Research in International Business and Finance 38 (2016) 246261

Contents lists available at ScienceDirect

Research in International Business

and Finance
j o u r n a l h o me p a g e : w w w . e l s e v i e r . c o m / l o c a t e / r i b a f

Full length article

Dynamic transmissions between Sukuk and bond markets

Aktham I. Maghyereh a, , Basel Awartani b
a
b

Department of Economics and Finance, United Arab Emirates University, United Arab Emirates
Plymouth Business School, Plymouth University, Drake Circus, Plymouth, United Kingdom

a r t i c l e

i n f o

Article history:
Received 3 November 2015
Received in revised form 26 April 2016
Accepted 29 April 2016
Available online 30 April 2016
JEL classication:
C1 F3 G1
Keywords:
Islamic nance
Sukuk
Diversication
Volatility spillovers
Variance decomposition

a b s t r a c t
In this article we contribute to the recent debate on the difference between Islamic bonds
(Sukuk) and conventional bonds by investigating returns and volatility spillovers of Sukuk
and global bonds with equities. The dynamic spillover index methodology proposed by
Diebold and Yilmaz (2012) indicates different transmission mechanisms of Sukuk compared
to bonds. The main distinctive features of the Sukuk market are the higher transmission
of information from equities, and the weaker transmission of information from the Sukuk
market to other markets. Thus, this paper highlights the importance of Sukuk in the strategic
asset allocation and hedging of international investors.
2016 Published by Elsevier B.V.

1. Introduction
The recent years have witnessed growth and interest in the Sukuk market. In terms of overall volumes, the market grew
from less than $14.8 billion in 2001 to more than $472.68 billion by the end of 2013. This year alone, there has been more
than $118.8 billion of Sukuk sales.1 Moreover, the market has started to attract non-Muslim issuers in Europe, Asia and
Africa.2 For instance, the UK government have structured and sold 200 million of Sukuk last June. It was followed by the
governments of Hong Kong, South Africa and Luxembourg.3
The debate on Sukuk is mainly concerned about their similarities and differences to conventional bonds. The major distinguishing features of Sukuk compared to bonds is the prohibition of a xed interest payment, the exclusion of transactions
involving extreme uncertainty or a deliberate lack of transparency (gharar), the exclusion of gambling (maysir), short selling,
arbitrage and excessive speculation (Aloui et al., 2015a,b; Bouslama and Lahrichi, 2016). Moreover, the return on Sukuk
should depend upon the return of the underlying investments that have to be compliant with Islamic rules. Hence, Sukuk
should not be driven by interest rate changes in the same way as conventional bonds. Unlike bonds which constitute a

Corresponding author.
E-mail addresses: a.almaghireh@uaeu.ac.ae (A.I. Maghyereh), basel.awartani@plymouth.ac.uk (B. Awartani).
1
See Global Sukuk report (2015). Published by Malaysias Islamic Finance Marketplace. Accessed at http://www.mifc.com/index.php?ch=28&pg=
72&ac=124&bb=uploadpdf.
2
Sukuk are currently traded in more than 12 securities markets across the world. In many other markets they are traded as private issues.
3
Note that the demand for Sukuk is strong and that these issues were oversubscribed and sold by these governments without breaking a sweat. See
Financial Times Special Report, for more details information about Sukuk market. Accessed at www.ft.com/reports@ftreports.
http://dx.doi.org/10.1016/j.ribaf.2016.04.016
0275-5319/ 2016 Published by Elsevier B.V.

A.I. Maghyereh, B. Awartani / Research in International Business and Finance 38 (2016) 246261

247

nominal debt that the issuer has to re-pay at maturity, Sukuk represents a share in the project and its value at maturity has
to reect the current market value of the underlying investment.4
However, to make them familiar and their returns and risks easier to assess by investors, issuers structure Sukuk in a
similar way to bonds in terms of features such as rating, issuance, redemption procedures, coupon payments and default
clauses. Therefore, many studies claim that Sukuk are not innovative nancial products, and that they are simply another
version of bonds in terms of their lack of ownership, right to xed return and guaranteeing repayment of the principal by
the issuer (Miller et al., 2007; Wilson, 2008).
To distinguish Sukuk from conventional bonds Cakir and Raei (2007) investigated Sukuk price behavior and they found a
genuine difference in their association with other bonds. In particular, they assessed the impact of Sukuk issued on the cost
and risk structure of investment portfolios by using the Value-at-Risk (VaR) framework. Their results indicated that Sukuk
returns are weakly correlated with other bond returns, and that their inclusion in a bond portfolio reduces its risk beyond
the reduction that is implied by adding another issuer different maturity bond. They also found that Sukuk are diversifying
in a bond portfolio despite it being similarly structured.5 Similarly, and to expose their difference to bonds, Godlewski et al.
(2013) compared stock market reaction to bond issuance with its reaction to the issuance of Sukuk. They applied an event
study methodology over a sample of Malaysian listed issuers and they found that stock prices react differently. In particular,
a bond sale announcement was considered as no news by equity investors with no inuence on stock prices. However, Sukuk
issuance was regarded as bad news and they triggered price declines. Sukuk sales were also found to be negatively correlated
with market returns. They attributed this result to the excess demand for Sukuk and to an adverse selection mechanism that
favors Sukuk issuance by lower quality debtor companies.6
A group of recent studies have investigated the association between Islamic equities and Sukuk. For instance, Aloui
et al. (2015a,b) studied the co-movement between Islamic stocks and Sukuk using the time frequency domain analysis and
they nd a strong dependence particularly in the longer term. In a related research, Kim (2013) have also analyzed the
transmission mechanism of volatility between Islamic stocks and Sukuk in Malaysia using a VAR-bivariate GARCH model.
Their results indicate a unidirectional volatility spillover from the stock market to the Sukuk market.
In this paper we contribute to the current debate over whether Sukuk are nancing products that own distinctive features
to bonds by comparing returns and volatility information transmissions of the Sukuk market with those of the bond market.
The empirical study of return (volatility) spillovers is interesting from the particular perspective of portfolio diversication
and hedging strategies. In particular, knowledge of directional spillovers may help in portfolio management, in strategic
asset allocation and market selection. Furthermore, accounting for directional volatility may improve risk prediction, which
is crucial for stock valuation models, option pricing models, value at risk and optimal hedging. Thus, this paper highlights
the importance of Sukuk in the strategic asset allocation and hedging of international investors.
The empirical literature on Sukuk is rare. In some of the previously mentioned studies, the samples were restrictive. They
contained only some of the Sukuk issues and there was not a time series that is long enough to be analyzed. For instance, Cakir
and Raei (2007) analyzed only selected issues of sovereign Sukuk for which secondary market data is available.7 Similarly,
the distinction between Sukuk and conventional bonds in Godlewski et al. (2013) was done on the basis of a sample of
Malaysian listed companies over the period from 2002 to 2009. The rest of studies have focused on the spillover between
Islamic bonds and equities in a specic region. For example, the study of Kim (2013) has considered Malaysian stock and
bond indices. Similarly, the focus of Aloui et al. (2015a,b) was on GCC Sukuk and equities.8 On the contrary, in this study we
focus on the transmission mechanism in a system that includes global bonds, global equities, global Sukuk and global Islamic
stocks. The Sukuk index used in this study represents all issues in the Sukuk market and across all countries. It contains a
long enough series of daily observations that extends from September 2005 to February 2014. This has given us the chance
to study the transmission mechanisms in the Sukuk market and to compare it to the bond market. Unlike previous studies,
this research reveals the dynamics of the spillovers between Sukuk and equities across time. Therefore, this paper adds to
the literature regarding the volatility transmissions between Sukuk and conventional bond markets. It also quanties the
consequences of cross-asset market linkages on the diversication benets of international investors who are venturing into
the global Sukuk market.
The information transmission of returns and volatilities are measured by using the Diebold and Yilmaz (2012) spillover
indices that are based on normalizations of forecast error variance decompositions derived from a generalized vector autoregressive (GVAR) model. These indices are suitable as they can be used to study the information transmission to and from
the Sukuk (bond) market and other markets.9

4
Note that while shares have no maturity, Sukuk do. Moreover, they are issued on specic assets. The Sukuk value should be fully paid to the investors
at maturity.
5
To the contrary of Cakir and Raei (2007) pointed out that Sukuk issued in different markets suffer from high levels of long term correlations with bonds
and that their diversication benets are limited.
6
On the other side, Alam et al. (2013) found similar stock market reaction. The stock market responded in the same way to either the issuance of bonds
or Sukuk. However, their results were similar to Godlewski et al. (2013) in the pre-crisis period.
7
Sukuk markets are in its infancy and most trading occurs in the primary market.
8
GCC refers to the Gulf Co-operation Council Countries, namely; Saudi Arabia, Kuwait, Oman, The Unites Arab emirates and Qatar.
9
For the application of this approach, refer to, for example Awartani et al. (2013) and Maghyereh et al. (2015).

