t1 + i D03 (2)
j=1 j=1

which is a generalization of the Bollerslev (1990) univariate GARCH process, where hi,t is the conditional
variance at time t, hj,t1 refers to ownpast2variance for i = j, and past conditional variances of the sectors
in the market or system for i = / j, ij j,t1 is the short run persistence (or the ARCH effects of past

shocks), and ij hj,t1 is the contribution to the long run persistence (or the GARCH effects of past
volatilities). We will use statistical techniques to test the signicance of including this dummy in this
model.
From (2), the conditional variance for the ith sector is impacted by past shocks and past conditional
variances of all the sectors in the market, thereby capturing interdependencies or spillovers. Therefore,

5
We have followed a large body of the literature that adds dummy variables to multivariate GARCH models. For example,
Conrad, Gultekin, and Kaul (1991), to our knowledge, is the rst paper that added exogenous dummies to those models. Those
authors included a dummy in Table 7 of their paper to control for the effect of the rst week of January.
834 S.M. Hammoudeh et al. / The Quarterly Review of Economics and Finance 49 (2009) 829842

this specication allows for the cross-sectional dependency of conditional volatilities among all the
sectors. The past shock and volatility of one sector are allowed to impact the future volatilities not only
of itself but also of all the other sectors in the system.
The parameters of the VAR(1)GARCH(1,1) system dened above are obtained by using the
maximum likelihood estimation (MLE) when the distribution of i,t is standard normal, and by quasi-
maximum likelihood estimation (QMLE) when the distribution is not standard normal. Ling and
McAleer (2003) established the structural and statistical properties of VARGARCH, using the second
and fourth moments. The i.i.d. property of i,t implies that conditional correlation matrix of t = [1,t ,
2,t , . . ., n,t ] may be modeled as constant over time (see Engle, 2002 and McAleer et al., 2008 for
dynamic extensions of the constant conditional correlation model). The constant correlation matrix is
given by  = E(t t ), where t = [1,t , 2,t , . . ., n,t ] .

4. Empirical results

We will discuss the empirical results in terms of own sector volatility and shock dependence, inter-
sector volatility, shock spillover and political risk for the three sectors in each of the four GCC countries.
As is the case with the BEKK version of the multivariate GARCH model, we are also constrained by the
number of sectors that can be included in the system to achieve computational convergence.

4.1. Volatility and shocks dependence

Most of the three sector indices in the individual GCC countries show signicant and positive sen-
sitivity to past own conditional volatility in the long run, but to considerably varying degrees. This
implies that past own conditional volatility can be used in predicting future volatility for those sectors.
The Banking/Financial sector seems to be the least volatile among the sectors for most countries. This
should not be surprising as Banking/Financial is a dominant sector in most GCC economies, ooded
with petrodollars and ushed with liquidity. Given the current global nancial crisis, this relative
Banking/Financial stability is a crucial strength of the GCC economies.
For Saudi Arabia, own conditional volatilities (s) are, to some extent, similar across sectors, with
the Service sector exhibiting the greatest relative volatility dependency (0.686), as displayed in
Table 2. The situation is different in Kuwait. In contrast to Saudi Arabia, the conditional volatility
discrepancy across sectors in Kuwait is relatively high (Table 3). While the Kuwaiti Service sector
is the most volatile (0.868), higher even than in Saudi Arabia, Banking shows the highest stability
(0.400), which is more stable than in Saudi Arabia (0.664). This implies that the Saudi Monetary Agency
(SAMA) should pay closer attention to its banks. The banks in Kuwait are strongly supervised by its
central bank. In Qatar, only the Industrial sector shows signicant positive volatility, which is close to
its counterparts in Saudi Arabia and Kuwait (Table 4). In the UAE, which does not have an index for
the Industrial sector, Insurance has the highest conditional volatility, while Financial has the lowest
conditional volatility (Table 5). While insurance companies provide insurance for different kinds of risk
that involve individuals and businesses, insurance company stocks are themselves highly sensitive to
natural, economic and political events. In the Gulf, those events are dominated by political and military
events such as the 19811988 Iran Iraq war, the 1991 Kuwait war and the 2003 Iraq war which is
included in our sample. These events increase the potential for risk and freight prices thorough the
Gulf, leading to higher conditional volatility in their stock prices. Moreover, the Insurance sector suffers
from illiquid stocks and thin markets.6
The sensitivity to past own shocks or news is also positive for all sector indices in the short run. But
this sensitivity is much smaller for all sectors across countries than the own volatility, suggesting
that past own volatilities are more important in predicting future volatility than past shocks or news.
Among the individual GCC countries, Qatar has the highest shock sensitivity in the Service and Banking

