The Stock Market Responses of MENA Countries to
Major Global Shocks
A thesis submitted to
The School of Economics and Finance
University of Western Sydney
College of Law and Business
In fulfilment of the requirements for
the degree of
Doctor of Philosophy
November 2011
Hisham Mohammad Sharif Alrefai
Statement of Authentication
The work presented in this thesis is, to the best of my knowledge and belief, original
except as acknowledged in the text. I hereby declare that I have not submitted this
material, either in whole or in part, for a degree at this or any other institution.
Hisham Mohammad Sharif Alrefai
University of Western Sydney
November 2011
i
Dedication
To all of my family: you have been my inspiration and source of support, and I m
grateful for your love and continuous encouragement.
ii
Abstract
This thesis investigates whether the Middle East and North Africa (MENA) stock
markets are susceptible to various global shocks, in particular of financial and
geopolitical crises. The purpose of this research is to improve our understanding of
the effect and consequences of turbulent global events on the MENA equity markets.
The recent attempts of emerging markets, including the MENA equity markets to
integrate have profound implications for asset pricing, diversification opportunities,
and susceptibility to future global market turbulence and geopolitical unrest.
Existing studies on MENA equity markets are characterized by a limited number of
countries under examination; this thesis in contrast examines eleven equity markets
from the MENA region. Furthermore, the majority of studies have failed to examine
the country risk over time and their exposure to external shocks. This failure may
stem from the fact that these markets are completely segmented and, therefore,
considered safe haven for international investors. This raises a question of why the
empirical literature does not provide comprehensive results to justify this supposed
immunity. To solve this problem, this thesis employs several econometric models
such as the Kalman filter, the Bai and Perron (2003) model, and the Johansen
cointegration model, in addition to panel regression analysis.
The first chapter examines the country beta instability using unconditional ICAPM.
The chapter also examines the impact of the war on Iraq of 2003 using the statespace model and the Kalman filter approach. The outcomes reveal that the country
betas are not stable and that the effect of the war on Iraq is limited to only four
markets in the MENA region. The second chapter employs the multiple
iii
structural breakpoints model of Bai and Perron (2003). The results show that the
country betas have coped with regional and global crises such as the Asian, Russian,
and Turkish financial crises, and geopolitical unrest such as that brought on by the
September 11, 2001 attacks. The Bai and Perron (2003) model shows that more
MENA markets have experienced structural breaks due to the war on Iraq than what
is found from the Kalman filter.
The third chapter examines whether economic variables can explain the variation of
country beta in a panel analysis. The results show that money supply and inflation
have significant positive impacts on the country beta of MENA equity markets,
whereas the accumulation of foreign currency reserves helps to alleviate it over time.
In the fourth chapter, the thesis examines the integration of MENA equity markets
with four major developed markets: Germany, Japan, UK, and USA, using the
Johansen cointegration. The outcomes show that the integration of the MENA equity
markets has significantly increased since the global financial crisis, indicating more
market exposure to the global financial system.
iv
Acknowledgments
I must first thank my supervisors, Dr. Girijasankar Mallik and Associate professor
Partha Gangopadhyay. I would like to thank them for their patience during my study.
I am very grateful for their helpful recommendations and suggestions.
I am grateful to my parents in every aspect of my life, for their encouragement and
support. They have been the source of my inspiration. Also, my thanks go to each of
my brothers and sisters for their continuous encouragement.
I wish to thank Professor Raja Junankar and Dr. Roger Ham for their assistance
during the early stages of my study, and Professor John Lodewijks for providing me
with teaching opportunities in the Principles of Economics and the Australian
Macroeconomy. I am also grateful to the administrative staff at the School of
Economics and Finance of UWS: Trish O Brien, Craig Berry, Carolyn Love, Amelia
Younane, and Lina Gong for their friendly support.
I would like to extend my appreciation to my colleagues in the PhD program at the
School of Economics and Finance of UWS. Special thanks go to my best friends
Gazi Manul Hassan, Zahid Hasan, Ammar Jreisat, Mustafa Rahman, Alex Pham,
Soner Teknikeller, Robert Wells, and Ahmad Muzaffar for their help and
encouragement. I would like to express my appreciation to Dr. Margaret Johnson of
The Book Doctor for her excellent editing of the thesis.
v
Table of Contents
Abstract
1
Introduction ........................................................................................... 1
1.1
Motivation of the study ................................................................................... 3
1.3
Thesis structure ............................................................................................... 6
1.2
2
iii
Objective and significance .............................................................................. 4
Literature Review .................................................................................. 8
2.9. Conclusion ............................................................................................................... 64
3.1. Introduction.............................................................................................................. 67
3.2. The war on Iraq and equity markets ........................................................................ 69
3.3. Empirical framework ............................................................................................... 72
3.3.1. Data............................................................................................................................... 72
3.3.2. CAPM and beta............................................................................................................. 73
3.3.3. Testing for parameter stability ...................................................................................... 73
3.3.4. The conditional beta test ............................................................................................... 75
3.3.5. The conditional ICAPM and the impact of the war on Iraq.......................................... 76
3.4. Empirical results ...................................................................................................... 77
3.4.1. Descriptive statistics ..................................................................................................... 77
3.4.2. Correlation analysis ...................................................................................................... 82
3.4.3. The stability of country beta using the static ICAPM................................................... 88
3.4.4. The conditional market model and the impact of the war on Iraq ................................ 93
3.5. Conclusion ............................................................................................................... 96
Further Evidence on the Impact of Financial and Geopolitical Crises on the Country
Beta of MENA Markets ...................................................................... 98
4.1. Introduction.............................................................................................................. 99
4.2. Description of the unstable crises periods ............................................................. 105
4.3. Data and methodology ........................................................................................... 108
4.3.1. Data............................................................................................................................. 108
4.3.2. The Kalman filter model............................................................................................. 108
4.3.3. Structural breaks analysis ........................................................................................... 109
4.4. Empirical results .................................................................................................... 110
4.4.1. Descriptive statistics of the country betas................................................................... 110
4.4.2. Structural breaks results.............................................................................................. 111
vi
4.5. Conclusion ............................................................................................................. 119
Determinants of Country Betas of MENA Equity Markets: A Panel Analysis ....... 121
5.1. Introduction............................................................................................................ 122
5.2. Empirical framework ............................................................................................. 127
5.2.1. Description of equity markets ..................................................................................... 127
5.2.2. Selection of the economic variables ........................................................................... 128
5.2.3. The unanticipated components of the economic variables.......................................... 133
5.2.4
Country beta in panel model ....................................................................... 135
5.2.5. Hypothesized relations................................................................................................ 137
5.6. Empirical results .................................................................................................... 137
5.6.1. Descriptive statistics ................................................................................................... 137
5.6.2. Correlation analysis .................................................................................................... 138
5.6.3. Panel unit root test ...................................................................................................... 139
5.6.4. Panel regression results............................................................................................... 140
5.7. Conclusion ............................................................................................................. 149
The Integration of MENA Stock Markets with Germany, Japan, UK, and USA .... 150
6.1. Introduction............................................................................................................ 151
6.2. The recent global financial crisis ........................................................................... 155
6.3. Empirical framework ............................................................................................. 157
6.3.1. Data............................................................................................................................. 157
6.3.2. Unit root test ............................................................................................................... 158
6.3.3. Multivariate cointegration analysis ............................................................................. 159
6.3.4. Error Correction (EC) model ...................................................................................... 161
6.4. Empirical results .................................................................................................... 161
6.4.1. Descriptive statistics ................................................................................................... 161
6.4.2. Correlation analysis .................................................................................................... 162
6.4.3. Unit root results .......................................................................................................... 165
6.4.4. Lag length criteria....................................................................................................... 166
5.4.5. Cointegration results ................................................................................................... 168
6.4.6. Structural break analysis ............................................................................................. 169
6.4.7. Adjustment to the shocks ............................................................................................ 175
6.4.8. Error Correction (EC) results...................................................................................... 180
6.5. The regional stock market integration of MENA countries................................... 188
6.5.1. Stock market development in the MENA region ........................................................ 189
6.5.2. Empirical framework .................................................................................................. 192
6.5.3. Cointegration results ................................................................................................... 193
6.5.4. Error Correction (EC) Model...................................................................................... 194
vii
6.6. Empirical results on the regional integration of MENA markets .......................... 194
6.6.1 Error correction results ................................................................................................ 196
6.7. Conclusion ............................................................................................................. 197
Summary and Concluding Remarks ........................................................................ 201
7.1. Introduction............................................................................................................ 201
7.2. Concluding remarks............................................................................................... 202
7.3. Implications of the results...................................................................................... 204
7.4. Limitations and weaknesses of the thesis .............................................................. 206
7.5. Areas of future research......................................................................................... 208
Appendix
References
209
211
viii
List of Tables
Table 3-1. Descriptive statistics of weekly return series ........................................... 79
Table 3-2. Correlation analysis of MENA market returns and the global MSCI
returns .................................................................................................. 84
Table 3-3. Analysis of country beta stability using equation (3.2) ............................ 89
Table 3-4. Conditional ICAPM with the war on Iraq dummy using equation (3.6) .. 95
Table 4-1. Crisis timelines ....................................................................................... 106
Table 4-2. Descriptive statistics of the country betas series .................................... 112
Table 4-3. Bai and Perron s structural breakpoints with their 95% confidence
intervals ............................................................................................. 115
Table 5-1. Selection and description of the macroeconomic variables.................... 133
Table 5-2. Decomposition of the economic variables.............................................. 135
Table 5-3. Descriptive statistics of the economic variables and country betas........ 138
Table 5-4. Correlation analysis of the economic variables ...................................... 139
Table 5-5. Unit root test for panel data variables..................................................... 140
Table 5-6. Results for the fixed effects (FE) model using equation (5.5)................ 141
Table 5-7. Results of the fixed effects (FE) model using equation (5.5) ................. 144
Table 5-8. Results of the expanded fixed effects (FE) model using equation 5.6 ... 147
Table 6-1. Descriptive statistics of the MENA and the developed markets weekly
returns ................................................................................................ 163
Table 6-2. Correlation matrix of MENA equity markets with the developed markets
of Germany, Japan, UK, and USA .................................................... 164
ix
Table 6-3. Unit root results ...................................................................................... 165
Table 6-4. Optimal lag length in VAR for models (6.6) to (6.9) ............................. 167
Table 6-5. The results of the cointegration technique using equations (6.6) to (6.9)
........................................................................................................... 171
Table 6-6. Results of Zivot and Andrews (1992) model for the timings of the
structural breakpoint.......................................................................... 174
Table 6-7. Results of the cointegration model augmented with the dummy variables
of the structural breakpoints .............................................................. 176
Table 6-8. Normalized cointegrating vectors before and after the crisis ................. 179
Table 6-9A. EC model with Germany ..................................................................... 182
Table 6-9B. EC model with Japan ........................................................................... 183
Table 6-9C. EC model with UK............................................................................... 184
Table 6-9D. EC model with USA ............................................................................ 185
Table 6-10. MENA equity markets indicators, 2009................................................. 190
Table 6-11. Optimal lag length in VAR................................................................... 193
Table 6-12. Empirical results of Johansen s cointegration test................................ 195
Table 6-13. Number of cointegration vectors .......................................................... 196
Table 6.14. EC model results ................................................................................... 198
x
List of Figures
Figure 3.1: Conditional country betas for Tunisia, Egypt, Kuwait and Morocco
showing the impact of the war on Iraq. ............................................... 96
Figure 4.1: Country betas estimated by the Kalman filter approach........................ 113
Figure 6.1: Natural logs of stock indices and returns for Germany, Japan, UK, and
USA. .................................................................................................. 170
Figure 6.2: Natural logs of MENA stock indices..................................................... 192
xi
Abbreviations
ADRs
American Depositary Receipts.
AIC
Akaike Information Criterion.
APT
Arbitrage Pricing Model.
ADF
Augmented Dickey-Fuller.
ARCH
Autoregressive Conditional Heteroscedasticity.
ARIMA
Autoregressive Integrated Moving Average.
BIC
Bayesian Information Criterion.
BERI
Business Environment Risk Intelligence.
CAPM
Capital Asset Pricing Model.
CPI
Consumer Price Index.
CRIS
Control Risks Information Services.
CUSUMSQ
Cumulative Sum of Squares.
DCC
Dynamic Conditional Correlation.
EGARCH
Exponential GARCH.
EWS
Early Warning System.
EIU
Economist Intelligence Unit.
EC
Error Correction.
FDI
Foreign Direct investment.
GARCH
Generalized Autoregressive Conditional Heteroscedasticity.
GMM
Generalized Method of Moments.
GDP
Gross Domestic Products.
GCC
Gulf Cooperation Council.
HQIC
Hannan-Quinn Information Criterion.
ICAPM
International Capital Asset Pricing Model.
xii
ICSS
Iterative Cumulative Sum of Squares.
ICRG
International Country Risk Guide.
IFS
International Financial Statistics.
IMF
International Monetary Fund.
IPS
Im, Pesaran, and Shin.
LR
Likelihood Ratio.
KIW
Kuwait Investment Company.
KPSS
Kwiatkowski, Phillips, Schmidt, and Shin.
MENA
Middle East and North Africa.
MSCI
Morgan Stanley Capital International.
OLS
Ordinary Least Squares.
OPEC
Organization of Petroleum Exporting Countries.
PCA
Principle Component Analysis.
PP
Phillips-Perron.
SBIC
Schwarz Bayesian Information Criterion.
SUR
Seemingly Unrelated Regression.
SEMIFAR
Semi-Parametric version of Fractional Autoregressive.
S&P
Standard & Poor s.
TARCH
Threshold ARCH.
VAR
Vector Autoregression.
VECM
Vector Error Correction Model.
UAE
United Arab Emirates.
xiii
1 Introduction
The last two decades have witnessed an increase in financial crises and geopolitical
unrest that has affected a large number of emerging and developed equity markets,
either directly or indirectly. These turbulent events have been accompanied by
episodes of financial market volatility and contagion and threatened the financial and
economic stability of both developed and emerging equity markets around the world.
The real impact of such events on stock markets has varied, depending on the degree
of market integration. According to Sharma (2010), deep financial integration means
more rapid and powerful spillover across economies through both traditional trade
and more new types of financial channels. Empirical studies have defined three
types of stock market integration: full integration, full segmentation, and
mild/partial segmentation or integration (see for example; Bilson et al., 2001;
Fedorova & Vaihekoski, 2009). Because of the developed markets integration with
the global capital market, international investors have been drawn to less integrated
markets, mainly emerging markets, due to their low correlation with the global
market, rapid economic development, high returns, potential for diversification, and,
most importantly, their supposed immunity to the global market volatility (see for
example Arouri et al., 2009; Harvey, 1995c).
It has been argued that because of the liberalization of emerging markets, the
globalizing economy and financial interlinkages have been partially responsible for
financial crises around the world because the propagation of external shocks may lie
in the presence of financial linkages. According to Sharma (2010), the recent global
financial crisis in 2008 unambiguously and painfully underscored that fact that in
1
today s globalized and interconnected world, no nation is an island. The ferocious
contagion or the exogenous shocks from advanced economies, threw the global
economy into its most serious crisis since the 1930s.
The MENA markets are relatively new, and small in size compared to other
emerging and developed markets. They are considered attractive diversification
opportunities by international investors due to their low correlation with the global
market and their lack of integration. During the last two decades, most MENA
countries have liberalized their stock markets and introduced structural reforms to
their financial sector, and as this liberalization transition has continued, the MENA
stock markets have became more connected than before with the global financial
system.
The liberalization process means that their financial markets are more likely to be
affected by external shocks, making their financial system potentially more
vulnerable to global shocks. According to the World Economic Forum Report
(2007), MENA countries supposed immunity to regional and global events has
varied, apparently because small events can have increasingly disproportionate or
asymmetric effects. In addition, the vulnerabilities of the different economies and
political structures, geographies, and religious compositions across the region
inevitably mean that global risks play out differently there.
One of the most crucial challenges of academics and researchers is the assessment of
a country s market exposure in the international arena. From an econometric
perspective, by using the international version of the CAPM, one can infer that
systematic risk or beta is an indicator of country risk or country beta . Within this
2
framework, the country beta is measured as the sensitivity of a country s returns to
the global capital market (Harvey, 1991). From other econometric perspectives such
as cointegration analysis, the link between long-run relationships with major
international stock markets and how these links vary through time can be inferred to
indicate increased market integration and therefore increased market exposure to the
global financial system.
1.1 Motivation of the study
This thesis is motivated by a number of factors: the first is the evidence that
systematic risk (or beta) has been well-researched in the developed and emerging
markets at both macroeconomic and microeconomic level but that few studies have
directly addressed this issue at the country level. The second is that the responses of
stock market risk to external shocks such as financial crises and geopolitical unrest
have received little attention. The third is that the literature has documented evidence
that external shocks represent a risk aversion for international investors for certain
portfolio and individual betas. None of the studies, on the other hand, draws the same
conclusion for the country beta. Therefore, the country beta variable will reflect the
level of risk aversion of investors during financial and geopolitical unrests.
Moreover, there are little or no study has been undertaken in the context of MENA
countries and this thesis will address the gap in the existing literature.
The consequences of major financial and geopolitical events on stock markets have
gained increased attention given recent increased global market instability. In
particular, financial crises such as the Asian financial crisis, the recent subprime
crisis in the USA, and regional and global geopolitical events such as the September
11, 2001 attacks and the war on Iraq in 2003, have impacted on regional and the
3
world economies, and such events could have major implications for stock market
performance. However, their impacts have varied remarkably across the countries
around the world. A large number of studies have attempted to uncover how global
shocks are transmitted to stock markets. Most of the empirical literature has
examined equity market volatility, spillover effects, and contagion for both regional
and international markets. Few studies, however, have examined their impact
specifically in the MENA region.
1.2 Objective and significance
The main objective of this thesis is to examine the responses of the MENA equity
markets to major financial and geopolitical crises. To attain this objective, the thesis
seeks to present the following:
A comprehensive literature review of the notion of systematic risk, the
country risk or
country beta
in the quantitative and qualitative
framework. In addition, the thesis will review the global stock markets
responses to major financial crises and geopolitical events, and review the
outcomes of stock market integration of emerging equity markets. In
addition, the study will explore the literature relating to stock market
responses to global unrest using various econometric models from various
regions.
An examination of the statistical properties of the MENA equity markets,
their country beta instability within the framework of the International
Capital Asset Pricing Model (ICAPM), and the impact of the 2003 war on
Iraq on the country beta of MENA equity markets. The question here is
4
how the betas of MENA countries have evolved over time and whether
the geopolitical shock of the war on Iraq had an effect on them.
Further evidence on the responses of country beta of MENA equity
markets to major regional and global financial and geopolitical unrest,
using the structural break model of Bai and Perron (2003). This model
has the advantage of locating several structural breaks in time series with
no knowledge of the breakpoint a priori. The chapter will try to answer
the question of whether the unstable time periods of financial and
geopolitical crises have caused any significant structural breaks in
country beta over time.
The determinants of country beta of MENA markets using panel data
regression analysis. The question of this chapter is to investigate whether
economic variables have any considerable positive/negative effect on the
country beta. This empirical investigation is significant because policy
makers and international investors may need to pinpoint which economic
variable(s) may increase the country beta or help cushion it from further
external shocks. Furthermore, the current focus of empirical finance
research has shifted to the study of equity markets performance especially
at regional or global level.
An examination of the integration of MENA stock markets with the
developed stock markets of Germany, UK, Japan, and USA using
Johansen s cointegration analysis. This model is widely used for
examining the level of stock market integration and the degree of co-
5
movements between stock market indices. The question in this chapter is
whether the level of MENA equity market integration has increased
during the recent turbulent period of the global financial crisis.
1.3 Thesis structure
The remainder of the thesis is organized as follows:
Chapter 2 presents the literature review on the topic of systematic risk, country risk
(beta) and the integration of both developed and emerging markets, and the various
empirical analyses of regional and global unrest and their impact on stock market
performance for both developed and emerging markets.
Chapter 3 presents the analysis of country beta instability using the unconditional
International CAPM, for equity markets in the MENA region. The chapter also
examines the impact of the war on Iraq on the country betas of MENA equity
markets using the dynamic state-space model and the Kalman filter approach.
Chapter 4 presents a structural breaks analysis using the novel Bai and Perron (2003)
model for the country beta series estimated by the Kalman filter approach, and
relates the structural breaks to the major global shocks such as financial and
geopolitical crises.
Chapter 5 examines the determinants of country beta of MENA markets using local
factors and the oil price in a panel data regression analysis. The reason for this
chapter is to consider the effects of financial, economic, and geographical location,
and the possible linkage of stock markets in the region.
Chapter 6 adopts Johansen s cointegration technique and the vector error correction
6
model to examine the level of stock market integration between MENA countries
and the developed markets of Germany, UK, Japan, and USA, taking into account
the period of the global financial crisis.
Chapter 7 presents the summary and conclusions of the thesis. The findings of the
previous chapters are presented along with their implications. The limitation and
weaknesses of the work are discussed, and suggestions are made about the direction
of future research that might minimize current limitations.
7
2 Literature Review
2.1 Introduction
The introduction of the Capital Asset Pricing Model (CAPM) by Sharpe (1964) and
Lintner (1965) has shaped the central development of finance theory. The traditional
measure of risk, or beta, which measures an asset s responsiveness to the market, is
assumed to be constant over time. Although the CAPM does not require the beta
coefficient to be constant over time, this assumption has undermined the predictive
ability of the model for future periods (Dotan & Ofer, 1984).
The regression technique used to estimate the CAPM betas is the Ordinary Least
Squares (OLS). This method, which assumes the constancy of beta, has been widely
rejected, and the unconditional CAPM has been criticized for this limitation as much
empirical evidence suggests that the beta coefficient is time-varying in most
developed and emerging markets. Early empirical studies showed that beta displays
stochastic behaviour (e.g. Alexander & Benson, 1982, Blume, 1975; Bos &
Newbold, 1984; Fabozzi & Francis, 1978) on the US market. This has prompted
many researchers to describe the behaviour of systematc risk "beta" over time in
other developed and emerging markets. Studies in favour of time-varying betas over
the static betas in both developed and emerging markets include the works of
Bodurtha and Mark (1991), Berglund and Knif (1999), Brooks et al. (1998), Brooks
et al. (1992), Choudhry (2001, 2002), Ebner and Neumann (2005), Grieb and Reyes
(2001), Moonis and Shah (2003), Park and Kim (2007), and Wells (1994), amongst
others.
8
Portfolio managers and international investors seek the benefits of increasing their
expected returns and reducing their overall portfolio risk through international
diversification, and this has encouraged practitioners and academics to develop
accurate estimates of portfolio risk and expected returns of their equity holdings.
Investment in less-developed emerging equity markets is attractive because of their
high average returns and low correlations with world markets (see for example
Harvey, 1991, Harvey & Zhou, 1993). Solnik (1974) introduced the International
CAPM (ICAPM) for assessment of international equity market pricing on the basis
of Sharpe s (1964) and Lintner s (1965) domestic model. However, the ICAPM has
been scrutinized in the literature along lines similar to the local CAPM. According to
Ferson and Harvey (1999a), Hwang and Pedersen (2004), and Nummelin and
Vaihekoski (2002), the ICAPM is not suitable for emerging and smaller developed
markets that are not fully integrated to the global capital markets.
2.2. International CAPM and country beta
A large number of studies have sought to find appropriate international asset pricing
models under the assumptions of integration and segmentation. According to Clark
and Kassimatis (2006), the test of the ICAPM depends crucially on the index or
indices chosen to proxy for the true market portfolio. On the other hand, Fama and
French (1998) state that the global market risk is unable to capture the crosssectional return spreads among portfolios sorted by book-to-market ratio in the
global markets. They conclude that international models are specified under the
market integration hypothesis, which implies that only global risks are priced and
that prices are uniform across countries. Zhang (2006) states that it is appropriate for
international asset pricing models to choose national market indices as the base
9
assets; she argues that if the world markets are integrated, then assets with the same
risk characteristics should receive the same prices, irrespective of their provenance.
Most financial decision-making in the international setting needs to apply a
framework for estimating a country-level risk. This strategy is particularly important
in the assessment of investment projects (Wdowinski, 2004). Numerous approaches
to measuring country risk (qualitative and quantitative) have been examined in the
international finance literature. The best-known qualitative methods are offered by
the Bank of America World Information Services, the Business Environment Risk
Intelligence (BERI), the Control Risks Information Services (CRIS), the Economist
Intelligence Unit (EIU), Euromoney, the Institutional Investor, Standard & Poor s
Rating Group, the International Country Risk Guide (ICRG) (see for example: Erb et
al., 1996b; Hoti, 2005; Hoti & McAleer, 2005; Kaminsky & Schmukler, 2002;
Khoury, 2003). The best well-known quantitative measure is the country risk, based
on the International CAPM of Solnik (1974). Bekaert at al. (1997) conclude that the
traditional beta risk paradigm is problematic in emerging markets because a number
of them are not fully integrated with the global capital market. According to Saleem
and Vaihekoski (2008), if equity markets are completely integrated into the world
economy, then the ICAPM suggests that the only systematic source of risk is global
market risk, assuming that investors do not hedge against exchange rate risk and that
risk free assets exist. Bekaert et al. (1998) suggest that empirical relationships
between risk and stock returns in emerging markets are not appropriately described
by the ICAPM to estimate the country beta. On the other hand, Khoury (2003) argues
that there is no model that is best to specify the country risk in international
diversification.
10
The concern is that systematic risk at the international level tends to be unstable and
needs to be tested and modelled (Brooks et al., 2002; Gangemi et al., 2000). The
urgent question raised is whether to consider the country risk of the ICAPM as
constant or time-varying. Harvey (1991) argues that the price of risk in global
markets is not constant over time, and that the world price of risk is not an
appropriate measure of covariance risk. Ferson and Harvey (1993) apply both
constant and time-varying beta models to a sample of emerging and developed
markets and found that both models are misspecified. Khoury (2003) states that if
markets are not completely integrated into the world economy then there is no reason
to put much faith in the ICAPM. He notes that country betas
against a single factor model or multiple sources of risk
whether measured
appear to have less than
convincing ability to discriminate between expected returns from different assets or
financial markets. According to Bekaert et al. (1996b), the failure of the ICAPM to
explain emerging market returns may be interpreted in a number of ways: first, the
benchmark world portfolio may not be mean-variance efficient; second, perhaps a
multifactor representation is more appropriate for emerging markets; third, an
examination of average returns and average risk could be misleading if the risk and
expected returns are not changing over time; and fourth, the ICAPM is not the
appropriate framework if these emerging markets are not fully integrated into world
markets.
The time-varying country risk has become a persistent concern in international
finance especially the world markets have reached to rapid business environment,
and assessment has become important for international investors (Brooks, 2003;
Brooks et al., 2002). The estimation of country beta requires rigorous analysis if
11
international investors are to consider international diversification and incorporate
country risk in their decision-making. The country risk assessment can save investors
from massive losses when diversifying internationally (Khoury, 2003). Brooks and
Del Negro (2002) argue that diversification across countries is more effective for risk
reduction than is diversifying across industries. In this sense, the country risk in
terms of the ICAPM is defined as the contribution of the variance of a well
diversified portfolio; however, this kind of test requires the international market to be
integrated. According to Bekaert and Harvey (1995), when markets move from a
state of segmentation (a market effectively closed to outside investors) to a state of
integration (a market accessible to outside investors), the economic sources of risk
change, and in the integrated markets the risk is measured with respect to the world
economy. Chan et al. (1992) conclude that it is difficult to reject the ICAPM for
integrated markets. Studies investigating stock market integration into developed and
emerging markets using the ICAPM include those of Arouri (2004), Arouri et al.
(2006), Bekaert and Harvey (1995), Bruner et al. (2008), Dominguez (2007), and
Gerard et al. (2007). The general conclusion of these studies varies as to the degree
of market integration, especially for emerging markets.
The new studies on country beta instability have shaped the asset pricing models at
the international framework, where the country beta is defined relative to a global or
regional market portfolio. Khoury (2003) states that there is an urgent need to
measure the country risk of international markets before investing internationally.
Harvey (2000) argues that the ICAPM is a powerful model to measure the country
beta, which has met with some success when applied to developed market returns.
However, he notes that the same model may fail when applied to emerging markets.
12
He offers many reasons why this may be so; the leading reason is the lack of market
integration in some emerging markets. Bekaert and Harvey (2002) state that much of
the research on time-varying country beta has focused on the efficient developed
markets, which are most likely to be consistent with the theoretical models; but this
consistency does not exist in emerging markets. They suggest a modified method
that redefines the beta as a ratio of local to world standard deviation rather than the
covariance divided by the world variance; however, they note that neither method
has any theoretical foundation. Harvey (2000) concludes that risk measures implied
by asset pricing theory
in particular, the world beta and coskewness
work
reasonably well in capturing the cross-section of average returns in world markets.
Fama and French (1998), examining portfolio risk, conclude that country portfolios
leave plenty of room for international asset pricing models to fail.
Although the assessment of the country risk in terms of the ICAPM has found some
objections in the literature, especially with regard to emerging markets, there is
substantial evidence that at the international level, the time variation in country risk
is still very meaningful to foreign investors. Numerous studies have shown that
country risk relative to the world market portfolio is changing over time, and this
especially holds true for emerging markets. According to Johansson (2009a) it is
difficult to model the country risks of emerging markets as stable processes, even if
they are not fully integrated. He advocates a conditional approach to country risk
estimation when looking at investment possibilities in developing countries. Harvey
(1995b) argues that the rejection of the country beta could be caused by the
assumption that the country beta and the expected returns are constant over time.
2.3. Methodologies of country beta estimation
13
Given the increasingly international investment environment, the need for an
accurate model of country risk lies high on the research agenda. One of the issues is
the extent to which the techniques used to model risk in a domestic context can be
readily applied to the international context (Brooks, 2003). Johansson (2009a) states
that the time-varying country beta estimation may allow investors to more efficiently
apply dynamic hedging of their portfolios, and may also be used to better understand
what factors influence market risk in different countries. Country beta has shown
evidence of instability similar to that of individual stock or industry portfolios, and
finance literature examines this instability using a number of techniques applied to
both individual and industrial betas.
There are several possible methodological approaches to account for the time
variation of beta. The most popular models in recent studies are the Schwert and
Seguin model (1990), the Kalman filter, the multivariate GARCH, and the stochastic
volatility model. Brooks et al. (2002) argue that time-varying beta approaches do not
provide equal utility, as different conditional approaches may generate different
estimates of conditional betas. Schwert and Seguin (1990) attempt to correct the
misspecification problem of the CAPM by extending the single factor model to
overcome what they call this pervasive phenomenon . In their seminal article, they
show that the single factor model has failed to account for the heteroscedasticity of
equity returns, and find evidence that the market risk betas of both small and large
US market capitalization firms are differently sensitive to the aggregate market
volatility. Episcopos (1996) uses the Schwert and Seguin (1990) model to examine
the daily time-varying portfolio betas from Toronto stock exchange. The author finds
significant increased spread between betas of safe and risky sub-index portfolios
14
during periods of increased aggregate portfolio volatility. Other studies that use this
model include the work of Brooks, et al. (1998) on a sample of nine Australian
industry portfolios, Reyes (1999) on a two major small and large-cap UK indices,
Faff et al. (2000) on 32 UK industrial sectors, Ibrahim (2004) on 12 portfolios on the
Kuala Lumpur stock exchange, and Wang and Di Iorio (2007) on all A-shares listed
on the Shanghai and the Shenzhen stock exchange markets.
At the international level, this model has been successfully applied to small and large
market capitalization of international equity returns. Koutmos et al. (1994) use the
Schwert and Seguin (1990) model to estimate the time-varying country betas for ten
developed countries. The data used in their study are the weekly stock market returns
from January 16, 1976 to December 27, 1991. They find that world market volatility
has a statistically significant positive impact on the country risk of Australia,
Germany, and Switzerland and a statistically significant negative impact on the
country risk of Japan and USA. They argue that the markets with high volatility
persistence possess higher systematic risk during periods of higher world market
volatility. Brooks et al. (2002) compare the Kalman filter, the Multivariate GARCH,
and the Schwert and Seguin (1990) models for generating the best time-varying
country betas for a sample of 12 European countries plus Canada, USA, Australia,
Japan, and Hong Kong. They find that the GARCH and the Schwert and Seguin
(1990) models produce lower forecasts errors compared to the Kalman filter model.
The second method is the Kalman filter approach, which has become a very popular
model for estimating the time-varying beta for both asset and portfolio betas. Brooks
et al. (1998) compare the Kalman filter, the Multivariate GARCH, and the Schwert
and Seguin (1990) models for a sample of monthly returns on the Australian industry
15
portfolios over the period 1974 1996. Based on in- and out-of-sample forecast
errors, they find evidence that the Kalman filter is the superior of the two methods.
Mergner and Bulla (2005) investigate the time-varying behaviour of beta for 18 panEuropean sectors using the bivariate GARCH (1,1), the Kalman filter, and the
stochastic volatility models. The results of their study indicate that time-varying
sector betas are not stable over time and can be best described by a random-walk
process estimated by the Kalman filter approach. They conclude that while the insample results overwhelmingly support the Kalman filter approach, its superiority is
only partly maintained in out-of-sample forecasts. Marshall et al. (2009) assess the
time-varying country betas of 20 emerging equity markets from Latin America, the
Middle East, Eastern Europe, and Asia using the Kalman filter, the DCC-GARCH,
and Schwert and Seguin (1990) models and taking daily data from the period January
1995 to December 2008. They find that the Kalman filter model outperforms the
GARCH and the Schwert and Seguin (1990) models based on the in- and out-ofsample forecasts.
The third method used to estimate the time-varying beta is the multivariate GARCH
model, which has become very popular in the literature. Brooks et al. (2002) apply
the multivariate GARCH specification, the Kalman filter, and the Schwert and
Sequin (1990) models. They conclude that the GARCH approach produces the
lowest error as well as the Schwert and Seguin (1990) models. Giannopoulos (1995)
applies the Multivariate GARCH at the international level for a sample of 13
countries in Europe plus the US and Japan using weekly data for the period of 1984
1993. His results suggest that for most countries, market volatilities can be attributed
to country-specific events. In addition, he finds that systematic risk is unstable over
16
time, and in certain periods, such as the stock market crash in 1987, is a primary
driving factor of total risk. A third finding is that the paramatization of the model
handles the time variation of country beta reasonably well.
The fourth model used in the literature is the stochastic volatility model. Johansson
(2009a) applies multivariate stochastic volatility and multivariate GARCH models to
estimate the time-varying country betas for a set of 27 emerging equity markets from
1994 2006 from different regions such as Asia, Latin America, Africa, the Middle
East, and Eastern Europe. When comparing the outcomes of both models, his results
show that the mean absolute errors and the mean squared errors indicate that the
multivariate stochastic volatility model results in significantly lower forecast errors
than the multivariate GARCH model.
These available models have created a strong ground of competition among
researchers and academics to choose and to recommend the right model based on
their studies. In general, Brooks et al. (2002) note that all these models perform
adequately in explaining the time variation of risk, compared to traditional timeinvariant methods.
2.4. The risk exposure of national equity markets
The next challenge to authors, academics, and practitioners is how to account for the
attributes of developed and emerging markets through international asset pricing
models, to measure and quantify risk and expected asset returns. Erb et al. (1996a)
have studied the economic content of five different measures for a sample of country
risk for 135 countries. They conclude that country risk measures are correlated with
future equity returns. Harvey and Zhou (1993) test the single world pricing model on
17
a group of developed and emerging equity markets, finding that the expected country
returns are related to movements in world market portfolios. Ferson and Harvey
(1993) study the equity returns for 18 emerging and developed markets and find a
direct contribution of time-varying betas to the model variation in expected country
returns.
It is widely accepted that current stock levels are positively related to future levels of
real activity as measured by GDP or industrial production. This understanding seems
intuitive since returns are a function of the future cash flow stream, which is highly
dependent upon future economic conditions (Bilson et al., 2001). Several economic
policies exist that might have an impact on the country risk. Gallup et al. (1998)
argue that stable macroeconomic policies lower risk for investors and increase their
investment, and in turn lead to more sound economic growth. Thus, lower financial
and economic risk may encourage both domestic and foreign investment and
contribute to economic development and equity markets. The Arbitrage Pricing
Model (APT) of Roll (1977) and Chen et al. (1986) establish new grounds for the
relationship between economic innovation and stock returns. They find a number of
priced factors in stock market returns, such as interest rates, inflation rates, bond
yield spreads, and industrial production. They conclude that economic forces have a
major impact on stock market prices, but they do not find any evidence that oil price
risk is a major factor impacting the stock market.
One of the most difficult tasks in tactical global asset allocation is to assess the risk
exposure of a national equity market. Traditional factor models, while reasonably
successful in characterizing the expected return/risk trade-off in developed markets,
have failed when applied to new emerging equity markets (Erb et al., 1995c). While
18
there are many reasons to believe that the sources of risk in emerging and developed
markets may differ, Verma and Soydemir (2006) argue that in a highly volatile
business environment, assessing a country s risk becomes a strategically important
task for global investors. By identifying relevant variables affecting market exposure,
investment managers may be able to improve their portfolio performance.
Using a GARCH approach, Patro, et al. (2002) estimate a time-varying two-factor
international asset pricing model for the weekly equity index returns of 16 developed
countries for the period January 1980 to December 1997. They find significant timevariation in the exposure (beta) of country equity index returns to the world market
index and in the risk-adjusted excess returns (alpha). They explain these world
market betas and alphas using country-specific, annual macroeconomic and financial
variables in a panel data approach. The fundamental macroeconomic variables that
they consider include the ratios of exports and imports to GDP, the CPI inflation rate,
the ratio of government surplus to GDP, the logarithm of the country s credit ratings,
the ratio of tax revenues to GDP, the growth of the monetary aggregate M1, and the
ratio of stock market capitalization to GDP. Their panel regression outcomes suggest
that several variables including imports, exports, inflation, market capitalization,
dividend yields and price-to-book ratios significantly affect a country s exposure to
world market risk. When these variables are lagged, they may be useful as predictors
of world market risks.
Using the one-factor model, Harvey (2000) examines average returns and risk in a
sample of 28 emerging markets and 19 developed markets. He concludes that risk
measures implied by asset pricing theory, in particular world beta and coskewness,
work reasonably well in capturing a cross-section of average returns, but many
19
emerging markets appear to be affected by total risk measures like variance and
skewness. He suggests that the reason is that these markets not completely integrated
in world capital markets.
To account for expected returns across countries, a number of authors have
developed international pricing models to approximate country risk exposures and to
determine whether variation in expected returns on the world market portfolio
induces significant hedging motives, which are reflected in average returns across
countries (see Hodrick et al., 1999). According to Ferson and Harvey (1999b), the
international equity market exposure has taken two directions: (1): sensitivities to
economic variables are used in the framework of a traditional, multi-beta, capital
asset-pricing model (CAPM) to explain the time series and cross-section of expected
returns, and (2) the fundamental characteristics approach of Fama and French (1992)
has been implemented in an international context. Harvey (1995a) uses factors such
as world inflation, world GDP, world oil prices, and a trade-weighted world
exchange rate, and finds that exposure to common world risk factors are low. In
addition, he argues that the implicit assumption of complete integration of emerging
markets is invalid. Erb et al. (1996a, b) concludes that in segmented capital markets
it is not appropriate to use the beta of the country with respect to the world market
portfolio as a measure of country risk; such methodology could lead to gross
underestimates of the cost of capital in segmented equity markets. Saleem and
Vaihekoski (2008) suggest that local sources of risk are important to consider when
investing internationally, especially in the context of emerging markets.
Using the fixed effects model in panel data analysis, Marshall et al. (2009) examine
the explanatory power of economic variables with regard to time-varying betas
20
generated by the Kalman filter approach on five monetary and real economic
variables for each country: domestic interest rates, domestic money supply,
Consumer Price Index (CPI), foreign exchange rates, and industrial production; and
one influential global economic variable: US interest rates. They find that country
betas are strongly associated with each nation s interest rates, US interest rates and
the CPU, and to a lesser extent with the exchange rates.
Using a multi-factor model, Ferson and Harvey (1993) examine predictability in 18
national equity market returns, using local and several global economic risk factors.
They allow both the betas and the expected risk premia to vary over time. The global
economic risk variables used are: (1) the MSCI world returns in excess of a shortterm interest rate; (2) the trade-weighted US dollar prices of the currencies of 10
industrialized countries; (3) the unexpected component of a monthly global inflation
measure; (4) the weighted average of the percentage changes in the consumer price
indices in the G-7 countries; (5) the monthly change in a measure of long-term
inflationary expectations; (6) the change in the spread between the 90-day Eurodollar
deposit rate and the 90-day US Treasury-bill yield; (7) the weighted average of shortterm interest rates in the G-7 countries; and (8) the change in the monthly average
US dollar price per barrel of crude oil. Their results suggest that the effect on country
returns is related to the country-specific betas. They find a small direct contribution
of time-varying betas to the model variation in expected country returns, where most
of the predictable variation that is captured by their model is attributed to movements
in the global risk premia.