248

A.I. Maghyereh, B. Awartani / Research in International Business and Finance 38 (2016) 246261

Furthermore, this framework enables us to directly assess the nature and the direction of spillovers and how they vary over
time. The time-varying volatility (returns) spillovers may can be computed from rolling windows and plotted against time.
These time plots can be used to capture any volatility (return) spillover regime changes.10 Thus, this framework provides
a rich source of information regarding magnitudes and directions of return (volatility) transfer, and this suits exactly the
objective of the paper.
Our empirical analysis shows clearly that Sukuk are distinctive from conventional bonds in terms of their transmission
mechanism. The returns spillovers between Sukuk and bonds are weak. However there exists a signicant volatility crossover
from bonds to Sukuk. Though returns and volatility transmissions from global stock markets to the Sukuk market are weak,
they are still much higher than transmissions to bonds. Therefore, Sukuk are more exposed to innovations transmitting from
global equities than conventional bonds. Unlike the bond market in which volatility information transmits to global stock
markets, the volatility transmission from Sukuk to either equities or even bonds is weak and thus in most transmissions
Sukuk is a net receiver of information spills. These results highlight the importance of the Sukuk market in the strategic
asset allocation of international investors. The Sukuk market is weakly associated with conventional bonds; and therefore
they offer great diversication potential.
The fundamental difference of Sukuk structures than those of conventional bonds,11 as well as the impact of shariacompliance on the set of common information and cross-assets hedging,12 may explain the lower returns and volatility
information transmissions between Islamic and conventional assets (see Akhtar et al., 2013; Aloui et al., 2015a,b). According
to Fleming et al. (1998), common information on expected returns and cross hedging are the two main channels of transmission that drive asset price co-movements. In Islamic nance, the prohibition of a xed interest payment (riba) may reduce
the set of common factors that drives both the expected returns on sharia-compliant and conventional assets. This results
in a stronger dependency of volatility on the rms performance, rather than on common macroeconomic factors such as
the interest rates, and it reduces co-movement. Other distinguishing features of investing according to Islamic rules are
the investment restrictions imposed on assets that are characterized by extreme uncertainty, lack of transparency (gharry),
gambling (may sir), short selling, and arbitrage.13 These constraints may reduce the need for cross-asset hedging and hence,
lower returns and volatility information cross over between Sukuk and conventional bonds can be expected.
The rest of the paper is organized as follows: in Section 2 contains the main features and structures of Islamic nancing.
In Section 3, we include the Diebold and Yilmaz measure of information transmission between Sukuk and other indices.
Section 4 describes the data set. Section 5 reports the empirical results and provide some robustness checks in Section 6.
The conclusion can be found in Section 7.
2. Islamic nancing structures
To be consistent with Islamic beliefs nancial products have to be structured such that they represent an ownership in a
tangible asset, usufruct or services of revenue-generating rm. For Sukuk to be compliant with Islamic rules, its payments
to investors should come from prots, and the Sukuk value repaid at redemption should reect the current value of the
underlying investments. This implies that Sukuk nancing resembles certain structures of asset backed securities and in
that sense they are closely related to bonds. In certain other Sukuk structures, the holders share the risk of the investment
proportionally as in equity nancing. In fact, it is the written contract between the parties that determine the responsibilities
of each party and how the loss and prots will be allocated in Sukuk nancing. Therefore their similarity to either debt or
equity depends ultimately on the written contract.14 However, there are six main types of Sukuk contracts. These are the
Jira, the Murayama, the Mudarabah, the Musharaka, the Istisna, and the Salam contract.15
Under the Ijarah contract, the Sukuk sponsor sells assets to a special purpose vehicle (SPV) and agrees to buy it back at
a predetermined price at the maturity of the issue. The SPV funds the purchase by issuing Sukuk certicates and sells them

10
By rolling generalized forecast error variance (FEV) decompositions in a Vector Autoregressive (VAR) model, this approach allows us to calculate the
gradually changing directional (return) volatility spillovers between any pairs of assets over the entire observation period.
11
Sukuk are asset-based securities and are independent of interest rate movements, such that the prot depends upon the return of the underlying asset.
This may affect the volatility information transmissions through a stronger dependency of volatility on the performance of the underlying asset rather than
through the exogenous effects of common macroeconomic factors like interest rates (Akhtar et al., 2013).
12
The effect of common information refers to the idea that common sources of information (i.e., macroeconomic news) inuencing expectations in more
than one market simultaneously. Cross-market hedging refers to the purchase or sale of an asset in order to insure position in another asset. Thus,
investors respond to shocks in one asset by optimally readjusting their portfolios in another asset.
13
While there is specic evidence from the Holy Quran and Sunna that clearly prohibit gharar and gambling (see for example Metwally, 2006), there is no
explicit statement on short selling and arbitrage. However, the consensus among Muslim scholars (Ijma) is that arbitrage and short selling are forbidden. In
the short selling, the investor gains as the underlying company loses value. This is considered as a clear violation of the ban of unjust deeds that is stated in
the Holy Quran, Suret Al Baqara, 2, 278279: Deal not unjustly, and ye shall not be dealt unjustly. The authors El-Din and Hassan (2007) have shown that
short selling contains Gamble, Foolishness, Riba and Gharar elements that are strictly forbidden in Islam. Regarding arbitrage Metwally (2006, p.16-17)
shows some indirect evidences from the Sunna in the following quotes of Prophet Muhammad: He who accumulates stocks of grain during shortages is a
great sinner, Evil indeed is that man who withholds transactions in anticipation of rise in prices and nally One who imports from outside and sells at
the market rate for his maintenance is blessed, while he who withholds transactions in view of estimated dearness in future, is thrown away from Gods
pleasure
14
The contracts are enforced by the ethical standards and the social values of the Sharia which guarantees that compliance occurs by all parties.
15
See Islamic Finance Recourses, for more details information about the types of Sukuk contracts. Available on the link: http://ifresource.com/.

A.I. Maghyereh, B. Awartani / Research in International Business and Finance 38 (2016) 246261

249

to investors.16 The assets sold are then leased back to the issuer, and the rent payments are passed to Sukuk holders in the
form of coupon payments. When the Sukuk matures, the SPV collects the agreed upon repurchase price, and pass it through
to Sukuk holders as a principal payment. This is similar to capital lease or to hire purchase agreement in western nance.
The purpose in the Murabaha Sukuk is to provide short term capital and consumption nancing. The investor buys the
working capital means of production (or durables) and resells it after adding a specied prot margin.17 The borrower has
then to repay in installments over the maturity of the nance to Sukuk holders. The maximum margin is usually determined
by the central bank. This type of nancing needs the active involvement of the investor in the process of surveillance,
monitoring and collection; therefore it is normally provided by banks (Badr-El-Din, 2003). The schedule of payments in the
Ijara and Murabaha Sukuk is normally xed and this has raised concerns on their validity as Sharia compliant and on whether
these two instruments are genuinely different from bonds.
Some other Sukuk are based on prot sharing. For instance, in the Mudarabah contract, capital is provided by one or more
investors and managed by the entrepreneur. This type of nancing is considered by many as equity nancing (Aggarwal
and Yousef, 2000). However, in Mudarabah the investor and the manager maintain close relationships instead of the distant
legally binding association that exists between managers and common shareholder in western nance.18
Two less common contracts are Istisna and Salam. Istisna can be briey dened as a contractual agreement for the sale
of goods or commodities to be produced in the future. The price is xed and specication of the goods is precisely described.
Furthermore, it is not necessary to pay in advance, but it may be deferred through a staged payment method and as agreed
between the parties. The Istisna contract is common use in infrastructure, large construction and manufacturing projects.
Salam contracting is common in agriculture where the agreement is made between two parties to exchange the underlying
at a xed future date. However, the value of contract is settled now in cash rather than at maturity.
The Sukuk market is growing fast. Sukuk are sold by many specialized exchanges around the globe. The similarity of some
of Sukuk structures with conventional bonds has raised the issue of whether they are genuinely different in practice. Some
scholars including Miller et al. (2007) and Wilson (2008) are strongly supporting this argument. Moreover, it is claimed that
Islamic nance is converging fast towards conventional nance, particularly in the case of Sukuk (Mirakhor, 2007). Thus,
distinguishing Sukuk from bonds is crucial as debt and bond nancing based on interest do not conform to Islamic Sharia
principles and this has long term implications on the development and progress of the Sukuk market.19
3. Methodology
Our main objective in this paper is to capture the dynamic patterns in spillovers occurring in the returns and volatilities
of the Sukuk and conventional bond markets. To achieve that, we used a group of spillover indices that was proposed by
Diebold and Yilmaz (2012, hereafter DY). The DY indices are based on generalized forecast error variance decomposition. In
the following, we provide a brief overview of the DYs methodology.20
Let market returns (or volatilities), xi to be modeled as a vector autoregressive process with N-variables and p lags. The
VAR(p) model (system) can be written as
xi =