6
In Dubai Financial Market (DFM), there are 13 insurance companies; only four stocks are actively traded. In Abu Dhabi
Stock Market (ADSM), there are 14 companies; only one stock is actively traded. We thank Aksel Kibar of Abu Dhabi Investment
Company for providing us with this information.
 S.M. Hammoudeh et al. / The Quarterly Review of Economics and Finance 49 (2009) 829842 835

Table 2
Estimates of VARGARCH for Saudi Arabia.

Variables Service Industrial Banking

Mean equation
C 0.0003 0.0002 0.0004
AR(1) 0.0503b 0.0511b 0.1530
D03 0.0009 0.0022a 0.0014

Variance equation
C 3.39E06b 8.97E06a 3.62E06
2service (t 1) 0.2398a 0.1176a 0.0344
2industrial (t 1) 0.1009a 0.2069a 0.0011
2banking (t 1) 0.0770c 0.0339 0.2152
hservice (t 1) 0.6862a 0.2266a 0.0965
hindustrial (t 1) 0.1956a 0.6049a 0.0723
hbanking (t 1) 0.0583 0.1148 0.6654
D03 1.47E05a 2.71E05a 1.56E06

Constant conditional correlations

Service 1.00
Industry 0.62a 1.00
Banking 0.53a 0.51a 1.00

Log likelihood 13,400.05

AIC 17.14
#Obs. 1561

Notes:
(1) 2j (t 1) represents the past unconditional shock of the jth sector in the short run, or news.
(2) j = Service, Industrial, Banking hjj (t 1) denotes the past conditional volatility dependency or interdependency.
(3) D03 is the dummy for the 2003 Iraq war.
(4) The superscripts a, b and c represents signicance at the 1%, 5% and 10% levels.

sectors. In Saudi Arabia, all three sectors have similar shock sensitivities, but with about one-third of
the sensitivity to past own volatility. Among the GCC sectors, the Industrial sector shows the least news
sensitivity for most GCC countries. This suggests that this sector is more sensitive to past volatilities
related to changes in the fundamentals such as the supply and demand for oil and natural gas, oil and
natural gas products, petrochemicals, energy-intensive goods and other commodities than to news or
noise. In contrast to past own volatility sensitivity, the Banking sector seems to be the most sensitive to
past news. This is not surprising because of this sectors interconnectedness with the global Financial
sector.

4.2. Long run volatility and shock interdependency

The inter-sector volatility results are signicant, as expected for the most sectors and countries. Still,
the results show moderate volatility spillovers between the sectors within the individual countries,
with the exception of the UAE, for which we have substituted the Insurance sector for the Industrial
sector due to non-existence of data on the Industrial sector. Interestingly, tiny Qatar, which exports
both oil and natural gas products, has the most volatility spillover from the Industrial sector to the
other two sectors. Moreover, Kuwait has more sector volatility transmission than Saudi Arabia, which
relatively has the least inter-sector volatility spillovers.
Among the sectors, the volatility results are generally signicant and as expected. Cross-volatility, or
spillover, is more widespread from the Industrial sector to the Service sector than the reverse. When a
GCC economy grows or contracts, it rst affects the demand for goods arising from the Industrial sector.
In turn, this requires services for hauling and distribution of the goods, thereby leading to uctuations
in the Service sector. On the other hand, the demand arising from the Service sector for goods produced
or generated in the Industrial sector is much less voluminous, giving rise to signicantly less volatility
spillover towards this sector. Surprisingly, the Banking sector shows cross-volatility independence.
836 S.M. Hammoudeh et al. / The Quarterly Review of Economics and Finance 49 (2009) 829842

Table 3
Estimates of VARGARCH for Kuwait.