Following the work of Fama and French (1998), Ferson and Harvey (1999b) explore
to what extent the price-to-book factors have explanatory power for asset returns.
21
They implement a multiple-beta ICAPM that is simplified by the introduction of the
Euro currency. Their study covers equity returns for 18 developed markets from
January 1975 to December 1997. The factors used in their study are: (1) the weighted
unexpected inflation in the G-7 countries; (2) the change in weighted expected longterm inflation in the G-7 countries; (3) the weighted real interest rate in the G-7
countries; (4) the change in the oil price; (5) the change in G-7 industrial production;
and (6) the GDP weights for the G-7 countries. Their results suggest that most of the
improvement over the CAPM comes in the model that includes the world market
portfolio and the two currency excess returns. After including the world market
factor, they find that the Euro currency risk factor is the next most important source
of global risk. They conclude that the price-to-book return offers some power in
explaining equity returns.
To examine the empirical validity of Fama and French s (1992) three-factor model,
like Ferson and Harvey (1999b) Griffin (2002) examines the usefulness of domestic,
world, and international versions of the model for five equity returns: USA, Japan,
UK, and Canada. He also examines international models that incorporate both
foreign and domestic factors. His findings do not support the notion that there are
benefits to extending the Fama and French (1992) model to a global context, but that
country-specific three-factor models are more useful in explaining average stock
returns than international models. He argues that cost-of-capital calculations,
performance measurement, and risk analyses using Fama and French-style models
are best done on a within-country basis.
2.5. Economic forces and stock markets
22
A large body of studies has established the importance of a country s economic
variables in explaining the variations in country beta measures for developed and
emerging markets. Using the Kalman filter, Verma and Soydemir (2006) investigate
whether global and local risk factors have any degree of influence on the country
betas in Latin America. They find that local and global factors have varying impacts
on the country risk of Latin America. Among the global factors, real interest rates
and inflation in G-7 countries have significant negative impact on the Latin
American country betas: the highest on Mexico, followed by Brazil, Chile and
Argentina. Among the local factors, they find that money supply and exchange rates
have a statistically significant effect on country beta, and that the effect of money
supply is most significant in the case of Mexico, followed by Chile and Brazil; the
exchange rate has the greatest impact on Mexico and Brazil and little on Argentina
and Chile. They argue that their results can be used to support the notion that Latin
American markets are becoming integrated with developed markets, but at different
speeds.
Wdowinski (2004) analyses the macroeconomic factors influencing the capital
market risk in Poland for the months of 1996 2002, estimating the country beta on
major foreign stock market indices. He uses monetary factors such as interest rates
and exchange rates, and real factors such as income, labour productivity, trade
balance, and budget deficit, as explanatory variables. He finds that monetary
variables are more influential than real variables, based on the analysis of ex post
forecast errors and ex ante models of checking predictive quality. He concludes that
the integration of the Polish capital market with other European and world markets
will be strengthened by the accession of Poland into the European Union.
23
Using a Semi-Parametric approach, Jeon (2001) tests the influence of
macroeconomic influences on country risks for 14 developed countries. The variable
set includes: (1) the US term premium; (2) the US default premium; (3) the trade
weighted exchange rate; (4) the consumer price index; (5) the money market rate;
and (6) the industrial production index. The results suggest that for the majority of
countries, consumer price index, industrial production index, and exchange rate have
a significant impact on betas.
Like Ferson and Harvey (1993), Gangemi et al. (2000) explore the relevance of the
country beta approach in the Australian context, in that Australia s country beta is
allowed to vary as part of a set of Australian macroeconomic variables. Their set of
variables is generally similar to those chosen in the existing literature, particularly
Chen et al. (1986). Their analysis covers the period of January 1974 to
December1994, using monthly data, with the Morgan Stanley Capital International
(MSCI) index as a proxy of the world stock market. The outcomes of their modelling
show that only the trade-weighted index of exchange rates is linked to variations in
Australia s country risk. They note that the insignificance of the balance on current
account is particularly interesting given its prominent role in the Australian external
balance debate and its link to Australia s increasing net foreign debt.
Similar to the work of Gangemi, (2000), Tourani-Rad et al. (2006) examine the New
Zealand s country risk using the country beta approach. They allow beta to vary
according to a set of macroeconomic variables plus the world market index for the
period from September 1985 to March 2000. The macroeconomic variables are: (1)
the commodity price index; (2) the exchange rate to the US dollar; (3) the money
supply (M3); (4) the exchange rate to the Australian dollar; (5) the 90-day bill yield;
24
(6) the 10-year government bond; (7) the food price index; (8) the monetary
conditions index; and (9) the trade-weighted index. They improve the explanatory
power of the model by including a dummy variable to account for the October 1987
stock market crash. They find that the estimated regression of the model including all
ten macroeconomic variables produces no significant coefficients on the explanatory
variables, and argue that the influence of macroeconomic variables is contained in
the world market index.
Similar to the work of Verma and Soydemir (2006), Andrade and Teles (2006) build
an econometric model to evaluate the explanatory power of the macroeconomic
variables on the Brazilian country risk. The variables used are: (1) foreign currency
reserves; (2) oil prices; (3) the nominal interest rate; (4) public sector borrowing; and
(5) the public debt index. Their results suggest that monetary policy was important
throughout the period. They also find that there was an increase in the significance of
the coefficient of foreign currency reserves during the Russian crisis of 1998: right
after the crisis; a small shock on the country risk intensified.
Using the same econometric model to evaluate the explanatory power of
macroeconomic variables to country risk, Andrade and Teles (2008) investigate the
effects of monetary policy shocks on the behaviour of country risk of six emerging
economies: Argentina, Mexico, Russia, Thailand, Korea and Indonesia, and attempt
to determine the extent to which aspects of fiscal dominance, and consequently loss
of inflationary control by monetary policy mechanisms, are significant in these
emerging economies, and the situations in which they occur. The fundamental
variables used in their study are foreign currency reserves, oil prices, nominal
interest rates, and the fiscal balance. Using the time-varying approach of the Kalman
25
filter, their results indicate that monetary policy is effective during foreign exchange
crises, but is slightly ineffective during periods of normality. They do not find that
the country risk of these countries reacts to external shocks, represented by the oil
price.
Goldberg and Veitch (2010) examine the importance of economic factors in a timevarying beta model of country risk before and after the occurrence of financial
integration for South Africa s stock market over the period 1993 2008. They find
that exchange rate and gold prices are significant economic variables that induced
significant volatility in South Africa s beta during the pre-financial integration
period. They also find that after financial integration, South Africa s country beta
rises and fundamental economic factors cease to be significant in determining its
variation. They conclude that these results are consistent with an integrated financial
market.
2.6. Country credit risk ratings as measure of country risk
Country risk ratings are important for lending and investment decisions by large
corporations and international financial institutions; rating agencies provide
qualitative and quantitative country risk ratings, combining information about
economic, financial and political risk ratings into a composite risk rating. This is
particularly important for developing countries, for which there is limited
information available. Country risk ratings help developing countries to enter capital
markets and provide economic, financial, and political officials with essential tools
to assess such risks (Hoti, 2005; McAleer et al., 2009). Khoury (2003) argues that
sociological, political, and economical factors result in rankings of countries from
26
the highest to the least risky. Erb et al. (1996b) find that factors that simultaneously
influence a country s credit rating are mainly political risk, inflation, exchange rate
variability, the industrial portfolio, economic viability, and sensitivity to global
economic shocks. They suggest that the reward for credit risk is no different across
emerging and developed markets. They also find a positive relationship between
political risk indicators and market volatility in emerging markets, including Middle
Eastern markets. Heaney and Hooper (1999) argue that country-specific factors
could be important in explaining returns and risk indices (economic, financial, and
political) in capturing idiosyncratic country effects.
Using the multivariate GARCH-M, Saleem and Vaihekoski (2008) employ two types
of risk factors in the international asset pricing model: global market risk and
exchange rate risk. They assess the version of international CAPM on a Russian data
set consisting of 417 weekly observations from January 1999 to December 2006.
Using a time-varying specification for the price of world and currency risk, they find
the global risk to be time-varying and the currency risk to be priced and highly timevarying in the Russian stock market; their results show that the local risk is not
priced for the US market. They state that the results confirm the need for an
appropriate model specification that takes into accounts both currency risk and local
market risk as independent pricing factors and conclude that the low level of
international integration of the Russian market and the implied high expected returns
make the Russian market very attractive for international investors.
Using the Multivariate GARCH, Hoti (2005) analyses the degree of economic,
financial, and interdependences between the countries in the Balkan region: Albania,
Bulgaria, Greece, Romania, Serbia and Montenegro, and Turkey, using monthly risk
27
ratings data from October 1985 to April, 2005. His results show that the selected
Balkan countries were closely related in terms of the shocks to their economic,
financial, political and composite risk returns. Country spillover effects were
observed in almost every country risk return across the six countries. His results
suggest that the risk return volatility (or uncertainty) of a country is negatively
related to the shocks to risk returns for the other countries in the region.
Heaney and Hooper (1999) examine the impact of regional and world effects upon
the equity market returns of the Asia-Pacific region using a model similar to Erb et
al. (1996a, b) and Bekaert and Harvey (1997). They find that the world returns are
less important, with exceptions in the more integrated equity markets. They conclude
that the country-specific political risk indices provide very little time-series
explanatory power over equity market returns, and note that there is some variation
in these results with the choice of world returns proxy; although in general the choice
of world return proxy is not critical. They also find consistency in the relationship
between the financial risk index and equity market volatility, where decreases in the
former are generally associated with increases in the latter.
Kaminsky and Schmukler (2002) examine the cross-country and cross-security
spillover effects of rating changes by examining the effect of domestic vulnerability,
measured by the ratings of credit agencies, on the extent of international spillovers.
They use data on sovereign bond yield spreads, stock prices, and US interest rates for
16 emerging markets including East Asian, Eastern European, and Latin American
economies. Their data covers the period from January 1990 to June 2000. Other
variables used are stock prices, US interest rates, and credit ratings. Using a fixed
effects model in panel regression, they find that the actions of rating agencies not
28
only affect the instrument being assessed but also have spillover effects across
financial markets and countries; furthermore, ratings changes trigger widespread
market instability during times of volatility, suggesting that they may act like a
wake-up call or a signal that directs investors towards a bad equilibrium. They also
find that ratings may contribute indirectly to financial market instability if the ratings
provide additional information on the fundamentals of each country. When
Kaminsky and Schmukler divide the sample into vulnerable and non-vulnerable
countries according to ratings, they find that vulnerable economies are the ones that
react most strongly to adverse world monetary shocks.
2.7. Stock market crashes and the integration of financial markets
The processes of globalization, liberalization, and deregulation of financial markets
have increased the correlation between stock markets over time, and this has led to
increased correlations in periods of financial crises (Bekaert & Harvey, 1997; 2003).
In the last two decades, several emerging economies have undertaken a significant
relaxation of capital controls, and investors have experienced greater accessibility to
foreign markets. The implication of this trend away from segmentation is that it
necessarily involves a change in the market s risk-return profile, both in terms of the
risk factors themselves, and of the risk exposures and/or risk premia (Harvey,
1995a). Cai et al. (2006) investigate the exposure of country-level conditional stock
return volatilities to the conditional global stock return volatility for seventeen
developed and seventeen emerging nations, sampled over the period of March 1993
to March 2000. They find that all the countries included in the study exhibit
significant and positive systematic exposures to global volatility. In addition, they
find that the emerging market exposures are considerably higher than of developed
29
market exposures.
Recent episodes of sharp declines in international financial markets have drawn
investors and academics attention to the consequences of international crises and
stock markets crashes. Harvey (1995b) notes that international investors care about
the correlation of international stock markets, to diversify risks by investing in
foreign stocks and reduce the total portfolio risk; high/low correlation among stock
markets will provide more/less risk when unexpected shocks happen. Harvey
(1995a) tests whether adding emerging market assets to a diversified portfolio
enhances the investment opportunity set, finding that by adding the emerging market
assets, the global minimum variance portfolio and the standard deviation
significantly reduce. He finds that most of the correlations between developed and
emerging markets are low and many are negative; some have effectively a zero
average correlation with developed markets. However, Harvey indicates that the
behaviour of the correlations and the relation between the correlations and market
betas is important because some observers interpret increased magnitude of
correlation as evidence of increased market integration. However, Harvey adds that
there is no necessary link between correlation and integration. A country can have
zero correlation with the world market and be perfectly integrated into the world
capital markets. Harvey notes that the low correlation could be caused by the
weighted average of the firm betas in the country index equalling zero. But,
importantly, the correlation measure, the ratio of volatilities, and the conditional
betas are not sufficient to make inference about the degree of integration in these
capital markets.
The degree to which international diversification can reduce risk depends upon the
30
relations among world stock returns: if these are not correlated, then international
diversification could eliminate risk. On the other hand, if world stock returns are
perfectly correlated, international diversification could not eliminate risk (Malliaris
& Urrutia, 1995). Bekaert et al. (2005) argue that negative news regarding world or
regional markets may increase the volatility factor more than positive news, and
hence lead to increased correlations between stock markets. They show that the
increased return correlation between two countries during a period of crisis could
simply be the consequence of their exposures to a common factor.
There has been a substantial number of works in recent years looking at the degree to
which stock markets are, or are becoming, integrated. If markets are more integrated,
we would expect to see their indices display common trends (Dickinson, 2000).
Early studies that examined levels of integration generally found that emerging
markets are partially integrated (Bekaert & Harvey, 1995). Bilson et al. (2001) argue
that the assumptions of both perfect integration and perfect segmentation are
unrealistic since global and local factors can be important in determining equity
return variation. Harvey (1995c) suggests that a country s integration level will
determine the relative importance of global and local factors; high/low integration
should result in high exposures to global/local factors.
There appears to be some room in emerging stock markets for diversification
benefits for international investors; however, these opportunities are shrinking
following market liberalization policies that facilitate greater integration, and country
betas have generally increased, reducing the potential for diversification benefits
(Shamila & Shambora, 2008). Several studies have examined the process of
liberalization of emerging markets, including Gangemi et al. (1999), who state that
31
liberalization has facilitated the integration of capital markets and created closer
links between markets. Thus national stock market movements have come to more
closely resemble global stock market movements. They state that the covariation
between national stock markets implies that country betas, which depend on the
covariation between the national stock market and the global market, are also likely
to exhibit a mean reversion tendency toward a grand mean of unity. Nguyen and
Bellah (2007) consider that the liberalization of stock markets is beneficial to
emerging countries in that it allows for international risk sharing between domestic
and foreign investors through capital market integration. However, liberalization
may be harmful to stock markets in newly liberalized countries: they state that
foreign trading and free capital mobility resulting from liberalization policies may
increase stock market volatility and instability, leading to market crashes.
There has been extensive research on international stock market crises. In the last
three decades, global financial markets have coped with several economic and
financial crises such as the stock market crash of 1987, the Mexican crisis in
December 1994, the Asian crisis in July 1997, the Russian crisis in August 1998, the
Brazilian crisis in 1999, the Turkish crisis in 2001, and Argentinean currency crisis
in 2002 (Dungey et al., 2002). The global financial crisis of 2008 has been the latest
series of global shocks that triggered a series of panics, and caused recession in the
global economy, where more economies
advanced and emerging
were caught.
Equity markets during global shocks are the first to respond, with massive plunges,
as part of financial innovation and linkages. During turbulent times, financial
markets interactions and inter-linkages play a crucial role, as the high level of
integration between countries are the major source of spillover effects that lead to
32
extreme volatility in the world s financial markets (Bernadette & O Callaghan,
2009).
Jayasuriya et al. (2009) and Saez et al. (2009) state that negative shocks have a more
significant effect on the equity market volatility than positive shocks, and this has a
significant impact on systematic risk. Bekaert and Harvey (1997) show that an
emerging market is more likely to be affected by world events if it is financially
integrated with the world capital markets. They argue that the increasing impact of
global risks on volatility in some countries is consistent with increased market
integration; in integrated world markets, shocks to the world market returns affect all
countries that have none zero covariance with the world markets. Bekaert and
Harvey (1997) also note that crises in developed markets economies can seriously
harm more liberalized emerging markets in integrated world markets.
The Asian financial crisis spilled over very quickly, via Russia to Brazil and then to
other Latin American countries. Pressure on emerging markets that peg their
currencies increases as capital starts to flow out, following the harsh lessons learned
from the Asian countries that pegged their currencies. Rigobon (2003) tests for the
stability of the transmission mechanisms among 36 stock markets during the
Mexican, Asian, and Russian financial crises. Using daily data from 36 stock market
indices from Latin America, eight from South East Asia, eighteen among the most
developed nations, and three from the rest of the world covering the period from
January 1993 to December 2001, he finds that the transmission mechanism during
the financial crises as relatively stable during the 1994 Mexican crisis. He also finds
evidence of some changes in the transmission mechanism after the 1997 Asian
financial crisis.
33
Very little is known about MENA markets vulnerability to financial contagion.
Contagion refers to the cross-country transmission of aggregate shocks that hit
different countries and lead to simultaneous negative co-movements (LagoardeSegot & Lucey, 2009). Depending on its severity, financial contagion may affect
portfolio allocation strategies, the direction and magnitude of capital flows and,
ultimately, economic stability. Shift-contagion refers to the influence of excess
returns in one country on the excess returns in another country after controlling for
the effects of fundamentals. The resulting shift in market expectations leads to an
observable structural break in market linkages. Underlying mechanisms include
financial cognitive dissonance, endogenous liquidity shocks, and perception of
political risk (Forbes & Rigobon, 2002; Lagoarde-Segot & Lucey, 2009).
Kamin et al. (2001) have developed an Early Warning System (EWS) approach to
identify the separate contributions of domestic and external variables to the
probabilities of crisis. They use the following domestic variables: (1) the deviation in
real GDP growth from its average in prior three years; (2) the ratio of public sector
fiscal deficit to GDP; (1) the three-year growth in ratio of M2 to international
reserves; (3) the deviation of external debt-to-exports ratio from long-term averages;
(4) the deviation of ratio of international reserves to short-term external debt from
long-term averages; (5) deviation in real effective exchange rates; (6) deviation in
export growth from its average in the prior three years; (8) the current account
balance to GDP ratio; (7) the ratio of foreign direct investment (FDI) to GDP; (9)
and the change in percentage growth in terms of trade from prior year. The externalshock variables were: (10) the change in percentage growth in terms of trade from
the prior year; (12) real US 3-month Treasury bill interest rates; and (13) changes in
34
industrial countries GDP growth from the prior year. Using a Probit model for 26
emerging countries from 1981 to 1999, they find that relative to domestic factors,
adverse external shocks and large external imbalances contribute little to the average
estimated probability of crisis in emerging market countries, but account for many of
the spikes in the probability of crisis estimated to occur during actual crisis years.
They conclude that on average, domestic factors have tended to contribute to much
of the underlying vulnerability of emerging market countries; adverse swings in
external factors may have been important in pushing economies over the edge and
into financial crisis.
Horta et al. (2008) adopt a copula methodology that allows the comparison of
dependence relationships in a period of relative financial stability, referred to as the
pre-crisis period, and in a turbulent phase, referred to as the crisis period. The precrisis period begins in January 2005 and ends before the burst of the subprime
bubble, assumed to have occurred on 1 August 2007. They use daily closing data for
the USA and the other G7 countries, plus Portugal. Their results suggest that markets
in Canada, Japan, Italy, France and the UK display significant levels of contagion,
but that this is less relevant in Germany. They observe that Canada appears to be the
country where the highest intensity of contagion is observed.
Flavin et al. (2008) develop a model that allows simultaneous testing for the
presence of both shift and bi-directional pure contagion. Their analysis concentrates
on the emerging financial markets of East Asia: Korea, Indonesia, the Philippines,
Singapore, Taiwan and Thailand, using weekly returns from July 1997 to February
2008. Their results showed a great deal of instability in these markets, with
widespread evidence of pure contagion in both directions. In addition, they find that
35
there is less evidence of shift contagion, with the transmission of common shocks
unchanged between regimes for the majority of countries.
The effects of contagion from financial crises are an issue that has gained renewed
attention with the recent onset of financial crisis in Asia, Russia and Brazil. Since
many Asian countries have been mired in recession since 1997, the impact of the
financial crisis in Asia continues to be of particular interest and importance. As the
recent experiences in those countries or regions mentioned above have shown, a
currency crisis can be associated with severe slumps in the equity markets, with the
international transmission of a financial crisis being consistent with the increasing
pervasiveness of economic and financial linkages across markets. In particular, any
small panic-trading signal can become a precipitating factor for investors, leading to
an overall loss of confidence and an increase in the perceived risk of holding a range
of investments in different equities. As the theoretical models of equity market
behaviour suggest, information asymmetry can cause herding behaviour or
contagion, pushing these countries into poor equilibrium and financial distress
(Chung, 2005).
Lagoarde-Segot and Lucey (2009) investigate the MENA stock markets
vulnerability to shift-contagion, focusing on the recent major financial crises on the
MENA markets, namely the Asian financial crisis, the Russian financial crisis, the
Brazilian financial crisis, the Turkish financial crisis, the September 11, 2001
attacks, and the Argentinean financial crisis. They find a positive association
between market development and financial vulnerability in the MENA region
MENA equity markets sporadically move together in synchronized fashion during
36
periods of external stress (Asian financial crisis, September 11, 2001 attacks, and the
Brazilian financial crisis). They also find that the equity markets of Turkey, Israel
and Jordan were the most vulnerable over the 1997 2009 periods, followed by
Tunisia, Morocco, Egypt, and Lebanon; they conclude that MENA-based
diversification strategies may be relatively inefficient in periods of global volatility.
They also observe increasing time-varying vulnerability levels in the MENA region,
with a possible structural break after the recent global financial crisis, suggesting that
the presence of mechanisms aimed at limiting price fluctuation in Israel, Turkey,
Tunisia, and Lebanon do not seem to prevent financial contagion.
Khallouli and Sandretto (2010) investigate if any of the developing stock markets of
MENA region have been affected by the financial contagion of the 2007 US
subprime crisis. They use daily stock market assets returns data ranging from
February 2007 to March 2009 in a Markov-Switching EGARCH model with timevarying transition probabilities in order to test contagion from the US stock market
(ground-zero country) to eight MENA stock markets. They argue that their model
can capture the existence of pure contagion, not shift contagion. Their model was
able to capture the persistent phase of a bear market in all the sample MENA stock
markets with the exception of Jordan; this bear market phase started in September
2008, coinciding with the third phase of the current global financial crisis. In
addition, they find both mean and volatility contagion in the Bahrain and Egypt stock
markets, plus mean contagion in Morocco and Turkey, while the contagion to Oman
and Dubai is explained only by US volatility (volatility contagion). They also find
evidence in line with that of Lagoarde-Segot and Lucey (2009) concerning the
contagion vulnerability of Morocco, Egypt and Turkey. Moreover, they claim to
37
have discovered, for the first time, the vulnerability of the GCC region to contagion
from the global financial crisis. They conclude from their results that the magnitude
of the current financial crisis is affecting a large variety of markets, regardless of
their characteristics (size of the market, liquidity, stage of liberalization, level of
international financial integration and so on). On the other hand, they claim that the
current financial crisis is not related to the type of specialization of the countries
under consideration (whether they are oil-exporting countries or more diversified
economies). They conclude that MENA countries should further strengthen their
domestic financial systems by enhancing prudential regulations and supervision, and
continue their efforts and progress in deepening capital markets to improve liquidity,
and in diversifying sources of financing.
Lagoarde-Segot and Lucey (2006) investigate vulnerability to financial contagion in
a set of expanding emerging markets of MENA during seven episodes of
international financial crisis. They find evidence of contagion for Israel and Turkey
during the Asian financial crisis in 1997, Egypt and Tunisia during the Russian
financial crisis and the Brazilian financial crisis in 1998, Israel, Morocco, and
Lebanon during the Turkish financial crisis in 2001, Tunisia, Turkey, and Jordan
during the September 11, 2001 attacks, and Morocco, Turkey, Jordan, and Lebanon
during the Enron scandals of 2000. They conclude that the time-increasing number
of countries suggests that the probability of being affected by contagion grows as
markets develop in size and liquidity and become more integrated with the world s
markets.
Lim et al. (2008) investigate the effects of the 1997 Asian financial crisis on the
market efficiency of eight Asian stock markets, Hong Kong, Indonesia, Korea,
38
Malaysia, Philippines, Singapore, Taiwan, and Thailand, by applying a rolling bicorrelation for the three sub-periods of pre-crisis, crisis, and post-crisis. Their results
demonstrate how the Asian financial crisis affected the efficiency of most Asian
stock markets, with Hong Kong being the hardest hit, followed by the Philippines,
Malaysia, Singapore, Thailand, and Korea. However, they found that most of these
markets recovered in the post-crisis period in terms of improved market efficiency.
They also found evidence of nonlinear serial dependencies for equilibrium deviation
resulting from external shocks; from this they argue that higher inefficiency during
the crisis meant that investors tended to overreact not only to local news, but also to
news originating in the other markets, especially when the reports were adverse.
Arestisa et al. (2005) examine whether during the 1997 Asian financial crisis there
was any contagion from the four largest economies in the region (Thailand,
Indonesia, Korea and Malaysia) to a number of developed countries (Japan, UK,
Germany and France). They tested for contagion as a significant positive shift in the
correlation between asset returns, taking into account heteroscedasticity and
endogeneity bias. Their results show that the impact of the Asian financial crisis on
developed financial markets was small. They found some evidence of contagion, in
particular from Japan, which, although the main international lender in the region,
drastically cut its credit lines to other Asian countries in 1997.
Yang et al. (2006) examine the long-run price relationship and the dynamic price
transmission among the US, German, and four major Eastern European emerging
stock markets, with particular attention to the impact of the 1998 Russian financial
crisis, using daily data covering the period from January 1995 to June 2002. Using
the Johansen cointegration approach and the generalized VAR, their results show
39
that both the long-run price relationship and the dynamic price transmission were
strengthened among these markets after the crisis; and that the influence of Germany
became noticeable on all Eastern European markets after, but not before the crisis.
Saez et al. (2009) analyse the transmission of the 14 exogenous shocks from the 14
emerging markets economies to 15 mature economies equity markets, covering 12
Euro area countries, USA, and Japan. They find that shocks in the emerging markets
have not only a statistically but also an economically significant impact on the global
equity markets. They find that emerging markets economies influence global equity
markets just about as much in good times as in bad times: that is, during crises or
periods of financial turbulence. They document a large degree of heterogeneity in the
transmission of emerging markets shocks to individual countries equity markets,
stressing different degrees of financial exposure.
Beirne et al. (2009) examine the contagion running from mature to emerging equity
markets using 41 emerging market economies form Asia, Europe, Latin America, the
Middle East, and South Africa, using daily data from September 1993 to March
2008. Using the trivariate GARCH model, their results suggest that spillovers from
mature markets do influence the dynamics of conditional variances of returns in
many local and regional emerging stock markets. They also document evidence of
changes in the spillover parameters during turbulent episodes in mature markets.
They conclude that changes in conditional correlations between mature and emerging
markets during turbulence in the emerging markets appear to be driven in many
cases by a relatively large rise in mature market volatility, with beta coefficients
either unchanged or lower compared to those of non-turbulent periods.
40
The time variation in correlation suggests that the sensitivity of many emerging
markets to world markets is increasing. If the risk of these markets is increasing, this
has implications for international investment strategies: a low correlation of
emerging markets with the world portfolio is important if global investors are to
enhance portfolio performance and to reduce their overall portfolio risk (Harvey,
1993). Bekaert and Harvey (2002) state that during financial crises, the world
responds with higher volatility, and this will increase the correlation with global
capital markets: the local and global volatilities governed by increased correlation
will increase the dynamic exposure of the country beta. According to Forbes and
Rigobon (2002), stock markets of very different sizes, structures, and geographic
locations can exhibit a high degree of co-movements after a shock from one market
to another. They argue that a high degree of co-movements suggest the existence of
mechanisms through which domestic shocks are transmitted internationally.
Bekaert and Harvey (2003) state that the correlation and beta of emerging markets
with the world market has increased since equity market liberalizations, even after
introducing control variables. They note that when correlation increases the benefit
of diversification decreases, but they argue that the current correlation is still
sufficiently low to provide important diversification benefits. Arouri et al. (2009)
conclude that market liberalization for emerging stock markets and the resultant
increase in globalization do increase the co-movements of capital markets around the
world, and thus reduce significantly the benefits of international diversification. They
argue that the co-movements of stock markets in times of crisis can be considered
evidence of contagion effects.
Emerging markets are becoming more sensitive to the international co-movements of
41
financial markets and tend to increase during crises; and the dynamic price
transmission was strengthened among these markets recently (Lim et al., 2008).
After the stock market crash of October 1987, there was considerable interest in
empirical and theoretical investigations of the links between capital markets. A
decade later, the financial crisis in Asian markets renewed this interest. This issue is
an important concern for investors, since greater integration among world markets
implies stronger co-movements between markets, reducing opportunities for
international diversification; furthermore, market co-movements can lead to market
contagion as investors incorporate into their trading decisions information about
price changes in other markets, in an attempt to form complete information set
(Fernandez-Serranoa, and Sosvilla-Riverob, 2001). Bekaert and Harvey (1997)
consider that as an economy becomes more integrated with the world portfolio, the
country s volatility components increase significantly. Fontaine (2005), thinking
along similar lines, states that as international co-movements tend to increase during
crises, emerging markets become more vulnerable to large shocks than do developed
markets, due to their integration efforts. Dowling and Verbiest (2002) claim that the
world seems to be more crisis-prone now than it was couple of decades ago;
globalization and the array of new financial instruments have contributed to greater
vulnerability in the world economic system.
International financial markets have grown rapidly during the recent decades, and
their links have increased significantly. This increase in international financial
linkages is of major importance for research in macroeconomics and finance, as
international financial linkages can play an important role in the propagation of
shocks from one economy to another (Park & Kim, 2007). Soydemir (2000) argues
42
that as emerging markets open their markets to achieve more financial integration,
they are more likely to be vulnerable to external shocks. Goldberg and Veitch (2010)
argue that country betas change systematically across liberalization episodes: it is
likely that the beta of a liberalizing country exhibits a structural break, changing to
reflect the higher correlations of local financial market returns to movements in the
world index after liberalization.
A large number of studies have concluded that the co-movements of stock returns in
most emerging and developed markets have increased, particularly after financial
crises. Fernandez-Serranoa and Sosvilla-Riverob (2001) examine the linkages
between stock markets in Asia, using daily data from 1977 1999 in the cointegration
technique, which allows for structural shifts in the long-run relationship. They
suggest that these shifts may be due to changes experienced in the countries during
the sample period, such as a movement from relative isolation to their stock markets
being opened up and exchange rates floated, or turmoil in the equity and foreign
exchange markets. Their focus on geographic integration, examining long-run
linkages among Asian stock markets, reveals that a cointegration test without
structural breaks provides no evidence of a long-run relationship between the Asian
and the Japanese stock markets. When they introduce the possibility of structural
breaks, they find a strong evidence in favour of such relationship between the
Taiwanese and Japanese indices from October 1987, while some marginal
cointegration is found between Singapore and Japan until February 1992 and
between Korea and Japan from April 1987. They conclude that it is possible in the
Asian region to derive portfolio diversification in the short-run but not in the longrun; and that the gains from international diversification for investors with long
43
holding periods may be limited. Leong and Felmingham (2003) assess the degree of
interdependence among the stock markets of Singapore, South Korea, Japan, and
Taiwan using daily data from July 1990 to July 2000. They find that the correlations
of these five indices have strengthened following the Asian financial crisis of 1997.
Gupta & Guidi (2012) test the evidence of the long-run relationship, the dynamic
comovements, and the test of contagion between the Indian and stock markets of
Hong Kong, Japan and Singapore. The cointegration of Johansen's and Gregory and
Hansen's tests reveal that the Indian stock market show no sign of long-run
relationship. When using the DCC-GARCH model, the results indicate that the
conditional correlations between the Indian and the Asian stock markets have risen
dramatically during the periods of September 11, 2001 attacks as well as the global
financial crisis. However, they found that after both crises, the correlations returned
to their initial levels. Similarly, Kenourgios & Padhi (2012) use the cointegration and
the DCC-GARCH to examine the contagion effect of the Asian crisis, Russian and
Argentine turmoil, and the subprime crisis of 2007, using data from equity and bond
markets in Latin America, Asia, Europe, Middle East and Africa, as well as USA and
two global indices. The cointegration results show that the long and short run
dynamics exist only for stock markets during the Russian and Asian crises, for both
stock and bond markets during the subprime crisis, while the Argentine crisis has no
impact on any of the examined financial markets. The DCC-GARCH results show
that the emerging markets are vulnerable to both emerging and global shocks
Especially in Asia and Latin America.
Siklos and Ng (2001) investigate stochastic trends among national stocks of the
USA, Japan, and five Asia-Pacific stock markets: Hong Kong, Korea, Singapore,
44
Taiwan, and Thailand from January 1976 to August 1995. Their results reveal that
stock market integration is largely a feature of the post-October 1987 stock market
crash, and that it intensified during the 1990s. They conclude that stock markets
seem to be attached to each other, but only after an idiosyncratic shock (such as the
1987 stock market crash or the 1990 Gulf War). They argue that the diversification
gains do persist when the longer-term (20 years or more) history of Asian-Pacific
markets is considered. They also found that before the 1987 crash, the seven stock
markets analysed did not share a common trend, which means that investors in the
Asian-Pacific stock markets did not exploit diversification opportunities; however,
after the 1987 crash, investors in Asian-Pacific stock markets appeared to behave as
if trends in the US and Japanese stock markets as well as those in neighbouring stock
markets did influence their markets.
Neaime (2002) investigates the integration of seven MENA markets with the US,
UK, and French stock markets, using the Johansen cointegration test. The MENA
stock markets are Bahrain, Kuwait, Saudi Arabia, Jordan, Egypt, Morocco, and
Turkey. His empirical results show that the stock markets of Egypt, Turkey, Jordan
and Morocco have matured and are cointegrated with world financial markets. He
also finds strong sensitivity to unidirectional shocks from the USA and UK in
MENA countries.
Narayan and Smyth (2005) examine whether the New Zealand equity market is
integrated with the equity markets of Australia and the G7 economies, applying both
the Johansen (1988) and Gregory and Hansen (1996) approaches to cointegration for
the period of January 1975 to April 2003. The results of applying the cointegration
test without a structural break, suggest no evidence of a long-run relationship
45
between the New Zealand stock market and the other stock market indices. However,
when structural breaks are introduced, there is some evidence of cointegration
between the New Zealand and US stock markets. They argue that when modelling
long-run relationships between stock price indices, it is important to consider the
effects of structural breaks rather than just using the conventional specifications.
Maghyereh (2006) investigates interdependence among the daily equity market
returns for four major Middle Eastern and North African (MENA) emerging markets,
Jordan, Egypt, Morocco, and Turkey, taking daily data from November 1997 to
December 2002 and applying the Vector Autoregression (VAR) model. The
empirical evidence indicates that none of the MENA markets is completely isolated
and independent, although the links between them are relatively small. Maghyereh s
results suggest that the degree of sensitivity of MENA markets to shocks is related to
their degrees of openness. He notes that his results are highly consistent with existing
literature, which argues that markets that are more liberalized and have fewer
restrictions and regulations on international capital flows are more affected by other
markets.
Measuring the risk of financial returns has been the centre of attention of several
studies. The subject that has become particularly relevant in light of recent
worldwide events is the long-term effect of political instability on the fluctuations of
stock markets (Fernandez, 2008). The accurate measurement and control of financial
risk in international markets is crucial to institutions exposed internationally. Over
recent years Asian markets have been characterized as particularly risky, with
extremely volatile returns on equity markets (Pownall & Koedijk, 1999).
46
Choudhry (2005b) examines the impact of the Asian financial crisis on the risk of
selected Asian firms. He indicates that when stock returns are more volatile, market
risk is expected to be larger, with greater stock price responses for firms that have
greater exposure to the source of risk. His results indicate that the Asian financial
crisis in 1997 may have increased the market risk of individual firms in the region.
McKenzie et al. (2000) test the integration of Australian industrials relative to the
world market. They generate the time-varying estimates of 24 industrial betas
relative to the Australian and world market indices, using the Kalman filter model,
and find that the conditional industrial betas significantly increase at the time of the
1987 crash when they are measured relative to the world market, but not when the
domestic market index is used. They claim that whether the integration hypothesis
holds depends on the choice of benchmark.
Pownall and Koedijk (1999) investigate the implications of non-normality for risk
management in general, and the estimation of Value-at-Risk using daily data from
January 1993 until January 1998 for the IFC Asia 50 index. Their results show that
the use of an additional parameter to account for additional downside risk can
provide a more accurate tool for risk management. Johansson (2009b) analyses the
time-varying beta using the bivariate GARCH model for the greater China equity
markets: the Shanghai composite, the Shenzhen composite, the Hong Kong, and the
Taiwan weighted indices, using weekly data for the period from April 1991 to
December 2007 and using the MSCI world index as a proxy for the market portfolio.
His results indicate that the Shanghai and Shenzhen markets have a very low average
systematic risk when measured against the world market, but is mean reverting. He
notes that the Hong Kong and the Taiwanese stock markets experienced a strong
47
increase in systematic risk during the Asian financial crisis in 1997, indicating that
the market revalued the risk at that time.
Johansson (2009a) in his study on time-varying country beta has documented a
significant increase in the country beta of Thailand following the eruption of the
Asian financial crisis in 1997. He concludes that the way the market perceives the
riskiness of investing in the Thailand market country risk has changed considerably
over the years following the crisis. Lie and Faff (2003) investigate the potential
impact of the October 1987 crash on the global industry portfolios betas. When
examining the stability of global industry betas over the sample period in the light of
the 1987 crash, their results show that 15 of the 34 industries produced a statistically
significant crash dummy.
Miller (2006) estimates the daily time-varying country betas for eighteen emerging
equity markets using the rolling regression model. He finds that all country betas
rose considerably during the Asian financial crisis and the country betas for most
countries peaked around the time of the Russian financial crisis. Marshall et al.
(2009), who ran a rolling window of fifty-two weeks for the time-varying country
betas for 20 emerging equity markets from Latin America, Middle East, Eastern
Europe, and Asia using the Kalman filter, the DCC-GARCH, and Schwert and
Seguin (1990) models, find that all countries exhibited time variation in their beta
coefficients and most of countries exhibited an increase in their country risk in 1997
(reflecting the Asian crisis) and in 2008 (reflecting the global credit crisis).
Goldberg and Veitch (2002a) examine the contagion effects in Argentina s country
risk by assessing the time variation of the beta for Argentina using monthly data
48
from January 1992 to September 1999. Their assessment focuses on how the beta for
a country following a currency board (a hard fixed exchange rate regime) is
affected by changes in the exchange rate behaviour of its trading partners. The
fundamental variables included in their study are the CPI, the exchange rate of
Argentina to the US dollar, the money supply, reserve money, and international
reserves, as well as the exchange rate for Brazil, the exchange rate for Chile, and the
exchange rate for Mexico. Using the country beta model, they find that the contagion
effects of currency crises play an important role in the time variation of Argentina s
country beta, and conclude that exchange rate surprises of the peer countries Brazil
and Mexico, and not Argentine macroeconomic variables, are the economic factors
that matter for variations in Argentina s country beta.
Following their work on Argentina, Goldberg and Veitch (2002b) investigate the
importance of economic factors in a time-varying beta model of country risk for
Mexico, using monthly data period from January 1990 to September 1999. The
fundamental variables include inflation, exchange rate, reserve money, net exports,
and international reserves. They find that exchange rate surprises are the main
determinants of how Mexico s country beta varied over the period of 1990 to 2000.
They also find that the type of exchange rate regime, managed float versus floating
rates, affects the influence of exchange rate surprises on Mexico s country risk, and
that Mexico s country risk trends are lower over the managed rate period (1990
1994), but shift up and become more volatile in the floating period (1995 1999).
Arouri et al. (2009) investigate the co-movements of emerging Latin American
markets with the global market, proxied by the Morgan Stanley Capital International
(MSCI) world stock market index over the period January 1985 to August 2005. The
49
Latin American emerging markets are Argentina, Brazil, Chile, Colombia, Mexico,
and Venezuela. They find a clear upward trend in the time-varying correlation from
1994 and onwards as a result of the liberalization of the Latin American markets.