p


i=1

xti + t

(1)

wherext = x1,t , x2,t , . . ., xN,t is a vector of N endogenous variables, is an N N matrix of parameters to be estimated,
t = 1, . . ., T is the time index and n = 1,...,N is the variable
index. Also assume that the vector of error terms is independently

and identically distributed with zero mean, and
covariance matrix. In this study, we included the returns of four indices:
the Sukuk index, the world bond index, the Islamic stock market index, and the world stock market index. Hence, our VAR
specication is made up of N = 4 variables.
When the VAR system in Eq. (1) is covariance stationary; this specication can be rewritten in its innite order moving
average form as
xt =

Ati

(2)

i=0

where the N N coefcient matrices Ai obey a recursion of the form Ai = 1 Ai1 + 2 Ai2 + . . . + p Aip with A0 is the N N
identity matrix and Ai = 0 for i > 0. A transformation of coefcients in Eq. (2) can be used to identify variance decompositions
(or impulse responses). The aggregation of these decompositions will be used subsequently to compute the gross spills from a
particular market to other markets, and also in the opposite direction. However, for the measure to be stable, decompositions

16

These assets could be real estate assets or high value industrial equipment such as aircrafts.
The corner stone is not to use a xed interest (riba) for the nance to be compliant with Islamic beliefs. Various types of durables are nanced under
these contracts: car nancing and electrical appliances nancing.
18
In mudarabah and musharaka, Sukuk are more related to private equity nancing rather than to common equity. There is some inuence and control
over management.
19
The more Sukuk are similar to bonds or the more they are driven by interest rates, the less likely they will conform to Islamic rules.
20
A more detailed explanation on the construction of this index can be found in Section 2 of Diebold and Yilmaz (2012).
17

250

A.I. Maghyereh, B. Awartani / Research in International Business and Finance 38 (2016) 246261

should be independent of the ordering of the markets. Thus, a standard Cholesky decomposition is not suitable as it is sensitive
to the ordering of markets.
To control for this problem, Diebold and Yilmaz (2012) used the generalized VAR framework of Koop et al. (1996), and
Pesaran and Shin (1998) (the KPPS hereafter). The KPPS decompositions are robust to the ordering of markets by allowing
correlated shocks and using the historically observed distribution of the errors to account for the shocks. However, unlike
Cholesky, the KPPS decompositions do not sum to one and hence, normalization is needed in this case.
The KPPS, forecast error variance decomposition (H step ahead) is computed as
g
ij

(H) =

jj1

H1 

ei hh
h=0
H1
ei hh
h=0

 2
ej

 

(3)

ej

where
is the variance matrix of the vector of errors , and jj is the standard deviation of the error term of the jth market.
Finally, ei is a selection vector with one on the ith element, and zero otherwise.
To ensure that the allocated decompositions sum to one, we normalize each entry of the variance decomposition matrix
in Eq. (3) by the row sum as21
g

g
 ij (H) =

ij (H)

N

g

j=1 ij

(4)

(H)

Hence in the equation above the decomposition including own shocks in each market sums to one, i.e.,

N g
and the total decomposition over all markets sums to N, i.e.,

= N.
j=1 ij (H)

N

g

j=1 ij

(H) = 1,

Using the decomposition above, the total spillover index is then dened as:

N

i, j = 1
i=
/ j

S g (H) =

N

g
 ij (H)

N

i, j = 1
i=
/ j

100 =

g

i,j1 ij (H)

g
 ij (H)

100

(5)

The index in Eq. (5) measures the total information ow among the markets included in the system. It is an off-diagonal
addition of the proportions of the forecast error variance of all xi s due to shocks to xj , for alli =
/ j.
In addition to the total spillover index, the information from KPPS variance decomposition allows to obtain the directional
spillovers. Specically, the spillover transmitted from the Sukuk to other markets can be expressed as follows

N

j=1

g
 ij (H)

i=
/ j

100
N g
 (H)

Si (H) =

(6)

j=1 ij

Similarly, the directional spillovers received by the Sukuk (and/or the world bond market) from all other markets can be
calculated as

N

j=1

g
 ij (H)

i=
/ j

100
N g
 (H)

S i (H) =

(7)

j=1 ij

It is also possible to calculate the net directional spillover from Sukuk (and/or the world bond market) to other markets
by calculating the difference between Eqs. (6) and (7) as,
g

Si (H) = Si (H) S i (H)

(8)

This net directional spillover provides summary information about whether Sukuk (and/or the world bond market) is a
receiver or transmitter of shocks in net terms.
Finally, the net pairwise volatility spillover which shows how much each market i contributes to another market j is given
by

g
Sij

(H) =

ij (H)

N

g

k=1 ik

(H)

ji (H)

N

g

k=1 jk

(H)

100

(9)

21
The normalization of the sum will enable an intuitive computation of the contribution of a particular market, and an intuitive sum of contributions
across markets.

A.I. Maghyereh, B. Awartani / Research in International Business and Finance 38 (2016) 246261

251

Fig. 1. Weekly returns and volatilities.

The pairwise net spillover measure in Eq. (9) indicates whether the Sukuk market is a net giver or a net receiver of
information to the bond market or other markets that include world stock indices. We now turn to discuss the data set and
the empirical results.
4. Data set and descriptive statistics
The empirical analysis covers the period from the 30th of September 2005 to the 24th of February 2014. This study period
covers over a ten-year of nancial history, which includes the global nancial crisis from September 2008 to December 2009,
the European sovereign debt crisis from April 2010 to June 2012, and the Arab Spring and political revolutions in the Middle
East that began on December 2010-present. Four dollar indices were included in the analysis. These are the Dow Jones
Citigroup Sukuk Index, the Dow Jones Citigroup Global bond index, the Dow Jones Islamic stock market index and nally
the Dow Jones Global stock market index. The data for these indices were retrieved from Bloomberg database.
The Sukuk Index includes U.S. dollar-denominated Sukuk issued in the global market that has been screened and considered as Shariah compliant investments. To be included in this index, a bond must pass screens and meets the standards
issued by the Accounting and Auditing Organization for Islamic Financial Institutions (AAOIFI). The bond has also to have at
least a minimum maturity of one year and a minimum outstanding size of at least $200 million. All bonds included in the
index are of investment grade rating and all are rated by leading rating agencies. A parallel and unrestricted counterpart of
the Sukuk Index is the Global bond index, which is a comprehensive benchmark of global xed income universe.
The Dow Jones Islamic stock market index provides a global investable universe of stocks that comply with Islamic
principles. The index excludes unethical or highly indebted companies or companies engaged in gambling, alcohol, tobacco,
pork, and other prohibited activities in Islam (Charles et al., 2015). This index is a subset of the Dow Jones Global stock
market index that includes data for over 12,000 companies from 77 countries. These indices are designed to provide portfolio
managers with a comprehensive set of world indices that are useful in benchmarking and asset allocation.
From the daily indices, we measure weekly continuously compounded rates of returns as the change in log price indices,
Friday to Friday and over the whole sample.22 This has resulted in 438 weekly return observations. To measure the latent
volatility we use a simple range based estimator that was rst proposed by Garman and Klass (1980). Specically, we use
weekly high, low, opening and closing prices from Monday open to Friday close. For any market i, the volatility in week t is
computed as

2
i,t
= 0.511 Hi,t Lt

2

0.019



Ci,t Oi,t



Hi,t + Li,t 2Oi,t 2 Hi,t Oi,t



Li,t Oi,t

0.383 Ci,t Oi,t

2

where H, L, O and C are the Monday to Friday open, high, low, the Monday open and the Friday close respectively. In Fig. 1, we
draw a time series of computed weekly returns and weekly returns volatilities of the four indices. As can be seen in the gure,
all indices suffered large losses and exhibited extreme swings during the global nancial crisis in 2008 and after the collapse

22

We use weekly data to overcome the potential non-synchronous trading bias characterizing daily data and to avoid the day-of-the-week effect problem.