Variables Service Industry Banking

Mean equation
C 0.0015a 0.0005 0.0008c
AR(1) 0.1569a 0.0785a 0.0071
D03 7.94E04 2.61E04 3.13E05

Variance equation
C 4.63E06b 3.93E06c 3.15E06
2service (t 1) 0.1132a 0.0039 0.1315a
2industrial (t 1) 0.0214 0.1167a 0.0182
2banking (t 1) 0.0432b 0.0316 0.1977a
hservice (t 1) 0.8683a 0.0818 0.5835a
hindustrial (t 1) 0.3297b 0.6515a 0.2782
hbanking (t 1) 0.3913b 0.2178 0.4001b
D03 4.50E07 2.63E06 4.33E06

Constant conditional correlations

Service 1.00
Industry 0.64a 1.00
Banking 0.45a 0.48a 1.00

Log likelihood 16,023.28

AIC 20.59
#Obs. 1553

Notes:
(1) 2j (t 1) represents the past unconditional shock of the jth sector in the short run, or news.
(2) j = Service, Industrial, Banking hjj (t 1) denotes the past conditional volatility dependency or interdependency.
(3) D03 is the dummy for the 2003 Iraq war.
(4) The superscripts a, b and c represents signicance at the 1%, 5% and 10% levels.

This may be due to the supervision of the GCC central banks and this sectors inter-connection with
the global nancial markets.
In terms of inter-sector shock spillovers, the shock contagion is weak, and even weaker than own
sector shock sensitivity. As in the own shock case, Saudi Arabia has the weakest inter-sector shock
spillover links, and Qatar has the strongest, excluding UAE, which does not have the same three sectors.

4.3. Geopolitics

With regard to sensitivity of the major sectors indices to geopolitical events, as represented by
the 2003 Iraq war, geopolitics partly elevated the mean return of the Industrial sector in Saudi Arabia
and the Insurance sector in UAE. The war increased prices of oil, rened products, petrochemicals,
energy-intensive goods and other commodities, which Saudi Arabia produces the most. This country
also received more than its share of domestic political violence during the sample period. Shipping
premiums on freights going through the Gulf also increased substantially because of the war. It is not
surprising that the Insurance sector in UAE is sensitive to the 2003 Iraq war. Dubai is the third largest
re-exports hub in the world and it houses the giant D World Ports. Moreover, increased crowdedness
and congestion increased insurance rates in Dubai and Abu Dhabi, the two major emirates in the
seven-emirate UAE.
If the 2003 Iraq war dummy, D03, is suppressed in the models for the four countries, the removal
of this powerful event will make the impacts of the past sector-specic shocks on current conditional
volatilities more signicant, particularly in the case of Saudi Arabia and UAE. Among the sectors, the
signicance of the released impacts are particularly evident for the Banking or Financial sector. Thus,
strong geopolitical shocks can mute sector-specic shocks, especially for banks.7 However, although

7
Results are not reported but available from the authors.
 S.M. Hammoudeh et al. / The Quarterly Review of Economics and Finance 49 (2009) 829842 837

Table 4
Estimates of VARGARCH for Qatar.

Variables Service Industry Banking

Mean equation
C 0.0009b 0.0009b 0.0006
AR(1) 0.2113a 0.1860a 0.2820a
D03 6.84E04 6.11E04 8.83E05

Variance equation
C 8.92E06c 6.52E06b 3.36E05a
2service (t 1) 0.3170a 0.0546a 0.0211
2industry (t 1) 0.0649b 0.1010a 0.1002b
2banking (t 1) 0.0690c 0.1091a 0.3536a
hservice (t 1) 0.0730 0.1593 0.2306
hindustry (t 1) 0.9179a 0.6068a 0.8242a
hbanking (t 1) 0.3697b 0.9164a 0.0056
D03 8.70E06 7.59E06c 7.96E06