They also find that using the Bai and Perron structural break test, reveals structural
changes in conditional correlation following the Asian and Brazilian financial crises
in 1997 1998 and, to a lesser extent, the stock market crash in 1987 and the Latin
American market crises in 1994 and 2001.
Groenewold and Fraser (1999) estimate the time-varying beta for 23 Australian
industry portfolios using monthly data for the period of 1979 1994, using recursive
estimation, rolling regression, and the Kalman filter approaches. They assess the
properties of betas by testing the series against a time trend and the structural break
of the October 1987 crash. Their results show considerable variability of betas over
time when accounting for the structural break dummy.
2.8. Financial markets and geopolitical uncertainty
Empirical studies have come to the conclusion that exposure of emerging equity
markets are increasingly sensitive to global sources of risk (see Harvey & Zhou,
1993; Bekaert & Harvey, 1997). Recently, empirical finance literature has paid
increasing attention to the impact of socio-political events on stock market
behaviour. In particular, major political events such as war and terrorist attacks are
reflected in asset prices (Athanassiou et al., 2006). Chen & Siems (2004) argue that
the prices of individual stocks reflect investors hopes and fears about the future, and
that stock price movements can generate a tidal wave of activity. Because of stock
market liquidity, unforeseen disastrous occurrences like terrorist attacks or military
50
invasions can have serious implications for stocks and bonds. Decisions to buy and
sell can quickly, easily, and inexpensively be reversed. He argues that when
information about a cataclysmic event becomes known, investors often flee in search
of safer financial markets.
Blass et al. (2004) examine fluctuations in the risk premium on Israel s sovereign
debt traded in the USA between 1996 and 2000. They find that during this period,
Israel s risk premium was affected predominantly by global events, most notably the
crises in Asia and Russia, while domestic and regional events like the peace process,
political changes, terrorist attacks, and economic reforms had a miniscule immediate
impact on the risk premium: only in the year 2000 were Israeli bond prices more
affected by internal affairs, as a result of dramatic events in that year; possibly due to
the absence of major global emerging-market crises.
Haddad and Hakim (2008) estimate the determinants of sovereign risk in five MENA
countries: Egypt, Lebanon, Morocco, Tunisia, and Turkey; and two non-MENA
countries: Brazil and South Africa; in a panel setting using cross-sectional and time
series data on credit spreads derived from Eurobond issues. They attempt to test
whether a fundamental repricing of risk in specific MENA countries has occurred as
a result of turbulent events such as the murder of Lebanese Prime Minister, Rafiq
Hariri, and Israeli Palestinian conflicts. They find that the Sovereign risk have
increased due to the prolonged war in Iraq, the political unrest in Lebanon, the
tension between Israel and Palestine, and the spectre of an attack on Iran. They
conclude that these events have significantly increased business uncertainty in
several countries in the MENA region.
51
Mei and Gau (2004) examine the impact of political uncertainty or risk on the recent
financial crises in nine emerging equity markets. Using a combination of probit and
switching regression analysis, they find that there is a significant relationship
between political election and financial crisis, after controlling for differences in
economic and financial conditions. They also document increased market volatility
during political elections and transition periods; they find some evidence that
political risk is more important in explaining financial crisis than market contagion.
They conclude that political uncertainty could be a major contributing factor to
financial crisis.
Zeileis et al. (2003) examine multiple structural changes in oil prices using the Bai
and Perron (2003) model. They find that a number of breakpoints coincide with
historical, political or economic events, which might have caused these breaks: such
as the Arab oil embargo after the Yom Kippur war, the start of the Iranian revolution
followed by the war between Iran and Iraq in 1979, and a joint effect of various
minor events such as a worldwide slowdown of demand for oil, Great Britain,
Norway and Mexico becoming major suppliers in international oil markets, and
internal quarrels in the OPEC cartel which led Saudi Arabia to increase its
production.
Chen and Siems (2004) assess the effects of 14 terrorist/military attacks dating back
to 1915 on US capital markets, including the September 11 attacks and Iraq s
invasion of Kuwait in 1991. Using an event-study methodology, they find that US
capital markets are more resilient now than in the past, and recover sooner from
terrorist attacks than other global capital markets do.
52
Malliaris and Urrutia (1995) assess the responses of international equity prices to the
Persian Gulf crisis of 1991. Their outcomes show that the Persian Gulf crisis has had
a negative effect on equity market prices, and conclude that the Gulf war had a
greater negative impact on European, Asian, and Australian markets than on
American and Canadian markets. They find that national stock markets seem to have
responded to all the crisis periods, according to each country s dependence on oil;
from this they argue that oil, its prices, production, and distribution, has significant
consequences not only as an industry-specific effect but as a global effect.
Erdal and Gunduz (2001) investigate the interdependence of the Istanbul stock
exchange with the G-7 stock markets and those of Israel, Jordan, Egypt and Morocco
before and after the Asian financial crisis, using the cointegration technique. They
find only Turkey showing one cointegrating vector with the G-7 markets.
Athanassiou et al. (2006) examine the impact of external national security-related
shocks, military tension, and conflict (mainly due to Greek-Turkish bilateral
relations) on the Athens stock exchange. They include an external security-related
shock variable and a military crisis dummy in the GARCH (1,1), TARCH (1,1) and
EGARCH (1,1) volatility models. Their empirical findings show that border tensions
and the threat of military escalation that could have led to armed conflict adversely
affected the Athens stock market behaviour, and that the impact of such threats
appear to be partly responsible for persisting volatility. They argue that a stable
external security environment and the absence of geopolitical tensions are significant
factors in reducing stock market volatility.
Equity market volatility can be linked to uncertainty associated with geopolitical
concerns, but it is still very difficult to explain (Schwert, 1989). Karakatsani & Bunn
53
(2004) claim that the volatility of equity returns reflects a convolution of economic
fundamentals, technical characteristics, agent behaviour, and aspects of market
design, often confounded by environmental constraints and political interventions;
the analysis of these various drivers of volatility is crucial to understand the sources
of pricing risk. Fernandez (2007), assessing the impact of recent political conflicts in
the Middle East on stock markets worldwide and in particular, how political
instability affects long-term volatility of stock markets, finds that most of the stock
market indices in her sample experienced volatility shifts around September 11, 2001
in both regional and global markets.
2.8.1. September 11, 2001 attacks
The September 11, 2001 attacks are considered an event that affected the global
financial market in the USA, as were also the following declaration of the war on
terror and the invasion of Iraq in 2003. In the wake of the terrorist attacks on the
USA, the Bush administration launched an all-out military, diplomatic, political, and
financial campaign against all terrorists, with a global reach. As a result, stock
market investors were suddenly faced with risks beyond those associated with
normal economic or financial fundamentals. Stock prices became increasingly
sensitive to developments on the diplomatic and military fronts in the US campaign
(Jones, 2002). The 9/11 attacks and the subsequent declaration of the war on terror
were considered a major cause of political and regional instability, and created major
disruption in the global economy and the world market (Fernandez, 2006). The shortterm impact was felt in global tourism, airline industries, and financial markets.
While the global economy has recovered and is adjusting to new global realities, the
longer-term impact of heightened security risk across the world can be felt in the
54
form of higher risk premiums in asset markets as well as a shift of resources towards
dealing with terrorism (Koh, 2007). Nikkenin et al. (2008) compare the volatility
behaviour of stock markets in six different regions before and after the September
11, 2001 attacks to examine whether the impact of this event is pervasive across
regions. They find that the attacks prompted little significant decline in MENA
equity markets compared with the rest of the world.
The event ushered in a period of greater global uncertainty on various fronts,
political, economic, military and technological. In the aftermath of the attack, there
was concern that the flow of investment from major industrial countries to less
developed countries might slow down, as companies evaluated the country risk
profiles of overseas locations. The dampening of consumer and investor confidence
and the accompanying reassessment of operational and security risks led to a sharp
withdrawal from capital markets. The shrinkage of cover for terrorism-related risks
left many businesses dangerously exposed. Fears of further terrorist acts continued to
affect global asset markets in late 2001 and into 2002. In terms of its economic
impact, the event exacerbated the weakness already seen in the global economy in
2001. In the aftermath of September 11, consumer and business surveys showed falls
in the overall confidence measures in the United States and other countries, similar in
magnitude to those observed in the wake of the Iraqi invasion of Kuwait in 1990
(Koh, 2007).
Hatemi-J et al. (2005) investigate whether the September 11 attacks have
significantly changed the links between the US equity market and those of China,
Taiwan, Indonesia, Singapore and Japan, using leveraged bootstrap causality tests.
They also investigate whether the attacks caused spillover effects from USA to these
55
markets. Their data are daily prices, covering the period from March, 2001 to March,
2002; they designate a six-month period before the attacks and another after the
incident as two sub-periods of study. Their results reveal that the September 11
attacks have significantly changed the interaction of the US market with the Asian
financial markets covered in the study. They find that the USA became more highly
linked with Japan and Singapore but less with Taiwan, ceased to be linked with
China and continued not to be linked with Indonesia. They conclude that the
September 11 attacks diverted US influence to the developed markets of Japan and
Singapore and away from the less developed markets of China and Taiwan. They did
not find any contagion effects in the increased interaction of the US with Japan and
Singapore. They claim that in spite of the climate of uncertainty created by the
September 11 attacks, the US market continued to deal with other markets. They also
conclude that the results offer no explanation for which of the possible transmission
channels of contagion or spill-over was effective in the wake of equity market
collapses after September 11.
Hammoudeh and Li (2008) examine sudden changes in volatility in five Gulf area
Arab stock markets, using the Iterative Cumulative Sum of Squares (ICSS)
algorithm. They find that most Gulf Arab stock markets are more sensitive to major
global events than to local and regional factors. The 1997 Asian crisis, the collapse
of oil prices in 1998 after the crisis, the adoption of the price band mechanism by
OPEC in 2000, and the September 11 attacks are found to have consistently affected
the Gulf markets
unlike the results found by Nikkenin et al. (2008) using GARCH
models.
Like Hammoudeh and Li (2008), Fernandez (2006) considers whether the Asian
56
financial crisis and the September 11 attacks have caused permanent volatility shifts
in global markets, using the ICSS algorithm and Wavelet-Based Variance analysis.
Her results show that the ICSS algorithm fails to reveal any volatility shifts over the
period 1997 2002, whereas the Wavelet analysis provides evidence of volatility
breakpoints at the lower scales of the data. Choudhry (2005a) investigates the impact
of September 11 on the time-varying beta of selected US companies, using the
Multivariate GARCH; he finds that most firms betas were not affected by the crisis
dummy and some firms experienced a lowering of their beta after the event. His
results also reveal that the market volatility of the Standard and Poor s 500 (S&P
500) has had more effect on the beta after the event.
Richard et al. (2005) examine the short- and long-term effects of September 11 on a
comprehensive sample of stock market indices from 33 industrial and emerging
economies, using the ICAPM with the estimation method of Seemingly Unrelated
Regression (SUR) analysis. The sample consists of stock indices covering the
regions of the Americas, Europe, Asia/Pacific, and Africa/Middle East. They use
daily stock market indices for the period spanning April 2001 to February 2002, and
document statistically negative short-term stock market reactions to the 9/11 event
for 28 countries. Moreover, they find increases in the level of systematic risk for ten
stock markets, attesting to the presence of negative permanent effects emanating
from 9/11. However, they also find that many capital markets (including USA,
Canada, Japan, China, Russia, and the largest European economies) did not
experience statistically significant increases in systematic risk post-9/11. They
conclude that the decisiveness of the evidence clearly points in the direction of
resilience and flexibility in the world capital markets.
57
Kim and Gu (2004) investigate the impact of September 11 on the returns and
systematic risks of airline companies in the USA. They use weekly data before and
after the event, and apply the standard CAPM to examine changes in risks and
returns. Their results show that the stock returns and their systematic risks were
significantly different after September 11, and conclude that risk has increased
significantly since the event, regardless of the firm size.
Similar to the work of Kim and Gu (2004), Drakos (2004) investigates the effects of
the September 11 attacks on a set of airline stocks listed at various international stock
markets, using the standard CAPM. He uses daily closing prices for thirteen airline
stocks covering the period 12 July 2000 to 26 June 2002, decomposing the risk into
two components: systematic and idiosyncratic. His results show that the systematic
risks of these airline stocks have significantly increased. His results document a
major structural break in systematic risk for the majority of the airline stocks. He
also finds that volatility has dramatically increased in the post-September 11 period,
reflecting increased uncertainty surrounding the airline industry.
2.8.2. The war on Iraq
The war on Iraq is considered a major event that impacted negatively on world
economies and their stock markets, like the Gulf War of 1991. The countries
surrounding Iraq are faced with greater risk uncertainty, and the impact of the war is
of immediate concern (Shachmurove, 2003). As the war and regional instability in
the Middle Eastern region have had serious outcomes on the industries of world
economies, oil exporting countries like those of the Gulf Cooperation Council (GCC)
have gained substantially from the rise in oil prices, and this has led to structural and
58
strengthening prospects for many of the macroeconomic indicators of their
economies (Nell & Simmler, 2006; Onour, 2007; Hong et al., 2007; Al-Otaibi &
Slywester, 2007; Razzak, 2007; Soussa et al., 2008). However, oil importing
countries have suffered seriously as direct impact of the war, due to the economic
burden of oil imports that has led to increased spending on fuel relative to the GDP
rather than a relatively increased consumption of oil (Williams, 2006).1 Wolfers and
Zitzewitz (2009) stated that negative effects of the war are larger for regional
countries that depend heavily on oil. Shachmurove (2003) stated that ongoing
outcomes of the war has undermined regional stock markets and caused further
political, regional, economical, and financial risks, and affected many significant
components of regional and international economic activities.
The recent war in Iraq was accompanied by dramatic fluctuations in oil prices, which
had adverse outcomes on economies and financial markets worldwide (Weiner,
2003). Rigobon and Sack (2005) study the impact of the war on several financial
variables in the USA. They find that in the ten weeks before the start of the war, the
risk (expectation) of war explains between 13 and 63 percent of the variance of
financial variables such as the S&P 500 index, oil prices, gold prices and the US
dollar. Wolfers and Zitzewitz (2009) use the Saddam Security
2
to estimate the
expected cost of the war on Iraq, finding that the US stock market was extremely
Before the war, Jordan depended upon Iraq, not only as an export market under the UN supervision,
but more importantly for the subsidized oil supplies that gave Jordan fuel at a price it could afford.
Until 2003, Iraq was Jordan s largest export market: for example, in 2001 22% of Jordan s total
domestic exports were exported to Iraq. The country s overall trade deficit increased dramatically
from JD 1.8 billion to JD 4.5 billion, possibly due to suspended exports to Iraq, the end of heavily
subsidized fuel, dramatically rising international fuel prices, and the increase in population (from the
influx of over 800,000 thousand Iraqi refugees) who demanded additional goods; see Faris (2003) and
Saif and DeBartolo (2007) for more details.
1
An asset whose payoffs depend on the ousting of the Iraqi leader by a specific date; these securities
were traded on the Iowa Electronic Markets (Wolfers & Zitzewitz, 2009).
2
59
sensitive to changes in the probability of war; a 10% rise in probability was
accompanied by a 1.5% decline in the S&P 500. They state that in the period leading
up to the war, spot oil prices were strongly positively correlated, while equity prices
were negatively correlated. The war would lower the value of US equities by around
15 percent (or $1.1 trillion in market value of all stocks in the S&P 500 index). The
same instrument used by Amihud and Wohl (2004) finds that the likelihood of
Saddam Hussein s fall from power, as reflected in the movement of Saddam
securities , was related to the movement in stock market returns of the US Indices
and oil price fluctuations.
Fernandez (2008) studies whether the long-term volatility of worldwide stock
markets has undergone permanent shifts due to the current political instability in the
Middle East, primarily caused by the invasion of Iraq and the ongoing IsraeliPalestinian conflict, using Semi-parametric version of a Fractional Autoregressive
(SEMIFAR) model. The sample covers the period from January 2000 to June 2006.
Her study focuses on four geographic regions: the Americas, Africa/ Middle East,
Europe, and Asia/Pacific. Her findings show that the greatest impact of such political
instability occurred around the beginning of the war on Iraq, and that major
international stock markets were relatively more volatile around that time. She
concludes that volatility worldwide experienced transitory increments from June
2000 to June 2006, primarily associated with volatility clustering. Leippold and
Lohre (2010) test for dispersion effects in many companies, in a sample of 16
emerging equity markets and 15 European markets and the US, covering the period
from 1987 to 2007; they find that one of the breakpoints coincides with the dawn of
the war in Iraq in 2003.
60
Attia (2004) investigates the impact of the Iraqi invasion of Kuwait, the Gulf War of
1990, and the fluctuations of OPEC oil production companies during these events on
the stock prices of petroleum companies operating in the Gulf countries. His data
consist of monthly closing stock prices plus monthly OPEC oil production for the
period January 1988 to December 1995; he divides these into three periods: a pre
event period , an event period , and a post event period . Using multiple regression
analysis, he finds that OPEC oil production has had statistical significant influence
on the average monthly index of the multinational petroleum companies operating in
the Gulf countries during both the month of August and the overall war period. In
addition, he finds that the average stock prices of multinational petroleum companies
during August 1991 are significantly different from those during the months before
and after August.
Damar (2007) examines whether a large geopolitical event, such as the 2003 war in
Iraq, can affect the lending of foreign banks in developed countries to emerging
MENA markets, specifically Israel, Jordan, Morocco, Tunisia, and Turkey, for the
period 2000 2005. Using fixed effects in panel data, he finds the war has had a nonuniform effect on foreign banks but did discourage British and Italian banks in
particular from lending to the region. He concludes that the stability and reliability of
foreign bank credit in the face of increased geopolitical risks should be identified and
discussed, although the aggregate impact of the war in Iraq on the MENA economies
is hard to determine.
Agus and Erbil (2010) investigate the impact of instability on Turkish and Iraqi
bilateral foreign trade with regard to Turkish exports to Iraq and their fluctuations
before, during, and after periods of regional and global instability. Using the fixed
61
effects model in panel data with a set of exports in 60 sectors in the years 1980
through 2004, they find the effects of both global and regional instability to be
positive on Turkish exports to Iraq. They assert that a stable and growing economy in
the region is more beneficial to Turkey than a country crippled by war and conflict,
while conceding that Turkey may find some advantage to trade during the process of
stabilization in Iraq.
Filis et al. (2011) provide evidence that time-varying correlation of oil and stock
prices do not differ for oil-importing and oil-exporting economies. They attribute the
findings to the precautionary demand-side oil price shocks which tend to influence
oil-importing and oil-exporting countries. They conclude that the non-economic
crises trigger a stronger negative link between oil prices and stock markets and on
the economic crises or booms trigger a stronger positive link between oil prices and
stock markets. The non-economic crises (negative link) includes the Iraqi invasion to
Kuwait and Second war in Iraq in 2003, and the economic crises (positive link)
includes the Asian economic crisis in 1997 and the global financial crisis in 2008.
Paleari et al. (2005) assess the financial impact of global shocks such as the war in
Iraq on the indices of five of the most capitalized stock markets in the world, USA,
UK, France, Germany, and Italy, using event study methodology. They group the
countries that formed the coalition in the war in Iraq (USA and UK) and those that
publicly opposed the military solution (France and Germany) and left Italy as
independent case. Their analysis is carried out by considering both a one-day event
period and a two-day event period (the event day and the day before), using daily
data from 17 October 1994 to 31 July 2003. Their event methodology employs five
62
statistical tests used to check if the returns are significantly different from zero. 3
Their findings show that in the first event (Bush s statement to the United Nations
General Assembly on 12 September 2002) shows a negative return when the indices
are examined together on a one-day basis. In addition, they found that the negative
return is less significant when the indices are considered separately, especially on the
US, UK, and Italian markets. When they consider a two-day event period, they find
that the second event (the adoption of Resolution 1441 by the United Nations) had a
negative impact on the aggregate index and on a single market level, while the fourth
event (the Blix report) did not appear to have any significant impact. They also find
that the fifth event (the US ultimatum) showed a positive impact on aggregate levels
in both the one-day and on two-day event periods. When the US-led coalition forces
entered Iraq (the sixth event), the markets did not react significantly. In contrast, they
find an aggregate positive return when the coalition forces entered Baghdad (seventh
event) in both one-day and two-day event periods. At a country level, they find that
Germany and France are the only indices to show a statistically positive return with
reference to this event. Finally, they find that the official announcement of the end of
the conflict (the eighth event) did not present any effect on the indices. When they
analyse the time-varying correlation between these markets using the Kalman filter
methodology, their findings suggest that the correlation increased between Italy, UK
The first test is the standard parametric test introduced by Fama, Fisher, Jensen, and Roll (1969),
which relies on the assumption that the returns are independent drawings from an underlying normally
distributed population. The second test is the crude dependence adjustment test introduced by Brown
and Warner (1980) to overcome the cross-sectional dependence problem. The third test is the
standardized test suggested by Patell (1976) that surmounts the heteroscedasticity problem. The fourth
test is the generalized sign test developed by Cowan (1992), which compares the number of positive
sign returns over the estimation period with that over the event period. The fifth test is the rank test
developed by Corrado (1989), which involves placing in order all the returns, over both the estimation
and the event period, and testing whether the actual rank of event day returns is significantly above or
below zero.
3
63
and France but has decreased between these markets and both USA and Germany.
They attribute this behaviour to two different factors: the official position and
military exposure related to the war, where USA and UK had the higher war
exposure and Germany and France had the lowest military exposure; and the
economic interest in Iraq. They measure this factor by considering the oil imports
from Iraq to the five countries in the study, and conclude that Germany was the
country least affected by the war, whereas USA was the most affected. In addition,
they conclude that UK, France, and Italy showed an intermediate level of Iraq
Exposure which seems to explain the correlation pattern.
2.9. Conclusion
This chapter outlines and summarizes the main relevant studies on the concept of
systematic risk, country risk in quantitative and qualitative frameworks, the
responses of stock markets to geopolitical and financial crises, and the integration of
developed and emerging markets. The majority of these studies conclude that
systematic risk is time-varying at individual, industrial, portfolio and even country
level for both developed and emerging markets. The empirical evidence of these
studies suggests that time variation in volatility, systematic risk, and country risk is
meaningful to local and foreign investors.
Economic stability plays a major rule in the performance of financial markets, and a
vast number of studies conclude that financial markets around the world have reacted
negatively to major events of the last two decades. These findings have major
implications for volatility, risk and correlation with the global capital system. For
example, many studies have shown that financial markets worldwide react negatively
to geopolitical concerns and financial
contagions and show signs of extreme
64
volatility. Therefore, international investors will respond by changing their strategies
and decisions regarding their asset pricing, diversification decisions, and portfolio
allocation. In addition, many studies have concluded that the co-movements of stock
returns in developed and emerging equity markets have increased during financial
crises besides other geopolitical risks suck as wars, political unrest and terrorist
attacks.
The dramatic effects of global events vary across financial markets around the world.
The severity of the impact of troubled times on financial markets depends on the
level of integration and the market exposure of the equity market to the global
financial system. Most of the previous studies examine volatility, spillover effects,
and contagion effects in both regional and international markets; in addition, the
literature documents evidence those unstable periods represents risk aversion of
international investors. Therefore, the following four chapters will attempt to answer
the thesis questions on the impact of the geopolitical and financial crises on the stock
markets of MENA countries.
The first three chapters examine the stock market exposure of MENA countries,
more specifically, using the country beta. This endogenous variable will reflect the
level of risk aversion to external shocks such as financial crises and geopolitical
unrests. In the third chapter, the panel regression will investigate how the changes in
the state of economies in the MENA region have varying impact on the country beta.
The last chapter will address the link between the major international stock markets
and the MENA equity markets and how the link vary through time can be inferred to
indicate increased market integration and therefore increased market exposure to the
global financial system.
65
The Stability of Country Beta of MENA Markets and the
Impact of the War on Iraq4
The aim of this chapter is twofold. First, the chapter examines the stability of the
country betas of the MENA equity markets using the static International Capital
Asset Pricing Model (ICAPM). Second, it aims at examining the impact of the war
on Iraq on the country betas of MENA markets using the dynamic ICAPM. The use
of the ICAPM especially for emerging markets requires strong assumptions
especially on stock market integration; however, the use of the ICAPM is useful even
for segmented markets for dynamic hedging and investment possibilities (Johansson,
2009a).
A paper derived from this chapter has been accepted for publication in International Journal of
Economics and Finance. The contents of this chapter has benefited from the comments of two
anonymous referees.
4
66
3.1. Introduction
Empirical evidence shows that the systematic risk or beta is unstable over time in
most equity markets in the world.5 Moreover, the evidence of beta instability has
strongly influenced the asset pricing research conducted in the international setting,
where the beta risk is defined relative to a global market proxy (Brooks et al., 2002).
The concern is that the beta risk at the international level tends to be unstable and,
hence, needs to be tested and modelled (Brooks, 2003; Gangemi et al., 2000,). Given
the high volatility especially of emerging markets, it could be expected that betas in
these markets would be time-varying, and that this variation would have major
implications for international investors in these markets (Marshall et al., 2009).
The time-varying country beta has become a persistent issue in international finance,
especially as world markets have become a rapid business environment; an
assessment of this kind has become an important need for international investors
(Brooks, 2003; Brooks et al., 2002). Although the assessment of the country beta in
terms of the ICAPM has found some objections in the literature, especially for
emerging markets, substantial evidence suggests that in an international context the
country beta measure is very meaningful to foreign investors (Brooke, 2003; Brooks
et al., 2002, Johansson, 2009a, b).
Early studies by Ferson and Harvey (1993; 1995) and Harvey (1995c) have shown
that the rejection of the country beta as a measure of international beta could be
Early empirical studies have shown that the beta coefficient using the CAPM follows a stochastic
behaviour (e.g. Fabozzi and Francis, 1978; Bos and Newbold, 1984; Wells, 1994; Brooks et al., 1998;
Berglund and Knif, 1999; Brooks et al., 1992; Choudhry, 2001, 2002; Grieb and Reyes, 2001; Moonis
and Shah, 2003; Ebner and Neumann, 2005; and Park and Kim, 2007). Thus, the unconditional
CAPM has been criticized due to the static framework in which it was originally derived (Johansson,
2009a).
5
67
caused by the assumption that betas and the expected returns of equity markets are
constant over time. Johansson (2009a) argues that it is difficult to model emerging
markets betas as stable processes, even if they are not fully integrated to the world
markets. A conditional approach is more appropriate to estimate the country beta
when looking at investment possibilities in developing countries. Johansson (2009a)
also states that the time-varying country beta estimation may allow investors to apply
dynamic hedging of their portfolios more efficiently, and that estimated time-varying
country betas may be used to better understand what factors influence market risk in
different countries.
One issue in the estimation of the country beta is the extent to which the techniques
used to model the market beta in a domestic context can be readily applied to an
international context (Brooks, 2003). To answer this question, a number of asset
pricing models have been applied to the international context to estimate the country
beta over time. Koutmos et al. (1994) use the novel Schwert and Seguin (1990)
model to estimate the time-varying country betas for ten developed countries.
Giannopoulos (1995) employs the Multivariate GARCH at the international level for
a sample of 13 countries. Brooks et al. (2002) use the Kalman filter, the Multivariate
GARCH, and the Schwert and Seguin (1990) models in a sample of 17 developed
markets. Johansson (2009a) applies the multivariate stochastic volatility and the
multivariate GARCH model to estimate the country betas for 27 emerging markets.
Johansson (2009b) applies the multivariate GARCH model for the four China s
equity markets: the Hang Seng, the Shanghai composite, the Shenzhen composite,
and the Taiwan Weighted Indices. Marshall et al. (2009) assess the time-varying
country betas of 20 emerging equity markets from Latin America, the Middle East,
68
Eastern Europe, and Asia using the Kalman filter, the DCC-GARCH, and the
Schwert and Seguin (1990) models. These studies clearly show that the country betas
are not stable over time, and that accounting for the time variation in country betas
has major implications for investment and risk management, especially in emerging
markets.
3.2. The war on Iraq and equity markets
It is well documented that in terms of returns and volatility, international capital
markets react quickly and simultaneously to major events, although the timing and
magnitude of changes in stock returns and volatilities differ across markets around
the world (Nikkinen et al., 2008). This has become particularly relevant in light of
recent worldwide events, given the long-term effects of political instability on the
fluctuations of stock markets (Fernandez, 2008).
The empirical finance literature has paid increasing attention to the impact of sociopolitical events on stock market behaviour. In particular, major political events are
reflected in asset prices (Athanassiou et al., 2006). Chen & Siems (2004) assess the
effects of 14 terrorist and military attacks on US equity markets including the
September 11 attacks and Iraq s invasion of Kuwait in 1991, finding that the US
equity markets are more resilient now than in the past and recover sooner than other
global capital markets. Malliaris and Urrutia (1995) analyse the responses of
international equity prices to the Persian Gulf crisis of 1991. Their outcomes show
that the Persian Gulf crisis had a greater negative impact on the European, Asian, and
Australian markets than on the US and Canadian equity markets, and those national
stock markets seem to respond according to each country s dependence on Gulf oil.
69
Athanassiou et al. (2006) examine the impact of external national security-related
shocks, military tension, and conflict on the Athens stock exchange, with particular
reference to Greek-Turkish bilateral relations, using the volatility models of GARCH
(1,1), TARCH (1,1) and the EGARCH (1,1) with an external shock variable and a
military crisis dummy. Their empirical findings show that border tensions and the
threat of military escalation adversely affected stock market behaviour. They also
find that the impact of such threats appears to be partly responsible for persisting
volatility.
The war on Iraq is considered a major event that has negatively affected world
economies and global equity markets. According to Salameh (2008), the war has
impacted on global oil production capacity by creating instability in the Middle East
and increasing the risk of investing in the region. Shachmurove (2003) indicates that
the countries surrounding Iraq are faced with greater uncertainty than usual, and the
impact of the war is an immediate concern. He argues that the ongoing outcomes of
the war have undermined regional stock markets and caused further political,
regional, economic, and financial risks, and affected many significant components of
regional and international economic activities.
The impact of the war and the regional instability of the Middle Eastern region have
had serious outcomes on the industries of world economies (Soussa et al., 2008).
Wolfers and Zitzewitz (2009) argue that the negative effects of the war in Iraq are
larger for the regional countries that depend heavily on oil imports, such as Israel,
Jordan, Pakistan, and Turkey. Fernandez (2008) studies the volatility of worldwide
stock markets, primarily the result of the invasion of Iraq in 2003. Her findings show
that the greatest incidents of political instability coincided with the beginning of the
70
war in Iraq, and that major international stock markets became relatively more
volatile around that time. Rigobon and Sack (2005) find that in the ten weeks before
the start of the war in Iraq, the risk of the war explained between 13 and 63 percent
of the variance of financial variables such as the S&P 500 index, oil prices, gold
prices, and the US dollar. Leippold and Lohre (2010) test for the dispersion effects in
many companies from emerging equity markets, Europe, and the US, finding that
one of the breakpoints coincided with the dawn of the war in Iraq in March, 2003.
Wolfers and Zitzewitz (2009) using the financial instrument of Saddam Security
6
find that the US stock market was extremely sensitive to changes in the probability
of war in Iraq. They conclude that a 10% rise in the probability of war was
accompanied by $1 increase in the spot oil price, decreasing the S&P 500 by 1.5%.
Nell and Simmler (2007) note that the war in Iraq was a major cause of uncertainty
throughout the world and seriously damaged many weak economies that could not
afford an increase in their imports, although as Damar (2007) concludes, the
aggregate impact of the war on the MENA economies is hard to determine. Rigobon
and Sack (2005) state that the lack of formal evidence in large part indicates that the
risks associated with the war are unobservable, which makes it difficult to estimate
their effects. Agus and Erbil (2010) find the effects of both global and regional
instability to be positive on Turkish exports to Iraq, and conclude that Turkey may
have some advantages in trade during the process of stabilization in Iraq.
These studies indicate that financial markets around the world have definitely reacted
to major events of the last two decades. Many dramatic events, both regional and
6
An asset whose payoffs depended on the ousting of the Iraqi leader; see Footnote 2.
71
global, have had varied effect on the equity markets around the world, and most
empirical literature recognises this, examining equity volatility for both regional and
international markets; Several studies have also examined the effects on individual
and portfolio returns. This body of literature offers strong evidence that external
shocks represent risk aversion for international investors.
This chapter contributes to the literature in different ways: first, the literature has not
examined the impact of geopolitical risks on the country beta of equity markets.
Second, the war on Iraq is considered the most prominent geopolitical risk in the
MENA region
one that has created unprecedented instability. The impact of the
war on Iraq, especially on the country beta of regional MENA equity markets, has
not previously been examined.
3.3. Empirical framework
3.3.1. Data
This chapter uses weekly data to estimate the country beta. Two main data groups
are considered. The first group is from the Morgan Stanley Capital International
(MSCI) for the period 3 January 1995 to 30 December 2008 for the markets of
Egypt, Israel, Jordan, Morocco, and Turkey: a total of 731 observations. For Qatar,
the data spans 3 June 2002 to 30 December 2008: a total of 344 observations. The
second group is from Standard & Poor s (S&P) for the period of 24 April 2000 to 30
December 2008 for the markets of Lebanon, Oman, Saudi Arabia, and Tunisia: a
total of 454 observations. The Kuwait stock market data is sourced form the Kuwait
Investment Company (KIW) for the period of 3 January 1995 to 30 December 2008:
a total of 731 observations. All data are sourced from Thomson DataStream. The
72
world MSCI is used as a proxy for the world index. The market indices are
denominated in US dollars.
3.3.2. CAPM and beta
The ICAPM is employed to examine the country beta as a measure of market risk
exposure to the world portfolio. The static ICAPM model takes the form:
Rit
i
R ft
i
i
Cov ( Ri , Rm )
( Rmt
2
m
R ft )
it
,
it
~ N (0,
2
)
(3.1)
Where Rit [ ln( Pit / Pit 1 ) 100] is the market return on country i at time t , and, R ft
is return for the 3-month US Treasury-bill rate as a proxy for the international riskfree rate at time t .7 Rmt is the world portfolio return represented by the MSCI index
at time t , ( Rmt
R ft ) measures the market risk premium at time t .
i
is the measure
of the market beta for country i , which is assumed to be constant over time.
the error term normally distributed with zero mean and constant variance
2
it
is
.
3.3.3. Testing for parameter stability
In practice, the variance
it
~ N (0,
2
t
2
is not constant over time and heteroscedastic; therefore
) . The stability of the country beta-
i
and the variance
2
in the
market model can be assessed using a variety of tests. One popular test is the
ARCH( q ) model of Engle (1982). Engle et al. (1987) indicate that the presence of
heteroscedasticity can lead to inaccurate rejection or non-rejection of hypotheses on
The assumption of the international risk-free rate does not exist in the theory; however, we assume
that it exists and follow the literature in using the 3-month US Treasury-bill rate (see for example
Harvey & Zhou, 1993; Ferson & Harvey, 1999a; Johansson, 2009a, b).
7
73
regression coefficients. Brooks et al. (2002) argue that the models of time-varying
beta continually evolve due to heteroscedasticity in the data, and can be detected
using the ARCH( q ) test. The variance in the market model in equation (3.2) can be
expressed as:
2
t
c0
where
2
t
c1
2
t
2
t 1
c2
2
t 2
...... cq
(3.2)
2
t q
is not assumed to be constant but depends on the q lags of squared errors
. A test for determining whether the ARCH effects are present in the residuals of
the market model may be conducted by choosing the number of q lags. The null
hypothesis is that all q lags of the squared residuals are not significantly different
from zero. If the test statistic is greater than the critical value from the
2
distribution, then we reject the null hypothesis. In addition, the test of
heteroscedasticity will suggest that the parameter
i
may be time-varying, but will
not indicate the specific form of time variation. The test will indicate whether the
coefficient of the market model fluctuates randomly through time or if it develops a
random walk (Wells, 1996).
Another test that can be used is the Cumulative Sum of Squares (CUSUMSQ) of
Brown et al. (1975), which is based on the recursive residuals
t
of one-step-ahead
prediction errors calculated from the regression model. Brooks et al. (2002) argue
that this test is likely to perform well when the parameters experience a discrete
jump. The CUSUMSQ test statistics ( st ) are estimated as follows (Wells, 1996,
P.55; Brooks et al., 2002):
74
st
t
j k 1
2
i
/
T
j k 1
2
i
,
(3.3)
t k 1,....., T .
The CUSUMQ test provides a plot of the recursive residuals
2
i
to period t divided
by the total sum of squared residuals against a pair of 5% critical lines for the
number of k regressors. The null hypothesis is that the test statistics ( st ) do not
show any movement outside these critical lines. If the critical value is exceeded, this
supports the alternative hypothesis against the null hypothesis as to the instability of
the regression parameters in the market model (see Brooks et al., 2002). The
CUSUMSQ test of parameter stability on the international domain was also proposed
by Brooks et al. (2002). According to Cashin and McDermott (1998), these stability
tests are used to examine whether any exogenous shocks or policy changes might
significantly affect equity markets and cause a structural change in the beta
coefficient. Examples of such policy changes include the revision of investment laws
or the opening of financial markets to foreigners.
3.3.4. The conditional beta test
If evidence suggests the instability of the parameters in the market model, the
assumption of time invariance can no longer be held. In this case, it is more
appropriate to specify the country beta
i
as a time-varying parameter (
it
). A more
dynamic estimation technique is required in this case, such as the Kalman filter,
which uses the state-space model to estimate the time-varying coefficient.8 The timevarying market model can be expressed as:
See Appendix for the state-space model representation and the derivation of the Kalman filter
approach.
8
75
Rit
i
it
Rmt
it 1
it
it
,
it
,
it
it
~ N (0,
~ N (0,
)
(3.4)
)
where Rit is market returns on country i in excess of the 3-month US Treasury-bill
rate. Rmt is the excess on world portfolio returns.
zero mean and covariance matrix
mean and covariance matrix
, and
it
it
is a vector of error terms with
is a vector of disturbances with zero
. Both error terms (
it
,
it
) are assumed to be
Gaussian, serially uncorrelated, and mutually independent of each other.
3.3.5. The conditional ICAPM and the impact of the war on Iraq
The dynamic technique of the Kalman filter will be employed to further examine the
impact of the war on Iraq on the time-varying country beta. The conditional market
model in (3.4) then takes the following form:
Rit
it
where
it
it
it 1
2
Rmt
2
it
,
DIraq 03
it
,
it
it
~ N (0,
~ N (0,
)
)
(3.5)
is the coefficient of the dummy variable of the war on Iraq. In this sense,
the transition equation in the Kalman filter allows an explicit incorporation of the
war on Iraq dummy with the evolution of the country betas over time. The time that
defines the war dummy must be carefully approximated to model the changes in the
country beta over time. The dummy variable - DIraq 03 , takes the value of 1 from
76
March 20 to May 1, 2003;9 0 otherwise. A significant positive (negative) coefficient
of the war dummy reflects an increase (decrease) in the country s beta after the war.
3.4. Empirical results
3.4.1. Descriptive statistics
Table 3-1 provides the summary statistics for the markets returns and world portfolio
returns. The table provides a battery of descriptive measures to examine the returns
series, such as average returns, standard deviation, skewness (the large positive or
negative movements in return series), kurtosis (the likelihood of big returns, positive
or negative) and the Jarque-Bera normality test statistic.10, Table 3-1 also reports the
Ljung-Box test statistics of autocorrelation in return level for all markets under
study.11
It is important to observe the properties of these emerging equity markets over
certain periods of time to gain further insights. During the 1990s, many of the
emerging markets were liberalized (see Bekaert et al., 2003; Ben Naceur et al.,
2008). The sample is therefore partitioned into three non-overlapping sub-periods.
Panel-A provides the first period from 3 January 1995 to 17 April 2000 with a total
of 277 observations for Egypt, Israel, Jordan, Kuwait, Morocco, and Turkey. Panel-B
provides the second period from 24 April 2000 to 16 August 2004 with a total of 226
Although the war lasted longer than two months, this latter date is when the then president of the
United States, George W. Bush, declared mission accomplished on the aircraft carrier USS Abraham
Lincoln; see Paleari et al. (2005).
9
Bekaert et al. (1998) and Harvey and Siddique (2000) argue that investors prefer positive skewness
over negative skewness and take it into account when making portfolio decisions, contributing to the
skewness of the overall portfolio.
10
Bekaert and Harvey (1995) indicate that the returns of emerging equity markets can be predicted
from past returns. Harvey (1995a) argues that predictability can be driven by market imperfections
such as infrequent trading of component securities.
11
77
observations for all markets, with the exception of Qatar where the weekly
observations total 115. Panel-C provides the third period from 23 August 2004 to 30
December 2008 with a total of 228 observations. Panel-D, presents descriptive
statistics for the whole sample period.