252

A.I. Maghyereh, B. Awartani / Research in International Business and Finance 38 (2016) 246261

Table 1
Descriptive statistics, market returns and volatilities.
Global Bond Index

Islamic Stock Index

Global Stock Index

Panel A: Summary statistics of weekly return series

Mean
0.000610
Median
0.000934
Maximum
0.075990
Minimum
0.154360
Std.Dev.
0.010646
Skewness
5.98*** (0.0000)
Kurtosis
116.23*** (0.0000)
Jarque-Bera
249188.60*** (0.0000)
Q(10)
26.37***
(0.0000)

Sukuk

0.000217
0.000031
0.034850
0.064640
0.008104
1.41*** (0.0000)
11.59*** (0.0000)
2599.39*** (0.0000)
37.34*** (0.0000)

0.001213
0.003325
0.216140
0.216900
0.031758
0.536117*** (0.0005)
10.09*** (0.0000)
1881.17*** (0.0000)
10.73 (0.3786)

0.000711
0.002740
0.146620
0.159840
0.027632
0.707359** (0.0000)
6.35*** (0.0000)
772.44*** (0.0000)
9.44 (0.4902)

Panel B: Summary statistics of weekly return volatilities

0.000016
Mean
0.000001
Median
0.002870
Maximum
Minimum
0.000000
0.000148
Std.Dev.
Skewness
17.04*** (0.0000)
322.21*** (0.0000)
Kurtosis
1915948.64*** (0.0000)
Jarque-Bera
9.72 (0.4650)
Q(10)

0.000022
0.000012
0.000455
0.000000
0.000037
6.03*** (0.0000)
52.77*** (0.0000)
53492.65*** (0.0000)
289.27*** (0.0000)

0.000260
0.000119
0.009210
0.000000
0.000639
8.91*** (0.0000)
103.71*** (0.0000)
202132.88*** (0.0000)
410.47*** (0.0000)

0.000220
0.000101
0.005580
0.000000
0.000455
6.64*** (0.0000)
59.10*** (0.0000)
66971.09*** (0.0000)
572.81*** (0.0000)

Notes: Q(10) is the LjungBox statistics for serial correlation in raw series. The values in parentheses are the actual probability values.
**
Signicance at 5% level.
***
Signicance at 1% level.

Table 2
Correlation matrix.

Sukuk
Global Bond Index
Islamic Stock Index
Global Stock Index

Sukuk

Global Bond Index

Islamic Stock Index

Global Stock Index

1.000
0.055
0.256
0.222

1.000
0.031
0.010

1.000
0.855

1.000

of Lehman Brothers on the 15th of September. However, compared to the global bond index, the losses experienced by Sukuk
investors were larger in magnitude. Moreover, the volatility of the Sukuk index was much higher than the volatility of the
world bond index during the crisis. Both the Islamic and the global stock indices have shown similar return and volatility
patterns during the whole period.
To compare the statistical properties of the Sukuk with bonds, we compute a variety of summary statistics of returns.
Panel A of Table 1 reports the mean, the standard error, the skewness, the excess kurtosis coefcients, and the Jarque-Bera
statistics to test the null hypothesis that daily returns are normally distributed.
The Panel shows that relative to the world bond index, Sukuks returns and their standard deviation are higher on
average. They also exhibit wider range, larger negative skeweness and higher leptokurtosis compared to bonds.23 The Panel
also reports the Ljung-Box portmanteau statistics for the last 10 days. The statistics shows that unlike bond returns, Sukuk
returns are serially correlated.24 The Panel nally reports the statistics of the weekly returns of the global stock index and
the Islamic stock index. As can be seen, the Islamic stock index displays higher average returns, volatility, kurtosis and it is
more negatively skewed compared to the global stock index. Both indices are not serially correlated.
Table 2 display pairwise simple correlations between Sukuk returns and returns of global bonds and equities over the
sample periods. Panel A of the table presents the daily return correlation matrix. It shows clearly that the correlation between
Sukuk and the world bond index is tiny and negligible over the sample period (0.055). This implies that Sukuk may have
the potential to diversify and improve the tradeoff of risk and returns in a global bond portfolio. Similarly, the table shows
that Sukuk appears to be weakly correlated with stocks. Its correlation with either the world stock index or even the Islamic
stock index is relatively small and less than 0.25.25 The weak correlation with global stocks and bonds makes Sukuk a good
candidate for diversifying a global bond portfolio, stock portfolio and a bond-stock portfolio.

23
As can be seen in the panel, the returns to stock indices are also negatively skewed and leptokurtic, and the Jarque-Bera statistics rejects the null
hypothesis of normality very strongly.
24
One possibility is that Sukuk valuation suffers from stale pricing and smoothing, which may introduce serial correlation into Sukuks daily return series.
For longer horizon returns Sukuk is not serially correlated.
25
Note that the global bond was very low correlated with global stocks over the sample period.

A.I. Maghyereh, B. Awartani / Research in International Business and Finance 38 (2016) 246261

253

Table 3
Return and volatility spillover matrices.
To index i

From index j
Sukuk

Global Bond Index

Islamic Stock Index

Global Stock Index

Contribution from others

Panel A: Return spillovers

Sukuk
Global Bond Index
Islamic Stock Index
Global Stock Index
Contribution to others
Contribution including own

74.6
0.2
6.5
4.3
11
100

3.9
91.3
2.4
2.0
8
85

10.2
4.5
52.9
39.4
54
107

11.3
4.1
38.2
54.4
54
108

25
9
47
46
127
Total spillover index: 31.7%

Panel B: Volatility spillovers

Sukuk
Global Bond Index
Islamic Stock Index
Global Stock Index
Contribution to others
Contribution including own

30.0
0.8
0.4
1.3
2
32

27.7
74.9
27.2
29.6
84
159

23.7
5.6
37.0
22.6
52
89

18.7
18.7
35.4
46.5
73
119

70
25
63
53
212
Total spillover index: 52.9%

Notes: The underlying variance decomposition is based on a weekly VAR system with two lags. The (i, j) value is the estimated contribution to the variance of
the 10 step ahead index return (volatility) forecast error of index i coming from innovations to returns volatility of index j. The decomposition is generalized,
and thus it is robust to the ordering shown in the column heading.

To summarize, the preliminary evidence reveals that Sukuks correlation with stocks is higher than its correlation with
bonds for all hoding periods. Similarly the cross correlation between world bonds and stocks is lower than between Sukuk
and stocks. Hence, Sukuk appears to have a more important role to play in a bond portfolio rather than in a stock portfolio.
5. Empirical results
As mentioned previously, in order to analyze the extent of transmissions between markets, we rst compute return and
volatility spillover indices as proposed by Diebold and Yilmaz (2012).26 In particular, the forecast error variance decompositions excluding own shocks are aggregated and divided on aggregate variance decompositions including own shocks to
nd a meaningful percentage that indicates the extent of spillovers among all markets in the system.27
To study and compare the information transmission between Sukuk and other markets, we aggregate variance decompositions. The variance decomposition of Sukuk attributed to a particular market is a measure of how much this market
gives to the Sukuk index and vice versa. The sum across markets will then give the extent to which other markets take from
Sukuk, or give to it.
Table 3 reports the spillover index matrix of returns and volatilities that includes the four indices. All results are based on
vector autoregressions of order 2, and generalized variance decompositions of 10 week ahead forecast errors. The (i, j) entry
in each panel is the estimated contribution to the forecast error variance of market i coming from innovations to index j.
Contribution to others at the bottom of the panel, is the directional spillovers from an index to all other indices. Contribution
from others in the last column, is the directional spillovers from all indices to a particular index. The total spillover index as
in Eq. (5) is reported in the lower right corner of each panel.
The directional spillovers from Sukuk to other indices is reported in the rst column. The total contribution of Sukuk
across all stock markets as in Eq. (6) is reported just before the last row in the rst column (parallel to item Contribution to
others). Similarly, the directional spillovers from other indices to Sukuk is reported in the rst row. The total spills as in Eq.
(7) is reported as the last observation in the rst row of Panels A and B (down the item Contribution from others).
Column 1 and Row 1 of Panel A in Table 3 shows returns bi-directional spills of the Sukuk market. The Panel shows
that return spillovers between Sukuk and international bonds is weak. The Sukuk receives 3.9% spills from the world bond
market and it gives it only 0.2%. This indicates that the association between Sukuk and bonds is weak. The matrix also shows
that Sukuk is more exposed to return spillovers compared to bonds. It receives 25% contribtion from others, whereas bonds
receives only 9%. The contribution of both Sukuk and bonds to others in terms of returns spills is weak. Sukuk contributes
11% to others while bonds only contributes 8%. Overall there is a higher level of return spillovers between Sukuk and equities
compared to bonds. However, in every pairwise comparison of directional spillovers, Sukuk is a net receiver. For instance,
the transmission from Sukuk to world stocks is 4.3%; while it is more in the opposite direction (around 11.3%). This holds true

26
Note that inference on association could have been done using dynamic correlations. However, in this case the analysis could be biased and inaccurate
due to the dependence of the correlation measure on the state of volatility. Specically, during market stress, volatility is high and correlation estimates tend
be upwardly biased. If volatility is not accounted for, the rise in correlations may reect only an increase in market volatility, and not higher interdependence.
Moreover, the dynamic conditional correlations are not informative of the direction and of the volatility transmissions across markets.
27
Note that forecasts are generated by VAR (2) Model. Unlike the dynamic conditional correlations which assume a GARCH volatility process to compute
dynamic correlations, the spillover analysis assumes an auto regressive model.