Constant conditional correlations

Service 1.00
Industry 0.52a 1.00
Banking 0.53a 0.50a 1.00

Log likelihood 143,12.67

AIC 18.43
#Obs. 1557

Notes:
(1) 2j (t 1) represents the past unconditional shock of the jth sector in the short run, or news.
(2) j = Service, Industrial, Banking hjj (t 1) denotes the past conditional volatility dependency or interdependency.
(3) D03 is the dummy for the 2003 Iraq war.
(4) The superscripts a, b and c represents signicance at the 1%, 5% and 10% levels.

this dummy variable is only individually signicant in 2 out of 12 mean equations, and 6 out of 12
variance equations, its parameters are jointly signicant, and its inclusion signicantly increases the
log likelihood for each estimation. The standard log likelihood ratio test, the Wald test and the joint
F-test all prefer the model with dummy included. The results of the three tests are provided in Table 6.

4.4. Constant conditional correlations (CCC)

As expected, all the CCCs between the three sectors for all GCC markets are positive, reecting
simultaneous growth in the overall economy. They are all below 0.64, echoing different advantages and
varying-roles played by those sectors in the economy. The estimates demonstrate that the highest CCC
for all the countries is between the Service and Industrial sectors, suggesting more mutual responses
economic factors between those two sectors than other sectors. The Industrial and Service sectors are
highly complementary to each other. While services create demand for other services, the industries
initiate original services. The CCC between Industrial and Banking, and between Services and Banking,
are very close, reecting banks mutual ties to all sectors in the economy.
It is worth noting that when the models are re-estimated over the whole sample period but without
the Iraqi war dummy, D03, the CCCs did not change much. One of implications of this result is that
association between the sectors for any of the GCC economies was affected by the same powerful force
and they all responded the same way to the same force.

4.5. Post-war subperiod

If the models of the four countries are estimated over the subperiod April 3, 2003 to the end
of the sample, where the start of the subperiod date coincides with the war dummy break, the
838 S.M. Hammoudeh et al. / The Quarterly Review of Economics and Finance 49 (2009) 829842

Table 5
Estimates of VARGARCH for UAE.

Variables Service Insurance Financial

Mean equation
C 0.0001 0.0003 0.0005a
AR(1) 0.1488a 0.0045 0.1527a
D03 4.32E05 9.12E04a 3.34E05

Variance equation
C 4.10E07 1.16E07b 1.43E06a
2service (t 1) 0.2076a 0.0024 0.0303
2inusrance (t 1) 0.0136 0.0387a 0.0224
2banking (t 1) 0.1247a 0.0456a 0.2581a
hservice (t 1) 0.4490a 0.1903a 0.0133
hinsurance (t 1) 1.3025a 0.8767a 2.0340a
hbanking (t 1) 0.5616a 0.3400a 0.3216a
D03 4.80E06a 5.27E07a 2.25E06a

Constant conditional correlations

Service 1.00
Insurance 0.14a 1.00
Financial 0.41a 0.18a 1.00

Log likelihood 17,720.78

AIC 20.68
#Obs. 1711

Notes:
(1) The table uses the data for the NBAD indices, which are superior to the NBAD Emirates indices. The capitalizations percentages
are based on those equity shares included in the NBAD index are actively traded. Those percentages are published in National
Bank of Abu Dhabis monthly Bulletin of Economics and Finance. We should also not that the UAE does not have an equity index
for the Industrial sector.
(2) 2j (t 1) represents the past unconditional shock of the jth sector in the short run, or news.
(3) j = Service, Industrial, Banking hjj (t 1) denotes the past conditional volatility dependency or interdependency.
(4) D03 is the dummy for the 2003 Iraq war.
(5) The superscripts a, b and c represents signicance at the 1%, 5% and 10% levels.

Table 6
Tests for joint signicance of including D03 in the models.