Panel-A of Table 3-1 shows that Egypt, Israel, Kuwait, Morocco, and Turkey yield
average positive returns, whereas Jordan yields average negative returns. Turkey
yields the highest gap between maximum and minimum, and Jordan yields the
lowest gap. Turkey has the highest volatility judged by the standard deviation,
whereas Morocco is the least risky over the sample period. Egypt, Jordan, and
Morocco provide positive skewness whereas Israel, Kuwait, and Turkey provide
negative skewness. Kurtosis is higher than 3 for all the market returns, which
indicates that the returns are leptokurtic. The Jarque-Bera test rejects the null
hypothesis of the normal distribution for all return series. These results are consistent
with the argument in Bekaert et al. (1996a) that there are significant deviations from
normality in the distributions of many of the emerging market returns. On the other
hand, the Ljung-Box test of autocorrelation shows signs of predictability for the
market returns of Egypt, Jordan, and Morocco, whereas the equity returns of Israel,
Kuwait, and Turkey cannot be predicted.
The second period in panel-B of Table 3-1 shows that Jordan, Kuwait, Lebanon,
Morocco, Oman, Qatar, and Saudi Arabia yield positive returns whereas Egypt,
Israel, Tunisia, and Turkey yield negative returns. Turkey and Lebanon yield the
highest/lowest gap between maximum and minimum. Turkey and Kuwait have the
highest/lowest standard deviation. Egypt, Kuwait, Qatar, Saudi Arabia, and Turkey
yield negative skewness whereas Israel, Jordan, Lebanon, Morocco, Oman, and
78
Table 3-1. Descriptive statistics of weekly return series
Panel-A. From 3 January1995 to 17 April 2000
Egypt
Israel
Jordan
Mean
0.250
0.337
-0.132
Maximum
13.251
8.383
6.833
Minimum
-10.355 -12.852
-4.444
St. Dev.
3.095
3.292
1.9176
Skewness
0.846
-0.631
0.767
Kurtosis
6.008
4.574
4.473
J-B prob.
0.000
0.000
0.000
Q-stat(36)
-0.048*
-0.015
0.022*
Values
227
227
227
Panel-B. From 24 April 2000 to 16 August 2004
Egypt
Israel
Jordan
Mean
-0.064
-0.092
0.278
Maximum
11.968
16.242
6.419
Minimum
-13.648 -13.783
-5.916
St. Dev.
4.159
4.088
1.953
Skewness
-0.181
0.001
0.132
Kurtosis
3.182
4.568
3.812
J-B prob.
0.472
0.000
0.033
Q-stat(36)
0.023
-0.102
-0.089
Values
226
226
226
Kuwait
0.080
7.757
-7.722
1.859
-0.120
6.486
0.000
-0.022
227
Lebanon
NA
Morocco
0.209
5.897
-6.562
1.765
0.276
4.333
0.000
0.037***
227
Oman
NA
Qatar
NA
Saudi Arabia
NA
Tunisia
NA
Turkey
0.569
22.686
-23.756
7.411
-0.135
3.651
0.003
0.008
227
MSCI
0.296
6.514
-7.605
1.943
-0.498
4.543
0.000
0.043*
227
Kuwait
0.477
5.132
-8.352
1.796
-0.809
6.187
0.000
-0.070**
226
Lebanon
0.039
21.316
-9.820
3.217
1.555
15.061
0.000
0.001
226
Morocco
0.006
9.332
-6.751
2.438
0.382
4.699
0.000
0.042
226
Oman
0.209
15.439
-4.982
2.057
1.9532
15.726
0.000
0.029*
226
Qatar
0.566
9.497
-9.326
2.974
-0.270
4.369
0.006
0.061
115
Saudi Arabia
0.419
6.166
-9.307
2.058
-0.673
5.884
0.000
0.156**
226
Tunisia
-0.115
12.798
-8.793
1.874
0.676
17.778
0.000
-0.005
226
Turkey
-0.370
24.089
-32.332
8.021
-0.523
4.989
0.000
-0.082*
226
MSCI
-0.140
9.327
-9.567
2.579
-0.081
4.467
0.000
-0.097
226
79
Table 3-1. Descriptive statistics of weekly return series (continued)
Panel-C. From 23 August 2004 to 29 December 2008
Egypt
Israel
Jordan
Kuwait
Mean
0.514
0.119
0.123
0.148
Maximum
14.077
6.168
11.927
7.171
Minimum
-24.883 -12.225
-18.186
-14.812
St. Dev.
5.008
2.624
3.784
2.747
Skewness
-1.184
-0.757
-0.850
-1.237
Kurtosis
7.347
4.993
6.579
8.296
J-B prob.
0.000
0.000
0.000
0.000
Q-stat(36)
-0.035
-0.056
0.015
0.037**
Lebanon
0.313
37.345
-24.416
5.710
1.604
17.076
0.000
-0.016
Morocco
0.380
10.304
-14.278
3.451
-0.918
6.476
0.000
-0.080
Oman
0.162
8.925
-19.729
3.298
-1.511
10.495
0.000
Qatar
0.146
19.884
-19.335
5.397
-0.194
5.395
0.000
-0.008
Saudi Arabia
-0.085
14.557
-17.465
5.045
-0.879
4.562
0.000
0.057
Tunisia
0.309
9.819
-12.455
2.061
0.559
15.465
0.000
-0.026
Turkey
0.0860
17.832
-27.826
6.575
-1.103
5.588
0.000
-0.052
MSCI
-0.044
12.170
-12.864
2.399
-0.944
10.459
0.000
0.022
-0.032*
Values
228
228
228
228
228
228
228
228
228
228
228
228
Panel-D. Full sample
Egypt
Israel
Jordan
Kuwait
Lebanon
Morocco
Oman
Qatar
Saudi Arabia
Tunisia
Turkey
MSCI
Mean
0.245
0.123
0.077
0.224
0.177
0.209
0.185
0.287
0.166
0.098
0.134
0.054
Maximum
14.077
16.242
11.926
7.757
37.350
10.305
15.439
19.883
14.557
12.798
24.089
12.169
Minimum
-24.883 -13.783
-18.185
-14.812
-24.416
-14.278
-19.729
-19.335
-17.465
-12.455
-32.332
-12.863
St. Dev.
4.087
3.401
2.652
2.163
4.639
2.592
2.750
4.724
3.866
1.979
7.328
2.297
Skewness
-0.579
-0.340
-0.620
-0.993
1.803
-0.496
-0.915
-0.267
-1.1679
0.630
-0.522
-0.539
Kurtosis
6.766
5.037
9.172
9.07
21.515
7.599
13.384
6.341
7.1293
16.246
4.785
6.952
J-B prob.
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
Q-stat(36)
0.004*
-0.050
0.004***
0.001*
0.005
-0.035***
-0.014*
0.008**
0.086*
-0.015
-0.035***
-0.002*
Values
730
730
730
730
454
730
454
343
454
454
730
730
Notes: *, **, *** denote significant at 1%, 5%, and 10% respectively. Returns are in US dollars. The sample for Egypt, Jordan, Israel, Kuwait, Morocco, and Turkey spans 1/03/1995 to
12/30/2008 with 731 values. The sample for Lebanon, Oman, Saudi Arabia, and Tunisia spans 4/17/2000 to 12/30/2008 with 454 values. The sample for Qatar spans 6/03/2002 to
12/30/2008 with 354 values. Panel-A spans 1/03/1995 to 4/17/2000 with 227 values, Panel-B spans 4/24/2000 to 8/16/2004 with 226 values. Panel-C spans 8/23/2004 to 12/29/2008 with
228 observations. Std. Dev., J-B prob., Q-stat(36) stand for Standard Deviation, Jarque-Bera test of normality, and the Ljung-Box test statistics of autocorrelation respectively.
80
Tunisia yield positive skewness. As in the first period, kurtosis is higher than 3 for
all the equity return series, which also indicates that the return series are leptokurtic.
With the exception of Egypt, the Jarque-Bera test of normality is rejected for all the
equity markets, with 95% level of confidence. The Ljung-Box test shows that the
equity returns of Oman, Saudi Arabia, and Turkey returns can be predicted.
Panel-C of Table 3-1 shows that with the exception of Saudi Arabia, all the markets
yield positive average returns. Lebanon, Turkey, and Qatar have the highest gap
between maximum and minimum while Israel and Tunisia yield the lowest gap.
Turkey and Lebanon have the highest standard deviations for this period, and Tunisia
and Israel are the least volatile judged by their standard deviations. With the
exception of Lebanon and Tunisia, all markets provide negative skewness. As in the
previous period, kurtosis is high especially for Lebanon, Tunisia, and Oman. The
Jarque-Bera test rejects the null hypothesis that the returns series are normally
distributed, while the Ljung-Box test reveals that for this period, only the market
returns of Egypt, Kuwait, and Oman are predictable.
Panel-D of Table 3-1 provides the descriptive statistics for the whole sample period.
The table shows that all the equity markets yield positive average returns. The
average mean ranges from the highest in Qatar and Egypt to the lowest in Jordan and
Tunisia. The markets of Turkey and Lebanon are the most volatile in the sample, and
Tunisia is still the least volatile market in the sample, with the lowest standard
deviation. Kurtosis is high for Lebanon, Tunisia, and Oman, as in the previous
periods. In addition, only Lebanon and Tunisia provide positive skewness. The
Jarque-Bera test rejects the normality for all the return series.
81
The world returns provide negative average returns in the second and the third
periods, and positive returns in the first and full sample periods. In addition, the
world returns provide negative skewness in all the sub-samples and the full sample
period. The highest kurtosis in the sample period can be found for the world returns
in the third period. The Jarque-Bera test rejects the assumption of normality for the
MSCI world portfolio returns for the sub-samples and the whole sample period.
The general conclusions from these statistics are that the markets of Kuwait,
Lebanon, and Morocco yield consistent positive returns in the three sub-samples and
the full sample period. Turkey has the highest standard deviation in the sub-sample
and the full sample periods; in fact, the Turkish equity market is known to be one of
the most volatile markets in the region because of the volatile nature of its economic
and financial systems. With the exception of Egypt in the first sub-sample, the
weekly returns are not normally distributed in sub-samples or the full sample period.
3.4.2. Correlation analysis
The correlation analysis within these markets and correlations with world returns are
also examined. The correlation is divided into three sub-samples similar to the subsample periods used for the descriptive statistics.12 Table 3-2 presents the correlation
analysis of the MENA market returns and world returns. The results from the first
period reported under Panel-A show that the inter-correlations are low or
insignificant. The highest correlations can be found between Israel and Turkey, and
Bekaert and Harvey (2000) present evidence that suggests that emerging markets have become
more correlated with the world portfolio. However, Harvey (2000) argues that if correlations with
emerging market returns are sufficiently low, they will provide a low or negative beta and
economically significant portfolio diversification.
12
82
Egypt and Turkey. Israel has the highest correlation with world portfolio returns at
0.397, followed by Turkey at 0.252 and Jordan at 0.130.
Panel-B of Table 3-2 examines the correlation analysis for the second period. Similar
to the results from the first sub-sample in panel-A, the inter-correlations are low or
insignificant. The highest correlations can be found between Israel and Turkey at
0.295 and Jordan and Saudi Arabia at 0.269. There is a negative and significant
correlation between Lebanon and Tunisia, while Israel, Turkey, and Egypt have
significant correlations with the global market.
Panel-C of Table 3-2 provides some interesting contrasts in the third sub-sample
period. Most of inter-correlations are now significant. The highest correlations can
be found between Israel and Turkey at 0.509 and Egypt and Turkey at 0.508. All the
MENA markets have significant correlations with the world returns. As in the
previous period, the highest correlations with world returns can be found for Israel at
0.631 and Turkey at 0.613 and Egypt at 0.541.
Panel-D of Table 3-2 provides the correlation analysis for the full sample period.
Overall, the correlation is higher among equity markets, compared to the subsamples results. As in the three sub-samples, the highest correlations can be found
between Israel and Turkey at 0.368, Turkey and Egypt at 0.342 and Israel and Egypt
at 0.267. The correlations among all the markets and the MSCI are significant similar
to the same conclusion for the third period. The highest correlations with the world
returns can be found for Israel at 0.597 and Turkey at 0.426 and Egypt at 0.363.
The general conclusions are that the correlations among these countries have
substantially increased, especially in the 2004-2008 period. Egypt, Israel, and Turkey
83
Table 3-2. Correlation analysis of MENA market returns and the global MSCI returns
Panel-A. From 3 January1995 to 17 April 2000
Egypt
Egypt
Israel
Jordan
Kuwait
Lebanon
Morocco
Oman
1.000
Israel
Jordan
Kuwait
Lebanon
Morocco
NA
0.162*
(0.007)
1.000
0.088
(0.146)
-0.012
(0.849)
-0.099
(0.102)
1.000
0.193*
(0.002)
-0.093
(0.123)
0.073
(0.231)
-0.096
(0.114)
1.000
0.181*
(0.003)
0.233*
(0.000)
0.154**
(0.011)
0.015
(0.808)
-0.025
(0.681)
0.141** 0.099
(0.020) (0.103)
Oman
Qatar
NA
NA
Saudi
Arabia
NA
Tunisia
Turkey
MSCI
NA
1.000
Qatar
Saudi
Arabia
Tunisia
Turkey
MSCI
0.095
(0.116)
0.397*
(0.000)
0.130**
(0.031)
0.060
(0.323)
-0.065
(0.281)
84
1.000
0.252*
(0.000)
1.000
Table 3-2. Correlation analysis of MENA equity market returns and the global MSCI returns (continued)
Panel-B. From 24 April 2000 to 16 August 2004
Egypt
Egypt
Israel
Jordan
1.000
0.15**
(0.020)
Israel
Jordan
Kuwait
1.000
-0.023
(0.730)
0.033
(0.618)
0.037
(0.575)
-0.091
(0.176)
0.048
(0.465)
-0.099
(0.135)
0.044
(0.503)
0.104
(0.123)
Saudi
Arabia
0.083
(0.209)
0.043
(0.511)
Turkey
0.185*
(0.005)
Lebanon
Morocco
Oman
Qatar
Tunisia
MSCI
0.029
(0.659)
0.066
(0.318)
Oman
1.000
0.052
(0.437)
0.051
(0.449)
Morocco
0.025 0.123***
(0.714) (0.068)
0.061
(0.356)
0.013
(0.846)
0.051
(0.443)
-0.082
(0.219)
1.000
0.269*
(0.000)
-0.005
(0.947)
0.092
(0.166)
0.007
(0.914)
0.024
(0.712)
0.085
(0.199)
0.108
(0.106)
0.039
(0.558)
0.038
(0.568)
0.167** 0.596*
(0.012) (0.000)
0.045
(0.496)
Tunisia
Turkey
MSCI
1.000
0.006
(0.918)
0.060
(0.365)
Saudi
Arabia
1.000
0.080
(0.234)
-0.008
(0.901)
0.295*
(0.000)
Qatar
NA
0.093
(0.161)
Kuwait
Lebanon
0.071
(0.291)
-0.129***
(0.053)
0.073
(0.730)
0.045
(0.501)
1.000
1.000
0.095
(0.153)
-0.015
(0.812)
0.097
(0.144)
0.055
(0.404)
-0.052
(0.429)
0.068
(0.303)
85
-0.093
(0.164)
0.097
(0.145)
1.000
-0.082
(0.215)
0.037
(0.577)
1.000
0.281*
(0.000)
1.000
Table 3-2. Correlation analysis of MENA equity market returns and the global MSCI returns (continued)
Panel-C. From 23 August 2004 to 29 December 2008
Egypt
Israel
Jordan
Kuwait Lebanon Morocco Oman
Qatar
Saudi Tunisia
Arabia
Egypt
1.000
Israel
Jordan
Kuwait
Lebanon
Morocco
Oman
Qatar
Saudi
Arabia
Tunisia
Turkey
MSCI
0.437*
1.000
(0.000)
0.422*
0.193*
1.000
(0.000) (0.003)
0.116*** 0.084
0.156**
(0.082) (0.209) (0.019)
0.127**
0.075
0.197*
(0.056) (0.256) (0.003)
0.355*
0.166*
0.309*
(0.000) (0.011) (0.000)
0.312*
0.244*
0.241*
(0.000) (0.000) (0.000)
0.269*
0.206*
0.257*
(0.000) (0.002) (0.002)
0.335*
0.203*
0.312*
(0.000) (0.002) (0.000)
0.163** 0.127*** 0.279*
(0.014) (0.054) (0.000)
0.508*
0.509*
0.253*
(0.000) (0.000) (0.000)
0.541*
0.623*
0.408*
(0.000) (0.000) (0.000)
Turkey
MSCI
1.000
0.106
(0.113)
0.146**
(0.028)
0.458*
(0.000)
0.287*
(0.000)
0.333*
(0.000)
0.035
(0.602)
0.094
(0.158)
0.160**
(0.013)
1.000
0.155**
(0.019)
0.155**
(0.019)
0.105
(0.115)
0.131**
(0.047)
0.239*
(0.000)
0.117***
(0.077)
0.200*
(0.002)
1.000
0.286*
(0.000)
0.198*
(0.003)
0.261*
(0.000)
0.232*
(0.000)
0.162**
(0.013)
0.380*
(0.000)
86
1.000
0.432*
(0.000)
0.439*
(0.000)
0.101
(0.127)
0.225*
(0.000)
0.300*
(0.000)
1.000
0.290*
(0.000)
0.047
(0.481)
0.248*
(0.000)
0.289*
(0.000)
1.000
0.194*
1.000
(0.003)
0.124*** 0.082
(0.062) (0.199)
0.200* 0.251*
(0.002) (0.000)
1.000
0.613*
(0.000)
1.000
Table 3-2. Correlation analysis of MENA equity market returns and the global MSCI returns (continued)
Panel-D. Full sample period
Egypt
Israel
Jordan
Kuwait Lebanon Morocco
Oman
Qatar
Saudi Tunisia
Arabia
Egypt
1.000
Israel
0.267*
1.000
(0.000)
Jordan
0.315* 0.113** 1.000
(0.000) (0.017)
Kuwait
0.088*** 0.047
0.148*
1.000
(0.063) (0.319) (0.002)
Lebanon
0.080*** 0.051
0.161*
0.051
1.000
(0.089) (0.286) (0.001) (0.282)
Morocco
0.238* 0.106** 0.227* 0.120**
0.114*
1.000
(0.000) (0.025) (0.000) (0.011)
(0.015)
Oman
0.220*
0.069
0.194*
0.356*
0.128*
0.174*
1.000
(0.000) (0.146) (0.000) (0.000)
(0.007)
(0.000)
Qatar
0.269* 0.206* 0.257*
0.287*
0.105
0.198*
0.432*
1.000
(0.000) (0.002) (0.002) (0.000)
(0.115)
(0.003)
(0.000)
Saudi
0.254* 0.114** 0.305*
0.261*
0.121**
0.195*
0.351*
0.290*
1.000
Arabia
(0.000) (0.015) (0.000) (0.000)
(0.010)
(0.000)
(0.000) (0.000)
Tunisia
0.128*
0.050
0.200*
0.040
0.116**
0.184*
0.057
0.047
0.151*
1.000
(0.007) (0.290) (0.000) (0.402)
(0.014)
(0.000)
(0.226) (0.481) (0.002)
Turkey
0.342* 0.368* 0.173*
0.094* 0.081*** 0.104** 0.084*** 0.248* 0.099** 0.001
(0.000) (0.000) (0.000) (0.048)
(0.087)
(0.028)
(0.075) (0.000) (0.035) (0.988)
MSCI
0.363* 0.597* 0.263* 0.120**
0.136*
0.237*
0.154*
0.289*
0.145* 0.147*
(0.000) (0.000) (0.000) (0.011)
(0.004)
(0.000)
(0.001) (0.000) (0.002) (0.002)
Turkey
MSCI
1.000
0.426*
(0.000)
1.000
Notes: *, **, *** denote significant at 1%, 5%, and 10% respectively. P-values are in parentheses. For the sample periods of the correlation analysis of each
panel, see Table 3-1.
87
have the highest inter-correlations of the MENA countries. These results are not
surprising since these markets are first liberalized markets in the MENA region (see
Ben Naceur et al., 2008). However, the correlations are still reasonably low for
regional and international investors. In the first two sub-samples, the results suggest
that these equity markets would have provided international investors with
substantial diversification opportunities. On the other hand, Israel has the highest
correlation with the world returns in the three sub-samples and the full ample period
followed by Turkey and Egypt.
3.4.3. The stability of country beta using the static ICAPM
The analysis starts with the estimation of the unconditional market model using the
equation (3.1). To deal with autoregressive as well as heteroscedastic features in the
data, the analysis adopts the Newey-West (1987) robust standard errors. Table 3-3
provides the output of the unconditional ICAPM for the three sample periods and the
full period; the table also reports the CUSUMSQ and the ARCH( q ) tests.
Panel-A of Table 3-3 examines the results for the first period. None of these markets
outperformed the world market returns, as can be seen from the insignificant
intercepts of the market model. The beta coefficient is insignificant for Egypt and
Kuwait and negative for Morocco, but significant at 1% for Israel and Turkey and
5% for Jordan. Since the correlation tests from previous tables show that Egypt and
Kuwait are insignificantly correlated with the world market, and Morocco has
negative correlation, it is not surprising that the beta coefficient for Egypt is
statistically insignificant or negative. This confirms the findings of Harvey (1995c)
and Erb et al. (1996c), that negative or low correlations yield negative or low betas.
88
Table 3-3. Analysis of country beta stability using equation (3.2)
Panel-A. From 1/03/1995 4/17/2000
CUSUMSQ
D-W
Adj- R 2
i
i
test (5%)
Egypt
0.153
0.152
0.006
1.582
Rejected
(0626)
(1.589)
1996-1999
Israel
0.067
0.700*
0.156
1.851
Rejected
(0.325)
(7.234)
1998-2000
Jordan
-0.250** 0.126**
0.013
2.080
Rejected
(-2.216)
(2.220)
1997-1998
Kuwait
-0.026
0.058
-0.000
2.118
Rejected
(-0.214)
(1.136)
1996-1998
Morocco
0.153
-0.059
0.000
1.547
Rejected
(1.088)
(-1.111)
1997-1997
Turkey
0.277
0.948*
0.059
2.025
Rejected
(0.623)
(3.543)
1996-1999
Panel-B. From 4/24/2000 8/16/2004
Egypt
-0.063
0.272**
0.024
2.047
Not rejected
(-0.230)
(2.230)
Israel
0.041
0.946*
0.358
2.145
Rejected
(0.193)
(6.178)
2002-2004
Jordan
0.247***
0.040
-0.002
1.981
Not rejected
(1.850)
(0.854)
Lebanon
0.003
0.060
-0.003
1.961
Not rejected
(0.012)
(0.764)
Kuwait
0.439*
0.052
0.002
1.772
Rejected
(3.118)
(0.862)
2003-2003
Morocco
-0.135
0.043
-0.002
1.924
Not Rejected
(-0.862)
(0.561)
Oman
0.155
-0.039
-0.002
1.697
Rejected
(1.008)
(-0.576)
2001-2003
Qatar
0.543***
0.014
-0.009
2.069
Rejected
(1.841)
(0.183)
2003-2004
Saudi
0.383***
0.057
0.001
1.561
Rejected
Arabia
(2.489)
(1.193)
2001-2004
Tunisia
-0.157
0.029
-0.003
1.984
Rejected
(-1.198)
(0.567)
2001-2003
Turkey
-0.219
0.887*
0.078
2.023
Rejected
(-0.421)
(3.225)
2001-2003
Panel-C. From 8/23/2004 12/29/2008
Egypt
0.571**
1.129*
0.289
2.095
Rejected
(2.055)
(8.260)
2007-2008
Israel
0.129
0.681*
0.384
2.158
Rejected
(1.078)
(8.736)
2007-2008
Jordan
0.129
0.645*
0.163
1.972
Rejected
(0.534)
(4.840)
2006-2008
89
ARCH (24)
test
1.361
2.011*
1.485***
2.711*
1.699**
1.173
0.900
0.894
1.760**
0.700
0.458
0.750
0.117
1.129
2.838*
0.868
1.924*
2.898*
1.474***
3.562*
Kuwait
0.105
(0.555)
Lebanon
0.301
(0.808)
Morocco
0.375***
(1.857)
Oman
0.143
(0.688)
Qatar
0.151
(0.370)
-0.102
Saudi
Arabia
(-0.311)
Tunisia
0.269**
(2.032)
Turkey
0.201
(0.713)
Panel-D. Full sample
Egypt
1.183
(1.117)
Israel
0.066
(0.617)
Jordan
0.012
(0.114)
Kuwait
0.156***
(1.678)
Lebanon
0.159
(0.731)
Morocco
0.142
(1.384)
Oman
0.156
(1.103)
Qatar
0.261
(0.853)
0.144
Saudi
(0.718)
Arabia
Tunisia
0.061
(0.633)
Turkey
0.083
(0.337)
0.180
(1.421)
0.476*
(3.063)
0.544*
(3.793)
0.410*
(4.691)
0.649*
(4.461)
0.422*
(3.074)
0.215*
(3.873)
1.683*
(10.574)
0.021
2.062
0.036
1.943
0.140
2.099
0.086
1.975
0.079
1.871
0.036
1.972
0.058
2.033
0.374
2.417
0.531*
(3.294)
0.787*
(11.664)
0.267*
(2.586)
0.092
(1.508)
0.255***
(1.730)
0.192**
(2.135)
0.170
(1.276)
0.389*
(2.741)
0.225**
(1.989)
0.117***
(1.720)
1.179*
(7.365)
0.089
1.916
0.282
2.019
0.050
2.009
0.008
1.998
0.017
1.961
0.028
1.960
0.022
1.883
0.041
1.885
0.020
1.901
0.020
1.983
0.142
2.094
Rejected
2007-2008
Rejected
2007-2008
Rejected
2005-2006
Rejected
2006-2008
Rejected
2005-2006
Rejected
2005-2006
Rejected
2006-2006
Rejected
2005-2008
Rejected
1996-2007
Rejected
2000-2006
Rejected
1996-2008
Rejected
1999-2008
Rejected
2001-2006
Rejected
1997-2008
Rejected
2002-2008
Rejected
2003-2006
Rejected
2001-2008
Rejected
2002-2006
Rejected
2000-2005
2.595*
1.118
0.639
5.912*
3.110*
1.492***
1.312
2.780*
2.985*
3.138*
15.216*
5.910*
2.277*
4.223*
5.265*
3.668*
4.945*
1.895*
3.600*
Notes: *, **, *** denote significant at 1%, 5%, and 10% respectively; t-statistics are in parentheses.
All returns are in excess of the 3-month US Treasury-bill rate of return. The parameters are estimated
using the OLS method with Newey-West (1987) correction of heteroscedasticity and autocorrelation.
CUSMSQ is the test of residual stability on 5% bound. ARCH (24) is the test of heteroscedasticity of
Engle (1982).
90
The CUSUMSQ test rejects the stability of beta for all the markets, while the test of
no ARCH effects rejects the null hypothesis at 1% for Israel and Kuwait, 5% for
Jordan, and 10% for Morocco; this can be interpreted as evidence of unstable
parameter coefficients. The adjusted R 2 is extremely low for Morocco and Egypt
and negative for Kuwait but reasonably high for Israel at 0.156.
Panel-B of Table 3-3 for the second period shows that the intercept is positive and
significant for Jordan, Kuwait, Qatar, and Saudi Arabia. This shows that these
markets have outperformed the world market for this period. The beta coefficient is
negative and insignificant for Oman, and positive and significant at 1% for Israel and
Turkey and at 5% for Egypt. The beta coefficients of Kuwait and Jordan are
insignificant compared to that of Jordan in the first period. The CUSUMSQ test
rejects the stability of beta for the markets of Egypt, Jordan, Lebanon, and Morocco,
while the test of no ARCH effects rejects the null hypothesis for Jordan, Saudi
Arabia, and Turkey.
The outcomes of the third period reported in Panel-C of Table 3-3 show some
contrasting results. The intercept of the market model is positive and significant for
three countries, Egypt, Saudi Arabia, and Tunisia, suggesting that these markets have
outperformed the world portfolio. With the exception of Kuwait, all the beta
coefficients are positive and statistically significant at 1%. Egypt and Turkey have
betas above the world market risk in this period. The CUSUMSQ test for this period
rejects the stability of beta for all countries, while the test of no ARCH effects rejects
the null hypothesis except for Lebanon, Morocco, and Tunisia.
Panel-D of Table 3-3 examines the market model for the full sample period. It can be
91
seen that all the coefficients of betas are statistically significant with the exception of
Kuwait and Oman. On average, Turkey has the highest beta in the region, greater
than the world portfolio risk at 1.179, while Tunisia has the lowest at 0.117. The
stability of market model coefficients using the CUSUMSQ have exceeded the 5%
bound of significance for all the markets in the full sample period. This implies that
the test has detected some structural breaks in the equation for the market
coefficients; the evidence is supported by the outcome of the ARCH test of
heteroscedasticity for all the markets. These results suggest that the country beta
parameters are not stable in the long-run, despite what the tests reveal in some cases
for the three sub-samples; this suggests a nonstationary country beta.
The adjusted R 2 as explanatory power ranges from a poor fit for Kuwait with 0.008
to a reasonable 0.285 for Israel and 0.142 for Turkey.13 The value of the DurbinWatson test statistics are reasonably close to 2 in the full sample period, indicating
an absence of serial correlation. However, the Durbin-Watson test indicates negative
serial correlation for Turkey in the third sample period with 2.417, and positive serial
correlation for Egypt in the first period with 1.582. This problem undermines the
model s usefulness when returns are serially correlated, and could lead to inefficient
estimates of the market coefficients.
The general conclusion is that the betas for the MENA countries have increased
considerably from the 1990s to the 2000s, suggesting that the MENA stock market
returns are becoming more affected by global market returns; this is, to some extent,
The adjusted R 2 is negative for the markets of Kuwait, Jordan, Lebanon, Morocco, Oman, Qatar,
and Tunisia especially in the first and second sample periods and started to increase in the third
period, but still low for these markets.
13
92
consistent with the degree of market integration discussed by Bekaert et al. (2003)
and Arouri (2006). In fact, most of the MENA stock markets have been liberalized
during this period and this explains the gradual increase in the market risk exposure.
In addition, the CUSUMSQ and heteroscedasticity tests suggest that the assumptions
of stable market parameters over time are invalid, especially in the long-run.
3.4.4. The conditional market model and the impact of the war on Iraq
Since the evidence suggests the instability of the country beta
i
parameter in the
market model, it is more appropriate to adopt the dynamic market model. Table 3-4
reports the results of the conditional ICAPM using the Kalman filter approach
expressed in equation (3.6). As mentioned before, the transition equation in the
Kalman filter allows the direct incorporation of the war on Iraq dummy with
evolution of the country betas over time.
The explanation of the dummy variable is as follows: first, since the oil price has
increased exponentially and reached to $147 in 2008 after the Iraq war, we expect
that oil exporting countries will experience a decrease in their country betas while oil
importing countries will experience an increase. This is due to the fact that the effect
of the increased oil price and its volatility has different effect on each country.
Second, political and regional uncertainty has undermined the investment and
business environment in the region. The outcome of the war on Iraq depends on
whether it achieves its objectives of establishing a regime change and establishing
political stability in the region. The longer the objective takes to accomplish, the
longer it takes for trading and business activities to go back to normal. As argued by
Schneider and Troeger (2006), international investors trading on markets near a
93
conflict region have a tendency to react to increased confrontation, and may flee the
regional markets en masse. Third, liberalized stock markets are more affected by the
war compared to the closed markets.
The findings show that the war on Iraq has a positive impact on the country betas of
all MENA markets but is statistically significant only for Egypt, Morocco, and
Tunisia as oil importing countries and Kuwait as an oil exporting country. The other
MENA markets are not significantly affected, whether they are oil importing or oil
exporting countries. The implications are that the combination of higher oil prices,
weaker confidence of consumers and investors, local or international, have
contributed largely to the increase in the country betas for Egypt, Tunisia, and
Morocco. Besides, certain sectors in these countries have continued to suffer since
the war: tourism and civil aviation especially, in both Egypt and Morocco, have been
adversely affected. Most importantly, the markets of Egypt, Morocco and Tunisia are
liberalized markets whereas the Kuwaiti stock market is a closed one. This highlight
how vulnerable the liberalized stock markets to geopolitical unrests in the region.
Prior to the war s commencement, there was uncertainty about whether it would be
waged, when it would end and whether it would achieve its objectives. The
uncertainty was so great that the Kuwait stock market activities was suspended
following the start of the war on Iraq. In addition, local and international investors
were reluctant to undertake planned expenditure, asset allocation, and trading
activities in these equity markets. The unstable environment and the geopolitical risk
significantly affected the stock markets of these countries, and particularly of
Kuwait, reflected in a significant increase in its country beta.
94
Table 3-4. Conditional ICAPM with the war on Iraq dummy using equation (3.6)
Wald test
Log-likelihood
AIC
i
t
2
Egypt
Israel
Jordan
Kuwait
Lebanon
Morocco
Oman
Qatar
Saudi
Arabia
Tunisia
Turkey
0.190
(1.335)
1.128*
(7.334)
0.130*
(3.249)
10.552*
2014.53
5.50
2.290
1811.23
4.973
0.041
(0.428)
0.812*
(7.272)
0.062
(1.513)
0.370
1716.76
4.714
0.461*
(3.629)
0.023
(1.608)
0.064***
(1.659)
2.753***
1596.06
4.392
0.001
1341.41
5.940
0.266
(1.177)
0.136***
(2.167)
4.696**
1706.66
4.687
0.173
1088.62
4.833
0.636**
(6.560)
0.024
(0.417)
0.056
(1.035)
0.786
1018.25
5.967
0.006
(0.079)
0.006
1256.03
5.572
0.047*
(2.936)
8.617*
951.80
4.225
2.621
2436.72
6.687
0.069
(0.643)
0.147***
(1.720)
0.196
(0.818)
0.159***
(1.715)
0.694*
(4.015)
0.601*
(3.520)
0.170
(1.290)
0.509**
(2.307)
0.179
(0.875)
0.606**
(3.599)
0.274
(1.066)
0.046
(0.463)
0.097
(0.383)
0.278*
(3.085)
1.471*
(3.613)
0.000
(0.006)
0.194
(1.619)
Notes: *, **, *** denote significant at 1%, 5%, and 10% respectively; t-statistics are in parentheses.
The model is estimated by the maximum likelihood methods described by Berndt, Hall, Hall, and
Hausman s (1974) iterative algorithm. Eviews software does not report the R 2 for the Kalman filter
output.
These results correspond to the strength of geopolitical risks on the stock market
exposure. Starting in March 2003 with the expectation of a quick victory that would
rapidly topple the Iraqi regime and establish democracy and stability, the subsequent
prolonged fighting and the divisive aftermath not only destabilized equity markets in
the region, but also led to negative responses from both local and international
investors, which further aggravated the country betas of the MENA region. It is no
leap to assume that the war on Iraq not only disrupted regional trading activities but
95
also affected the profits of regional firms in the region.
The significant and positive dummy suggests a sudden surge in country betas, which
are evidenced in a major structural break and a dramatic increase of the country betas
of these regional markets. Figure 3.1 shows the common upward trend in country
betas for those markets that have been significantly affected by the war on Iraq:
Egypt, Morocco, Tunisia, and Kuwait. The magnitude of the coefficient of the
dummy variable is highest for Egypt and Morocco, followed by Tunisia and Kuwait.
The Wald test hypothesis fails to reject the null hypothesis that the war dummy
variable equals zero for these countries.
T u n is ia
Eg y p t
1.6
.4
1.2
.3
0.8
.2
.1
0.4
.0
0.0
-.1
2001
2002
2003
2004
2005
2006
2007
2008
-0 . 4
1996
1998
2000
K u w a it
2004
2006
2008
2004
2006
2008
Moroc c o
.8
2.0
1.6
.6
1.2
.4
0.8
.2
0.4
.0
0.0
-.2
-.4
2002
-0 . 4
1996
1998
2000
2002
2004
2006
-0 . 8
2008
1996
1998
2000
2002
Figure 3.1: Conditional country betas for Tunisia, Egypt, Kuwait and Morocco
showing the impact of the war on Iraq.
3.5. Conclusion
This chapter evaluates the stability of country beta of MENA equity markets and the
impact of the 2003 war on Iraq. The results indicate that most of the MENA markets
have low average country betas measured against the world market portfolio and that
96
these are mostly unstable during the sample period. In addition, the betas for MENA
countries have increased considerably from the 1990s to the 2000s. This suggests
that MENA stock market returns are increasingly influenced by global market
returns; this is consistent with the degree of market integration.
When examining the effect of the war on Iraq on the country beta, the results show a
positive impact for four MENA markets: Egypt, Kuwait, Morocco, and Tunisia. The
test of significance of the dummy variable indicates that the war has indeed produced
a fundamental shift in the perception of risk in MENA equity markets, and this is
reflected in the sudden upward trend in country betas. These results confirm the
evidence for the impact of geopolitical risks on equity markets in the region. As the
magnitude of the war is strongest on the market beta of Egypt, followed by those of
Tunisia and Morocco, it is logical to conclude that investors are likely to suffer
substantial losses and lowered returns when unexpected shifts in beta occur, and may
demand higher average returns to compensate for investing in such risky markets.
International investors are likely to exercise a high degree of caution when
reinvesting in such markets, due to the increased risks.
For policy or investor perspective, the investors should take extreme caution before
they invest in the MENA countries because of increased risk. It is logical to conclude
that as country beta increases, investors may suffer substantial losses and lowered
returns when beta experience unexpected shifts, and there is a reasonable expectation
that international investors may demand higher average returns to compensate for
reinvesting in such risky markets.
97
Further Evidence on the Impact of Financial and
Geopolitical Crises on the Country Beta of MENA Markets
The previous chapter examined the impact of the war on Iraq on country betas, using
the dummy variable specified in the transition equation. This chapter undertakes a
multiple structural breaks analysis of the country betas of MENA markets, using the
Bai and Perron (2003) model, and relates them to key financial and geopolitical
events. Over the sample period, there are several significant breakpoints in country
betas series that may be attributed to financial and geopolitical crises.
98
4.1. Introduction
The assessment of country beta is central to international investment, especially in
emerging markets. Country risk, measured by the coefficient of beta relative to the
world market portfolio, has rapidly grown in importance due to the continuing
integration of the world s capital markets. International finance literature dealing
with this subject has been driven by international portfolio diversification. Country
betas are seen to have practical implications, especially in emerging markets, where
allowing for variation in country betas over time helps to develop appropriate weight
asset allocations (Brooks et al., 2002; Johansson, 2009a; McKenzie et al., 2000).
There is concern that systematic risk at the international level has the tendency to be
unstable, and therefore the need to estimate country beta has become persistent and
important, particularly in light of recent worldwide events and regional political
instabilities (see for example Gangemi et al., 1999, 2000). The focus in accounting
for variations of country beta in both developed and emerging markets has been on
the choice of modelling techniques.14 The application of different models has raised
questions on a number of issues other than the central one
country betas over time
the stability of the
such as the integration of both developed and emerging
markets into the global capital market.
It has been demonstrated that international capital markets become more volatile
after major exogenous shocks across regional and international arenas. Studies like
those of Athanassiou et al. (2006), Blass et al. (2004), Fernandez (2006, 2007, 2008),
Hammoudeh and Li (2008), Leippold and Lohre (2010), Malliaris and Urrutia
See Koutmos et al. (1994), Grieb and Reyes (2001), Brooks et al. (2002), Giannopoulos (1995),
Johansson (2009a, b), Marshall et al. (2009).
14
99
(1995), Nikkinen et al. (2008), Pownall and Koedijk (1999), Rigobon and Sack
(2005), and Wolfers and Zitzewitz (2009), amongst others, show that equity markets
react to major regional and global events. In terms of the increase in systematic risk,
Choudhry (2005a, b), Drakos (2004), Faff and Lie (2002), Groenewold and Fraser
(1999, 2000), Kim and Gu (2004), and McKenzie et al. (2000) document a sudden
increase in systematic risk after major global events such as the crash of 1987, the
Asian financial crisis in 1997, the September 11, 2001 attacks, and the 2003 war on
Iraq.
A number of studies examine the country beta of equity markets and document a
significant change in time paths caused by global market volatility. Giannopoulos
(1995) finds that the stock market crash of 1987 was a primary driving factor of
country beta for a number of developed markets, especially USA and Japan.