254

A.I. Maghyereh, B. Awartani / Research in International Business and Finance 38 (2016) 246261

Table 4
Tests for structural breaks.
Sukuk

Global Bond Index

Islamic Stock Index

Global Stock Index

Test statistic

p-value

Test statistic

p-value

Test statistic

p-value

Test statistic

p-value

Panel A: Return
AI. Individual Series
Sup-F
Exp-F
Mean-F
Breakpoint dates

48.34991*
19.76564*
12.62413*
(9/27/2008)

0.0000
0.0000
0.0000

8.89871**
1.09283
0.95608
(10/27/2008)

0.0438
0.1718
0.3608

3.71993
0.64947
1.07889
(3/09/2009)

0.4192
0.3405
0.3105

16.10121**
11.08013**
11.82768**
(09/17/2008)

0.0153
0.0175
0.0134

AII. VAR equation

Sup-F
Exp-F
Mean-F
Breakpoint dates

9.95721*
1.64053**
2.33404**
(11/03/2008)

0.0000
0.0134
0.0148

14.00715*
2.36201**
2.05116**
(10/27/2008)

0.0000
0.0008
0.0321

11.9645*
2.72904*
3.21093*
(12/08/2008)

0.0000
0.0002
0.0012

7.50490*
1.89943*
2.98544*
(1/19/2009)

0.0002.
0.0049
0.0023

Panel B: Volatility
BI. Individual Series
Sup-F
Exp-F
Mean-F
Breakpoint dates

45.19441*
18.63405*
18.53153*
(9/26/2009)

0.0000
0.0000
0.0000

65.85208*
28.91205*
38.25681*
(12/03/2007)

0.0000
0.0000
0.0000

75.09920*
33.23924*
41.36326*
(4/13/2009)

0.0000
0.0000
0.0000

70.49575*
30.81730*
32.41256*
(12/22/2008)

0.0000
0.0000
0.0000

BII. VAR equation

Sup-F
Exp-F
Mean-F
Breakpoint dates

27.64607*
8.505546*
2.996688*
(11/03/2008)

0.0000
0.0000
0.0023

37.21141*
13.47722*
3.88726*
(11/10/2008)

0.0000
0.0000
0.0002

74.02342*
31.28486*
6.388964*
(11/10/2008)

0.0000
0.0000
0.0000

11.41187*
3.109015*
4.110173*
(12/01/2008)

0.0000
0.0000
0.0001

Note: This table reports Sup-F, Exp-F and Mean-F structural breaks tests. Panels AI and BI report results for individual series. Panels BI and BII report results
for individual equations of the VAR-system. P-values are calculated using Hansens (1997) method.
*
Signicance at 10% level.
**
Signicance at 5% level.

for Sukuk relation with all markets included. These weak return transmissions between Sukuk and other markets implies a
potential diversication benets of including Sukuk in a global portfolio.
Panel B of Table 3 contains the spillover matrix of volatility. The volatility spillovers in the system are noticeably stronger
than the return spillovers. The total cross market transmissions of volatility as in Eq. (5) is around 53% compared to 31.7% of
total return spillovers. Nearly half of the volatility shocks in the whole system are driven by inter market spillovers.
The matrix also shows that the volatility transmission from the bond market to the Sukuk is substantial (27.7%), while
the spillovers in the opposite direction is very weak (0.8%). Similar spillover pattern of Sukuk with equities prevails. Sukuk
contributes 0.4% and 1.3% to Islamic and global equities respectively, while it receives 23.7% and 18.7% transmissions from
both indices in exchange.28 Similar pattern was not found for global bonds. On the contrary, bonds are a net giver of volatility
for all markets included in the system. Therefore we may conclude that while bonds are net transmitters of volatility, Sukuk
are net receivers. This indicates that Sukuk are vulnerable to increased global volatility of bonds and equities.
It is well known that at any point in time transmissions may change. For instance, when potential risks are substantial,
linkages among markets spike, and spillovers increase. The computaion over the whole sample will not capture these
important temporal changes. Hence, Diebold and Yilmaz (2012) perform rolling window analysis to obtain the dynamic
directional volatility (returns) spillovers over time. The rolling VARs analysis captures the gradual structural changes in
the estimated volatility (returns) spillovers over the sample period. This occurs without imposing any specic transition
structural breakpoint dates that are associated with the potential temporal changes which affect the transmission mechanism
in the VAR-system.
However, before conducting the rolling spillover analysis, we rst check parameter stability of the VAR system. If there
is adequate evidence of structural breaks in the VAR system, then it is appropriate to examine rolling sample volatility
spillovers.29 To test parameter stability, we implement the Andrews (1993) unknown breakpoint test on single time-series
of individual equations of the VAR system. The procedure is applied to detect structural breaks of the parameters using three
different test statistics: Sup-F, Mean-F and Exp-F. The test statistics are used toi test the null hypothesis of no structural
change against the alternative hypothesis of a single shift at unknown timing. Asymmetric trimming of 15% is imposed from
both sample ends and the test statistics is computed over the rest of the sample.
Table 4 reports the Sup-F, the Mean-F and the Exp-F structural break test statistics. The statistics on the individual series
shows clear evidence of discrete structural breaks in most of the returns and volatilities during the time period (see Panels

28
This results is consistent with Kim (2013) and Aloui et al. (2015a,b) who both recorded a weak dynamic correlations between sharia-stocks and Sukuk
in Malaysia and in GCC countries respectively.
29
The results of rolling spillover analysis become inappropriate with no structural breaks because of parameters stability.

A.I. Maghyereh, B. Awartani / Research in International Business and Finance 38 (2016) 246261

255

Fig. 2. Spillovers Indices.

AI&BI). The results also reveal that all of the series share one common break date in the third quarter of 2008 (i.e., during
the global nancial crisis).30 Similarly, it is evident that the null hypothesis of no structural breaks is rejected in all of the
cases when individual equations of the VAR-system are considered (see Panels AII & BII of Table 4). For both returns and
volatitilties equations, we have also identied a signicant structural change that corresponds to the fourth quarter of 2008.
These ndings support the notion that parameters of the VAR-system in each of the bivariate VARs are unstable and hence,
the use of rolling VARs is appropriate.
To obtain the dynamics of the information transmission, we estimate and forecast using 100 week rolling windows,
and we assess over time by computing the corresponding time series of spillovers as in Eqs. (5), (8) and (9).31 Fig. 2 shows
the dynamics of Eq. (5) which is the information transmissions among the system of markets that include Sukuk, global
bonds, Islamic stocks and global equities. Panel A of the gure shows the dynamic return transmissions, while Panel B shows
volatility transmissions. The panels clearly show that the degree of information crossover among markets is moderate
(2030%) around the beginning and the end of the sample. They also show that the volatility transmissions are higher than
return transmissions. The gure also depicts an intense transmission of information during and after the global nancial
turmoil in 2008. These transmissions are around 50% and 70% for returns and volatilities respectively.32 After maintaining
and uctuating around these levels till December 2009, the (return) volatility spillovers effect declined sharply to below 30%
and 20% respectively. Starting from January 2010, the (return) volatility transmissions moved in upward trends, till they
reach their second peak around 45% and 50% respectively in April-June 2012 largely in response to the European sovereign
debt crisis. After August 2012 information spillovers declined and went through some small cycles to revert back to their
initial levels.
The volatility transmission from Sukuk to other markets was aggregated and subtracted from the sum of transmissions
from other markets to the Sukuk market as in Eq. (8). Because the pattern of returns transmission is similar to the patterns
of volatility spillovers, we focus on volatilities, but the analysis also applies for returns. The dynamics of the net volatility
transmission is plotted in Fig. 3. For the sake of comparison we have also displayed net transmissions with global bonds. The
gure shows that while Sukuk is a net receiver of volatility spillovers from other markets, global bond are a net giver. The
extent of transmissions into the Sukuk market from other markets has increased during the global nancial crisis.
Fig. 4, displays the pairwise interaction between Sukuk and global bonds as well as between Sukuk and global stocks as
in Eq. (9). The pairwise net volatility spillovers between Sukuk and global bonds is depicted in (a) and with global equities
in (b). The gure shows that most of the time Sukuk is a net receiver of volatility transmissions from either global bonds
or global equities. The linkages are higher during stress as can be seen from the increased transmission during the 2008
nancial crisis. The evidence on increased linkages between bond Sukuk and global equity markets is consistent Aloui et al.