Saudi Arabia Kuwait Qatar UAE

Log L
Unrestricted (dummy included) 13,400.05 16,023.28 14,312.93 17,720.78
Restricted (dummy excluded) 13,362.01 16,015.00 14,305.13 17,706.47

LR ratio stat. 76.08a 16.56b 15.60b 28.62a

Wald stat. 51.43a 16.15b 18.90a 46.25a
F-stat. 8.57a 2.69b 3.15a 7.71a

Notes:
(1) H0 : The dummies D03 are jointly equal to zero.
(2) The superscripts a, b and c represent signicance at the 1%, 5% and 10% levels.
(3) Both LR ratio and Wald statistics follow the Chi2(6) distribution. The critical values are 16.81 at 1%, 12.59 at 5% and 10.65 at
10% levels. F-statistics follow F(6,Nobs Nparamaters ) distribution. The critical values of F-stat. are 2.80 at 1%, 2.10 at 5% and 1.77 at
10% levels.

results become mixed.8 While among the sectors, some estimates changed signicantly, particu-
larly for the Banking sector, others did not change much. For the individual countries, the CCCs
improved for Saudi Arabia and Qatar somewhat, dropped for UAE and remained almost the same
for Kuwait.

8
Results are not provided but they are available upon request.
 S.M. Hammoudeh et al. / The Quarterly Review of Economics and Finance 49 (2009) 829842 839

We do not know whether the changes in the estimation are caused by the small sample bias, the on-
going Iraq war, the high commodity prices or greater volatility in the commodity markets, particularly
oil. Nevertheless, the results overall are still closer to those of the whole sample with dummy than to
those the whole sample without the dummy. The exception is the Banking/Financial sector.

5. Implications for portfolio designs and hedging strategies

We now provide two examples using the estimates of the GCC equity sector markets for portfolio
design and hedging strategies.

5.1. Portfolio weights

The rst example follows Kroner and Ng (1998) by considering a portfolio that minimizes risk
without lowering expected returns. In this case, the portfolio weight of holdings of two equity sector
indices in the same market is given by:
h22,t h12,t
w12,t =
h11,t 2h12,t + h22,t

and

0, if w12,t < 0
w12,t = w12,t , if 0 w12,t 1
1, if w12,t > 1

where w12,t is the weight of, say, the rst sector index in one dollar portfolio of the two sector indices
at time t, h12,t is the conditional covariance between sector indices 1 and 2, and h22,t is the conditional
variance of the second sector index. Obviously, the weight of the second sector index in the one dollar
portfolio is 1 w12,t .
The average values of w12,t for the sectors in each GCC country are reported in Table 7. For instance,
the average value of w12,t of a portfolio comprising the Service and Industrial sector indices in Saudi
Arabia is 0.48. [Hassan and Malik (2007) used the BEKK model and estimated the average weight
between the Financial and Technology sectors at 0.66, while the average risk-minimizing hedge ratio

Table 7
Optimal portfolio weights and hedge ratios.

Portfolio Weight (w12,t ) of rst sector in 1$ Short/long beta (t )

portfolio (Kroner & Ng, 1998) (Kroner & Sultan, 1993)

Saudi Arabia
Service/industrial 0.48 0.66
Service/banking 0.18 0.87
Industrial/banking 0.14 0.96

Kuwait
Service/industrial 0.41 0.72
Service/banking 0.49 0.47
Industrial/banking 0.56 0.46

Qatar
Service/industrial 0.70 0.43
Service/banking 0.59 0.50
Industrial/banking 0.38 0.58

UAE
Service/insurance 0.41 0.18
Service/nancial 0.24 0.63
Insurance/nancial 0.39 0.23

Notes: (w12,t ) is the portfolio weight of sector index 1 relative to sector index 2 in a two-asset holding at time t, while average
t is the risk-minimizing hedge ratio for the two sector indices.
840 S.M. Hammoudeh et al. / The Quarterly Review of Economics and Finance 49 (2009) 829842

between these sectors is 0.64.] This suggests that the optimal holding of the Service index in one
dollar of Service/Industrial index portfolio for Saudi Arabia is 48 cents, compared with 52 cents for
the Industrial index. These optimal portfolio weights suggest that investors in Saudi Arabia should
own more Industrial stocks than service stocks in their portfolios. This nding conrms the result in
Hammoudeh and Al-Gudhea (2006). The result is more pronounced in Kuwait, where the holdings
tilt more heavily towards Industrial stocks. The case is opposite for Qatar, where the Service sector
overwhelmingly dominates the Industrial sector, possibly because Qatar has the highest own volatility
and volatility and shock spillovers in the Industrial sector.
Additionally, investors in Saudi Arabia, Qatar and UAE should also possess much more Bank-
ing/nancial stocks than other sectors stocks to minimize risk without lowering the expected returns.
However, the current nancial crisis is an exceptional period that happens once in a lifetime. The
optimal portfolios in Kuwait favor Industrial stocks over Banking stocks.