Johansson (2009a) documents an increase in the country beta of Thailand following
the Asian financial crisis in 1997, and Gangemi et al. (2000) for the country beta of
Australia after the stock market crash of 1987. Miller (2006) finds that all the
country betas in his sample of emerging markets rose considerably during the Asian
financial crisis, and most of their country betas peaked around the time of the
Russian financial crisis. Marshall et al. (2009) find that 20 emerging equity markets
from Latin America, Middle East, Eastern Europe, and Asia increased during the
Asian financial crisis of 1997 and again during 2008, reflecting the global financial
crisis. Johansson (2009b) finds that the Hong Kong stock market experienced a
strong surge in country beta during the Asian financial crisis in 1997.
The MENA region countries have been subjected to multiple political and economic
shocks, and significant periods of financial and political instabilities. Wars, political
100
unrest, economic and financial instability, and institutional underdevelopment have
traditionally been powerful obstacles to increased access to MENA capital markets
(Girard & Ferreira, 2003). The empirical literature has paid increasing attention to
the impact of such geopolitical events on stock market behaviour. Malliaris and
Urrutia (1995) analyse the responses of international equity prices to the Persian Gulf
crisis of 1991, finding that the stock markets seemed to respond to the crisis period
according to the countries dependence on oil imports. Recently, Nikkenin et al.
(2008) have compared the volatility behaviour of stock markets in six different
regions before and after the September 11 attacks, to examine whether the impact of
this event is pervasive across regions. They find that the attacks had no significant
effect on the mean nations, compared to the rest of the world. Using the (ICSS)
algorithm and Wavelet-Based Variance Analysis, Fernandez (2006) considers
whether the Asian crisis in 1997 and the September 11 attacks caused permanent
volatility shifts in world stock markets; her results reveal a number of shifts since the
occurrence of these events in- some MENA markets, and Wavelet-Based analysis
provides evidence of volatility breakpoints at the lower scales of the data.
Fernandez (2007) uses the Iterative Cumulative Sum of Squares (ICSS) and the
Wavelet-Based Variance Analysis to examine the effect of political conflicts in the
Middle East on stock markets worldwide. She finds evidence that political instability
in the Middle East has increased the sensitivity of stock markets to exchange rate
risk. Fernandez (2008) concludes that the ongoing Israeli-Palestinian conflict has a
major impact on the volatility of equity markets in the world. Hammoudeh and Li
(2008) examine sudden changes in volatility for five Gulf stock markets using the
(ICSS) algorithm. They find that most Gulf stock markets are more sensitive to
101
major global events than to local and regional events like the 1997 Asian crisis, the
collapse of oil prices in 1998, the adoption of the price band mechanism by OPEC in
2000, or the September 11 attacks.
Lagoarde-Segot and Lucey (2009) investigate the MENA stock markets
vulnerability to shift contagion, focusing on recent major financial crises on the
MENA markets, namely the Asian, Russian, Brazilian, Turkish, and Argentinean
financial crises, plus the September 11 attacks,. They find a positive association
between market development and financial vulnerability in the MENA region. Their
outcomes indicate that the MENA equity markets sporadically move together in
synchronized fashion during periods of external stress (Asian financial crisis,
September 11 attacks, and the Brazilian financial crisis) and that the equity markets
of Turkey, Israel and Jordan were the most vulnerable over the 1997 2009 period,
followed by Tunisia, Morocco, Egypt and Lebanon; they conclude that MENA-based
diversification strategies may be relatively inefficient in periods of global volatility.
They also observe an increasing time-varying vulnerability levels in the MENA
region, with a possible structural break since the recent global financial crisis. They
conclude that the presence of mechanisms aimed at limiting price fluctuations in
Israel, Turkey, Tunisia, and Lebanon do not seem to prevent financial contagion.
A number of methods may be used to identify the dates on which a dramatic change
in time series takes place. There are two methods to detect this change. The first
approach is reflected in the search for structural breaks and the listing of good and
bad days signified by sharp changes in the data series, followed by an attempt to find
important events that occurred on those dates. One research strategy is to include
dummy variables for all such breaks (Blass et al., 2004). However, this strategy is
102
problematic because there is a danger of ex-post fitting if breakpoints are selected as
a result of an observed change in the variable of interest (Ender & Sandler, 2008).
The second approach has the opposite starting point: it is based on forming
hypotheses a priori about the kind of events that might influence the stock market
Most classical tests against changes in the coefficients of a linear regression model
assume that there is just a single change under the alternative, or that the timing and
the type of change are known. Recently there has been a surge of interest in
recovering the date of a shift if one has occurred, or in developing methods that
allow for several shifts at once (Zeileis et al., 2003). Multiple structural break models
are becoming more appealing to academics and practitioners, especially when
applied to emerging markets. Most structural breaks in equity markets returns and/or
market risk exposure can be traced to the liberalization episodes of emerging equity
markets or to major global or regional events such as currency crises (Goldberg &
Veitch, 2002a, b; Arouri et al., 2009). From this theoretical framework, one can
implement the classical Cumulative Sum of Squared (CUSUMQ) test for structural
changes, initially proposed by Brown et al. (1975). However, despite its simple
implementation, dating a structural break with the CUSUMQ test is difficult
(Nguyen & Bellalah, 2007).
The advances in time series econometrics have introduced several testing procedures
to account for the subject of structural changes of time series over time. The models
of Bai and Perron (1998) and Zivot and Andrews (1992) provide tests to
endogenously detect the timing of a structural break with no knowledge of the
breakpoint a priori, and to test for its statistical significance. Antoshin et al. (2008)
argue that the size and power of these tests can be significantly distorted by small
103
sample sizes, small break sizes, small segment sizes, break clustering, and the use of
heteroscedasticity and autocorrelation corrections. Nevertheless, the existence of
multiple structural breaks analysis is attractive, especially in long-run time series
analysis where different factors may affect the behaviour of tested data in different
time periods (Zhang et al., 2006).
Zeileis et al. (2003) examine multiple structural changes in oil prices using Bai and
Perron s (2003) model. They find a number of breakpoints that coincide with
historical, political, or economic events which might have caused these breaks, such
as the Arab oil embargo after the Yom Kippur war, and the start of the Iranian
revolution followed by the war between Iran and Iraq in 1979, or are the joint
product of various minor events such as a worldwide slowdown of demand for oil,
the emergence of Great Britain, Norway, and Mexico as major suppliers in
international oil markets, and internal quarrels in the OPEC cartel, such as the one in
which led Saudi Arabia to increase its productions in 1985.
Motivated to extend their previous work on multiple breakpoints analysis, Arouri et
al. (2009) examine the multiple structural breaks of the dynamic conditional
correlation (DCC-GARCH) of Latin American emerging markets using the Bai and
Perron (2003) procedure. They find that there were sudden increases in the dynamic
conditional correlation following the Asian and the Brazilian financial crises in
1997 1998 and, to a lesser extent, after the stock market crash in 1987 and the Latin
American crises in 1994 and 2001. Goldberg and Delgado (2001) attempt to identify
the breakpoints of emerging market returns to confirm whether they are indicative of
financial integration. Using Structural VAR for individual stocks from Latin
America and Asia, they find that not all the significant breaks dates are attributable
104
to financial integration but are mainly responses to regional or global events such as
currency crises. They conclude that tracing the evolution of changes in pricing
regimes with respect to political, economic, and financial events could prove to be
useful for future research.
The main goal of this chapter is to examine whether the country beta series contain
several structural breakpoints due to the reaction of MENA stock markets to major
global crises. Chapter 4 focused on the structural break in the country beta which
was predetermined by the crisis dummy of the war on Iraq. Over the sample period,
there could have been several significant breakpoints in the country betas series
caused by other global events. Due to the unique paramatization of the Bai and
Perron (2003) model to handle multiple structural breaks with short time spans, this
chapter examines the multiple structural breaks in the country beta series of MENA
markets to examine the impact of other key global events.
Recent studies suggest that international financial crises and other global risks have
intensified the links between developed and emerging markets as well as those
among emerging markets (Gunduz & Hatemi-J, 2004). This chapter follows
Lagoarde-Segot and Lucey (2009) in focusing on the major financial and geopolitical
risks and the crisis timelines that affected the global markets. 15
4.2. Description of the unstable crises periods
This section explores the timings of major financial crises, beginning with the Asian
financial crisis in 1997, the Russian crisis in 1998, the Brazilian crisis in 1999, the
The analysis in this chapter drops the Argentinean solvency crisis because of its local impact (see
Lagoarde-Segot & Lucey, 2006).
15
105
Turkish crisis in 2001, the September 11 2001 terrorist attacks, and the 2003 war on
Iraq. Table 4-1 summarizes the crisis timeline.
Table 4-1. Crisis timelines
Crisis Name
Asian financial crisis
Russian financial crisis
Brazilian financial crisis
Turkish financial crisis
September 11, attacks
War on Iraq, 2003
Origin
Asia
Russia
Brazil
Turkey
USA
Iraq
Stable crisis period
1997:10:1-1997:10:22
1998:6:06-1998:8:05
1998:11:01-1998:12:31
2000:12:05-2001:2:14
2001:6:27-2001:8:26
2002:12:09-2003:3:17*
Unstable crisis period
1997:10:23-1997:11:22
1998:8:06-1998:10:05
1999:1:01-1999:3:01
2001:2:15-2001:3:13
2001:9:14-2001:10:13
2003:3:18-2003:5:01*
Source: Lagoarde-Segot and Lucey (2009). * Taken from Paleari et al. (2005) and Chapter 3.
According to Erdal and Gunduz (2001), Choudhry (2005b), Miller (2006), Fernandez
(2006), and Hammoudeh and Li (2008), amongst others, the breakout of the Asian
financial crisis began on July 2, 1997. The crisis was caused by the decision of the
Thai government to float its currency and end its peg to the US dollar. In less than
two weeks, other currencies in Asian countries such as the Philippines, Malaysia, and
Indonesia came under strong pressure, raising fears of financial contagion. The crisis
intensified on October 23 1997, when the stock market of Hong Kong lost onequarter of its market value in four days on fears over the dramatic increase of shortterm interest and pressures on the Hong Kong dollar (Lagoarde-Segot & Lucey,
2009).16 The crisis was ameliorated when the International Monetary Fund (IMF)
signalled its approval bailout package to South Korea on 3 December (Timeline.com,
Miller, 2006).
The crisis quickly spilled over to Russia, with the initial shock to the Russian bond
market occurring on August 6, 1998. The Russian government devalued the currency
16
The short-term interest rates increased by 300% to fend off speculative attacks on the Hong Kong
dollar. The stock market plunges in Hong Kong stock exchange wiped out nearly $29.3 billion
dollar of its market value (pbs.org).
106
and signalled the government s default on its debts. The stock market reacted one
week later, and the fallout continued until the end of September, 1998 (Alper &
Yilmaz, 2004; Lagoarde-Segot & Lucey, 2009). The pressure on emerging markets
that pegged their currencies to the US dollar increased as capital started to flow out,
following the harsh lessons of the Asian financial crisis.17 The Brazilian financial
crisis (which is often associated with contagion from the Russian financial crisis)
caused a severe devaluation of the nation s currency and caused further
destabilization of its stock market, but the main repercussions for the capital market
were felt from the end of November 1998 to March 1999 (Lagoarde-Segot & Lucey,
2009).
In the late 1990s the Turkish financial system experienced highly volatile inflation
problems that caused a severe currency crisis. The IMF engineered a stabilization
plan, and in December 1999, the Turkish central bank adopted a peg-stabilization
system to ease the banking problems. The program did not work, and the central
Bank of Turkey abandoned the inflation targeting and the fixed exchange rate regime
five months prematurely, under less than optimal circumstances. The heated public
debate over the handling of the crisis created panic when the Prime Minister of
Turkey clashed with the Turkish President over the banking reforms. Despite a
massive injection of $7.5 billion by the IMF in December 2000, the Turkish
government was forced to devalue the Turkish Lira in February 2001.18
The total capital flows declined from a large inflow of US$132.2 billion in 1996 to an outflow of
S$44.9 billion in 1998 (see Das, 2010; Arouri et al., 2009).
17
The cost of the crisis and the economic restructuring of the Turkish economy were significant. The
GNP fell by 8.5 percent in 2001 while inflation rose to 86 percent, and public sector borrowing
jumped to 19.6 percent of the GDP (Akyuz and Boratav, 2003; Yesilada et al., 2004; Lagoarde-Segot
and Lucey, 2006).
18
107
The terrorist acts on US soil took place on September 11, 2001. The events of
September 11 exacerbated an already very difficult situation in the global economy,
and had a significant impact on demand and general economic activities. The attacks
resulted in a loss of US$13 billion of destroyed private and government equity, an
estimated US$35 billion to US$50 billion of insured losses, and extreme stock
market volatility that wiped out US$1.2 trillion. Trading in the US stock market was
suspended for a week, resuming on 17 September (Choudhry, 2005a). LagoardeSegot and Lucey (2009) date the unstable crisis period as extending until October 13
2001, due to ongoing volatility of the US stock markets.
On 12 September 2002 US President George W. Bush announced to world leaders
gathered at the United Nations that the Iraq regime of Saddam Hussein posed a
grave and gathering danger to peace, and urged world leaders to move deliberately
and decisively to hold Iraq to account (Paleari et al., 2005). On 20 March 2003, the
war on Iraq commenced with President Bush s delivery of a speech, signalling the
start of the US-led campaign to overthrow Iraqi leader Saddam Hussein. The war
was declared accomplished on May 1, 2003.
4.3. Data and methodology
4.3.1. Data
The data for this chapter are the weekly observations identical to the data and groups
that were used for Chapter 3, which contains a description of the data.
4.3.2. The Kalman filter model
The country beta series used for this analysis are estimated from the Kalman filter
approach within the framework of the state-space model without the crisis dummy
108
variable of the war on Iraq in 2003. A detailed description of the conditional market
model and the derivation of the Kalman filter technique appear in Chapter 3 and the
Appendix.
4.3.3. Structural breaks analysis
This section explores sudden changes in the time-paths of the estimated country beta
series using the Bai and Perron (2003) model. The following description of the Bai
and Perron (2003) model is heavily based on Arouri et al. (2009), Zeileis et al.
(2003), and Zeileis and Kleiber (2005). Suppose the time series (country beta) ( yt )
can be modelled as the following function:
yt
X tT
t
for (t
1,...n)
(4.1)
That contains ( m ) breaks ( n1 ,...nm ) in the mean of the series, then the problem of the
dating of the breakpoints ( n1 ,..., nm ) that minimize the function:
n1 ,..., nm
arg; min(n1 ,...nm ) RSS n (n1 ,...nm )
(4.2)
is ( 2 1 )
where the RSS is the residual sum of squares of the linear equation, and
vector of the corresponding coefficients that may vary over time, X t
(1, yt 1 )T is a
vector of independent variables with the first observation equal to 1, and
t
is the
disturbance term that assumed to be normally distributed with zero mean and
constant variance. The structural stability test assumes the null hypothesis of no
structural break: H 0 :
i
0
for (i 1,..., n) against the alternative that the series
contains ( m ) breaks. Bai and Perron (2003) suggest obtaining the estimators via the
OLS approach and they provide dynamic algorithm to compute the number of ( m )
109
breaks efficiently.
Since the sample size is large, the search for the ( m ) breakpoints will be for every 26
weeks as the shortest distance (or the minimal segment size) between two breaks
(number of regressors); this may be considered a bandwidth or trimming parameter,
with a maximum number of 1 breakpoint for the whole sample period (Zeileis &
Kleiber, 2005).19 Bai and Perron (2003) state that the Akaike Information Criterion
(AIC) may overestimate the number of breakpoints and that the Bayesian
Information Criterion (BIC) is more suitable in the selection procedure; therefore, the
BIC will be used as a guide for the optimal number of the breakpoints contained in
the country beta series. The lowest BIC corresponds to the optimal number of
breakpoints. If the maximum number of breakpoints is equal to one breakpoints
corresponding to the lowest BIC, then the Bai and Perron (2003) test will be run
again with higher number of breakpoints. The optimal number of breakpoints will be
chosen when no further breakpoints are found.
4.4. Empirical results
4.4.1. Descriptive statistics of the country betas
To give a better overview of the properties of conditional country betas, Table 4-2
provides the descriptive statistics for the country betas. Most MENA markets
produce low average betas, especially Kuwait, Jordan, Qatar, Saudi Arabia, and
Tunisia. Egypt, Morocco, and Israel yield higher average betas in general. The
average beta of Israeli is almost the world market risk (0.998), which suggests that
For long weekly time series, Zeileis and Kleiber (2005) recommend a minimal segment size of six
months.
19
110
the country beta of Israel has a tendency revert to its grand mean of world market
risk. Negative betas are seen except in Israel. The Turkish equity market is known to
be one of the most unstable markets in the region due to the volatile nature of its
economic and financial systems; its average country beta is over the world market
risk. Based on the standard deviation, the Turkish country betas are the most volatile
in the sample. It is important to note that, with the exception of Israel and Turkey, all
MENA markets exhibited negative betas until recently, which is considered evidence
that they are yet to be integrated with the global capital market. This evidence
supports the results for the static country beta results found in section 3.4.3 in
Chapter 3. This is also not surprising that the country beta patterns is different for
each country in this period because the MENA stock markets show signs of
increased market risk exposure after their stock market liberalization and the Iraq
war due to the increased oil prices in the last decade.
Before the Bai and Perron (2003) model is examined, the time series of country betas
are tested for stationarity. The Kwiatkowski, Phillips, Schmidt, and Shin (1992) test
(KPSS) is employed to examine the stationarity of the country beta series. The test
conducted including constant and trend. The choice of the lag length for both tests is
automatically determined by the Newey-West bandwidth. Table 4 reports the results
of the unit root tests of stationarity of the country betas. The null hypothesis for
KPSS, that country betas do not contain a unit root, is not rejected for all the country
betas. Based on these results, the country beta series are treated as stationary.
4.4.2. Structural breaks results
Table 4-3 reports the results of the Bai and Perron (2003) structural break test in the
country betas. The optimal number of breakpoints (m) is chosen as the one associated
111
Table 4-2. Descriptive statistics of the country betas series
Egypt
Israel
Jordan
Mean
0.468
0.998
0.136
Maximum
1.691
1.760
0.860
Minimum
-0.361 0.307 -0.460
Standard Dev. 0.493
0.266
0.144
Skewness
0.540
0.865
1.943
Kurtosis
2.446
3.586 11.413
Jarque-Bera
44.228 100.084 257.68
Probability
0.000
0.000
0.000
KPSS
0.129** 0.111* 0.203***
Kuwait Lebanon Morocco Oman
0.070
0.542
-0.318
0.192
0.362
2.607
20.365
0.000
0.077*
0.110
0.815
-0.151
0.191
1.606
5.757
332.855
0.000
0.162*
Qatar
Saudi Tunisia Turkey
Arabia
0.163
0.035
0.216
0.051
0.061
1.276
1.616
1.192
0.822
0.713
0.474
3.540
-0.610 -0.620 -0.085 -0.437 -0.170 -0.033
0.342
0.234
0.197
0.159
0.111
0.757
1.375
0.438
0.463
0.761
0.150
0.184
5.600
6.047
2.929
7.131
3.160
2.524
430.810 186.730 12.058 359.989 2.1353 10.846
0.000
0.000
0.003
0.000
0.344
0.005
0.051* 0.051* 0.093* 0.075* 0.122** 0.120**
Notes: *, **, *** denote significant at 1%, 5% and 10% respectively. The first 10 observations of the weekly estimated conditional betas are
dropped from the descriptive statistics. The Jarque-Bera test statistics are to test the null hypothesis of the normal distribution. KPSS denotes the
Kwiatkowski, Phillips, Schmidt, and Shin test. The lag length in the KPSS is automatically chosen by the Newey-West bandwidth. The critical
values for the KPSS with intercept and trend are: 0.216, 0.146, and 0.119.
112
Egypt
2.0
Israel
2.0
Jordan
1.2
1.5
1.6
0.8
.4
1.0
1.2
0.4
.2
0.5
0.8
0.0
.0
0.0
0.4
-0.4
-.2
-0.5
95
96
97 98 99
00 01
02 03
04 05 06
07
08
Morocco
2.0
96
97 98 99
00 01
02 03
04 05 06
07
08
-0.8
Turkey
0.0
-0.5
95
96
97 98 99
00 01
02 03
04 05 06
07
Saudi Arabia
.8
08
0.2
0
0.0
-0.2
95
96
97 98 99
00 01
02 03
04 05 06
07
08
Qatar
02
03
04
05
06
07
08
-0.8
08
03
04
05
06
07
-.2
08
95
96
97 98 99
00 01
02 03
04 05 06
07
06
08
Oman
1.0
0.5
0.0
-0.5
00
01
02
03
04
05
06
07
08
05
06
07
08
Tunisia
00
01
02
03
04
Figure 4.1: Country betas estimated by the Kalman filter approach.
113
-.4
1.5
.0
-0.4
01
07
.2
0.0
00
04 05 06
.4
0.4
-.4
02 03
Lebanon
.6
0.8
.0
00 01
0.4
1.2
.4
97 98 99
0.6
1
-1
96
0.8
2
0.5
95
1.0
3
1.0
-.8
95
4
1.5
-1.0
0.0
Kuwait
.6
-1.0
00
01
02
03
04
05
07
08
with the minimum BIC. Due to the initialization of the Kalman filter, the first ten
observations of the country betas are dropped from the analysis. The chapter reports
the year and month only for the breakpoint dates. In table 4-3, the null hypothesis of
the Bai and Perron (2003) clearly reject the stability of the country betas due to the
existence of the structural breaks in column 2 under the BIC.
Looking at the structural breaks results, the present breakpoints around the 1997
Asian financial crisis for Israel and Turkey are clear in the month of October. The
Asian financial crisis had severe impact not only on the Asian and many developed
and emerging equity markets but appears to have been impacting the stock market
exposure of Israel and Turkey in the MENA region. According to Pownall and
Koedijk (1999), the Asian financial crisis created further turbulence in markets all
over Asia and in the global markets. In addition, the Asian financial crisis slowed the
growth of the world economy due to the shortfall in demand from the Asian region
and caused both a severe regional recession and deterioration in the trade balance of
many Asian countries and their trade partners. The impact of the financial crisis was
an increase in the observed volatility of financial markets and capital flows around
the world (Choudhry, 2005b).
Although the MENA countries do not have much in common with Asian countries,
nevertheless as investors become worried about the prospects of emerging markets
around the world, their fears could have infected many other emerging markets,
including those of Turkey and Israel. It can be seen that the country betas of Egypt,
Jordan, Morocco, and Kuwait did not experience structural breaks during the months
from July through December (containing both stable and unstable crisis periods),
which suggests that these MENA equity markets were relatively immune to the
114
Asian financial crisis: findings similar to those of Girard and Ferreira (2003) using
cointegration analysis, and Lagoarde-Segot and Lucey (2006) using the vulnerability
test in fixed effects model. Using daily data, Gunduz and Hatemi-J (2004) find a
unidirectional Granger causality from exchange rates to stock prices for Israel and
Morocco, before and after the Asian financial crisis. They conclude that in the case
of emerging markets, changes in exchange rates will be more influential as the
market develops. The results of this chapter also confirm the results found in Blass et
al. (2004) on the breakpoint for the stock market in Israel.
Table 4-3. Bai and Perron s structural breakpoints with their 95% confidence intervals
Market
BIC Breakpoint 95% interval
Crisis
Egypt
20
[1998:08] [1998:05 1998:08] Russian financial crisis, 1998
Egypt
[1999:02] [1999:01 1999:04] Brazilian financial crisis, 1999
Egypt
[2001:09] [2001:08 2002:02] September 11, 2001 attacks
Egypt
[2003:03] [2003:01 2003:04] War on Iraq, 2003
Israel
18
[1997:10] [1997:09 1997:11] Asian financial crisis, 1997
Israel
[1998:10] [1998:09 1998:11] Russian financial crisis, 1998
Israel
[2001:09] [2001:09 2001:10] September 11, 2001 attacks
Israel
[2003:03] [2003:02 2003:05] War on Iraq, 2003
Jordan
13
[2001:10] [2001:10 2001:12] September 11, 2001 attacks
Jordan
[2003:03] [2003:02 2003:06] War on Iraq, 2003
Kuwait
17
[1998:08] [1998:08 1998:09] Russian financial crisis, 1998
Kuwait
[2001:09] [2001:08 2001:10] September 11, 2001 attacks
Kuwait
[2003:05] [2003:03 2003:06] War on Iraq, 2003
Morocco
13
[1998:08] [1998:02 1999:03] Russian financial crisis, 1998
Morocco
[2003:02] [2002:12 2003:05] War on Iraq, 2003
Oman
10
[2001:03] [2001:02 2001:12] Turkish financial crisis, 2001
Saudi Arabia 10
[2001:02] [2000:12 2001:04] Turkish financial crisis, 2001
Tunisia
10
[2001:09] [2001:09 2001:10] September 11, 2001 attacks
Tunisia
[2003:04] [2003:04 2003:05] War on Iraq, 2003
Turkey
14
[1997:10] [1997:09 1997:11] Asian financial crisis, 1997
Turkey
[1999:02] [1999:02 1999:07] Brazilian financial crisis, 1999
Turkey
[2001:02] [2001:02 2001:02] Turkish financial crisis, 2001
Turkey
[2003:03] [2003:01 2003:07] War on Iraq, 2003
Notes: The breakpoint selection procedure in the Bai and Perron (2003) model is based on the
Bayesian Information Criteria (BIC). The maximum number of breaks is set to be 1, if the number of
breaks is equal to 1, with a minimum span of 26 weeks; a higher number of breaks is chosen to
capture all the breakpoints until no further breakpoints are found.
115
According to Alper and Yilmaz (2004), volatility persistence and contagion were
rather strong during the Russian financial crisis infecting major stock markets all
around the world. The findings show that the country betas of Israel, Kuwait, and
Morocco responded immediately after the outbreak of the financial crisis in Russia in
the month of August, 1998. The result for Israel is noticeably inconsistent with the
findings of Blass et al. (2004) about the impact of the Russian financial crisis on the
market risk premium. Gunduz and Hatemi-J (2004) find that the MENA equity
markets were hit by the Russian crisis in 1998 following the Asian financial crisis;
especially in Israel and Morocco. However, they did not examine the adverse effects
of this crisis on the equity market of Kuwait. Lagoarde-Segot and Lucey (2006) find
evidence of contagion in Egypt and Tunisia during both the Russian and the
Brazilian financial crises.
The impact of the Turkish financial crisis can be seen from the results. First, the
country beta of Turkey experienced a structural break that coincided with the
devaluation of the lira and the end of the pegged exchange rate arrangement. This
highlights the effect of Turkey s economic and political instability on the country
beta, supporting the argument of Akyuz and Boratav (2003) that the Turkish
financial crisis aggravated the systematic risk of Turkey. Second, although the
Turkish financial crisis originated domestically, the findings suggest that the country
betas of Egypt, Jordan, Kuwait, Oman, Saudi Arabia, and Tunisia experienced
structural breaks because of it. With the exception of Egypt, these breakpoints
occurred in the first week of March; the breakpoints in the country beta of Turkey
occurred at the end of February 2001. With the exception of Lebanon, these results
support the findings of Lagoarde-Segot, and Lucey (2006, 2009) on the vulnerability
116
of the MENA equity markets to the Turkish financial crisis. The implications from
this evidence are that the Turkish financial crisis was far from being localized and
not only caused a structural break in the country beta of Turkey, but also caused
destabilization of the country betas of other MENA markets.
The impact of the September 11 attacks and the declaration of war on terror were a
major cause of political and regional instability in the Middle East, creating major
disruption to the regional and global economy. One of the major consequences of the
September 11 attacks is that it shook the global confidence of future investments in
the Middle East region. For this reason, assessing the impact of the terrorist attacks
on the stock market has become a key issue, in order to assess the risk and magnitude
of the direct and indirect effects of terrorist attacks (Carrera & Musso, 2009). When
examining the timing of this event, it can be seen that the country betas of Egypt,
Israel, Jordan, Kuwait, and Tunisia experienced cotemporaneous structural breaks,
implying that these equity markets adversely responded to the attacks. The other
MENA equity markets were not directly affected by this global shock. While a
number of studies have investigated the impact of the September 11 attacks on global
business activities, empirical evidence points to the influence of this impact on the
volatility of international capital markets and individual company or industrial betas.
The results of this chapter indicate that such a global shock can impact not only on
company or industry betas around the globe, but also on the market risk profile of a
particular capital market, reflecting a significant influence on the country beta. For
the other MENA countries, the results may imply that the impact of the September
11 attacks was temporary.
The war on Iraq began on March, 2003 and was declared accomplished on May 1,
117
2003. The results from the previous chapter show that the war had a positive impact
on all the country betas, but a statistically significant and evident impact on those of
Egypt, Kuwait, Morocco, and Tunisia. The findings of the structural breaks analysis
show that the war on Iraq caused further structural breaks in the country betas of
Egypt, Israel, Jordan, Kuwait, Morocco, Tunisia, and Turkey from March to May
2003. The results of the Bai and Perron (2003) multiple structural breaks confirm the
results of the Kalman filter and show that other countries suffered systematic shocks;
significant damage occurred to the country betas of Israel, Jordan, and Turkey. These
results were made possible by the unique paramatization of the Bai and Perron
(2003) model to locate significant structural breaks and capture the effect of the war
risk on the MENA stock markets. Consequently, this empirical evidence indicates
that the war on Iraq increased regional uncertainty and destabilized further equity
market risks in the MENA region.
Table 4-3 includes the detected breakpoints in the country betas with their 95%
confidence intervals. For example, the breakpoint for the Brazilian financial crisis
falls into the 95% confidence interval [1999:01 1999:04] and the war on Iraq
[2003:01 2003:04] for the country beta of Egypt. In addition, some of the confidence
intervals have tight or close bounds. For example, the 95% confidence interval for
the Turkish financial crisis spans [2001:02 2001:02] for the country beta of Turkey,
the 95% confidence interval for the Russian financial crisis spans [1998:08 1998:09]
for the country beta of Kuwait, and the September 11 attacks span [2001:09
2001:10] for the country betas of Israel, Kuwait, and Tunisia. On the other hand, the
widest 95% confidence intervals for the Russian financial crisis spans [1998:02
1999:03] for the country beta of Morocco, and the Turkish financial crisis spans
118
[2001:02 2001:12] for the country beta of Oman. Finally, although not all the
breakpoints are included in the table, other structural breaks are present that coincide
with other regional events, such as the pullout of South Lebanon on May 22 2000
(for Israel), and the Israeli war on Lebanon in 2006 (for Lebanon); these breakpoints
are not pervasive across other MENA markets.
4.5. Conclusion
This chapter analyses the structural breaks in the country beta of the MENA markets
using the Bai and Perron (2003) model. The results show that the country betas have
experienced several structural breaks that coincide with regional and global events.
The chapter finds that the MENA markets have coped with a number of global and
regional shocks that destabilized them.
The main results show that the war on Iraq had greater impact than any event in the
MENA region, compared with financial events like the Russian and the Turkish
financial crises and political events like the September 11 2001 attacks. In addition,
the analysis of this chapter finds that the Egyptian, Israeli, and Turkish markets are
the most receptive to these exogenous shocks. These MENA markets have become
more sensitive, a signal of their increased integration with the global market. In this
sense, deteriorations in the global economy, coupled with global political and
regional risks, could have a significant contribution in amplifying the country beta in
these markets.
The findings of this chapter deepen our understanding of the stock market
development process in the MENA region. It provides valuable information for
global investors seeking diversification in emerging markets. The empirical evidence
119
shows with no doubt that the MENA stock markets are becoming more sensitive to
major global and regional events, and possibly to other global risks. If the MENA
equity markets are becoming more sensitive to regional and global unrest, they will
no longer be considered safe havens for international investors since any shock to the
global system will reach these markets, depending on their degree of integration with
the global financial system.
For policy and investor perspective, a structural break in country beta suggests a
different level of risk aversion, meaning that local and international investors will
adopt a different strategy of asset pricing and asset allocation in the MENA stock
markets. Fore example, in a country or a region that experience economic and
financial instability due to financial and geopolitical instability, local and
international investors might require additional risk premium to compensate for
investing in risky markets.
120
Determinants of Country Betas of MENA Equity Markets:
A Panel Analysis
The previous two chapters examined the impact of major regional and global shocks
on the country beta, using the time series approaches of the Kalman filter and the Bai
and Perron (2003) models. The purpose of this chapter is to examine the explanatory
power of economic variables on the variation of country betas in the MENA region
using a panel data regression analysis. The variables used in this chapter are interest
rate, inflation, exchange rate with the US dollar, money supply, foreign currency
reserves, and oil price. A number of studies have suggested that the country beta may
fluctuate with changes in a country s fundamental economic variables (see for
example, Andrade & Teles, 2006, 2008; Gangemi et al., 2000; Goldberg & Veitch,
2002a, b; 2010; Marshall et al., 2009; Patro et al., 2002; Tourani-Rad et al., 2006;
Verma & Soydemir, 2006). This chapter contributes to the literature in two ways.
First, the analysis of the MENA country beta model has not received any previous
attention in the literature. Second, the analysis focuses on the country-by-country
analysis; only the works of Patro et al. (2002) and Marshall et al. (2009) have
previously examined the determinants of country betas in a panel data regression
analysis for both developed and emerging equity markets.
121
5.1. Introduction
There is a general consensus among researchers that macroeconomic news affects
security prices. Ultimately, returns on stocks and bonds reflect real economic
activity, so in the long-run we expect to see a relationship between macroeconomic
activity and equity returns (Patro et al., 2002). Earlier studies such as that of Bos and
Newbold (1984) attribute the variation of systematic risk to the influence of
microeconomic and macroeconomic variables on stock returns. Groenewold and
Fraser (2000) find that much of the time variation in industrial betas is explained by
changes in aggregate economic conditions. According to Abell and Kreuger (1989),
the importance of identifying the movements of betas resulting from macroeconomic
change has been raised by the increasing sophistication and awareness of market
participants to track macroeconomic descriptors and hedge against them.
The literature has paid considerable attention to the impact of financial and economic
variables on the returns of both developed and emerging stock markets. A strand of
this literature recognizes that small, developing countries do not fully support the
single factor model of the International Capital Asset Pricing Model (ICAPM) given
their the incomplete market integration (see for example, Ferson & Harvey, 1997,
1999b; Nummelin & Vaihekoski, 2002; Saleem & Vaihekoski, 2008). An early study
by Harvey (1991), for example, finds that the time-varying covariance of a country s
equity returns with global market returns explains some, but not all, of the behaviour
of the country s returns over time. He attributes these results to incomplete market
integration, the existence of more than one source of risk, and some other
misspecification problems.
122
According to Roll et al. (1986), economic variables exert a systematic influence on
stock returns, and the valuation of stock prices indicate that the expected returns are
related to changes in the state of the economy. Since then, a large body of literature
has documented how economic variables affect the stock returns in the short-run (see
for example, Abell & Krueger, 1989; Anderson et al., 2005; Bekaert & Harvey,
1997; Ferson & Harvey, 1993; Harvey, 1995c, 2000; Schwert, 1990). A number of
studies conclude that economic variables can affect the stock returns on the long-run
(see for example, Dickinson, 2000; Gjerde & Saettem, 1999; Groenewold & Fraser,
1997; Maysami & Koh, 2000; Mookerjee & Yu, 1997; Mukherjee & Naka, 1995;
Wassal, 2005). On the other hand, other studies have concluded that the economic
variables capture a large component of the predictable variation in the stock returns
(see for example, Ferson & Harvey, 1991, 1993, 1995; Harvey, 1995a; Hassan et al.,
2003; Johnson & Sakolis, 2008; Yao, et al., 2005). Moreover, a fair number of
studies have documented the degree of co-movements between stock returns and
several economic variables (see for example Andrade & Teles, 2006, 2008; Cheung
& Ng, 1998; McMillan, 2005).
Since multifactor models are crucial in explaining variation in equity returns, the
question is which variables should be considered in the equilibrium asset models
(Hassan et al., 2003; Saleem & Vaihekoski, 2008). The empirical evidence suggests
that emerging markets can still be segmented from world capital market and that
local information is often more related to the country-specific variables while using
global factors across countries require some strong assumptions (see for example
Bekaert & Harvey, 1995; Bilson et al., 2001; Fedorova & Vaihekoski, 2009; Ferson
& Harvey, 1993). Bruner et al. (2008), for example, find evidence that emerging
123
markets remain at least partially segmented and country-level variables dominate the
global variables in explaining cross-sectional variations of stock returns. However,
assumptions of perfect integration/segmentation could be unrealistic since both
global and country-level variables might be important even in a financially
segmented environment (Bilson et al., 2001; Bodurtha et al., 1989). According to
Bodurtha et al. (1989) and Bruner et al. (2008), failure to reject the market
integration hypothesis must not be taken as strong evidence against some degree of
market segmentation.
At the country level, the literature has investigated the predictive power of economic
variables on the country risk (country beta) of stock markets for both developed and
emerging equity markets. According to Wdowinski (2004), the examination of
country beta at the international level is particularly important in an assessment of
investment projects, especially for emerging markets. Verma and Soydemir (2006)
add that the analysis and assessment of country beta are considered one of the major
challenges facing both international investors and academics, in both emerging and
developed markets. Within the country beta model framework, the relationship
between the returns on domestic equities and the returns on the rest of the world
equities are related to the country s economic variables (Gangemi et al., 2000).
A large body of literature has documented how the country beta may fluctuate with
changes in the state of economy. Studies investigating this relationship have been
documented by Andrade and Teles (2006, 2008), Bilson et al. (2001), Gangemi et al.
(2000), Goldberg and Veitch (2002a, b, 2010), Harvey and Zhou (1993), Marshall et
al. (2009), Patro et al. (2002), Tourani-Rad et al. (2006), and Verma and Soydemir
(2006), among others. One issue, whether to consider the country beta as endogenous
124
or exogenous with regard to economic variables, is raised by Andrade and Teles
(2008), according to whom the endogeneity of country beta appears to be well
established; but which economic variables are able to significantly affect country
beta and, above all, how economic policies can affect the country beta over time, is
not yet settled.
Since the country beta models can be best characterized at the domestic level, the
natural question is whether the country beta model should also incorporate global
variables for developed and emerging markets. A number of studies have shown that
multifactor models using global factors have been reasonably successful in
characterizing equity returns for developed markets due to their integration to the
global market (see for example Arouri et al., 2004; Arouri, 2006; Bodurtha et al.,
1989; Bruner et al., 2008; Chan et al., 1992; Dominguez, 2007; Engle & Rodrigues,
1993; Gerard et al., 2007). However, the low exposure of emerging equity markets to
commonly used global factors such as, but not limited to, the world portfolio, world
interest rate, trade-weighted world exchange rate, international industrial production,
and world inflation are signs of incomplete market integration (see for example
Bilson et al., 2001; Bodurtha et al., 1989; Engle & Rodrigues, 1993; Harvey, 1995c;
Rouwenhorst, 1999). A study by Bilson et al. (2001) shows that if global factors are
added to the country s macroeconomic factors, the model loses much of its
significance compared to the results gained on country-level macroeconomic factors
only. However, Verma and Soydemir (2006) find that global factors such as world
interest rate and inflation among the G-7 countries can have varying impact on the
country beta of some Latin American countries such as Mexico, Brazil and Chile.
According to Bilson et al. (2001), the selection of the initial factors is ultimately
125
subject to criticism on the grounds of subjectivity and the arbitrary nature of the
selection process. This is an unavoidable problem; researchers can do little more than
look to prior research and form judgments as to the relevance of various factors. In
general, the literature indicates that the multifactor models for both country-level and
global factors are sensitive to the model choice, the choice of selected variables in
the regression, and how the variables are capable of significantly affecting equity
returns and country beta over time (see for example Andrade & Teles, 2006, 2008;
Bilson et al., 2001; Gangemi et al., 2000). According to Gangemi et al. (2000) and
Andrade and Teles (2006, 2008), the country beta models are generally explanatory
in nature and some of the theories on the relationship between the country beta and
the economic factors have difficulty in predicting the direction of the relationship.
The purpose of this chapter is to examine the extent to which economic variables are
able to explain the variation of country beta in the MENA region in a panel
regression analysis. A large number of studies have shown that the country beta may
fluctuate with changes in the country s economic variables (Andrade & Teles, 2006,
2008; Gangemi et al., 2000; Goldberg & Veitch, 2002a, b; 2010; Tourani-Rad et al.,
2006; Verma & Soydemir, 2006). Although these studies have investigated the
impact of the economic variables on the country beta for both developed and
emerging equity markets, the emphasis has been on the markets individually rather
than collectively. The understanding of economic policies and how they affect the
country beta is of great importance in attaining sustainable financial stability across
the MENA region. According to Cuyvers et al. (2008) and Heaney and Hooper
(1999), regional economic stability plays an important role in investment choice by
local and international investors: regional economic prosperity guarantees economic
126
benefits and encourages investment by firms in the short and long-run, promises
efficient resources and growth in the economy. Bilson et al. (2001) find evidence of
a common sensitivity of stock markets to country-level macroeconomic variables
when the markets are considered in aggregate or at regional level. Given this, it
would be of interest to policy makers and investors to pinpoint what economic
variables affect the country beta at the regional level. Since the focus of empirical
research has now shifted to the study of integration of equity markets into the
regional or the global market, it would be appropriate to adopt the panel regression
analysis to add more insight into how the economic policies can affect the country
beta of MENA equity markets.