30

This is consistent with the results reported by Charles et al. (2015) who nd that Islamic and conventional stock indices are affected by the same events.
To double check our ndings, forecast future volatility spillovers with a different lag lengths and forecast horizons using two alternative rolling window
horizons: 50-week and 150-week. None of the changes signicantly affected the patterns of volatility spillovers. The results are not reported but available
from the authors upon request.
32
We test for structural breaks in the mean level of the (return) volatility spillover series to identify statistically the dates of possible spillovers shifts.
Using the Andrews (1993) unknown breakpoint test, we found a structural break in the (return) volatility spillover series fall around October 2008 that
corresponds to the global nancial crises. The results are not reported but available from the authors upon request.
31

256

A.I. Maghyereh, B. Awartani / Research in International Business and Finance 38 (2016) 246261

Fig. 3. Dynamic net volatility spillovers.

Fig. 4. Net pairwise volatility spillovers.

Fig. 5. Dynamic conditional correlations.

(2015a,b) who recorded increased dynamic correlations between sharia-stocks and Sukuk in the GCC countries during the
recent global crisis.
6. Some robustness analysis
6.1. Dynamic conditional correlations
To double check ndings under another data generating process, we estimated a dynamic conditional correlation model
of multivariate generalized autoregressive conditional heteroskedasticity (DCC-GARCH).33 The model allows for measuring
temporal association between Sukuk and other markets over time.34 The model is estimated by incorporating the dummies of
the structural break dates in conditional variances. As mentioned previously, these dummies are identied by implementing
the procedure of the Andrews (1993) that test for unknown structural breaks and reported previously in Table 4.35
The result of the model is a series of correlations that demonstrate the changes in returns association over the sample
period. Table 5 contains the estimation results of the dynamic conditional correlation model. Fig. 5, presents a time series
graph of correlations between Sukuk and world bonds and equities. The gure displays weak co-movement between Sukuk
and world bonds. The correlations are less than 10% most of the times. It was only during the global nancial stress that the
correlations spike to around 0.4, but it also jumps down afterwards to negative 0.2 by the end of 2008. The graph re-conrms
the above results on that Sukuk are weakly correlated with bonds and equities and that they have a great diversication

33

This model was rst proposed by Engle (2002) and it is based on assuming a GARCH volatility process of the second moment of returns.
The dynamic conditional correlation model was estimated in two steps. In the rst step, we compute a time varying conditional variances using a
multi-variate GARCH (1, 1) process; while in the second step, the time varying correlation matrix is calculated using the standarized residuals from the
rst step GARCH model.
35
As stated in the empirical literature (e.g., Rapach and Strauss 2008), the accommodating structural breaks in the conditional variance of the return
series allows avoiding overestimation of persistence in the GARCH models and ensures adjusted parameters.
34

A.I. Maghyereh, B. Awartani / Research in International Business and Finance 38 (2016) 246261
Table 5
Estimation results of DCC-GARCH model.
rt = 0 + 1 rt1 + t andt |t1 N (0, Ht )

ht = 0 + 1 2t1 + 2 hit1 + d1 D1 + . . . + d4 D4

Sukuk (1)


Qt = 1 Q + t1 t1 + Qt1

1/2

t = diag Qt

Islamic Stock Index (3)

Global Stock Index (4)

Panel A: Univariate GARCH estimates and univariate diagnostic tests

Conditional mean equation
0.00898*** (0.0000)
0.00231** (0.0301)
0
0.29391*** (0.0000)
0.03921** (0.04017)
1

0.02765*** (0.0099)
0.18035** (0.0000)

0.02545*** (0.0007)
0.05115*** (0.0017)

Conditional variance equation

0.02072 (0.5437)
0
0.13236** (0.0454)
1
0.84643*** (0.0000)
2
0.00126** (0.0184)
d1
0.00014 (0.9781)
d2
d3
0.00024 (0.8522)
0.00008 (0.8842)
d4

0.00639** (0.0481)
0.03830*** (0.0000)
0.8901*** (0.0000)
0.00093 (0.2924)
0.00129*** (0.0014)
0.00024 (0.2658)
0.00142*** (0.0003)

0.00528** (0.0011)
0.11630*** (0.0000)
0.86832*** (0.0000)
0.00009 (0.2924)
0.00129*** (0.0014)
0.00142*** (0.0003)
0.00024 (0.2658)

0.00342*** (0.0004)
0.09026*** (0.0000)
0.89495*** (0.0000)
0.00030 (0.1478)
0.00086*** (0.0005)
0.00050* (0.0562)
0.00109*** (0.0000)

Univariate diagnostic
Q(10)
Q2 (10)

9.88313 (0.4508)
8.3636
(0.2570)

11.1020
7.4624*

11.0626 (0.3526)
10.06097 (0.3445)

7.38513 (0.6886)
0.9895 (0.7857)

Global Bond Index (2)

257

(0.34962)
(0.1647)

Panel B: Conditional correlation estimates and multivariate diagnostic tests

Multivariate DCC equation

0.00580*
(0.0850)
0.98630***
(0.0000)

Dynamic conditional correlations

0.02350
21
31
0.02338
0.01110
41
32
0.08729**
42
0.22653**
43
0.43969***

(0.6155)
(0.6924)
(0.7342)
(0.0368)
(0.0000)
(0.0000)

Multivariate diagnostic
Li McLQ (10)
Li McLQ 2 (10)

(0.0000)
(0.0000)

136.492***
343.176***

Notes: The conditional variances of return series augmented by accounting for structural break dummies (Di ) as identied by the Andrews (1993), Andrews
and Ploberger (1994) and Hansen (1997) test for structural breaks which are reported in Table 4. Q(10) and Q2 (10) are the univariate LjungBox test
statistics for serial correlation in standardized and squared residuals, respectively. Li-McL Q(10) and Li-McL Q2 (10) are the multivariate Li and McLeods
(1981) test statistics for serial correlation in standardized and squared residuals, respectively. The values in parentheses are the p-values.
*
Signicance at 10% level.
**
Signicance at 5% level.
***
Signicance at 1% level.

potential in the context of global portfolios. The only thing to note here is that in a crisis, Sukuk does not offer comparable
protection as their association with global markets increases.
The plot of correlations of returns over the entire sample period shows weak association between world bonds and
world equities relative to the association between Sukuk and world equities. It also shows strong connections between
Islamic equities and world equities. The jump in pairwise conditional correlations swept all indices during the 2008 nancial
meltdown and subsequently reversed to normal levels afterwards.36
6.2. Jump and co-jump tests
A visual inspection of the previous results, indicates that the recent nancial crisis has been accompanied by an increase
in the correlation between Sukuk and the other nancial assets. Theoretically, the increased transmission among asset
classes during stress periods may not reect the comovement in the continuous part of the stochastic price processes of
assets. On the contrary, it may be caused by the interdependence in the jump discrete component. During turbulent periods,
information-related shocks hit across all asset classes and it causes contemporaneously jumps. This co-jumping can be
observed as higher comovements and spillovers over the measurement period (See for instance, At-Sahalia and Xiu, 2015).
The study of co-jumping in nancial markets has important policy implication in portfolio management, strategic asset
allocation and market selection. In particular, co-jumps increases systematic risk and it substantially reduces the benet of
diversication (Das and Uppal, 2004).

36
As mentioned previously, the inference on market association using correlations is prone to bias and inaccuracy due to the dependence of the correlation
measure on the state of volatility.