5.2. Hedge ratios

As a second example, we follow the example given in Kroner and Sultan (1993) regarding risk-
minimizing hedge ratios and apply it to the GCC markets. In order to minimize risk, a long position of
one dollar taken in one sector index in a given GCC stock market should be hedged by a short position
of $t in another sector index in the same market at time t. The t is given by:

h12,t
t = ,
h22,t

where t is the risk-minimizing hedge ratio for two sector indices, h12,t is the conditional covariance
between sectors 1 and 2, and h22,t is the conditional variance of the second sector.
The second column of Table 7 reports the average values of t for the GCC markets. The values of
the hedge ratios for the GCC sectors are smaller than those for the US equity sectors (Hassan & Malik,
2007), reecting the possibility of greater hedging effectiveness in GCC markets than in the USA. By
following this hedging strategy, one dollar long in the Service index, for example, in the Saudi market
should be shorted by 66 cents in the Industrial sector in that market. The most expensive hedge in the
Saudi market and the other GCC markets is by hedging the Service index with short positions in the
Banking/Financial sector. However, the most (hedging) effective to hedge long positions is between
the Insurance and Financial in UAE, where a one dollar long position is the former can be hedged by a
23 cents short position in the latter.

6. Conclusions

This study examines own volatility, shocks and inter-shock and volatility transmissions in three
equity sectors of four Gulf Cooperation Council (GCC). The sectors are Service, Banking and Industrial
or insurance and the GCC countries are Kuwait, Qatar, Saudi Arabia and UAE. The study uses the popular
version, the VARGARCH model, of the VARMA-model to achieve the results. In turn the results are
used to estimate the risk-minimizing hedge ratios to assess the hedge effectiveness among the sectors
and to calculate the optimal portfolio weights for favoring sectors.
The results suggest that past own volatility and not past shocks is the stronger driver in deter-
mining future volatility for the GCC frontier stock markets. This is similar to literature on developed
markets. This implies that a sectors fundamentals for these markets have more inuence on volatility
than shocks (news) or long-term volatility persistence dominates short-term persistence. In coun-
tries like the oil-rich GCC countries, changes in the fundamentals of oil and natural gas, as well as
for their products and energy-intensive goods, matter more when it comes to equity sector volatility
than sector-specic shocks. This is not surprising, given these countries heavy dependence on oil and
natural gas exports. It is important for the GCC countries to accumulate foreign assets in boom times
and invest them prudently in the region to stave off the negative impacts of uctuations in bust periods
and the migration of foreign capital.
 S.M. Hammoudeh et al. / The Quarterly Review of Economics and Finance 49 (2009) 829842 841

Since the Banking sector gures prominently in all GCC economies, the result shows that this
sector is a good venue for risk-adverse investors. Saudi banks are run very conservatively and the
Saudi Arabia Monetary Authority (SAMA) provides strong oversight over those banks. The results also
demonstrate that Saudi Arabias sectors are the least volatile among the GCC sectors and have the least
inter-sector volatility spillovers over the sample period. This is not surprising because this country has
the largest economy in the MENA region and its market capitalization is more than 40% of the total
GCC capitalization. However, Saudi stocks are particularly sensitive to political and military conicts.
Results in this regard show that a powerful military event such as the 2003 Iraq war can mute sector-
specic shocks in Saudi Arabia. In this case, policy makers such as central bankers and investors should
be cognizant of major military conicts in the region.
Moreover, it is hard to have a diversied GCC equity portfolio because it will be dominated by
either Saudi stocks or GCC bank sector stocks. On the other hand, Qatar sectors are the most volatile
among the GCC countries. Therefore, those stocks as well as those of Dubai are not for the faint of heart
investors. In 2008, Dubais stocks dropped by 72% compared to 32% for the S&P 500.
The GCC markets differ in terms of optimal portfolio holdings that minimize risk without lowering
expected returns, thereby allowing investors to hold more stocks in certain sectors than others and
effecting some diversication between sectors and countries. For example, investors in Saudi Arabia,
Qatar and UAE should possess much more Banking stocks than Service, Industrial or insurance stocks,
while in Kuwait they should favor Industrial over Banking stocks. These results should be relevant for
the GCC countries, which have recently embarked on establishing equity funds for both individual and
institutional investors.
Since the values for ratios of hedging long positions with short positions in the GCC sectors are
smaller than those for the US equity sectors, which reect the possibility of greater hedging effec-
tiveness in the GCC markets than in the USA, the GCC countries should develop hedging techniques
and strategies, such as futures, options and swaps that reduce volatility. The high volatility of the GCC
markets in 2008 makes such a conclusion imperative.