5.2. Empirical framework
5.2.1. Description of equity markets
The stock market data are sourced from Thomson DataStream. The MENA stock
market indices are available from the Morgan Stanley Capital International (MSCI)
group for Egypt, Israel, Jordan, Morocco, Qatar, and Turkey, while the data for
Lebanon, Oman, Saudi Arabia, and Tunisia are available from Standard and Poor s
(S&P) group. The Kuwait stock market index is provided by the Kuwait Investment
Company (KIW). The world MSCI index is used as a proxy for the world market
index. The data used are monthly from January 2000 until December 2008, with the
exception of Qatar, where the index data starts from June 2002. The monthly returns
are calculated for the MENA stock indices and the world index as follows:
Rit
(5.1)
ln( Pit ) ln( Pit 1 )
where Pit and Pit
1
are the monthly
127
stock index for country i at time t and
t 1 respectively, and Rit represents the monthly logarithmic rate of return for the
monthly stock index for country i at time t 20.
5.2.2. Selection of the economic variables
This section describes the economic variables that are used in the empirical analysis.
The economic variables are as follows:21
(1) The interest rate proxied by the 3-month deposit rate, obtained from the
International Financial Statistics (IFS). The data are monthly frequency starting from
January 2000 to December 2008, whereas the data for Qatar start from September
2004.
The impact of the interest rate on the stock market has important implications for
asset valuation. In theory, an increase in interest rates raises the required rate of
return and negatively affects the stock returns. Chen et al. (1986), for example, find
that the spread between long- and short-term interest rates has a significant impact in
explaining the expected stock returns. The consideration of this variable is due to the
traditional role of the monetary policy and its possible impact on the country beta.
According to Marshall et al. (2009), an increase in interest rates reflects anticipation
of inflation growth, and in emerging markets economies inflation is generally
negatively perceived by financial markets as a risk for stable and sustained growth.
Andrade and Teles (2006) find that increases in interest rates tend to increase the
country beta of Brazil. Andrade and Teles (2008) find that the interest rate has had
20
Similarly, RMSCI , t is the return for the world MSCI index at time t .
There is no claim that all the fundamental factors which affect equity markets should be used, since
the identification of such factors would be a difficult task (see for example Bilson et al., 2001;
Gangemi et al., 2000; Panetta, 2002).
21
128
significant negative impact on the country betas of Thailand and Mexico and positive
effect in certain periods, especially during the Asian financial crisis.
(2) Money supply (M2), obtained from the International Financial Statistics (IFS).
The data for Qatar start from September 2004.
There is substantial empirical evidence that links money supply to stock markets.
According to Bilson et al. (2001), changes in money supply may impact on
economic variables, thereby having a lagged influence on stock returns. The theory
for this linkage is that the increased money supply leads to a portfolio rebalancing
towards other real assets such as stocks, and this reallocation results in an increase of
stock returns (Bodurtha et al., 1989; Dhakal et al., 1993). Mukherjee and Naka
(1995) suggest that the increase in money supply growth may lead to increase in
inflation, increase the discount rate, and reduce stock returns. The impact of money
supply on the country beta has been documented by Verma and Soydemir (2006) for
Mexico, Chile, and Brazil, while Patro et al. (2002) finds mixed evidence on the
relationship between money supply and country betas for 16 developed equity
markets. Gangemi et al. (2000) find that the money supply does not influence the
country beta of Australia, and Marshall et al. (2009) discover no significant
relationship between the country beta and money supply for a sample of 20 emerging
markets.
(3) Foreign currency reserves, obtained from International Financial Statistics (IFS).
Foreign currency reserves are considered a potential factor in identifying exchange
rate differences and capital flows, because changes in foreign currency reserves
indicate whether or not a trade balance deficit is being financed from abroad or by a
129
run-down of foreign currency reserves (Bodurtha et al., 1989). Mohanty and Turner
(2006) show that a large foreign currency reserve is associated with better financing
options and rapid growth in equity markets. According to Andrade and Teles (2006),
abrupt drops in foreign currency reserves are associated with a lack of sustainability
of the exchange rate regime. In addition, the effect of foreign currency reserves on
the country beta is related to its volatile nature and to the lack of credibility of the
fixed exchange rate regime. Andrade and Teles (2006) find a negative impact of
foreign currency reserves on the country beta of Brazil during the Russian financial
crisis, while Andrade and Teles (2008) find that foreign currency reserves are an
important variable in explaining the variations in country beta of Brazil and
Argentina, especially during their financial crises. A study by Goldberg and Veitch
(2002b) shows that foreign currency reserves have no significant impact on the
country beta of Mexico but that it was mainly the exchange rate variable that
influenced the country beta of Mexico over time.
(4) Exchange rate to the US dollar, obtained from International Financial Statistics
(IFS).
Deviation from the Purchasing Power Parity (PPP) among countries will be priced
and borne by investors (Bilson et al., 2001). Changes in exchange rates reflect capital
mobility from one financial market to another (Erdem et al., 2005). Jorion (1991)
indicates that the exchange rate is a significant variable for explaining equity returns.
Exchange rate fluctuations have been a major cause of crises in emerging markets,
such as the Asian financial crisis of 1997. While there is an economic linkage
between exchange rate and stock returns, determination of the sign of the
relationship remains an empirical issue (Bodurtha et al., 1989). Verma and Soydemir
130
(2006) find that the exchange rate has a significant impact on the country beta of
Mexico and Brazil. In addition, Goldberg and Veitch (2002b) found that the
exchange rate is the only variable that explains the variations in the country beta of
Mexico, especially after the Mexican currency crisis of 1994. Gangemi et al. (2000)
find that the exchange rate is the only variable that explains the variation in country
beta of Australia.
(5) Inflation rate (Consumer Price Index), obtained from DataStream. For Lebanon,
the data start from January 2000 until December 2006 (sourced from the Ministry of
Economics and Trade) and continue from January 2007 until December 2008 (from
DataStream).
According to the Fisher theory, if stocks provide a hedge against inflation, then the
relation between stock returns and inflation should be positive (Patro et al., 2002).
This is because equity returns represent a contingent claim on the realm assets of a
firm (Bilson et al., 2001). Fama s (1981) proxy hypothesis asserts that an increase in
inflation is expected to be followed by a decline in real economic activities and
corporate profits; thus stock returns will react negatively to an increase in inflation.
According to Panetta (2002), the negative impact of inflation surprises on stock
returns may be due to the anticipation of monetary policies. Boyd et al. (2001) argue
that high rates of inflation may cause exacerbating informational frictions in
financial markets, interfering with the efficiency of the financial system. The
sensitivity of the country beta to the inflation rate has been documented by Patro et
al. (2002), who find that it is positively related to the country betas of 16 developed
equity markets, while Marshall et al. (2009) obtain similar results for the country
betas of 20 emerging equity markets. Gangemi et al. (2000) do not find a significant
131
effect of inflation rate on the country beta of Australia.
(6) The oil price from the Brent index, obtained from DataStream.
The oil factor is the most important production input in the economy, and oil price
instability produces risk for the whole world economy (Jones et al., 2004; Nell &
Simmler, 2007). Oil price increases act like a tax collected by oil producers, and
transfers income from users to producers so that it increases the burdens on users;
when the users are businesses, it increases costs that are likely to be passed on to the
consumer (Richardson, 2003). According to Nandha and Faff (2008), rising oil prices
increase the cost of production of goods and services, and impact on inflation,
consumer confidence, and financial markets. Oil plays an important role in economic
activity, inflation and monetary policy, and so has implications for asset prices and
financial markets (Huang et al., 1996; Mussa, 2000). McSweeny and Worthington
(2007) find that oil price affects the Australian energy sector and is a significant
variable in explaining changes in stock returns. Kaneko and Lee (1995) find
evidence in favour of an oil price factor impacting on stock returns in the Japanese
equity market. Ferson and Harvey (1995) similarly find evidence that oil price has a
significant impact on the 18 equity markets. Recently, Basher and Sadorsky (2006)
have found that emerging equity markets are riskier when oil prices are high, and not
when prices lower.
The consideration of oil supply shocks has not received much attention in the
country beta model. Andrade and Teles (2006) examine the impact of oil supply
shocks on the country beta of Brazil and find no statistical significance. However, for
the MENA countries, the choice of the oil price variable can be justified because of
132
their dependence on oil.
In light of these variable considerations and theoretical propositions, this chapter will
consider a panel regression using a fixed effects model (OLS) rather than a countryby-country regression analysis similar to the works of Marshall et al. (2009) and
Patro et al. (2002)22. The start and the end of the economic factors are not of major
concern since the panel can be considered unbalanced. The variables will be
expressed in terms of the first differences of the natural logarithm series. The
description of the selected economic variables and their time series transformations
are presented in Table 5-1.
Table 5-1. Selection and description of the macroeconomic variables
Variable
Definition
3-month deposit rate (DEP)
3-month deposit rate as proxy for interest rate
Inflation (CPI)
Consumer Price Index
Money supply (M2)
Money supply defined in form of M2
Foreign currency reserves (FR)
Foreign currencies held in MENA central banks
Exchange rate (EXCH)
The exchange rate to the US dollar
Oil supply shock (OIL)
Oil price from Brett oil index
Time series transformation
Changes in the 3-month deposit rate
Ln DEPt Ln ( DEPt ) Ln ( DEPt 1 )
Realized inflation rate
LnCPI t Ln (CPI t ) Ln (CPI t 1 )
Ln M 2t Ln ( M 2t ) Ln ( M 2t 1 )
Ln FRt Ln ( FRt ) Ln ( FRt 1 )
LnEXCH t Ln( EXCH t ) Ln( EXCH t 1 )
LnOILt Ln (OILt ) Ln (OILt 1 )
Growth rate of money supply (M2)
Growth rate of foreign currency reserves
Changes in the exchange rate to the US dollars
Changes in oil price (oil supply shock)
Notes: All economic variables are converted into the natural logarithm. The symbol
first difference of the natural logarithm series.
denotes the
5.2.3. The unanticipated components of the economic variables
The number of years is less than the number of countries, therefore, models such as the System
GMM (Generalized Method of Moments) with instruments are not suitable for the analysis.
22
133
The market efficiency assumptions require the identification of unexpected changes
in the economic variables (Panetta, 2002).23 There are various methods used to
extract the unexpected component series of the economic variables. Bilson et al.
(2001) and Panetta (2002) use Principal Components Analysis (PCA) to extract
information from the economic variables, similar to the work of Bodurtha et al.
(1989) and Chen et al. (1986), whereas Andrade and Teles (2006, 2008), Gangemi et
al. (2000), Tourani-Rad et al. (2006), and Verma and Soydemir (2006) use the error
terms from the Autoregressive Integrated Moving Average (ARIMA) method as the
unanticipated components. Cheung (1993) uses the state-space model and the
Kalman filter approach to extract (ex ante) real interest rates from (ex post) real
interest rates as predictors of expected inflation. He shows that the (ex ante) real
interest rate is not constant and the equity market returns draw on information from
past inflation rates. Faff and Heaney (1999) and Yao et al. (2005) adopt the Kalman
filter approach to extract the unanticipated components of the economic variables in
their study. The use of any of these methods to extract the unanticipated component
series is arbitrary.
Here, the method of the state-space model and the Kalman filter approach is adopted
for the analysis. Let Z i be a vector of the economic variable. The signal of the
system is taken as the expected component of the economic variable, while the
prediction error is the unexpected component series. The decomposition can then be
defined as follows:
In efficient market hypothesis, the stock markets react to the unanticipated components of the
economic variables. In addition, the advantage to the use of unexpected economic variables will
insure the stationarity of the series and reduce the possibility of spurious regression as well as
multicollinearity (Gangemi et al., 2000; Andrade and Teles, 2006:2008).
23
134
Zt
Z tE1,t
1,t
Z Et
t
Z tE 2,t 1
1,t
where Z t
1,t
(5.2)
t
is the vector of n expected economic variable for the period t 1 to t
observed at time t , and Z tE1,t is the vector of expected economic variable for the
period t 1 to t observed at time t . As in the work of Faff and Heaney (1999) and
Yao et al. (2005), the transition equation is set to evolve as a random walk. Both
and
t
are normally distributed with zero mean and known variance.
t
t
is the
vector of the unanticipated component of the economic variables, which will be
considered as a proxy for the economic variables in the panel regression.
After the decomposition process of the selected economic variables using the
Kalman filter, the unexpected components of the economic variables are used for the
panel regression estimated by the fixed effects model. Table 5-2 shows their new
description.
Table 5-2. Decomposition of the economic variables
Variable
Definition
Unexpected change in the 3-month deposit rate
U Ln DEPt
U LnCPI t
U Ln M 2t
U Ln FRt
U LnEXCH t
U LnOILt
Unexpected change in inflation rate
Unexpected change in money supply in form of M2
Unexpected change in foreign currencies reserves
Unexpected change in the exchange rate to the US dollars
Unexpected change in oil price (oil supply shock)
Notes: this table reports the new definition of the economic variables after the
decomposition process using the Kalman filter approach. U denotes the unexpected
component.
5.2.4 Country beta in panel model
The empirical benchmark for the country beta model takes the following form:
135
Rit
0
i
RMSCI , t
(5.3)
it
where Rit denotes the equity market returns for country i at time t , and RMSCI , t is
the returns on the world market represented by the MSCI index at time t . The above
model excludes the international risk free rate as do the works of Andrade and Teles
(2006, 2008) and Gangemi et al. (2000). The parameter
i
is the static country risk
(or country beta). The time-varying country beta model will take the following form:
Rit
0
where
it
it
RMSCI , t
(5.4)
it
is the time-varying country beta generated from the Kalman filter
approach24. Once the time-varying country betas are obtained, the second stage
investigates their determinants in a panel regression estimated as follows:
it
Here
bU
LnDEPit
1
i
b5U LnEXCH it
it
b2U LnCPI it
b6U LnOILt
b3U LnM 2it
b4U LnFRit
(5.5)
it
is the country beta generated by the Kalman filter approach. The variables
U LnDEP,U LnCPI ,U LnM 2,U LnFR,U LnEXCH ,U LnOIL denotes unanticipated
components of interest rate, inflation rate, money supply, foreign currency reserves,
exchange rate with the US dollars, and the oil price, all at time t .
i
is the country s
specific constants, and b1 ,..., b6 measures the sensitivity of the country beta to the
unexpected change of the economic variables, while
it
is the error term which
represents the effects of the omitted variables. The above model will be estimated by
24
For more details on the derivation of the Kalman filter, see appendix.
136
the fixed effects in panel regression, similar to the works of Marshall et al. (2009)
and Patro et al. (2002).
5.2.5. Hypothesized relations
The significance and the sign of the coefficients will provide valuable information on
how economic variables affect country beta in the MENA region. However, the
empirical problem is their expected signs a priori since the betas are endogenous.
Gangemi et al. (2000) and Andrade and Teles (2006; 2008) address the notorious
difficulties of such predictions. According to Gangemi et al. (2000), the empirical
outcome of country beta models are exploratory in nature, and going through the
expected signs is problematic (also see Abell & Krueger, 1989). For example,
Andrade and Teles (2008) argue that currency reserves may have two, opposite
effects in determining the country beta: the first effect considers that a drop in
reserves, in a fixed exchange regime, leads to an increase in the probability of an
exchange rate depreciation, and thus and increase in risk, while the second
recognizes the effect of the entry of speculative capital that increases the currency
reserves, but increases the risk too, in that the economy becomes more vulnerable to
external shocks. Gangemi et al. (2000) note that formulating a single prediction for a
country s exchange rate would be heroic since the exchange rate depends on the
relative roles of the traded/non-traded goods sectors as well as of the importing/
exporting activities across the economy.
5.6. Empirical results
5.6.1. Descriptive statistics
Table 5-3 presents the descriptive statistics of the economic variables and the country
137
beta series. The economic variables in this table are the unexpected components
(signals) estimated by the Kalman filter approach. With the exception of the interest
rate variable, the average signals from the economic variables are positive. The
inflation variable yields the highest positive average. The standard deviation of the
interest rate signals is highest among the economic variables. Skewness is high for
interest rate and inflation rate, and kurtosis is high for interest rate, inflation rate, and
money supply. The average country betas of the MENA equity markets are positive,
at almost half of the world market risk. The Jarque-Bera test statistics show that all
the variables under study depart from normality with a 99% level of significance.
Table 5-3. Descriptive statistics of the economic variables and country betas
Normal
Mean
Max.
Min.
S.D
Skew.
Kurtosis
distribution
probability
-0.617
66.733
-789.22
24.130
-31.433
1028.059
0.000
U LnDEP
0.073 55.696 -6.623
2.016
19.244
528.373
0.000
U Ln CPI
U LnM 2
U Ln FR
U LnEXCH
U Ln OIL
t
0.01
0.488
0.487
0.353
-0.231
-7.966
0.033
0.137
5.497
-0.702
116.600
3.603
0.000
0.000
0.481
8.150
-7.969
1.093
1.774
16.645
0.000
0.000
0.016
0.262
0.799
-0.304
-0.390
0.034
0.078
-1.716
2.780
47.802
30.211
0.000
0.000
Notes: This table reports descriptive statistics for the panel data regression variables. Max., Min, S.D,
and Skew denote maximum, minimum, standard deviation, and skewness. The economic variables are
in form of unexpected components (signals) extracted by the Kalman filter approach. The economic
variables are monthly observations of interest rate, exchange rate, inflation rate, money supply, and oil
price. The country betas are estimated using the Kalman filter approach using the equation (5.4).
5.6.2. Correlation analysis
Prior to the panel model analysis, the correlation analysis among the economic
variables is examined. The correlation analysis is reported in Table 5-4. As can be
seen from the table, most of the correlations are either significantly low or
insignificant among the variables; these results are not surprising since the analysis
uses the unanticipated components.
138
Table 5-4. Correlation analysis of the economic variables
U LnDEP U LnEXCH U Ln CPI U LnM 2 U Ln OIL U Ln FR
U LnDEP
U LnEXCH
U Ln CPI
U LnM 2
U Ln OIL
U Ln FR
1.000
0.057***
(0.071)
0.006
(0.862)
0.093*
(0.003)
0.007
(0.830)
-0.095*
(0.003)
1.000
0.069**
(0.030)
0.074**
(0.020)
0.005
(0.871)
-0.057***
(0.070)
1.000
-0.024
(0.455)
0.098*
(0.002)
-0.032
(0.316)
1.000
-0.015
(0.639)
0.03**
(0.022)
1.000
-0.027
(0.389)
1.000
Notes:*, **, *** denote significant at 1%, 5%, and 10% respectively. P-values are in parentheses. All
the economic variables are expressed as unanticipated components.
The table shows a positive and significant correlation between interest rate and
exchange rate at 0.057; interest rate and money supply at 0.093; exchange rate and
inflation at 0.069; exchange rate and money supply at 0.074; inflation and oil price at
0.098; and money supply and foreign currency reserves at 0.030. However, there is a
negative and significant correlation between interest rate and foreign currency
reserves at -0.095, and exchange rate and foreign currency reserves at -0.057. The
correlations among the variables are low, and this reduces the possibility of serious
multicollinearity in the panel regression.
5.6.3. Panel unit root test
To further avoid spurious regression, the analysis runs a panel unit root test before
the panel model is estimated. Table 5-5 reports the results of the unit root tests using
Breitung (2000) and Im, Pesaran, and Shin (2003) (IPS hereafter) methods. The unit
root of Breitung and IPS methods are tested for the first level with individual
intercept and trend. In Table 5-5, the results show that all the panel variables are
stationary. Based on these results, the panel series are I(0).
139
Table 5-5. Unit root test for panel data variables
U LnDEP
U Ln CPI
Breitung
-27.591
-12.384
Probability
0.000
0.000
IPS
-35.486
-27.112
Probability
0.000
0.000
U LnEXCH
U Ln OIL
-21.013
-19.758
0.000
0.000
-31.837
-27.246
0.000
0.000
U LnM 2
U Ln FR
-8.916
-23.167
-5.872
0.000
0.000
0.000
-34.527
-34.378
-41.272
0.000
0.000
0.000
Notes: This table reports the panel unit root test using the Breitung and the IPS methods.
5.6.4. Panel regression results
Table 5-6 reports the fixed effects regression output expressed in equation (5.5)25.
The table shows that the coefficient of interest rate has no statistical significance,
which suggests that the interest rate does not explain variation in the country beta.
This is not consistent with Andrade and Teles s (2008) findings for countries like
Thailand and Mexico, and Marshall et al. (2009) for 20 emerging markets. This is
because MENA countries are in a different region than East Asia, and different from
other emerging markets like those in Latin America. According to Ghanem (2010),
central Banks in MENA do not follow inflation targeting explicitly, which is why the
interest rate variable does not affect the country beta. However, as Andrade and
Teles s (2008) results show, countries like Thailand and Mexico have experienced
severe currency crisis, and this would justify a high-interest monetary policy and
explain the impact of interest rate on their country betas.
The coefficient of the inflation variable is positive and significant at 5%. This
implies inflation increases the country beta in the MENA equity markets a finding
The analysis is also repeated using the 2SLS to address the possibility of endogeneity problem, the
results are not different from the OLS output and the results will be given upon request.
25
140
consistent with those of Verma and Soydemir (2006) in emerging markets like Brazil
and Marshall et al. (2009) in a group of 20 emerging markets including some MENA
markets (for instance, Egypt, Israel, and Turkey). Patro et al. (2002) also find that the
inflation rate is positively related to the country betas of 16 developed equity
markets. On the other hand, the results reveal that the coefficient of the money
supply variable is positive and significant at 10%. Based on the previous outcomes of
interest rate and inflation, the implications are that expansion of the money supply
induces positive effects on the country beta of MENA equity markets. The
significant impact of the money supply is also an indication that the monetary
policies of MENA markets are out of control, and support the results and argument
for interest rate and inflation targeting in the MENA region. These results reveal the
lack of consistent and structural monetary policies like the inflation targeting, to curb
the increase in the country beta for the MENA equity markets.
Table 5-6. Results for the fixed effects (FE) model using equation (5.5)
Variable
U LnDEP
Coefficient
-0.001
(-0.035)
0.011**
(2.288)
0.591***
(1.761)
-0.373**
(-2.323)
-0.507
(-1.350)
0.051
(0.597)
0.874
U LnCPI
U LnM 2
U LnFR
U LnEXCH
U LnOIL
Adjusted- R 2
D-W
F-statistics
Observations
1.463
402.443
990
Notes: **, *** denote significant at 5%, and 10% respectively, t-statistics are in parentheses.
The panel regression is estimated by the fixed
effects model.
141
Table 5-6 also shows that the coefficient of the foreign currency reserves is negative
and significant at 5%. A negative coefficient suggests that the higher the foreign
currency reserves, the lower the country beta. While extensive literature offers
conclusive evidence that equity markets responds to a wide number of economic
factors, the presence of foreign currency reserves is relatively new in the MENA
countries. This result supports the work of Andrade and Teles (2008), who note the
importance of foreign currency reserves in curbing an increase in the country beta of
Brazil and Argentina, especially during financial crises. This is strong evidence that
foreign currency reserves are a major factor in mitigating the increase in country beta
of MENA equity markets. However, Andrade and Teles (2006) also find that foreign
currency reserves decreased the Brazilian country beta only during a period of fixed
exchange rate: the implication is that at such a time the increase in foreign currency
reserves boosted the confidence of foreign investors, and ultimately lowered the
perception of increased country beta.
To isolate the effects of the economic variables, the panel analysis is estimated under
five sets of regressions. The first set in model 1 includes all the variables except the
interest rate; the second set in model 2 includes all the variables except inflation; the
third set in model 3 includes all the variables except money supply; the fourth set in
model 4 includes all the variables except foreign currency reserves; and the fifth set
in model 5 includes all the variables except the exchange rate. The analysis excludes
the oil price variable because of its insignificant impact in all models. The model
specifications show that the exclusion of oil price variable does not strongly affect
the results. These alternative models can be justified by possible interdependence
among the variables, in a manner similar to the model considerations of Gangemi et
142
al. (2000) and Andrade and Teles (2006, 2008).
The results reported in Table 5-7 show that the coefficients of inflation and money
supply variables are still positive and significant. Only when excluding the inflation
variable, the coefficient of money supply becomes insignificant, the adjusted- R 2
drops sharply from 0.868 to 0.376, and the significance of the foreign currency
reserves substantially increases. This is conclusive evidence of the effect of monetary
policies on country beta over time. Generally, inflation and money supply catch up
with the effect of the other although in some regions, higher money supply may lead
to higher stock returns while in others it can have negative effects. Verma and
Soydemir (2006), for example, find that money supply has a negative effect on
emerging markets of Mexico, Chile, and Brazil. Whether this is strength or a
weakness of the country depends on the perception of investors.
In the fourth case, where the analysis excludes foreign currency reserves, money
supply and inflation are significant. These results are similar when the analysis
excludes the exchange rate variable. The coefficient of the foreign currency reserve
is still negative and significant in all model specifications. The models presented in
Table 5-7 indicate the importance of economic variables in the determination of the
country beta of MENA equity markets. The first noticeable change is that inflation
rate and money supply adversely affect the country beta of the MENA equity
markets, which discourages international investment. The second outcome is that the
foreign currency reserves variable does not lose its significance in four model
specifications (1, 2, 3, and 5), which indicates the importance of this economic
variable for alleviating the country beta of the MENA equity markets in light of their
monetary policies and exchange rate systems.
143
In terms of exchange rate and oil price variables, these variables are insignificant;
this could be due to different exchange rate regimes and the dependence of the
MENA countries on oil. The MENA countries adopt different exchange rate regimes,
ranging from flexible managed-float, to peg-exchange. These systems are designed
to deal with price volatility and curb inflation (Ghanem, 2010). For the oil variable,
the MENA countries are either oil producers or oil importers.
Table 5-7. Results of the fixed effects (FE) model using equation (5.5)
Variable
U LnDEP
U LnCPI
U LnM 2
U LnFR
U LnEXCH
Adjusted- R 2
D-W
F-statistics
Observations
Model 1
Without
U LnDEP
0.011**
(2.343)
0.536***
(1.667)
-0.482*
(-3.097)
-0.520
(-1.388)
0.868
1.523
429.85
1039
Model specifications
Model 2
Model 3
Model 4
Without
Without
Without
U LnCPI U LnM 2 U LnFR
-0.002
0.001
-0.000
(-1.523)
(0.250)
(-0.0394)
0.010**
0.011**
(2.334)
(2.230)
0.774
0.548***
(0.692)
(1.673)
-1.816*
-0.328**
(-4.199)
(-2.192)
-1.373
-0.444
-0.426
(-1.259)
(-1.229)
(-1.136)
0.385
0.877
0.875
0.294
41.987
1023
1.453
501.190
1067
1.430
453.980
990
Model 5
Without
U LnEXCH
-0.000
(-0.204)
0.010**
(2.269)
0.563***
(1.680)
-0.355**
(-2.219)
0.875
1.461
455.889
990
Notes: *, **, *** denote significant at 1%, 5%, and 10% respectively. The panel results are estimated
using the fixed effects model. The first model excludes the inflation rate, the second model excludes
the money supply, the third model excludes the foreign currency reserves, and the fourth model
excludes the exchange rate variable.
To further gauge the impact of economic variables on the country beta, the fixed
effects model can be expanded to differentiate between the MENA countries that peg
their currencies and that that do not, and between countries that export oil and those
that import. To this end, two sets of interactive dummy variables are included in the
panel model. The exchange rate systems for Jordan, Kuwait, Lebanon, Oman, Qatar,
144
and Saudi Arabia operate under the peg-exchange system tied to the US dollar,
whereas Egypt, Israel, Morocco, Tunisia, and Turkey operate under more flexible
systems.26 Kuwait, Oman, Qatar, and Saudi Arabia are countries that export oil,
whereas Egypt, Israel, Jordan, Lebanon, Morocco, Tunisia, and Turkey are countries
that imports oil.27 Thus, equation (5.5) can be expanded as follows:
it
i
bU
LnDEPit b2U LnCPI it b3U LnM 2it b4U LnFRit
1
b5 D pegU LnEXCH it b6 D floatU LnEXCH it b7 DexportU LnOILt
b8 DimportU LnOILt
(5.6)
it
where D peg is a dummy variable which takes the value of 1 for the countries that peg
their currencies: Jordan, Kuwait, Lebanon, Oman, Qatar, and Saudi Arabia; 0
otherwise28. D float is a dummy variable which takes the value of 1 for the countries
that do not peg their currencies: Egypt, Israel, Morocco, Tunisia, and Turkey; 0
otherwise. For the oil dependence, Dexport is a dummy variable which takes the value
of 1 for the oil producing countries: Kuwait, Oman, Qatar, and Saudi Arabia; 0
otherwise. Dimport is a dummy variable which takes the value of 1 for the oil
importing countries: Egypt, Israel, Jordan, Lebanon, Morocco, Tunisia, and Turkey;
0 otherwise. This model allows the interaction of each dummy to reveal the specific
impact of the exchange rate and oil price variables.
The exchange rate systems range from managed-float in Tunisia to pegging against a composite
number of currencies in Morocco, to an independently floating system in Turkey (IMF, 2004;
Lahreche-Revil & Milgram, 2006).
26
Egypt is considered self-sufficient, but the country does import oil for its various petroleum needs;
therefore, this analysis considers Egypt an oil importing country (US Energy and Information
Administration, EIA).
27
Although these countries hard-peg their currencies to the US dollar, with the exception of Qatar and
Lebanon, the exchange system allows some variation in exchange rate between 0.001 and 0.0001
basis points for some periods. The Kalman filter has detected low positive and negative signals over
time for the exchange rate variable of these countries.
28
145
The model is estimated under three sets of regressions. The first set in model 1
includes all the variables; the second set in model 2 includes all the variables except
the exchange rate variable, the third set in model 3 includes all the variables except
the oil price. Results in Table 5-8 show that the interest rate variable is still
insignificant, while the inflation and the money supply variables are positive and
statistically significant, and foreign currency reserves are still negative and
significant. The coefficients of the countries that peg and do not peg their currencies
are not statistically significant; the implication is that the adopted exchange rate
regimes are not sufficient explanations of variations in country beta over time. This
is due to the fact that the responses are different for different countries because the
exchange rate regime for some countries are fixed (Kuwait, Qatar, Bahrain, Saudi
Arabia) and floating for Israel, Turkey, Morocco. This differs from the finding of
Gangemi et al. (2000) for a developed country like Australia, and Bilson et al.
(2001), and Verma and Soydemir (2006) for emerging markets especially in Asia
and Latin America. An examination of fixed exchange regimes adopted in the
MENA region by Ghanem (2010) finds that they do not help combat the inflation
rate. It is possible that the effect of this variable has been absorbed in other variables,
such as inflation and money supply.
When countries are divided into groups according to whether they are oil exporting
or oil importing, the oil price variable does not reveal any contrasting results.29 This
is also due to the fact that the variation in country beta is different due to the oil price
and its volatility. Other studies have documented an inverse relationship between an
The analysis is also repeated with Egypt considered an oil exporting country. The results are not
markedly different than in Table 5-8.
29
146
Table 5-8. Results of the expanded fixed effects (FE) model using equation 5.6
Variable
U LnDEP
U LnCPI
U LnM 2
U LnFR
U LnEXCH
D pegU LnEXCH
D floatU LnEXCH
U LnOIL
DexportU LnOIL
DimportU LnOIL
Adjusted- R 2
D-W
F-statistics
Observations
Model 1
0.001
(0.026)
0.010**
(2.154)
0.563***
(1.678)
-0.362**
(-2.254)
-0.166
(-0.012)
-0.512
(-1.367)
Model Specifications
Model 2
0.001
(0.026)
0.010**
(2.155)
0.563***
(1.678)
-0.362**
(-2.256)
-0.512
(-1.368)
-0.273
(-1.606)
0.153
(1.594)
0.876
-0.273
(-1.610)
0.153
(1.596)
0.876
1.474
361.350
990
1.473
381.818
990
Model 3
-0.001
(-0.036)
0.010**
(2.286)
0.591***
(1.760)
-0.373**
(-2.319)
0.256
(0.016)
-0.507
(-1.350)
0.050
(0.597)
0.886
1.463
379.696
990
Notes: *, **, *** denote significant at 1%, 5%, and 10% respectively, t-statistics are in parentheses.
The model is based on segmenting the countries according to their exchange rate regimes and
dependence on oil.
increase in the oil price and economic activity and attribute the transition of the
negative impact to the equity market returns (Bashar & Sadorsky, 2006; Jones &
Kaul, 1996). According to Soussa et al. (2008), higher oil prices bring some adverse
outcomes to importing countries, like inflation and higher interest rates. Verma and
Soydemir (2006) find oil price a significant variable in explaining the country beta of
Argentina and Chile, while Andrade and Teles (2008) find oil price does not explain
the variations in the country beta of Brazil. In Table 5-8, both coefficients are
147
statistically insignificant, although their signs can be theoretically justified. These
results imply that the effect of the oil price variable on the oil importing countries
may be neutralized by other economic and/or financial variables, while for the oil
exporting countries, the oil price variable may be neutralized by different factors or
absorbed in the fiscal variables which has not been considered in this analysis.
The implications of these results are important to decision makers (central banks) in
the MENA countries. First, these countries should adopt structural policies like
inflation targeting as these could help curb inflation and its impact on country beta.
Second, they benefit from the accumulation of foreign currency reserves, which
reduces their country beta over time and so may attract more foreign investment and
further increase the foreign currency reserves. Third, the central banks should
consider a better and more flexible exchange rate system to free their capital
movements across the MENA region.
Understanding the effect of economic policies on the country beta is of great
importance if there is to be sustainable financial stability across the MENA region.
These results point to the need to adopt monetary coordination within the MENA
region to reap the benefits of risk reduction. This coordination would likely offer
overwhelming benefits in reducing the country beta in the MENA equity markets.
The results also have some implications for regional and international investors: by
recognising the variables that affect the country beta, investors can update their
investment strategies by requiring additional risk premiums or hedging against them.
For example, since the money supply has a positive effect, investors can choose to
balance their strategies to benefit from an increased money supply and the increased
risk premium, to hedge against inflation.
148
5.7. Conclusion
A number of studies examine the relationship between economic variables and
equity market returns, in both developed and emerging markets. This chapter
examines the relationship between the country beta and a number of economic
variables in a panel regression framework for 11 MENA equity markets. The chapter
finds that monetary policies have profound effect on country beta, which suggests
that a destabilized monetary policy may negatively affect the country beta over time.
A second important finding is that the foreign currency reserves can alleviate country
beta in the MENA region. These results may have implications for policymakers
(central banks) in the region to look beyond domestic policies and to achieve
stability of their economic fundamentals, for future prosperity in MENA equity
markets. Given this, it is important to alert decision makers to the impact of these
variables on their financial markets to assist them in reflecting on how country beta
can be mitigated at a regional level. Finally, these results point to a need for
monetary policy coordination within MENA region to reap the benefits of risk
reduction and future regional integration. The current unstable and unsound
monetary policies, and the lack of cooperation and integration across the MENA
region, may explain this empirical irregularity.
149
The Integration of MENA Stock Markets with Germany,
Japan, UK, and USA
This purpose of this chapter is to examine the integration of MENA equity markets
with the developed markets of Germany, Japan, UK, and USA, taking into accounts
the global financial crisis. The investigation is carried out through the cointegration
technique of Johansen and Juselius (1990). A few months after the global financial
crisis, it became clear that the contagion was spreading globally, including the
emerging equity markets. The outcomes of the chapter have major implications for
policy making and international investment, since these have direct consequences on
levels of stock market integration and diversification benefits.
150
6.1. Introduction
One of the major themes of modern portfolio theory concerns the merit of
international diversification. It pays to diversify internationally as long as stock
returns in different national markets are imperfectly correlated. The arguments for
international portfolio diversification include risk reduction and return enhancement
opportunities (Chen et al., 2002; Narayan, & Smyth, 2005; Siklos & Ng, 2001). In
recent years, global markets have tended to become more integrated as a result of a
broad tendency toward liberalization and deregulation in the money and capital
markets of developed as well as developing countries. These changes raise the
possibility that greater correlations may now exist between national stock markets,
implying that there are now reduced benefits from international diversification
(Gilmore & McManus, 2002). Research on increased linkages and international stock
market integration has profound implications for international investors; full
integration of equity markets implies that there will be no long term gains for
international investors (Mallik, 2006). Furthermore, market co-movements can also
lead to market contagion as investors incorporate into their trading decisions
information about price changes in other markets, in an attempt to form complete
information set (Fernandez-Serranoa & Sosvilla-Riverob, 2001).
There are several reasons why different countries stock prices may have a
significant long-run relationship. The presence of strong economic ties and policy
coordination between countries can indirectly link their stock prices over time (Chen
et al., 2002). Huge empirical evidence exists of linkages between developed and
emerging markets, produced by using the cointegration technique of Johansen (1988)
and Johansen and Juselius (1990). Ahlgren and Antel (2002) state that if stock prices
151
are cointegrated, prices in different markets cannot move too far away from each
other. In contrast, a lack of cointegration suggests that stock markets have no longrun linkages, and stock prices in different markets can diverge without bounds.
Kanas (1998) finds no cointegration between the major European markets and the
US equity market before and after the crash of 1987, and argues that there are still
some long-run gains from risk reduction that may come from diversifying in the US
markets and the stocks of the major European countries. Gilmore and McManus
(2002) find that Eastern European equity markets were not cointegrated with the US
equity market during the period of 1995 to 2001. They conclude that US investors
can benefit from international diversification by investing in emerging markets both
in the short and the long-run. On the other hand, Voronkova (2004) shows evidence
of a long-run relationship between the German, Polish, and Hungarian stock indices,
using the cointegration technique. Recently, Mallik (2006) has found increased
integration between seven Asian markets and major international stock markets
(USA, UK, Japan, and Australia) especially since the Asian financial crisis.
Financial market integration among both developed and emerging markets has been
thoroughly investigated. Neaime (2002) argues that as emerging markets mature,
they are likely to become increasingly sensitive to the volatility of stock markets
elsewhere. Earlier tests of equity market integration for both developed and emerging
markets have focused on the use of international asset pricing models like the
ICAPM and the IAPM (Bekaert & Harvey, 1995; Bekaert et al., 2005; Bodurtha et
al., 1989; Bruner et al., 2008; Dickinson, 2000; Engle & Rodriguez, 1993). Stock
market liberalization introduced in the early 1990s has been one of the most
profound factors to affect both stock market integration among countries and
152
volatility spillover transmissions (Gunasinghe, 2005). Generally, focus has
concentrated on major stock markets and the emerging markets of Europe, Asia, and
Latin America; however, a few studies have been devoted to financial integration
between the developed and MENA equity markets (see for example Marashdeh,
2005).
Cointegration analysis has become a popular method for examining the integration of
capital markets over the long-run. Cointegration tests determine whether stock prices
of different national markets move together over the long-run, while allowing for the
possibility of short-run divergence. Evidence of-long-run co-movements would
suggest that, for investors with longer-term investment horizons, the benefits of
diversification indicated by short-run correlations are overstated (Gilmore &
McManus, 2002). The interesting question is whether co-movements of stock prices,
and cointegration, reflect the integration of stock markets. One would expect stock
prices to be cointegrated if stock markets are integrated, but it is possible that stock
prices are cointegrated for some reason not having to do with stock market
integration (Ahlgren & Antell, 2002). Narayan and Smyth (2005) argue that whether
stock markets are cointegrated carries important implications for portfolio
diversification: if markets are cointegrated, this implies that there is a common force,
such as arbitrage activity, which brings the stock markets together in the long-run.
Financial market integration has been a hot topic among academics throughout the
world, especially since the stock market crash of 1987, and the Asian financial crisis
of 1997 and the technology meltdown of 2000 only fuelled its intensity
(Madhusoodanan & Kumar, 2008). The concept of stock market integration is broad,
and greater degrees of co-movements generally reflect greater stock market
153
integration (Neaime, 2002). Stock market integration means that all markets are
exposed to the same set of risk factors, and the risk premia on each factor are the
same in all markets (Ahlgren & Antell, 2002; Harvey, 2000). The growing literature
on integration between the world s stock markets has considerable relevance to
investors in an era of globalization where there are increasing flows of capital across
countries (Narayan & Smyth, 2005).