258

A.I. Maghyereh, B. Awartani / Research in International Business and Finance 38 (2016) 246261

Table 6
Summary of signicant jumps detected.
Sukuk

Islamic Stock Index

Global Stock Index

Panel A: Whole sample period, 03 October 2005-24 February 20014

39
96
# jumps
1.78
4.38
P(jump) (%)
37
93
# jump weeks
P(jumpweek) (%)
8.45
21.23

Global Bond Index

44
2.01
39
8.90

63
2.87
54
12.33

Panel B: Pre-crisis period, 03 October 200512 September 2008

# jumps
10
13
2.22
2.89
P(jump) (%)
9
13
# jump weeks
10.00
14.45
P(jumpweek) (%)

9
2.00
9
10.00

11
2.44
7
7.77

Panel C: Crisis period, 15 September 200830 June 2009

13
# jumps
2.67
P(jump) (%)
13
# jump weeks
8.38
P(jumpweek) (%)

48
6.28
47
30.32

17
2.19
14
9.03

21
2.69
18
11.53

Panel D: Post-crisis period, 01 July 2009-24 February 20014

16
35
# jumps
1.25
3.63
P(jump) (%)
15
33
# jump weeks
P(jumpweek) (%)
7.77
17.09

18
1.86
16
8.29

31
3.23
29
15.10

Note: The table presents the jump statistics from the Andersen et al. (2007) and Lee and Mykland (2008) jump test. The table displays the number of
jumps (# jumps), jump ratio (in%) over the sample observations (P(jump) = 100 (number of jumps detected/number of observations)), number of weeks
with at least one signicant jump (# jump weeks), the proportion of periods with at least one signicant jump (P(jumpweek) = 100 (number of jump
weeks/number of weeks)). The jump detection methodology based on sigicance level = 1%.

To see how Sukuk co-jumps with other markets and to provide further results, we use a non-parametric statistics proposed
by Lee and Mykland (2008) to check for co-jumping in the sample. The daily return observations are used to test for the
presence of Jumps and co-jumps in weekly periods. The approach consists of two steps. First, for each asset market, we
compute the jump statistics using daily data within each available calendar week. Second, for each pair of asset markets we
identify co-jumps by assess whether they share a simultaneous jump.
Lee and Mykland (2008) test statistics that tests for a jump over the period ti1 to t is given by37
Jumpt,i =

|rt,i |

(10)

 t,i

where rt , i is intra-weekly return over the period t, and  t,i is the local volatility estimate at t. The  t,i is estimated by realised
bi-power variation (BPV) over a local K-window around or before ti as follows




 t,i = 

1
2 K 2

i1


|rt,i ||rt,i1 |

(11)

j=ik+2

Under the null of no jump and as t 0, Jumpt,i follows a asymptotically normal distribution. In the occurance of jump
as t 0, Jumpt,i becomes very large. Within this denition, a jump can be detected if
Jumpt,i > G1 (1 ) Sn + Cn

(12)

where G1 (1 ) is the 1 quantile function of the standard Gumbel distribution, Sn =

log +(log n)
2c

2 log n

1
c(2 log n)0.5

and Cn =

2 log n
c

with the number of observations n. At a signicance level of , one can reject the null hypothesis of no jumps if

Jumpt,i > Sn + Cn with such that P  = exp e = (1 ), i.e., = log( log (1 )).
In our analysis below, we identify a jump using the 1% signicance level. After jumps in each of the asset markets are
identied according to Eq. (12), the occurrence of a co-jump is determined when the detected jumps in the two indices
appear simultaneously.
Table 6 (Panel A) presents the results of the jump statistics across the four indices over the sample period. The results in
the table show that the four indices exhibit evidence of jump with smallest number of jumps occurring in Sukuk (39, corresponding of 1.78% of the total sample observations) and the largest number occurring in the world bonds (94, corresponding
4.38% of the total sample observations). Similarly, the number of weeks with at least one signicant jump varies between

37

Note that the text and notation in this sub-section are extensively quoted from Lahaye et al. (2011).

A.I. Maghyereh, B. Awartani / Research in International Business and Finance 38 (2016) 246261

259

Fig. 6. Jumps.
Table 7
Summary of cojumps.
Islamic Stock Index

Global Stock Index

Panel A: Whole sample period, 03 October 2005-24 February 20014

0.00 (0.00%)
9.00 (23.08%)
Sukuk
9.00 (9.38%)
0.00 (0.00%)
Global Bond Index
10.00 (22.73%)
4.00 (9.10%)
Islamic Stock Index
Global Stock Index
14.00 (31.82%)
11.00 (25.00%)

Sukuk

Global Bond Index

10.00 (25.64%)
4.00 (4.17%)
0.00 (0.00%)
8.00 (18.18%)

14.00 (35.90%)
11.00 (11.46)
8.00 (18.18%)
0.00 (0.00%)

Panel B: Pre-crisis period, 03 October 200512 September 2008

0.00 (0.00%)
Sukuk
2.00 (16.66%)
Global Bond Index
Islamic Stock Index
1.00 (11.11%)
2.00 (18.18%)
Global Stock Index

2.00 (20.00%)
0.00 (0.00%)
1.00 (11.11%)
3.00 (27.27%)

1.00 (10.00%)
1.00 (7.69%)
0.00(0.00%)
1.00 (9.09%)

2.00 (20.00%)
3.00 (23.08%)
1.00 (11.11%)
0.00 (0.00%)

Panel C: Crisis period, 15 September 200830 June 2009

0.00 (0.00%)
Sukuk
3.00 (6.25%)
Global Bond Index
6.00 (35.29%)
Islamic Stock Index
Global Stock Index
4.00 (19.05%)

3.00(23.08%)
0.00 (0.00%)
0.00 (0.00%)
3.00 (14.29%)

6.00 (46.15%)
0.00(0.00%)
0.00(0.00%)
2.00 (9.52%)

4.00(30.78%)
3.00 (6.25%)
2.00 (11.76%)
0.00 (0.00%)

Panel D: Post-crisis period, 01July 2009-24 February 20014

Sukuk
0.00 (0.00%)
Global Bond Index
4.00 (11.43%)
3.00 (16.67%)
Islamic Stock Index
Global Stock Index
8.00 (25.81%)

4.00 (25.00%)
0.00 (0.00%)
3.00 (16.67%)
5.00 (16.13%)

3.00 (18.75)
3.00 (8.57%)
0.00 (0.00%)
5.00 (16.13%)

8.00 (50.00%)
5.00 (14.29%)
5.00 (27.78%)
0.00 (0.00%)

Notes: This table presents the number of co-incidences of jumps market pairs. The numbers in parentheses are and the percentage of each markets jumps
in common. For example, for the top right box in Panel A, Sukuk and Global Stock Index have 2 co-incidences jumps and that number corresponds to 20%
of the total jumps for Sukuk.

37 weeks for Sukuk to 93 weeks for global bond. Thus, global bonds exhibit the most frequent occurrence of jumps, while
Sukuk tends to show the least.
We split the sample into three subperiods: pre-crisis (03 October 2005-24 February 20014), crisis (15 September 200830
June 2009) and post-crisis (01 July 2009-24 February 20014), the results on these three periods are reported in Panels B, C
and D of Table 6 respectively. As expected the panels show that there is a considirable increase in the number of jumps in
all indices during the crisis period. One can see that the largest percentage of jumps are found in global bonds, followed by
global stocks and Islamac stocks. This can also be seen clearly in Fig. 6 which plots the estimated jump times for each market
over the sample period. Prior to global nancial crisis the number of jumps in the four indices were relatively stable and
small. However, during the global nancial crises there is a considerable increase in the number of jumps across all markets.
Following the crises, the number of jumps started to decline and this has continued until the end of the sample period. The
global bonds and equties suffer from the largest amount of jumps.
Table 7 reports the number of co-jumps which is the number of simultaneous jumps for pairs of markets. It also reports
the percentage of total jumps of each index that occurs at the same time as in other markets. Panel A of the table shows that
the co-incident of jumps is relatively uncommon and that most jumps are not simultaneous. For example, out of all jumps in
Sukuk only 23.08%, 25.6% and 35.90% occur simultaneously and at the same time as jumps in world bonds, Islamic equties,
and global equties respectively.
The weak co-jumping of Sukuk with other markets especially with global bonds gives more support to the inference that
jumps in these markets is not systemic and it is mostly idiosyncratic. This strongly supports the our previous conclusion
regarding the segmentation of the Sukuk market. Comparing co-jumping across the three sup-samples (See Panels B, C and
D in Table 7) it can be seen that co-jumping between Sukuk and other markets have increased slightly but it is still weak
even in the post global crisis period.