Acknowledgements

The authors wish to thank Farooq Malik, Mark Thompson and two anonymous referees of this
journal for helpful comments. We also have a special thank for Aksel Kibar and Talal Al-Jandali for
providing information and insights on the equity sectors. The third author is most grateful for the
nancial support of the Australian Research Council.

References

Bollerslev, T. (1990). Modelling the coherence in short-run nominal exchange rates: A multivariate generalized ARCH approach.
Review of Economics and Statistics, 72, 498505.
Chan, F., Lim, C., & McAleer, M. (2005). Modelling multivariate international tourism demand and volatility. Tourism Management,
26, 459471.
Conrad, J., Gultekin, M., & Kaul, G. (1991). Asymmetric predictability of conditional variances. The Review of Financial Studies,
4(4), 597622.
Engle, R. F. (2002). Dynamic conditional correlation: A simple class of multivariate generalized autoregressive conditional
heteroskedasticity models. Journal of Business and Economic Statistics, 20, 339350.
Engle, R. F., & Kroner, K. F. (1995). Multivariate simultaneous generalized ARCH. Econometric Theory, 11, 122150.
Engle, R. F., & Lee, G. G. J. (1993). A permanent and transitory component model of stock return volatility. Department of
Economics, UCSD, Discussion Paper No: 92-44R.
Hammoudeh, S., & Al-Gudhea, S. (2006). Return, risk and global factors in Saudi equity sectors. Journal of Emerging Markets,
10(Fall/Winter), 5564.
Hammoudeh, S., & Choi, K. (2007). Characteristics of permanent and transitory returns in oil-sensitive emerging stock markets:
The case of the GCC countries. International Review of Economics and Finance, 17, 231245.
Hammoudeh, S., & Li, H. (2008). Sudden changes in volatility in emerging markets: The case of Gulf Arab stock markets.
International Review of Financial Analysis, 17, 4763.
Hassan, H., & Malik, F. (2007). Multivariate GARCH model of sector volatility transmission. Quarterly Review of Economics and
Finance, 47, 470480.
Kroner, K. F., & Ng, V. K. (1998). Modeling asymmetric movements of asset prices. Review of Financial Studies, 11, 844871.
Kroner, K. F., & Sultan, J. (1993). Time dynamic varying distributions and dynamic hedging with foreign currency futures. Journal
of Financial and Quantitative Analysis, 28, 535551.
Ling, S., & McAleer, M. (2003). Asymptotic theory for a vector ARMA-GARCH model. Econometric Theory, 19, 278308.
842 S.M. Hammoudeh et al. / The Quarterly Review of Economics and Finance 49 (2009) 829842

Malik, S., & Hammoudeh, S. (2007). Shock and volatility transmission in the oil, US and Gulf equity markets. International Review
of Economics and Finance, 17, 357368.
McAleer, M. (2005). Automated inference and learning in modeling nancial volatility. Econometric Theory, 21, 232261.
McAleer, M., Chan, F., Hoti, S., & Lieberman, O. (2008). Generalized autoregressive conditional correlation. Econometric Theory,
24, 15541583.
Zarour, B., & Siriopoulos, C. P. (2008). Transitory and permanent volatility components: The case of the Middle East stock markets.
Review of Middle East Economics and Finance, 4, 114.