The techniques to test the integration of these markets with the developed markets
may vary depending on the choice of markets, the sample period, the frequency of
observations (daily, weekly or monthly), and the different methodologies employed
to investigate the interdependence of stock markets (see for example Chen et al.,
2002; Fernandez-Serranoa & Sosvilla-Riverob, 2001; Maghyereh, 2006). The
essence of a cointegrating relationship is that the variables in the system share a
common unit root process. This makes cointegration testing a particularly suitable
methodology because it provides a flexible functional form for modelling stock price
behaviour under the condition of long-run equilibrium.
Recently, emerging markets linkages with developed markets have received more
attention from researchers. Major studies of possible international stock market
linkages have found evidence for equity market linkages in Asia, Europe, and Latin
America (see for example Andern & Kjellsson, 2005; Fernandez-Serranoa and
Sosvilla-Riverob, 2001; Gilmore & McManus, 2002; Gunasinghe, 2005; Johnson &
Soenen, 2002; Kanas, 1998; Kasibhatla et al., 2006; Phylaktis & Ravazzolo, 2002;
Voronkova, 2004). However, the integration or segmentation of the MENA equity
markets is an important question that remains unanswered. Many countries in this
region have suffered wars, political or economic instability; many have not yet
154
emerged as economic powers; they are rarely referred to as influential in global
finance. This may explain the lack of academic research on the capital markets of
this region (Ben Naceur et al., 2008; Maghyereh, 2005, 2006; Marashdeh, 2005;
Neaime, 2002).
6.2. The recent global financial crisis
The recent global financial crisis has unambiguously and painfully underscored that
in today s global and interconnected world no nation is an island. The ferocious
contagion, or exogenous shocks from the advanced economies has plunged the
world economy into its most serious crisis since the 1930s (Sharma, 2010). The
international transmission of turbulence is only one of the ways in which the global
financial crisis can and will affect developing countries (Ghosh, 2009).
The financial crisis has had a major impact on global equity markets and led to
volatility and contagion not seen since the Asian financial crisis of 1997 and the
crash of 1987. Increased innovation in structured finance products, willingness by
lenders to take excessive risks, low interest rates, and the greed of investors allowed
complex products to be sold to an extremely wide range of investors around the
world (Klinz, 2008). Not only did the crisis inflict a major blow on equity markets: it
also triggered an extremely deep contraction in the world economy. In the Arab
countries, the economic turbulence reached both oil-rich and oil-poor economies,
although the forces behind the contagion and the impact were varied and uneven
across the region (Sharma, 2010).
The Gulf Cooperation Council s (GCC) equity markets suffered much more than
those in the USA and other developed markets. In Saudi Arabia and the UAE
155
(Dubai), indices shed more than 40 percent during 2008 (Woertz, 2008). According
to Sharma (2010), significant withdrawals from emerging economies, including the
GCC equity and debt funds, confirm that investors in the advanced economies began
to withdraw from emerging economies in October 2008.
The gloomy economic prospects and loss of confidence in the financial sector had a
huge negative impact on global stock markets. Falling stock markets, a swaying
financial system, and an emerging credit crunch triggered a rapid and deep
contraction in global economy that persisted through the first months of 2008 (Basse
et al., 2009). The role of central banks in the regulation of global financial stability
came under close scrutiny in the aftermath of the crisis that was seen to have
originated with massive failures of the subprime mortgage markets in the USA and
to have quickly spilled over to other countries. In addition to the efforts of other
authorities such as governments and international regulatory institutions, it is
generally believed that policy interventions by central banks are essential to regulate
financial stability and repair the negative impact of financial crises (Arouri et al.,
2010). Despite the adoption of highly expansionary policies
in Egypt, Jordan,
Kuwait, Saudi Arabia, the UAE, and other Arab countries to mitigate the adverse
economic shocks (with the central banks in the region providing stimulus packages
and liquidity, besides lowering reserve requirements and interest rates), the region
saw growth contract from 6% in 2008 to 2.5% in 2009. The slowdown was broadly
similar in oil producing and non-oil producing countries, although their socioeconomic impact varied (Sharma, 2010).
The purpose of this chapter is to examine the long-run relationships between 11
MENA equity markets of Egypt, Israel, Jordan, Kuwait, Lebanon, Morocco, Oman,
156
Qatar, Tunisia, Saudi Arabia, and Turkey, and the equity markets of Germany, Japan,
UK, and USA, using the cointegration technique of Johansen and Juselius (1990).
Some of the MENA countries have made progress in terms of economic
liberalization of trade and capital flows (see for example Ben Naceur et al., 2008;
Maghyereh, 2006; Neaime, 2002). Earlier studies on the cointegration of MENA and
developed markets have focused on a limited number of MENA markets. The lack of
cointegration between MENA markets and developed markets, especially in the light
of recent data (up to May 2010) may shed the light on the integration of the MENA
equity markets with the developed markets over the gloomy period of the global
financial crisis.
The cointegration technique has the advantage of taking into account the error
structure of the underlying process. It enables the estimation and testing of the
equilibrium relationship among nonstationary series while abstracting from shortterm deviations from equilibrium. Thus, it provides relatively powerful tests when
the model is correctly specified (Chen et al., 2002). The outcome of this analysis will
be important for local, regional and international investors and policy makers for
diversification benefits and future decisions on liberalization and the degree of
openness of economic and financial systems.
6.3. Empirical framework
6.3.1. Data
The sample data used in this chapter are weekly stock market indices obtained from
DataStream. The data are available from the Morgan Stanley Capital International
(MSCI) group for Egypt, Israel, Jordan, Kuwait, Lebanon, Morocco, Oman, Qatar,
157
Saudi Arabia, Tunisia, and Turkey from June 2005 to March 2010: a total of 250
observations. Four developed markets are chosen for the analysis from the MSCI
group: Germany, Japan, UK, and USA. The raw weekly indices are transformed to
their natural logarithms so that the coefficients of the variables can be expressed in
elasticities and the first differences can be interpreted as growth.
6.3.2. Unit root test
Cointegration testing requires the time series to be integrated of the order I(1). This
requires that the order of integration of the natural log series be determined. To test
for a unit root, the analysis employs the Augmented Dickey-Fuller (1981) (ADF), the
Phillips-Perron (1988) (PP), and the Kwiatkowski, Phillips, Schmidt, and Shin
(1992) (KPSS) tests. The equation that underlies the ADF test is as follows:
(Aggarwal & Kyaw, 2005):
yt
a0
a2 t
yt
T
1
i 2
i
yt
i 1
et
(6.1)
Where yt is a random walk with drift and linear time trend if
0 , and et is the
residual term. The PP test allows the error disturbances to be weakly dependent and
heterogeneously distributed:
yt
a0
a1 yt
1
a2 t T 2
ut
(6.2)
Test statistics for the regression coefficients under the null hypothesis that the data
are generated by yt
yt
1
ut , where E (ut ) 0 . The difference between the two
unit root tests lies in their treatment of any nuisance serial correlation. The PP test
tends to be robust to a wide range of serial correlations and time-dependent
158
heteroscedasticity (Chen et al., 2002).
The KPSS test for unit roots is based on the null hypothesis that a series is trend
stationary or stationary around a level. More technically, the KPSS test is the
Lagrange multiplier test of the hypothesis that the random walk has a zero variance
(Aggarwal & Kyaw, 2005):
Xt
at
yt
(6.3)
t
where t is the deterministic trend, yt is a random walk, and
t
is the stationary
error. Random walk can be expressed as:
yt
yt
1
(6.4)
et
Where et is iid (0,
Since
2
e
) . The initial value of y ( y0 ) is nothing more than the intercept.
is stationary, a null hypothesis of trend stationarity implies that
2
e
is equal
to 0. If a is equal to zero, the null is stationary around a level.
6.3.3. Multivariate cointegration analysis
If the natural logs of the stock indices are integrated of order one, i.e. I(1), then it is
possible to proceed to the multivariate cointegration test to examine their long-run
relationship. The analysis first considers the following Vector Autoregression (VAR)
model:
yt
where
c
K
i 1
i
yt
1
yt
1
(6.5)
t
is the first difference lag operator and yt is a ( p 1 ) vector of the non-
159
stationary variables (here are the ln or the natural logs of the stock indices), c is a
( p 1 ) vector of constant,
is ( p p ) matrix of parameters,
i
t
is a sequence of
is ( p p ) matrix of parameters which
zero-mean p-dimensional noise vectors,
contains the information about the long-run relationship among the natural logarithm
of the index series. The information on the coefficient matrix between the levels of
is decomposed as
, where
is the adjustment coefficients and
is the
matrix that contains the cointegrating vectors. yt is described in the following four
combinations of models:
f (Germany, Egypt, Israel, Jordan, Lebanon, Kuwait, Morroco, Oman,
Qatar, Saudi Arabia,Tunisia,Turkey) 0
f (Japan, Egypt, Israel, Jordan, Lebanon, Kuwait, Morroco, Oman,
(6.6)
Saudi Arabia,Tunisia,Turkey) 0
f (USA, Egypt, Israel, Jordan, Lebanon, Kuwait, Morroco, Oman, Qatar,
(6.8)
Qatar, Saudi Arabia,Tunisia,Turkey) 0
f (UK, Egypt, Israel, Jordan, Lebanon, Kuwait, Morroco, Oman, Qatar,
Saudi Arabia,Tunisia,Turkey) 0
(6.7)
(6.9)
Johansen and Juselius (1990) suggest two statistics to estimate the number of
cointegrating vectors: the trace
trace
test and the maximum
max
Eigenvalue. These
two test statistics are expressed as:
max
(r , r 1)
trace
(r ) T
k
i r 1
T ln(1
ln(1
r 1
i
)
(6.10)
)
where r is the number of the cointegrating vectors under the null hypothesis and
is the estimated value for the ith ordered Eigenvalue from the
160
i
matrix. If the test
statistics are greater than their critical values, then we reject the null hypothesis
(r
0 ) that there are r cointegrating vectors, in favour of the alternative that there
exist ( r 1 ) cointegrating vectors.
6.3.4. Error Correction (EC) model
If the variables (natural logs of indices) are found to be cointegrated, there is the
implication of an error correction representation (EC) by which the change in at least
one of the variables is a function of the disequilibrium in a previous period. The
Vector Error Correction Model (VECM) is a VAR model in the first differences with
the addition of vector of cointegration residuals.30 The EC model is introduced to
correct a disequilibrium that may shock the whole system. In the short term, the
deviation from the long-run equilibrium will feed back on the changes in the
dependent variable in order to force the movement towards the long-run equilibrium
(Chen et al., 2002). In addition, the EC model restricts the long-run behaviour of the
endogenous variables while allowing for short-run dynamics. Deviations from the
long-run equilibrium are corrected through series of partial short-run adjustments
(Kasibhatla et al., 2006).
6.4. Empirical results
6.4.1. Descriptive statistics
Table 6-1 reports the summary statistics of weekly returns for both the MENA and
the developed equity markets. The table reports the average returns, standard
The significant and negative error correction term has several intuitive implications and is useful as
a further test of the cointegration hypothesis, that when the variables are cointegrated, then in the short
term, deviation from the long-run equilibrium will feed back on the changes in the dependent variable
in order to force the movement towards the long-run equilibrium (see for example Chen et al., 2002;
Siklos and Ng, 2001).
30
161
deviation, skewness, kurtosis, and normality test for the full sample period. The
statistics show that the average returns are positive for Egypt, Israel, Lebanon,
Morocco, Tunisia, and Turkey, while the rest of the MENA markets yield negative
average weekly returns. Among the developed countries, UK and USA yield
negative returns whereas Germany and Japan yield positive returns. Among the
MENA countries, the highest standard deviation can be found for Turkey and Saudi
Arabia, while Tunisia and Israel are the least volatile in the sample. Surprisingly,
USA is the least volatile among the developed markets and UK the most volatile,
although the standard deviations for the developed equity markets are almost similar.
With the exception of Lebanon, all returns are negatively skewed, which implies the
distribution has a longer right tail, while the kurtosis is high for Oman and Kuwait,
which implies that the distributions are peaked relative to the normal distribution.
Therefore, the weekly equity returns are not normally distributed.
6.4.2. Correlation analysis
The correlation matrix is presented in Table 6-2. Among the MENA equity markets,
the highest correlation can be seen between Egypt and Saudi Arabia at 0.489 and
Israel and Turkey at 0.483. Oman has significant and negative correlations with
Germany at -0.134 and with USA at -0.110. Oman, Qatar, and Saudi Arabia have
negative correlations with UK and USA. Israel, Turkey, and, surprisingly, Lebanon
and Tunisia have significant correlations with all of the developed markets, while
most of the MENA markets have insignificant correlations with the developed equity
markets. Turkey and Israel have the highest correlations with the developed equity
markets. These results imply that Israel and Turkey provide less diversification
benefits compared to the other MENA equity markets. Among the developed
162
Table 6-1. Descriptive statistics of the MENA and the developed markets weekly returns
Egypt
Israel
Jordan
Kuwait
Lebanon
Morocco
Oman
Qatar
Saudi
Tunisia
Turkey
Germany
Japan
UK
USA
Mean
Maximum
Minimum
0.23
0.20
-0.26
-0.08
0.33
0.33
-0.06
-0.12
-0.24
0.23
0.17
0.08
0.01
-0.04
-0.01
10.75
6.92
9.92
12.39
16.46
8.46
11.17
12.86
13.49
6.61
25.01
11.82
11.36
19.83
11.63
-21.93
-10.11
-13.90
-25.63
-17.17
-14.28
-22.31
-22.75
-23.49
-13.82
-27.36
-13.56
-15.30
-15.84
-16.15
Standard
Deviation
4.79
2.69
3.33
4.13
4.26
3.18
3.75
4.48
5.16
2.61
6.44
3.36
3.23
3.50
2.66
Skewness Kurtosis Normal distribution
probability
-1.35
7.05
(0.00)
-0.77
4.48
(0.00)
-0.96
5.90
(0.00)
-1.44
10.03
(0.00)
0.25
6.69
(0.00)
-0.87
5.50
(0.00)
-1.39
10.57
(0.00)
-0.87
7.17
(0.00)
-1.03
6.09
(0.00)
-0.57
7.53
(0.00)
-0.20
5.34
(0.00)
-0.76
5.02
(0.00)
-0.56
6.69
(0.00)
-0.15
8.86
(0.00)
-1.45
8.81
(0.00)
Notes: This table presents the descriptive statistics for the weekly returns from June 2005 to March 2010 with a total number of 250
observations.
163
Table 6-2. Correlation matrix of MENA equity markets with the developed markets of Germany, Japan, UK, and USA
Egypt
Israel
Jordan
Kuwait
Lebanon
Morocco
Oman
Qatar
Saudi Arabia
Tunisia
Turkey
Germany
Japan
UK
USA
Egypt
Israel
0.32*
(0.00)
1.00
1.00
0.413*
(0.00)
0.12**
(0.05)
Jordan
0.10***
(0.09)
0.30*
(0.00)
0.31*
(0.00)
0.15**
(0.02)
0.18*
(0.00)
0.46*
(0.00)
0.25*
(0.00)
0.43*
(0.00)
0.48*
(0.00)
0.25*
(0.00)
0.35*
(0.00)
-0.04
(0.51)
0.15**
(0.01)
0.04
(0.44)
-0.03
(0.57)
0.21*
(0.00)
0.23*
(0.00)
0.17*
(0.00)
0.15**
(0.015)
0.48*
(0.00)
0.22*
(0.00)
0.23*
(0.00)
0.22*
(0.00)
0.18**
(0.01)
Lebanon
Morocco
Oman
Qatar
Saudi Arabia
Tunisia
Turkey
Germany
Japan
UK
USA
1.00
0.30*
(0.00)
0.29*
(0.00)
Kuwait
1.00
0.21*
(0.00)
0.16***
(0.08)
0.31*
(0.00)
0.39*
(0.00)
0.39*
(0.00)
0.39*
(0.00)
0.11*
(0.08)
0.14***
(0.03)
-0.02
(0.65)
0.10***
(0.10)
0.03
(0.59)
-0.04
(0.39)
1.00
0.16**
(0.01)
0.14**
(0.03)
0.44*
(0.00)
0.20*
(0.00)
0.34*
(0.00)
0.10***
(0.09)
0.17**
(0.01)
0.21*
(0.00)
0.06
(0.30)
0.14**
(0.02)
0.01
(0.90)
0.12**
(0.05)
0.03
(0.60)
0.19*
(0.00)
0.04
(0.44)
0.02
(0.76)
0.21*
(0.00)
0.23*
(0.00)
0.12*
(0.06)
1.00
0.18**
(0.01)
0.10
(0.11)
0.10
(0.11)
0.30
(0.00)
0.27*
(0.00)
0.12**
(0.04)
0.20*
(0.00)
0.09
(0.16)
0.07
(0.29)
1.00
0.53*
(0.00)
0.40*
(0.00)
0.19*
(0.00)
0.089
(0.17)
-0.13**
(0.04)
0.03
(0.63)
-0.07
(0.27)
-0.11
(0.08)
1.00
0.48*
(0.00)
0.04
(0.49)
0.08
(0.20)
0.15**
(0.02)
0.00
(0.98)
0.12**
(0.04)
-0.08
(0.19)
-0.05
(0.49)
-0.03
(0.64)
0.03
(0.27)
0.02
(0.74)
Notes: *, **, *** denote significant at 1%, 5%, and 10% respectively. P-values are in parentheses.
164
1.00
0.06
(0.35)
-0.08
(0.21)
1.00
0.27*
(0.00)
1.00
0.12**
(0.04)
0.43*
(0.00)
0.14**
(0.03)
0.11
(0.04)
0.43*
(0.00)
0.11***
(0.08)
1.00
0.43*
(0.00)
0.58*
(0.00)
0.33*
(0.00)
0.81*
(0.00)
0.87*
(0.00)
1.00
0.67*
(0.00)
0.44*
(0.00)
1.00
0.78*
(0.00)
1.00
markets, the highest correlations can be found between Germany and UK at 0.86,
Germany and USA at 0.81, and UK and USA at 0.78.
6.4.3. Unit root results
Table 6-3 reports the results of the unit root tests in levels and first differences for all
the markets under study. The tests in levels include a constant and a trend, while tests
in first differences are conducted with a constant only. The choice of the lag length is
automatically determined by the Akiake Information Criterion (AIC) for the ADF
test, and the Newey-West bandwidth for the PP and the KPSS tests.
Table 6-3. Unit root results
Level
Egypt
Israel
Jordan
Kuwait
Lebanon
Morocco
Oman
Qatar
Saudi
Arabia
Tunisia
Turkey
Germany
Japan
UK
USA
ADF
Level
PP test
t-Stat.
-2.023
-1.577
-1.925
-2.270
-2.773
-1.081
-3.073
-1.494
-1.716
First
difference
t-Stat.
-4.697*
-15.565*
-15.156*
-7.332*
-7.095*
-8.939*
-3.793*
-14.25*
-15.157*
Adj. t-Stat
-1.973
-1.567
-2.070
-1.630
-2.809
-1.078
-1.523
-1.723
-1.907
First
difference
Adj. t-Stat
-15.499*
-15.576*
-16.161*
-15.646*
-13.013*
-15.828*
-17.368*
-14.295*
-15.152*
-2.214
-1.981
-1.565
-2.508
-1.500
-1.503
-15.738*
-10.115*
-15.853*
-8.641*
-17.64*
-16.334*
-2.285
-2.137
-1.578
-2.237
-1.566
-1.454
-15.739*
-15.990*
-15.583*
-17.958*
-17.530*
-16.326*
KPSS test
Level
First
difference
LM-Stat. LM-Stat.
0.260
0.094*
0.161
0.114*
0.171
0.073*
0.356
0.098*
0.079
0.094*
0.443
0.069*
0.183
0.111*
0.146
0.094*
0.109
0.086*
0.231
0.187
0.361
0.318
0.314
0.292
0.050*
0.089*
0.128*
0.147*
0.162*
0.166**
Notes: *, ** denote significant at 1% and 5%, respectively. ADF, PP, and KPSS denote the Augmented
Dickey-Fuller, Phillips-Perron, and Kwiatkowski, Phillips, Schmidt, and Shin tests. The lag length in the
ADF test is automatically chosen by the AIC. The bandwidths in PP and KPSS are automatically
determined by the Newey-West bandwidth. The critical values for ADF and PP with intercept and trend
are -3.963, -3.412, and -3.128. The critical values for the KPSS with intercept and trend are 0.216, 0.146,
and 0.119. The critical values for intercept only for the ADF and PP are -3.457, -2.873, and -2.573. The
critical values for the KPSS with intercept only are 0.739, 0.463, and 0.347, all at 1%, 5%, and 10%
respectively.
165
The ADF and PP tests show that the null hypothesis, that the natural log series in the
levels have a unit root, is not rejected, while the KPSS test rejects the null hypothesis
that the series do not contain a unit root. On the other hand, the ADF and the PP tests
show that the null hypothesis, that the first differences of the natural log series are
nonstationary, is rejected, while the KPSS does not reject the null hypothesis that the
first differences contain a unit root. These results suggest that the natural logarithm
of the stock indices contains a unit root, and thus should be first differenced to
achieve stationarity: that is, all series are I(1) in level and I(0) in first differences.
Therefore, the natural logarithm of the stock indices can be modelled by the
cointegration test.
6.4.4. Lag length criteria
Before the cointegration technique is examined, the analysis determines the
appropriate lag length in VAR model. The appropriate lag length is important
because if the lag lengths included are too few, the models may be misspecified,
whereas if the lag lengths included are too many, the degrees of freedom are wasted
(Hsiao, 1981). The Likelihood Ratio (LR), the Akaike Information Criterion (AIC),
the Schwarz Bayesian Information Criterion (SBIC), and the Hannan-Quinn
Information Criterion (HQIC) are used to test the lag order. Hatemi-J and Hacker
(2009) note that in applied studies, the choice of criteria used in determining the lag
order is usually arbitrary. Sometimes the information criteria do not agree on choice
the lag order; the question then is which information criteria one should be used.
The results of the optimal lag length in VAR model are presented in Table 6-4. The
highest lag length of 8 is automatically chosen by the software Eviews. The LR
selects a VAR model with the highest number of lags. The SBIC and the HQIC
166
select the first lag for the cointegration test, while the second lag is chosen for
Germany, UK and USA with the MENA markets using the AIC. Although higher lag
length can be considered as the maximum lag judged by the LR and the AIC tests,
Ahlgren and Antel (2002) warn that the cointegration technique is sensitive to lag
length, and may find more significant cointegrating vectors using a higher order of
lags. For example, Siklos and Ng (2001) find more cointegrating vectors when the
AIC is applied and fewer cointegrating vectors under the SBIC.
Table 6-4. Optimal lag length in VAR for models (6.6) to (6.9)
Model
Germany with MENA
Japan with MENA
UK with MENA
USA with MENA
Lag
k
0
1
2
8
LR
NA
6677.444
276893
189.779*
Criteria
AIC
SBIC
-19.387
-19.214
-46.920 -44.671*
-47.006* -42.681
-47.006
-29.426
HQIC
-19.318
-46.014*
-45.264
-39.448
0
1
8
NA
6619.860
240.555*
-19.302
-47.020*
-46.935
-19.129
-44.771*
-29.689
-19.233
-46.114*
-39.710
0
1
2
8
NA
6687.690
287.289
198.725*
-18.882
-46.896
-47.029*
-46.398
-18.709
-44.646*
-42.705
-29.617
-18.812
-45.989*
-45.287
-39.638
0
1
2
8
NA
6639.596
301.603
200.998*
-19.623
-47.427
-47.627*
-46.973
-19.450
-45.178*
-43.302
-30.192
-19.554
-46.521*
-45.884
-40.213
Notes: * indicates lag order selected by the criterion. See equations (6.6 to 6.9) for the groups of
endogenous series. LR, AIC, SBIC, and HQIC denote the Likelihood Ratio, the Akaike Information
Criterion, the Schwarz Bayesian Information Criterion, and the Hannan-Quinn Information
Criterion. k refers to the number of lags in VAR. NA stands for not available.
Previous studies show that the choice of criteria to determine lag length is sometimes
arbitrary for daily, weekly, and monthly stock market data. These studies include the
works of Kanas (1998), Gilmore and
McManus (2002), Gunasinghe (2005),
167
Kasibhatla et al. (2006), Madhusoodanan and Kumar (2008), and Siklos and Ng
(2001). According to a simulation study performed by Hatemi-J and Hacker (2009),
the SBIC has the best performance in many cases, but the HQIC has better
performance in other cases. Because the choice of optimal lag length is not robust,
the analysis adopts the optimal lag if two criteria (or more) agree on the same order
of lags. In Table 6-4, the SBIC and the HQIC agree on the first lag for all VAR
models, and in one case (Japan with MENA) the AIC agrees with the SBIC and the
HQIC on the first lag. In three cases (Germany with MENA, UK with MENA, USA
with MENA), only the AIC chooses the second lag. Therefore, the first lag is chosen
for all the cointegration tests.
5.4.5. Cointegration results
Since the natural logs are nonstationary, i.e. the series are I(1), the analysis proceeds
to the cointegration analysis using equations 6.6 to 6.9. The cointegration test
assumes an intercept and linear deterministic trend. The cointegration output
(maximum Eigenvalue,
max
, and
trace
) is presented in Table 6-5. Part-A presents
the cointegration of MENA markets with Germany, part-B presents the cointegration
of MENA markets with Japan, part-C presents the cointegration of MENA markets
with UK, and part-D presents the cointegration of MENA markets with USA. For
each part, the table shows the hypothesis of the number of cointegrating vectors, the
second column reports the maximum Eigenvalue, and the third and fourth columns in
each part present the
max
and
trace
statistics respectively.
The conclusion about the precise number of cointegrated vectors will be based on the
significance of
trace
and
max
statistics. If the null hypothesis with ( r
168
0 ) is not
rejected, then we conclude that the MENA markets are not cointegrated with the
developed markets. If the null hypothesis with ( r
0 ) is rejected against the
alternative ( r 1 ) cointegrating vectors, this suggests that the MENA markets are
experiencing a partial integration if 0 r 11 cointegrating vectors, or full
integration with the developed markets if there exist r 11 cointegrating vectors.
The results in Table 6-5 parts A to D reveal the presence of significant cointegrated
vectors in the cointegration analysis. The test statistics of
the null hypothesis ( r
max
and
trace
show that
0 ) is rejected against the alternative of one or more
cointegrating vectors ( r 1 ). The statistics of
max
and
trace
exceed the critical
values at 5%. This evidence reveals the presence of at least two cointegrating vectors
between the MENA equity markets and the developed equity markets under study.
Confirming the presence of long-term relationships among the MENA equity
markets with the developed markets; there also exists evidence of partial integration.
These results will have a profound effect on international investors. Such recent
stock market linkages can be interpreted as diminishing opportunities for
diversification benefits. Maghyereh (2006), who argues that international investors
might gain from diversification in the MENA equity markets because of their lack of
co-movements with the developed equity markets, based his conclusion on analysis
of earlier MENA equity markets data., This new empirical evidence suggests that the
diversification benefits for international investors are shrinking, at least in terms of
investment opportunities in the MENA equity markets.
6.4.6. Structural break analysis
Figure 6.1 depicts the weekly natural logs of the stock indices of the developed
169
equity markets and their returns. Certain facts are evident from these graphs. It is
very clear that the developed markets under study have experienced massive
plunges/fluctuations post-2008. It is likely that the global financial crisis has affected
the long-run relationship of the MENA stock markets with the developed markets
under study. Therefore, the results from the cointegration technique may not be
robust in absence of a structural break.
8.0
8.2
7.8
8.0
7.6
7.8
7.4
7.2
15
7.0
10
7.2
4
0
0
-4
-5
-8
-10
-15
7.4
8
6.8
5
7.6
12
-12
2006
2007
2008
2009
-16
2010
2006
2007
Germany
2008
2009
2010
Japan
7.6
7.4
7.4
7.2
7.2
7.0
7.0
30
6.8
20
6.6
6.6
10
6.4
5
6.4
10
0
-5
0
-10
-10
-20
6.8
-15
2006
2007
2008
2009
2010
-20
UK
2006
2007
2008
2009
2010
USA
Figure 6.1: Natural logs of stock indices and returns for Germany, Japan, UK, and
USA.
170
Table 6-5. The results of the cointegration technique using equations (6.6) to (6.9)
Part-A. Cointegration with Germany (VAR: k 1 )
Hypothesis
0
r=0
r 1
r 2
r 3
r 4
r 5
r 6
r 7
r 8
r 9
r 10
r 11
0
r>0
r>1
r>2
r>3
r>4
r>5
r>6
r>7
r>8
r>9
r > 10
r = 11
Eigenvalue
0.287
0.258
0.215
0.167
0.150
0.141
0.109
0.060
0.053
0.041
0.030
0.025
max
84.062**
74.142***
60.145
45.607
40.324
37.944
28.727
15.406
13.756
10.498
7.693
6.385
Part-B. Cointegration with Japan (VAR: k 1 )
trace
424.697*
340.633*
266.490***
206.345
160.738
120.413
82.468
53.741
38.334
24.578
14.079
6.385
Hypothesis
0
r=0
r 1
r 2
r 3
r 4
r 5
r 6
r 7
r 8
r 9
r 10
r 11
0
r>0
r>1
r>2
r>3
r>4
r>5
r>6
r>7
r>8
r>9
r > 10
r = 11
171
Eigenvalue
0.313
0.256
0.213
0.188
0.137
0.122
0.114
0.056
0.053
0.044
0.032
0.027
max
93.308*
73.422***
59.644
51.888
36.625
32.468
30.293
14.494
13.563
11.366
8.215
6.877
trace
432.166*
338.857*
265.435
205.793
153.904
117.279
84.810
54.517
40.022
26.459
15.092
6.877
Table 6-5. The results of the cointegration technique using equations (6.6) to (6.9) (continued)
Part-C. Cointegration with UK (VAR: k 1 )
Hypothesis
Eigenvalue
max
0
r=0
r 1
r 2
r 3
r 4
r 5
r 6
r 7
r 8
r 9
r 10
r 11
1
r>0
r>1
r>2
r>3
r>4
r>5
r>6
r>7
r>8
r>9
r > 10
r = 11
0.299
0.260
0.209
0.168
0.145
0.137
0.113
0.063
0.054
0.042
0.031
0.026
88.284*
74.939***
58.205
45.801
39.071
36.646
29.761
16.283
13.857
10.664
7.858
6.669
trace
428.041*
339.757*
264.817
206.612
160.810
121.739
85.093
55.331
39.048
25.191
14.527
6.669
Part-D. Cointegration with USA (VAR: k 1 )
Hypothesis
Eigenvalue
max
0
r=0
r 1
r 2
r 3
r 4
r 5
r 6
r 7
r 8
r 9
r 10
r 11
1
r>0
r>1
r>2
r>3
r>4
r>5
r>6
r>7
r>8
r>9
r > 10
r = 11
0.327
0.265
0.209
0.163
0.149
0.137
0.110
0.059
0.054
0.042
0.026
0.024
trace
98.486* 436.472**
76.439** 337.985**
58.304
261.546
44.419
203.241
40.249
158.822
36.808
118.573
28.947
81.764
15.124
52.8177
14.004
37.692
10.716
23.688
6.780
12.971
6.190
6.190
Notes:*, **, *** denote significant at 1%, 5%, and 10% respectively. 0 ( 1 ) refers to null (alternative) hypothesis. The null hypothesis is that
there is no cointegration if r 0 . For the choice of optimal lag in VAR, see Table 6-4. The cointegration model contains a constant and a
deterministic trend.
172
According to Narayan and Smyth (2005), when modelling the long-run relationships
between stock indices, it is important to consider the effects of structural breaks
rather than just using the conventional specifications. Juselius (2011), in the
aftermath of the global financial crisis, stresses that the strong evidence of (near) unit
roots and structural breaks in economic variables suggests that standard economic
models need to be modified or changed to incorporate these strong features of the
data. To ensure the robustness of the cointegration results, it is appropriate to locate a
structural break for the developed markets and examine whether the MENA markets
have experienced any contagious effects. Zivot and Andrews (1992) provide a test to
detect the timing of a single endogenous structural break with no knowledge of the
breakpoint date a priori. The Zivot and Andrews (1992) model modifies the ADF
model in the following form:
yt
t
0
yt
1
DU t
1
DTt
1
k
i 1
di yt
j
t
(6.11)
where DU t is a dummy variable that captures the shift in the intercept, and DTt is a
dummy variable representing a shift in the trend as follows:
DU t
and
DTt
1
if t
TB
0 otherwise
(6.12)
t TB if t TB
0
otherwise
where TB is the number of observations in data extending to potential breakpoint
and each TB is determined by the breakpoint searched within the range of (0.15t,
0.85t), and t is the number of observations. The breakpoint is selected by choosing
173
the value of TB for which the t-statistic of
is minimized. The null hypothesis is
that the series have unit roots with structural breaks in both intercept and trend. The
optimal lag length is determined by the AIC.
The results for the Zivot and Andrews (1992) model are reported in Table 6-6. The
findings of the test indicate the existence of four breakpoints in the month of
September 2008 for all the developed markets. The timing of the breakpoints
coincides with the collapse or near collapse of major investment banks in the US,
especially the Lehman Brothers, and the world s biggest insurance company
(American International Group). The collapse of these institutions instantly triggered
series of panics in the US stock market and propagated the shockwaves to the global
stock markets.
Table 6-6. Results of Zivot and Andrews (1992) model for the timings of the structural
breakpoint
Developed market
t-statistics
Probability
Breakpoint
k
Germany
0
-3.930
(0.000)
9/12/2008
Japan
3
-3.666
(0.000)
9/30/2008
UK
1
-3.716
(0.000)
9/09/2008
USA
0
-4.540
(0.000)
9/09/2008
Notes: k is the optimal lag length determined by the AIC. Probability values are calculated from a
standard t-distribution. The critical values are -5.57, -5.08, and -4.82 at 1%, 5%, and 10%
respectively.
To gauge whether the stock market plunges in the developed markets affected the
long-run relationship with the MENA markets, the cointegration tests are repeated
with the inclusion of a dummy variable. The dummy variable equals 0 before the
structural breakpoint reported in Table 6-6; 1 until the end of the sample period. If
adding the dummy variables increases the cointegration vectors, then we may infer
the plunges in the global stock market during the global financial crisis contributed
to stronger relationships in the MENA region on the long-run.
174
The results of the cointegration tests augmented with the dummy variables are
presented in Table 6-7. The outcomes reveal that the dummy variables do strengthen
long-run relationships between the MENA markets with the markets of Germany,
UK and USA. In particular, the null hypothesis of r
2 is now significant compared
to r 1 in the previous analysis, implying the presence of at least three cointegrating
vectors. For Japan, the results are not markedly different from those reported in
Table 6-7. It seems that the global financial crisis has increased the cointegrating
vectors with the markets of Germany, UK and USA when accounting for their
structural breakpoints. This is a clear indication that the US financial conditions, the
contagious effects in the German, and the UK stock markets, have influenced the
long-run relationships with the MENA markets. The results imply that the integration
of the MENA equity markets with the developed markets of Germany, UK, and USA
increased after the global financial crisis, whereas the level of integration with Japan
remained the same; therefore, accounting for the structural break in the cointegration
relationship is important to capture any drift in the stock market relationships on the
long-run.
6.4.7. Adjustment to the shocks
Table 6-8 reports the normalized cointegrating relationship for the whole sample
period and after the period of the global financial crisis. The normalized
cointegrating relationship can be interpreted as a long-run relationship. The figures
represent the elasticities, since we have considered the natural logarithms of the stock
indices of each market. The coefficients that are statistically significant will
contribute to the long-run relationship.
175
Table 6-7. Results of the cointegration model augmented with the dummy variables of the structural breakpoints
Part-A. Cointegration with Germany (VAR: k 1 )
Hypothesis
0
r=0
r 1
r 2
r 3
r 4
r 5
r 6
r 7
r 8
r 9
r 10
r 11
Eigenvalue
1
r>0
r>1
r>2
r>3
r>4
r>5
r>6
r>7
r>8
r>9
r > 10
r = 11
0.283
0.272
0.238
0.176
0.163
0.143
0.124
0.108
0.060
0.041
0.028
0.014
max
82.660**
78.597**
67.386***
48.047
44.030
38.415
32.859
28.347
15.403
10.340
7.062
3.557
Part-B. Cointegration with Japan (VAR: k 1 )
trace
456.703*
374.043*
295.446*
228.060***
180.012
135.982
97.568
64.709
36.362
20.959
10.619
3.557
Hypothesis
0
r=0
r 1
r 2
r 3
r 4
r 5
r 6
r 7
r 8
r 9
r 10
r 11
176
1
r>0
r>1
r>2
r>3
r>4
r>5
r>6
r>7
r>8
r>9
r > 10
r = 11
Eigenvalue
0.304
0.290
0.216
0.206
0.141
0.123
0.114
0.069
0.054
0.045
0.043
0.016
max
89.975*
84.770*
60.269
57.290
37.654
32.622
30.132
17.728
13.717
11.367
10.982
3.907
trace
450.412*
360.436*
275.666**
215.398
158.108
120.454
87.833
57.700
39.972
26.255
14.889
3.907
Table 6-7. Results of the cointegration model augmented with the dummy variables of the structural breakpoints (continued)
Part-C. Cointegration with UK (VAR: k 1 )
Hypothesis
0
1
Eigenvalue
max
Part-D. Cointegration with USA (VAR: k 1 )
Hypothesis
trace
0
1
Eigenvalue
max
trace
r=0
r>0
0.298
87.728**
462.012*
r=0
r>0
0.342
103.814*
476.543*
r
r>2
0.244
69.302**
295.351*
r
r>2
0.247
70.351**
292.970*
r
1
r
3
r
r
r
r
r
r
r
r
2
4
5
6
7
8
9
10
11
r>1
r>3
r>4
r>5
r>6
r>7
r>8
r>9
r > 10
r = 11
0.273
0.179
0.154
0.143
0.116
0.110
0.063
0.042
0.031
0.012
78.932**
374.284*
48.929
226.050***
38.265
135.699
41.422
30.665
28.934
16.237
10.724
7.901
2.973
177.121
97.434
66.770
37.836
21.598
10.875
2.973
r
1
r
3
r
2
4
r
5
r
7
r
r
r
r
r
6
8
9
10
11
r>1
r>3
r>4
r>5
r>6
r>7
r>8
r>9
r > 10
r = 11
0.275
0.177
0.156
0.139
0.121
0.109
0.059
0.042
0.028
0.009
79.759**
372.729*
48.229
222.619***
37.007
132.468
41.922
31.876
28.619
14.988
10.702
6.998
2.277
174.390
95.462
63.585
34.966
19.978
9.275
2.277
Notes: *, **, *** denote significant at 1%, 5%, and 10% respectively. The cointegration technique is augmented with the structural break dummies
using Zivot and Andrews (1992) model. For the timing of these dummies, see Table 6-6.
177
The results show that most of the MENA equity markets had significant
positive/negative relationships with the developed markets, and thus contribute to the
long-run
relationship,
especially
after
the
global
financial
crisis.
An
increase/decrease of the stock indices of MENA markets resulted from the
increase/decrease in the stock markets of Germany, Japan, UK, and USA. The
empirical results show that Egypt, Israel, Jordan, Lebanon, Morocco, Oman, Qatar,
and Saudi Arabia have significant relationships with the market of Germany. Israel
adjusts about two-to-one in the long-run while Egypt, Jordan, and Morocco adjust
about one-to-one.
The adjustment is smaller for Lebanon, Oman, Qatar, and Saudi Arabia. The results
with Japan are similar: the markets of Egypt, Israel, Jordan, Lebanon, Morocco,
Oman, Saudi Arabia, and Turkey have significant long-run relationships with the
market of Japan. Again, Israel adjusts two-to-one in the long-run while Jordan
adjusts about 1.5-to-one. In the case of UK, Israel, Jordan, Kuwait, Lebanon,
Morocco, Oman, Qatar, Saudi Arabia, and Tunisia have significant long-run
relationships with UK. Surprisingly, Israel adjusts about 2.5-to-one in the long-run
and Jordan adjusts two-to-one, while Morocco, Oman, and Saudi Arabia adjust about
one-to-one. In the case of USA, Israel, Jordan, Lebanon, Morocco, Oman, Qatar,
Saudi Arabia, and Tunisia have significant long-run relationships. Israel adjusts 1.5to-one in the long-run while Jordan adjusts about two-to-one. The outcomes show
significant long-run relationships between the MENA markets and the developed
markets, particularly since the global financial crisis. It is noticeable that the market
adjustments of Israel and Jordan are high, while the rest of the MENA markets show
smaller degree of adjustment (below one-to-one adjustment).