260

A.I. Maghyereh, B. Awartani / Research in International Business and Finance 38 (2016) 246261

7. Conclusions
This paper contributes to the recent literature on the difference between Sukuk and bonds. In particular, we study the
transmission of returns and volatility information to (from) the Sukuk market from (to) other markets and we compare that
to the ow of information in the bond market. To examine the interactive linkage between Sukuk and bonds we utilize a
new spillover index approach introduced by Diebold and Yilmaz (2012). The empirical analysis covers the period from the
30th of September 2005 to the 24th of February 2014. Four dollar indices are included in the analysis. For the Sukuk market,
we include the Dow Jones Citigroup Sukuk Index, while for bond market, we included the Dow Jones Citigroup Global bond
index. The Dow Jones Islamic stock market index and nally the Dow Jones Global stock market index are also included as
proxy for Islamic and world stock markets respectively.
Our results indicate that Sukuk has a transmission mechanism that is different from bonds. Among the two markets (Sukuk
and bonds) the transmission of information is not high and it is tilted towards global bonds. Both markets are exposed to the
information transmission from global equities. However, Sukuk are more exposed as the net transmissions from equities
to Sukuk are higher than the spillovers from equities to bonds. Moreover, returns and volatility transmissions from the
Sukuk market to other markets including the bond market is tiny and negligible; while the transmissions in the opposite
direction may not be ignored specially from equities. Therefore, we conclude that Sukuk are a net receiver of returns and
volatility information from other equity and bond markets. Furthermore, compared to Sukuk, there is substantial volatility
information ows from the bond market to other markets including the Sukuk market. Hence, in terms of volatility the bond
market is a net giver of information to other markets, while Sukuk is a net receiver.
We double checked the robustness of these ndings to the data generating process by estimating a multi-variate GARCH
(1,1) dynamic conditional correlation model as in Engle (2002). The same inference on association with either the global
bonds, equities or Islamic stocks was drawn. However, there was a clear jump in the computed dynamic conditional
correlations during the global nancial crisis across all markets including the Sukuk market.
The paper have also investigated jumps and cojumps in the Sukuk, bonds and equity markets using a non-parametric
estimation procedure. The empirical investigation shows that the frequency of jumps has considerably increased in Sukuk
and in other markets since the global nancial crisis. It also indicates weak cojumps between Sukuk and global bonds.
The ndings of this paper highlight the importance of Sukuk in the strategic asset allocation and hedging of portfolio fund
managers. However, to quantify the consequences of this important conclusion on the diversication benets of international
investors, the inuence of adding Sukuk on the minimum variance efcient frontier of global portfolios needs to be measured
and depicted. This however would be best considered in another study using rm level data over varying holding periods
and therefore we leave it for future research.
References
At-Sahalia, Y., Xiu, D., 2015. Increased Correlation Among Asset Classes: Are Volatility or Jumps to Blame, or Both? Chicago Booth Paper No. 14-11. The
University of Chicago, Accessed at http://faculty.chicagobooth.edu/dacheng.xiu/research/OVERNIGHT.pdf.
Aggarwal, R.K., Yousef, T., 2000. Islamic banks and investment nancing. J. Money Credit Bank. 32, 93120.
Akhtar, S., Jahromi, M., John, K., Moise, C.E., 2013. Intensity of volatility linkages in Islamic and conventional markets. AFA 2012 Chicago Meetings Paper.
Alam, N., Hassan, H., Haque, M.A., 2013. Are Islamic bonds different from conventional bonds?: International evidence from capital market tests. Borsa
Istanb. Rev. 13, 2229.
Aloui, C., Hammoudeh, C., ben Hamid, H., 2015a. Global factors driving structural changes in the comovement between sharia stocks and sukuk in the Gulf
Cooperation Council countries. N. Am. J. Econ. Finance 31, 311329.
Aloui, C., Hammoudeh, C., ben Hamid, H., 2015b. Co-movement between sharia stocks and sukuk in the GCC markets: a time-frequency analysis. J. Int.
Financ. Mark. Inst. Money 34, 6979.
Andersen, T.G., Bollerslev, T., Diebold, F.X., Vega, C., 2007. Real-time price discovery in global stock, bond and foreign exchange markets. J. Int. Econ. 73,
251277.
Andrews, D.W.K., 1993. Tests for parameter instability and structural change with unknown change point. Econometrica 61, 821856.
Andrews, W.K., Ploberger, W., 1994. Optimal tests when a nuisance parameter is present only under the alternative. Econometrica 62, 13831414.
Awartani, B., Maghyereh, A.I., Al Shiab, M., 2013. Directional spillovers from the U.S. and the Saudi market to equities in the Gulf Cooperation Council. J.
Int. Financ. Mark. Inst. Money 27, 224242.
Badr-El-Din, A.I., 2003. Poverty Alleviation via Islamic Banking Finance to Micro-enterprises (MEs) in Sudan: Some Lessons for Poor Countries. Universitt
Bremen, Institute for World Economics and International Management (IWIM), Sudan Economy Research Group, Discussion Paper No. 35.
Bouslama, G., Lahrichi, Y., 2016. Uncertainty and risk management from Islamic perspective. Res. Int. Bus. Finance (in press).
Cakir, S., Raei, F., 2007. Sukuk vs. Eurobonds: Is there a difference in Value-at-Risk? International Monetary Fund Working Paper WP/07/237.
Charles, A., Darn, O., Pop, A., 2015. Risk and ethical investment: empirical evidence from Dow Jones Islamic indexes. Res. Int. Bus. Finance 35, 3356.
Das, S.R., Uppal, R., 2004. Systemic risk and international portfolio choice. J. Finance 59, 28092834.
Diebold, F.X., Yilmaz, K., 2012. Better to give than to receive: predictive direction measurement of volatility spillovers. Int. J. Forecast. 28, 5766.
El-Din, E.T., Hassan, M.K., 2007. Islam and speculation in the stock exchange. In: Kabir Hassan, M., Lewis, M. (Eds.), Handbook of Islamic Banking. Edward
Elgar, Cheltenham, UK, pp. 240250.
Engle, R.F., 2002. Dynamic conditional correlation: a simple class of multivariate generalized autoregressive conditional het-eroskedasticity models. J.
Bus. Econ. Stat. 20, 339350.
Fleming, J., Kirby, C., Ostdiek, B., 1998. Information and volatility linkages in the stock, bond, and money markets. J. Financ. Econ. 49, 111137.
Garman, M.B., Klass, M.J., 1980. On the estimation of security price volatilities from historical data. J. Bus. 53, 6778.
Godlewski, C., Turk-Ariss, R., Weill, L., 2013. Sukuk vs. conventional bonds: a stock market perspective. J. Comp. Econ. 41, 745761.
Hansen, E.B., 1997. Approximate asymptotic P values for structural-change tests. J. Bus. Econ. Stat. 15, 6067.
Kim, H.B., 2013. The volatility spillover effects in Islamic nancial market. Int. Rev. Bus. Res. Pap. 9, 1227.
Koop, G., Pesaran, M.H., Potter, S.M., 1996. Impulse response analysis in nonlinear multivariate models. J. Econom. 74, 119147.
Lahaye, J., Laurent, S., Neely, J., 2011. Jumps: cojumps and macro announcements. J. Appl. Econom. 26, 893921.
Lee, S., Mykland, P., 2008. Jumps in nancial markets: a new nonparametric test and jump dynamics. Rev. Financ. Stud. 21 (6), 25352563.

A.I. Maghyereh, B. Awartani / Research in International Business and Finance 38 (2016) 246261

261

Li, W.K., McLEOD, A.I., 1981. Distribution of the residual autocorrelations in multivariate ARMA time series models. J. R. Stat. Soc. Ser. B (Methodol.) 43,
231239.
Maghyereh, A.I., Awartani, B., Al Hilu, K., 2015. United dynamic transmissions between the U.S: and equity markets in the MENA countries: new evidence
from pre- and post-global nancial crisis. Q. Rev. Econ. Finance 56, 123138.
Metwally, M., 2006. Economic consequences of applying Islamic principles in Muslim societies. J. Islamic Bank. Finance 23 (1), 1133.
Miller, N.D., Challoner, J., Atta, A., 2007. UK welcomes the Sukuk. Int. Financ. Law Rev. 26, 2425.
Mirakhor, A., 2007. Islamic nance and globalization: a convergence? J. Islamic Econ. Bank. Finance 3, 1172.
Pesaran, M.H., Shin, Y., 1998. Generalized impulse response analysis in linear multivariate models. Econ. Lett. 58, 1729.
Rapach, D.E., Strauss, J.K., 2008. Structural breaks and GARCH models of exchange rate volatility. J. Appl. Econom. 23, 6590.
Wilson, R., 2008. Innovation in structuring Sukuk securities. Humanomics 24, 170181.