178
Table 6-8. Normalized cointegrating vectors before and after the crisis
In GER
Developed markets
In GER
In JAP
Whole sample After crisis
1.000
1.000
In UK
In JAP
Whole sample
After crisis
1.000
1.000
In USA
MENA markets
In EGY
In ISR
In JOR
In KUW
In LEB
In MOR
In OM
In QAT
In SAU
In TUN
In TUR
-0.199
1.941*
- 1.377*
-0.322
-0.576*
1.058*
0.781*
-0.493**
0.637*
- 0.169
-0.029
-0.926*
0.216
-0.212
0.067
-0.173**
-0.299***
-0.391*
-0.316
-0.073
-0.108
-0.024
0.136
2.168*
-1.536*
-0.237
-0.535*
0.776*
-0.141
0.131
0.654*
0.398
-0.525*
0.458***
1.571*
-1.356*
-0.085
-0.362*
0.323***
-0.518*
0.280
0.400*
0.411
-0.507*
In UK
Whole sample
After crisis
1.000
1.000
-0.553
2.411*
-1.945*
-0.261
-0.787*
1.074*
0.957*
-0.645**
1.130*
0.401
-0.181
0.071
0.616*
-0.763*
0.314**
-0.315*
-0.200
0.005
-0.272**
0.458*
0.737*
-0.001
In USA
Whole sample
After crisis
1.000
1.000
-0.092
1.450*
-0.938*
-0.106
-0.291*
0.343**
0.481**
-0.282**
0.303*
0.211
-0.057
0.089
0.675*
-0.490*
0.203**
-0.153*
-0.246**
0.067
-0.175***
0.154**
0.417*
-0.029
Notes: *, **, *** denote significant at 1%, 5%, and 10% respectively. All the cointegrating vectors are normalized on the developed markets of Germany, Japan,
UK, and USA. The dummy variables are endogenously determined by the Zivot and Andrews (1992) model, see table (6-6) for more details.
179
6.4.8. Error Correction (EC) results
Table 6-9, parts A to D, presents the long- and short-run equilibrium relationship
among the developed and the MENA markets. The EC model will include the
dummy variables for the structural breakpoints found in Table 6-6. The lag lengths of
the series are chosen similar to the cointegration test. According to Chen et al.
(2002), the long-run relationship is implied through the significance of the lagged EC
term which contains the long-term information since it is derived from the long-run
cointegrating relationship. According to Kasibhatla et al. (2006), if the EC term is
not significant then the market is exogenous, and will be the first receptor of external
shock.
Part-A presents the EC model for Germany with the MENA markets. In the long-run,
the EC terms are negative and statistically significant with the exceptions of Egypt,
Jordan, Qatar, and Saudi Arabia. This suggests that the four named markets might be
exogenous to the system and the first receptor of external shocks, with a tendency to
transmit the shock to other markets. The majority of the MENA markets exhibit
significant contributions to the long-run equilibrium relationships. The speed of
adjustments following a shock is quite slow: more specifically, the EC term explains
about 4.4% to 15.2% of the weekly movement of MENA markets to equilibrium.
The interpretation of the empirical findings in the long-run implies that when there is
a deviation from the equilibrium cointegrating relationship as measured by the EC
term, it is the stock market changes in Israel, Kuwait, Lebanon, Morocco, Oman,
Tunisia, and Turkey that adjust to clear the disequilibrium. In other words, these
markets bear the burden of short-run adjustments to the long-run equilibrium. In the
short-run, the lagged price changes (by one week) can be used to predict current
180
price changes in the cointegration system. With the exception of Egypt, it can be
shown that fluctuations in the German stock market seem to explain the movements
in all the MENA equity markets. This outcome is significant since it sheds light not
only on the responses of the MENA markets to the German stock market, but also on
their degree of integration with the global capital market. In addition, it can be seen
that stock market fluctuations in Qatar can explain the stock market movements in
Jordan and Turkey, while Turkey can explain stock market movements in Egypt and
Jordan can explain them in Lebanon. Israel can explain the stock market movements
in Qatar.
Part-A reports the dummy variable for the structural break. It can be seen that the
dummy variable is negative and statistically significant at 1% for the markets of
Germany, Kuwait, Morocco, and Oman, at 5% for Israel, Lebanon, and Turkey, and
at 10% for Tunisia. The negative sign of the structural break dummy suggests that
the MENA equity markets experienced reduced average weekly returns due to the
German stock market plunges, especially after the global financial crisis in 2008.
The results with Japan in part-B show that with the exception of Israel, Saudi Arabia,
and Tunisia, the EC term is negative and statistically significant. The speed of
adjustments to equilibrium following a shock is about 4.4% to 10.9% a week. These
findings suggest that when there is a deviation from the equilibrium relationship with
Japan as measured by the EC term, the changed of stock markets in Egypt, Jordan,
Kuwait, Lebanon, Morocco, Oman, Qatar, and Turkey adjust to clear the
disequilibrium in the long-run. In the short-run, the stock market fluctuations of
Japan explain the stock market movements in Egypt, Israel, Jordan, Morocco, Oman,
Qatar, Saudi Arabia, Tunisia, and Turkey, while Qatar explain the stock market
181
Table 6-9A. EC model with Germany
EC term
In GER
In EGY
In ISR
In JOR
In KUW
In LEB
In MOR
In OM
InQAT
In SAU
InTUN
InTUR
-0.152*
0.013
-0.006
0.701
-0.044**
0.304*
-0.016
0.291*
-0.081*
0.319*
-0.059**
0.211**
-0.070*
0.246*
-0.069*
0.392*
-0.001
0.582*
-0.045
0.488*
-0.035**
0.207*
-0.119*
0.822*
In EGYt 1
-0.071
-0.055
-0.037
-0.020
-0.101
-0.078
-0.087
0.062
0.041
-0.018
0.060
-0.137
In ISRt 1
-0.068
0.035
-0.160
0.041
-0.143
-0.025
-0.028
-0.129
-0.204***
-0.124
-0.010
-0.019
In JORt 1
-0.001
-0.146
0.085
-0.077
-0.073
-0.148***
0.054
-0.149
-0.027
-0.038
-0.069
0.122
InUWt 1
0.040
-0.023
0.027
-0.071
-0.036
-0.013
0.021
-0.085
0.071
-0.132
-0.037
-0.010
In LEBt 1
-0.082
0.088
-0.054
0.067
-0.051
0.171**
0.035
0.053
0.041
0.060
0.007
-0.102
In MORt 1
-0.038
0.113
-0.044
0.054
0.058
0.015
-0.047
-0.052
0.040
-0.021
-0.012
-0.096
InOM t 1
0.008
0.141
0.073
0.046
0.104
0.153
0.058
-0.076
0.012
0.044
0.053
0.123
InQATt 1
0.097
0.129
0.071
0.106***
-0.010
0.055
-0.053
0.055
0.076
0.042
-0.026
0.291**
In SAU t 1
-0.062
-0.061
-0.010
-0.047
0.060
0.061
0.013
0.021
-0.039
0.063
-0.039
-0.099
-0.022
-0.106
0.018
0.057
-0.133
-0.015
-0.017
0.067
0.048
-0.048
-0.058
0.011
InTURt 1
0.017
0.084***
0.020
0.018
0.065
-0.015
-0.015
0.033
0.044
0.082
-0.010
-0.203*
0.025*
-0.074*
0.133
0.001
-0.003
0.377
0.008*
-0.017**
0.181
(0.00)
-0.013
0.138
0.014*
-0.044*
0.102
0.012**
-0.031**
0.071
0.017*
-0.041*
0.107
0.010***
-0.036*
0.193
-0.001
-0.001
0.230
0.003
-0.019
0.112
0.007**
-0.017***
0.065
0.019**
-0.051**
0.202
InGERt 1
InTUN t 1
Dummy
Adj. R 2
F-statistic
3.716
11.697
4.899
3.836
2.997
2.357
182
3.113
5.215
6.268
3.215
2.218
5.479
Table 6-9B. EC model with Japan
EC term
In JAPt 1
In EGYt 1
In ISRt 1
In JORt 1
In KUWt 1
In LEBt 1
In MORt 1
InOM t 1
InQATt 1
In SAU t 1
InTUN t 1
InTURt 1
Dummy
Adj. R 2
F-statistic
In JAP
In EGY
In ISR
In JOR
In KUW
In LEB
In MOR
In OM
InQAT
In SAU
InTUN
InTUR
-0.014
-0.084**
-0.109*
-0.057**
-0.013
-0.061*
-0.081*
-0.088*
-0.044*
-0.101*
-0.093*
-0.025
-0.137**
0.515*
0.175*
0.144**
0.144
0.074
0.174**
0.218*
0.374*
0.333*
0.118**
0.376*
-0.101***
-0.183**
-0.061
-0.084
-0.137***
-0.113
-0.101***
0.001
-0.088
-0.075
0.043
-0.227***
0.018
0.038
-0.130***
0.017
-0.155
-0.061
-0.018
-0.162***
-0.245**
-0.091
0.010
0.023
-0.021
-0.148
0.067
-0.054
-0.060
-0.119
0.046
-0.121
0.006
-0.060
-0.083
0.103
0.069
-0.007
0.050
-0.063
-0.008
0.002
0.048
-0.058
0.079
-0.105
-0.022
0.051
-0.043
0.073
-0.047
0.065
-0.047
0.170**
0.036
0.051
0.028
0.062
0.003
-0.082
-0.156**
0.094
-0.038
0.046
0.054
0.002
-0.046
-0.063
0.017
-0.019
-0.011
-0.090
-0.124***
0.019
0.036
-0.005
0.068
0.121
0.042
-0.135***
-0.102
-0.022
0.030
0.013
0.011
0.177***
0.083***
0.097
-0.049
0.006
-0.071
0.018
0.090
0.070
-0.022
0.290**
-0.031
-0.087
-0.023
-0.057
0.050
0.057
0.003
0.008
-0.059
0.044
-0.046
-0.127
-0.027
-0.044
0.046
0.065
-0.135
-0.026
-0.014
0.072
0.085
-0.003
-0.043
0.066
0.088**
0.006*
-0.017*
0.177*
0.004
-0.009
0.140*
0.005***
-0.018*
0.051
0.008**
-0.016**
0.021
0.008*
-0.015*
0.119*
0.003
-0.016*
0.151*
0.002
-0.012***
0.139**
-0.001
-0.002
0.018
0.003
-0.004
-0.059
0.006
-0.010
0.175
4.751
0.268
7.472
0.059*** 0.085**
0.002
0.001
0.001
-0.013*
0.062
2.158
0.112
3.223
0.058
2.088
0.083
2.606
183
0.040
1.732
0.159
4.338
0.166
4.501
0.048
1.895
0.002
1.037
0.047
1.876
Table 6-9C. EC model with UK
EC term
InUK t 1
In EGYt 1
In ISRt 1
In JORt 1
In KUWt 1
In LEBt 1
In MORt 1
InOM t 1
InQATt 1
In SAU t 1
InTUN t 1
InTURt 1
Dummy
Adj. R 2
F-statistic
In UK
In EGY
In ISR
-0.228*
0.599*
-0.068
-0.156*
In KUW
In LEB
In MOR
In OM
InQAT
In SAU
InTUN
InTUR
-0.026
-0.057**
-0.066**
-0.083*
-0.067*
-0.061**
-0.005
-0.012
-0.101**
0.256*
0.240*
0.270*
0.088
0.140**
0.447*
0.528*
0.428*
0.154*
0.533*
-0.125***
-0.042
-0.042
-0.089
-0.081
-0.087
0.068
-0.02
-0.03
0.059
-0.170
-0.001
0.050
-0.139***
0.052
-0.120
-0.018
-0.024
-0.118
-0.208***
-0.078
0.012
0.034
0.003
-0.121
0.085
-0.070
-0.062
-0.117
0.094
-0.130***
0.010
-0.066
-0.078
0.141
-0.028
-0.039
0.032
-0.074
-0.031
-0.020
0.011
-0.088
0.046
-0.11
-0.028
0.004
-0.029
0.075
-0.054
0.065
-0.047
0.183
0.047
0.042
0.027
0.055
0.003
-0.084
-0.104
0.147
-0.027
0.069
0.071
0.010
-0.054
-0.036
0.059
0.022
0.002
-0.061
-0.120***
0.065
0.051
0.017
0.079
0.1113
0.011
-0.102
-0.065
0.028
0.043
0.034
0.132**
0.147***
0.066
0.109***
-0.027
0.048
-0.063
0.037
0.092
0.041
-0.030
0.288**
-0.105**
-0.069
-0.015
-0.051
0.053
0.049
0.001
0.018
-0.047
0.062
-0.042
-0.119
-0.022
-0.118
0.008
0.051
-0.156
-0.033
-0.043
0.032
0.027
-0.048
-0.060
-0.008
0.097**
0.021*
-0.065*
0.1216**
0.009**
-0.025**
0.077
0.010**
-0.032**
0.016
0.013**
-0.031**
0.013
0.018*
-0.043*
0.024
0.009**
-0.033*
0.072
0.008***
-0.029**
0.095
-0.001
-0.002
0.001
0.004
-0.007
-0.131
0.016**
-0.041**
0.204
5.528
-0.056** -0.030***
In JOR
0.328
9.598
0.033
0.033
0.007**
0.002
-0.010 -0.017***
0.127
3.558
0.116
3.323
0.066
2.254
0.056
2.039
184
0.085
2.642
0.227
6.168
0.222
6.024
0.079
2.506
0.018
1.327
0.094
2.828
Table 6-9D. EC model with USA
EC term
InUSAt 1
In EGYt 1
In ISRt 1
In JORt 1
In KUWt 1
In LEBt 1
In MORt 1
InOM t 1
InQATt 1
In SAU t 1
InTUN t 1
InTURt 1
Dummy
Adj. R 2
F-statistic
In USA
In EGY
In ISR
In JOR
In KUW
In LEB
In MOR
In OM
InQAT
In SAU
InTUN
InTUR
-0.175*
-0.073***
-0.003
-0.054***
-0.103*
-0.078
-0.119*
-0.120*
-0.113*
-0.047
-0.006
-0.068
-0.065
0.887*
0.472*
0.426*
0.479*
0.286*
0.175**
0.590*
0.715
0.713*
0.236*
1.044*
-0.015
-0.097
-0.016
-0.027
-0.074
-0.062
-0.087
0.077
-0.006
-0.011
0.069
-0.111
-0.157**
0.038
-0.134***
0.036
-0.141
-0.024
-0.030
-0.137
-0.225**
-0.108
0.013
0.039
0.024
-0.099
0.082
-0.049
-0.043
-0.127
0.087
-0.105
0.037
-0.017
-0.075
0.132
0.073***
-0.051
0.037
-0.086
-0.044
-0.019
0.010
-0.103
0.032
-0.138
-0.029
0.011
-0.061
0.081
-0.055
0.061
-0.054
0.171**
0.044
0.047
0.033
0.056
0.005
-0.095
-0.062
0.145***
-0.021
0.065
0.071
0.022
-0.045
-0.041
0.056
0.007
0.003
-0.042
-0.123**
0.086
0.089***
0.023
0.080
0.130
0.001
-0.114
-0.064
0.034
0.054
0.110
0.044
0.121***
0.046
0.095
-0.041
0.036
-0.062
0.027
0.077
0.016
-0.039
0.244**
-0.034
-0.056
-0.005
-0.043
0.066
0.063
0.009
0.028
-0.036
0.071
-0.039
-0.092
-0.053
-0.144
-0.008
0.032
-0.171
-0.038
-0.028
0.025
0.016
-0.088
-0.068
-0.040
0.037
0.016*
-0.049*
0.141*
0.009***
-0.022***
0.025
0.003
-0.001
0.035
0.004
-0.021**
0.079
0.012**
-0.036*
-0.002
0.011**
-0.026**
0.024
0.018*
-0.043*
0.053
0.011*
-0.038*
0.101
0.009**
-0.033*
0.104***
0.002
-0.013
0.004
0.003
-0.003
-0.154**
0.009
-0.018
0.181
4.898
0.379
11.785
0.210
5.697
0.171
4.641
0.116
3.314
0.072
2.360
0.078
2.494
0.251
6.918
0.245
6.725
0.133
3.716
0.031
1.572
0.167
4.546
Notes for tables (6-9A, B, C, D): *, **, *** denote significance at 1%, 5%, and 10% respectively. The EC model reports the first cointegrating vector for the cointegration
equations. The EC model is augmented with the dummy variable corresponding to the breakpoint date found in the Zivot and Andrews (1992) test. The dummy variable takes the
value of 0 before the breakpoint date, 1 until the end of the sample period.
185
movements in Egypt, Israel, and Turkey, and Egypt can explain the movements in
Morocco. Israel explains the stock market movements of Oman and Qatar, while
Egypt, like Qatar, can explain the movements of Turkey.
Part-B also show that with the exception of Israel, the dummy variable is negative
for all the MENA markets but statistically significant for Japan, Jordan, Kuwait,
Lebanon, Morocco, Oman, and Qatar. The negative sign of the structural break
dummy suggests that the MENA equity markets experienced a reduction of average
weekly returns after the stock market plunges of Japan due to the global financial
crisis in 2008.
The results for UK in part-C show that, with the exception of Jordan, Saudi Arabia,
and Tunisia, the EC term is negative and statistically significant. When there is a
deviation from the equilibrium relationship as measured by the EC term, the stock
prices in Egypt, Israel, Kuwait, Lebanon, Morocco, Oman, Qatar, and Turkey adjust
to clear the disequilibrium in the long-run. The speed of adjustment to equilibrium
following a shock is about 3.0% to 15.6% a week. It can be seen that most of the EC
terms are smaller in magnitude, similar to the results found with Japan. Turkey leads
the MENA countries in adjusting the disequilibrium, similar to the results found with
Germany. In the short-run, fluctuations of the UK stock market explain movements
in all the MENA stock markets, similar to the results found with Germany. The stock
market fluctuations in Qatar can explain the stock market movements in Egypt,
Jordan, and Turkey, while Jordan can explain the movements in Oman, and Israel
can explain the movements in the stock market of Qatar.
Part-C also shows that the dummy variable is negative and statistically significant at
186
1% for UK, Morocco, and Oman, at 5% for Egypt, Israel, Kuwait, Lebanon, Qatar,
and Turkey, and at 10% for Qatar. The negative sign of the dummy variable suggests
that these MENA equity markets have experienced a reduction of average weekly
returns after the global financial crisis in 2008.
The results with USA in part-D show that with the exception of Israel, Lebanon,
Saudi Arabia, Tunisia, and Turkey, the EC term is negative and statistically
significant. The EC term explains an adjustment of about 5.4% to 17.5% a week
toward their equilibrium. The coefficients on the significant EC terms indicate a
relatively high speed of adjustment towards the long-run equilibrium compared to
those in parts A, B, and C. The findings suggest that when there is a deviation from
the equilibrium cointegrating relationship as measured by the EC term, stock price
changes in Egypt, Jordan, Kuwait, Morocco, Oman, and Qatar adjusts to clear the
disequilibrium in the long-run. In other words, these markets bear the burden of
short-run adjustment to the long-run equilibrium. In the short-run, stock market
fluctuations in USA can explain the stock market movements in the MENA stock
markets with the exception of Qatar. This outcome is significant since it also sheds
light on the possible integration of the MENA markets with the US stock market.
With the exception of Israel, Saudi Arabia, Tunisia, and Turkey, the dummy variable
is negative and statistically significant. The negative sign of the structural break
dummy also suggests that these MENA equity markets experienced a reduction of
average weekly returns after the stock market plunges of USA. It can also be seen
that the fluctuations in Morocco, Qatar, and Turkey can explain the stock market
movements of Egypt, while Qatar can explain movements in Turkey, Oman can
explain movements in Israel, and Israel can explain movements in Qatar.
187
In general, most of the MENA equity markets, especially Egypt, Kuwait, Lebanon,
Morocco, Oman, Qatar and Turkey, show a tendency for convergence with the
developed markets to adjust back to equilibrium in the long-run, while Saudi Arabia
and Tunisia show evidence of deviation from the equilibrium on the long-run and
will be the receptor of further external shocks. Although the financial systems in the
MENA region are thought to be immune to the global market turbulence because of
their limited integration, the empirical evidence suggests that this integration has
significantly increased, implying that the MENA financial systems are not totally
segregated from the global finance and consequently show signs of higher market
exposure to the global capital market.
6.5. The regional stock market integration of MENA countries
In this section, the cointegration analysis is repeated excluding the developed
markets under study. Previous studies of the long-run relationship among MENA
stock markets have not also provided consistent results. The objective here is to
investigate the degree of regional stock market integration of the MENA countries:
Egypt, Israel, Jordan, Kuwait, Lebanon, Morocco, Oman, Qatar, Saudi Arabia,
Tunisia, and Turkey, using the Johnson cointegration analysis. The sample data used
in this section are similar to the MENA stock market data used in the previous
sections. For more details on the descriptive statistics and correlation analysis, see
section 6.4.1 and Table 6-1, section 6.4.2 and Table 6-2.
Regional market integration is a major concern for local, regional and international
investors planning capital budgeting decisions and diversification strategies, since an
increased interrelationship will increase market efficiency and provide investors with
more opportunities to reduce portfolio
risk. In particular, a higher degree of
188
market integration among the regional MENA equity markets may enhance their
growth and liquidity, since the presence of more liquid capital markets offers lower
costs of capital to firms wishing to raise funds locally. For policy makers (central
banks and governments), the degree of relationship among equity markets on a
geographical basis may provide evidence of the degree of intraregional trade and
macroeconomic coordination (see for example Maghyereh, 2006; Neaime 2002).
The benefit of international diversification is limited when national equity markets
are cointegrated because the presence of common factors limits the amount of
independent variation (Chan et al., 2002). A lack of cointegration, on the other hand,
suggests that the stock indices have no long-run linkages and can drift far from each
other. According to Gelos and Sahay (2001), there are several reasons why stock
indices may have a significant long-run relationship. Strong economic ties and policy
coordination between countries can indirectly link their stock prices over time. With
technological and financial innovation, the advancement of international finance and
trade, and deliberate regional and global co-operation, the geographical divides
between various national stock markets become less obvious.
6.5.1. Stock market development in the MENA region
Over the last two decades, the MENA equity markets have experienced a wave of
financial sector liberalization. According to Ben Naceur et al. (2008), some MENA
equity markets are fairly well advanced, whereas a few others have significant room
for improvement. As a group, the MENA countries have performed relatively well in
the areas of regulation and supervision as well as in financial openness, but more
efforts are needed in order to reinforce the institutional environment and promote
non-bank financial sector development.
189
The MENA stock markets differ tremendously in terms of size. As of 2009, Table 610 shows, Saudi Arabia leads the region in terms of market capitalization at $319
billion, followed by Turkey at $234 billion and Israel at $189 billion. These figures
are substantial when compared to Lebanon at $2.2 billion and Tunisia at $2.4 billion.
However, in comparison to each country s market capitalization to the GDP, the
apparent size looks completely different. Jordan leads with a market capitalization of
139 percent of GDP, followed by Qatar at 104.7 and Israel at 96.9 percent. Lebanon
and Tunisia are the smallest at 11.39 and 13.85 percent respectively.
Table 6-10. MENA equity markets indicators, 2009
Market
Egypt
Israel
Jordan
Kuwait
Lebanon
Morocco
Oman
Qatar
Saudi Arabia
Tunisia
Turkey
Liberalization
date
Number of
companies
Market
capitalization
1992
1993
1995
No
No
1988
1999
No
1999
1995
1989
306
622
272
207
16
78
125
44
135
44
315
91
189
32
96
2.2*
64
17
88
319
2.4*
234
Market
capitalization
(% of GDP)
48.50
96.90
139.10
86.50
11.39
71.00
32.40
104.70
86.20
14.85
38.10
Source: Saadi and Williams (2011); Ben Naceur et al. (2008). * are quoted from Bely (2007), as of
the year 2007. The market capitalization figures are in billions of US dollars.
In terms of the number of listed companies, Israel is the leading market with 622
listed companies, followed by Turkey, Egypt, and Jordan with 315, 306, and 272
respectively. The rest of the MENA equity markets are characterized by a small
numbers of listed companies; especially Lebanon, with 16 listed companies.
However, according to Ben Naceur et al. (2008), most of the listed companies in the
MENA region are not actively traded but closed or family-owned companies.
190
Table 6-10 also lists the official liberalization dates for the MENA stock markets as
presented in Bekaert et al. (2003) and Ben Naceur et al. (2008). The table shows that
most of the MENA equity markets began to liberalize their stock markets in the late
1980s and early 1990s. Morocco and Turkey were the first to do so, while Saudi
Arabia and Oman were the last to attempt the liberalization process. During the
liberalization, the MENA countries embarked on structural programs, the centre of
which was the financial sector reform that has since resurrected their stock markets
(Maghyereh, 2006). According to Bekaert et al. (2003), stock market liberalizations,
if effective, lead to important changes in both the financial and real sectors as the
economy becomes integrated into world capital markets. These reforms involve (for
the first time) the removal of foreign restrictions on domestic equity holdings. The
stock markets of Kuwait, Lebanon, and Qatar have not been officially liberalized yet.
Bekaert et al. (2003) and Ben Naceur et al. (2008) indicate that many emerging
markets, including the MENA equity markets, were already indirectly open to
foreign investment before official reforms, by way of American Depositary Receipts
(ADRs). ADRs enable mutual funds, pension funds, and other US institutions to hold
securities that are fungible with foreign shares. Ben Naceur et al. (2008) indicates
that the notable progress of the capital markets in the MENA region since the
adoption of financial liberalization policies in the 1990s does not represent a solid
vehicle for real investment opportunities: several stock markets in the region are in
need of more transparency, particularly the promotion of timely disclosure and
dissemination of information to the public.
191
7.5
7.0
6.5
6.0
5.5
5.0
4.5
2005
2006
EGY
MOR
TU R
2007
IS R
OM
2008
JO R
QAT
2009
K UW
SAU
LE B
TU N
Figure 6.2: Natural logs of MENA stock indices.
6.5.2. Empirical framework
Since the cointegration test requires the time series to be integrated of the order I(1),
section 6.3.2 shows that natural logs of the MENA indices are I(1) in level and I(0)
in first differences. The natural logs of the MENA stock indices can therefore be
modelled by the cointegration test. Like the analysis presented in section 6.4.4 on the
lag length in VAR, this analysis determines the appropriate lag length in a VAR
model. The results of the optimal lag length are presented in Table 6-11 and reveal
that AIC, SBIC, and the HQIC agree on the first lag for all VAR models. Therefore
the first lag is chosen for the cointegration test.
192
Table 6-11. Optimal lag length in VAR
Number of lags
k 0
k 1
k 2
k 8
LR
NA
6629.263
173.646
169.470
AIC
-16.633
-42.717*
-42.509
-41.724
SBIC
-16.474
-40.813*
-38.862
-27.610
HQIC
-16.569
-41.950*
-41.040
-36.039
Notes: * indicates lag order selected by the criterion. ** indicate higher lags up to 8. See equation
(6.10) for the groups of endogenous series. LR, AIC, SBIC, and HQIC denote the Likelihood Ratio,
the Akaike Information Criterion, the Schwarz Bayesian Information Criterion, and the HannanQuinn Information Criterion. k refers to the number of lags in VAR. NA stands for not available.
6.5.3. Cointegration results
Since the natural log series are found to be I(1), the analysis proceeds to the
cointegration analysis. Similar to the work of Chan et al. (2002), the countries are
entered into the VAR model based on their market capitalization.31 The analysis
examines the following cointegration relationship:
f (Saudi Arabia, Turkey, Israel, Kuwait, Egypt, Qatar, Morocco, Jordan,
Oman,Tunisia, Lebanon) 0
If the null hypothesis with ( r
(6.10)
0 ) is not rejected, then we conclude that the MENA
markets are not cointegrated. If the null hypothesis with ( r
0 ) is rejected against
the alternative ( r 1 ) cointegrating vectors, this suggests that the MENA markets are
experiencing a partial integration if 0 r 10 cointegrating vectors, or full
integration if there exists r 10 cointegrating vectors.
The results are not markedly different when the order of countries is entered as in equation (6.6) to
(6.9) without the developed markets of Germany, Japan, UK, and USA.
31
193
6.5.4. Error Correction (EC) Model
Once the variables included in the VAR model are found to be cointegrated, the
analysis in this section adopts the EC model, as described in section 6.3.4.
6.6. Empirical results on the regional integration of MENA markets
The cointegration outputs (maximum Eigenvalue,
max
,
trace
) are presented in table
6-12. The table shows the hypothesis of the number of cointegrating vectors: the
second column reports the maximum Eigenvalue, and the third and fourth columns
present
max
and
trace
test statistics respectively.
The results in Table 6-12 show significant cointegrated vectors in the cointegration
analysis. The test statistics of
max
and
trace
show that the null hypothesis ( r
0 ) is
rejected against the alternative of one or more cointegrating vectors ( r 1 ). This
evidence reveals the presence of at least two cointegrating vectors; hence, common
stochastic trends lie behind the long-term co-movements of equity markets in the
MENA region, and signs of regional integration. Earlier studies such as Girard and
Ferreira (2003), Maghyereh (2006), and Neaime (2002), using earlier stock market
data, do not find signs of cointegration among the MENA stock markets. Girard and
Ferreira (2003) for example, conclude that the MENA stock markets offer
diversification potential for the global investor because of the lack of regional
integration, and note that this evidence could be beneficial in global asset allocation
strategies. The more recent empirical evidence presented in this section indicates that
such gains from international diversification across MENA markets can no longer be
assumed.
To test whether the global financial
crisis has affected the degree of stock
194
market integration in the MENA region, the cointegration analysis is also repeated
with the inclusion of a dummy variable. In Table 6-6, all the structural breaks
coincide with the month of September, 2008; therefore, the dummy variable equals 0
before the month of September, 2008; 1 until the end of the sample period.
Table 6-12. Empirical results of Johansen s cointegration test
Hypothesis
0
r=0
r 1
r 2
r 3
r 4
r 5
r 6
r 7
r 8
r 9
r 10
0
r>0
r>1
r>2
r>3
r>4
r>5
r>6
r>7
r>8
r>9
r =10
VAR: k 1
Eigenvalue
max
0.270
0.257
0.199
0.150
0.136
0.109
0.059
0.056
0.047
0.031
0.024
78.024**
73.723**
55.002
40.294
36.378
28.573
14.966
14.168
11.959
7.694
5.940
trace
366.722*
288.698*
214.975
159.973
119.679
83.300
54.728
39.762
25.594
13.634
5.940
Notes: *, ** denote significant at 1% and 5%, respectively.
0 (
1 ) refers to null
(alternative) hypothesis. The null hypothesis is that there is no cointegration if r 0 . For
the choice of optimal lag in VAR, see table (6-10). The cointegration model contains a
constant and a deterministic trend.
Table 6-13 shows the number of the cointegrating vectors when the exogenous
variable of the structural break dummy is added to the cointegration relationship. The
results reveal a significant increase of cointegrating vectors, suggesting that the
global financial crisis proxied by the structural break has had a great influence on the
common stochastic trends of the MENA stock markets. This empirical evidence
suggests that the stock market integration has significantly increased, once the
control of a global shock such as the global financial crisis is considered.
195
Table 6-13. Number of cointegration vectors
Cointegration model
Without structural breakpoint
With structural breakpoint
Cointegrating vectors
2
3
Notes: the cointegration test is augmented with a common structural break
found for the developed markets of Germany, Japan, UK, and USA. The
dummy variable equals 0 before August 2008, 1 until the end of the sample
period.
6.6.1 Error correction results
Table 6-14 presents the long- and short-run equilibrium relationships among MENA
markets. The lag lengths of the series are chosen as they were for the cointegration
test. In the long-run, the EC terms are negative and statistically significant for
Turkey, Kuwait, Egypt, Qatar, Morocco, Jordan, Oman, and Lebanon. These markets
exhibit significant contributions to the long-run equilibrium relationship, but the
speed of adjustment to equilibrium following a shock is slow. Specifically, the EC
term explains movement of about 4.0% to 9.5% a week toward equilibrium. The
interpretation of the empirical findings in the long-run implies that when there is a
deviation from the equilibrium cointegrating relationship, as measured by the EC
term, stock market of Egypt, Jordan, Kuwait, Lebanon, Morocco, Oman, Qatar, and
Turkey adjust to clear the disequilibrium. The markets of Saudi Arabia, Israel, and
Tunisia are exogenous and thus the first receptors of any external shock: in other
words, they will be the first to pick up the panic, and will then transmit the shock to
other stock markets.
In the short-run, the fluctuations in the stock market of Turkey can explain market
movements in Saudi Arabia, Israel, Kuwait, Egypt, Qatar, Jordan, and Oman. This
implies that the Turkish stock market leads the MENA markets in the short-run,
while Qatar leads in the long-run (based on the EC term). This is a clear indication of
the leadership roles of the Turkish and
Qatari stock markets in the region. Israel
196
can explain the stock market movements in Qatar, while Egypt and Qatar can explain
them in Turkey, Qatar can explain them in Egypt and Jordan, and Jordan and
Morocco can explain them in Oman. Surprisingly, Lebanon, the smallest market in
the MENA region, can explain the stock market movements in Egypt, Jordan, and
Oman.
To summarize, the regional integration of MENA stock markets seems to be a recent
phenomenon, found only since the markets under study started to liberalize their
stock markets to foreign investors. Since the analysis covers the period of the global
crisis of 2008 which began in the US and rapidly spread across global financial
markets, it is obvious that this will be reflected in the MENA region. Indeed, MENA
stock markets, in line with the international trends, have been hit by falling stock
prices and slowdown of commercial, residential, and industrial projects. The
empirical evidence suggests that the global market instability has been transmitted to
the MENA stock markets. This implies, in turn, that further shocks to the global
financial system may increase stock market integration in the MENA region.
6.7. Conclusion
This chapter investigates the co-movements of MENA markets with four developed
markets of Germany, Japan, UK, and USA separately, using the Johansen s
cointegration technique. The chapter first investigates the properties of the MENA
stock markets along with the developed markets under study. The descriptive
summary shows that the some of the MENA equity returns as well as UK and USA
have negative average weekly returns, and all the markets deviate from the normal
distribution. The chapter also examines the correlation analysis and finds that Israel,
197
Table 6.14. EC model results
EC term
In SAU
InTUR
In ISR
-0.013
0.046
-0.046***
-0.126
In ISRt 1
-0.072
0.057
In EGYt 1
-0.071
-0.197***
-0.059
InOM t 1
-0.003
-0.095
-0.038
In MORt 1
-0.034
-0.026
0.026
In LEBt 1
0.106
-0.019
-0.026
1.396
392.073
-3.057
1.379
336.686
-2.610
1.662
554.654
-4.368
In SAU t 1
InTURt 1
0.196*
In KUWt 1
-0.111
InQATt 1
0.063
In JORt 1
-0.055
InTUN t 1
-0.001
Adj. R
F-statistic
Log likelihood
AIC
2
-0.003
0.019
In JOR
In OM
InTUN
In LEB
-0.095*
-0.056
-0.040*
-0.001
-0.054*
-0.060
-0.073*
0.008
-0.015
-0.048
-0.048*
0.053
-0.123
-0.170
0.010
-0.268**
-0.028
0.007
-0.155
0.007
-0.042
-0.163**
-0.048
-0.083
0.049
-0.073
0.054
-0.053
-0.056
-0.013
-0.021
-0.070
-0.205**
-0.001
0.014
0.080
-0.001
0.153**
0.100
0.077
2.674
453.971
-3.556
6.991
439.793
-3.442
0.045
0.003
0.018
In MOR
-0.077*
-0.084
0.046
0.096
InQAT
-0.071*
0.047
0.090*
0.094
In EGY
-0.010
-0.020
-0.013
0.299*
In KUW
0.077
0.145*
-0.006
-0.099
0.267*
-0.019
-0.042
0.150***
0.070
-0.053
-0.111
0.049
-0.090
-0.015
0.003
0.031
0.010
-0.005
0.000
0.075
0.001
0.225
0.204*
0.069
0.081
0.024
0.137
-0.002
0.178
5.442
449.209
-3.518
0.041
0.096*
0.053
-0.063
-0.073
0.101***
0.062
0.008
0.038
-0.060
-0.023
0.032
-0.027
-0.048
-0.128***
-0.075
0.099
0.122
-0.037
-0.002
0.143
0.002
-0.004
-0.056
0.014
-0.061
0.045
-0.108
-0.206*
0.104** 0.106***
0.004*** -0.004***
0.022
0.123
1.467
512.267
-4.026
0.133*
3.880
511.069
-4.017
4.439
488.883
-3.838
0.037
0.008
0.031
-0.137
0.014
0.024
0.205*
0.913
557.964
-4.395
2.454
445.587
-3.489
0.003
0.066
Notes: *, **, *** denote significant at 1%, 5%, and 10% respectively. The EC model reports the first cointegrating vector for the cointegration equations.
198
Turkey, Lebanon and Tunisia have significant correlations with all the developed
markets while most of the other MENA markets have insignificant correlations.
The chapter examines the unit root tests using the ADF, PP, and KPSS tests for the
natural logs and their first differences. The outcomes show that the natural logs of the
stock indices are I(1) in level and I(0) in first differences. Before the cointegration
technique is examined, the analysis determines the appropriate lag length in the VAR
model. The outcomes show that the first lag is chosen as the appropriate lag for all
the cointegration tests. The next step conducts the multivariate cointegration test
using Johansen s cointegration technique. The results show that the MENA stock
markets share at least two cointegrating vectors, which indicates that they have longrun equilibrium relationship with the developed markets. When the analysis adopts
the structural break by Zivot and Andrews (1992) test, four breakpoints among the
developed equity markets are detected in the month of September 2008. These
coincide with the onset of the global financial crisis. The findings show that, after
accounting for the structural break, the number of cointegrating vectors increase for
Germany, UK and USA. The possibility of structural breaks in the cointegration test,
offer strong evidence of the long-run relationship between the MENA markets and
the developed markets. These findings imply that the integration of the MENA
markets has increased; and seem to follow a trend of market openness. The analysis
also adopts the EC model augmented with the dummy variables found in Zivot and
Andrews (1992) model. The EC model results show that most of the MENA
markets show signs of convergence toward long-run equilibrium. They also show
that Saudi Arabia and Tunisia deviate from the long-run equilibrium and will likely
be the receptor of further external shocks.
199
The regional integration of the MENA markets excluding the developed markets
under study is then examined. The results show that the MENA stock markets are
cointegrated and, therefore, show signs of regional financial integration. The
structural breaks for the timing of the global financial crisis in September 2008
indicate that regional stock market integration has significantly increased. The
implication of this empirical evidence is that the stock market integration in the
MENA region can increase in the future, depending on the occurrence of further
shocks to the global financial system.
The implications for policy makers would suggest that the MENA stock markets may
adopt certain precautious measures during turbulent periods such as the imposition of
price limit on transactions to prevent fluctuation of stock prices, margin requirements
for investors to prevent high-frequency trading, and circuit breakers to prevent stock
market plunges.
200
Summary and Concluding Remarks
7.1. Introduction
The aim of this thesis is to examine the responses of MENA stock markets to major
financial and geopolitical unrests, using various econometrics techniques: the
Kalman filter, the multiple structural breaks of Bai and Perron (2003), and
cointegration. The thesis attempts to examine the stock market exposure by way of
economic variables, mainly monetary policy variables. The analysis sheds light on
MENA stock markets exposure to the global financial system and their signs of
integration into the global capital market.
This chapter summarizes the main findings of the previous chapters, discusses some
of the possible implications and lessons, outlines the major limitations that could
undermine the results, and identifies areas of future research to minimize these
limitations.
201
7.2. Concluding remarks
Chapter 3 investigates the country beta instability using the unconditional ICAPM
for equity markets in the MENA region. The results show that the country betas of
MENA equity markets are unstable, especially in the long-run, and have increased
considerably since the 1990s. The implication is that the MENA stock markets are
becoming more affected by global capital markets; this finding is to some extent,
consistent with extant literature on the degree of stock market integration. The degree
of financial integration points increased market exposure to various episodes of
global financial and geopolitical unrest. Taking the 2003 war on Iraq as the most
prominent geopolitical crisis, one that created unprecedented instability in the
MENA region, the chapter examines the impact of the war on the country beta of
MENA equity markets, using the conditional ICAPM to capture the war effect in the
region.
The state-space model and the Kalman filter is employed and modified to allow a
direct incorporation of the effect of the war on Iraq with the evolution of the country
beta over time. The results show that although the war had a positive impact on the
country beta of MENA equity markets, it was statistically significant for Egypt,
Kuwait, Morocco and Tunisia. The signs of significant and positive impact reflect a
sudden surge in country beta since the war. Although the war s impact was limited to
four equity markets in the region, this result corresponds to those found concerning
the impact of others incidents of geopolitical unrest on global stock markets.
Chapter 4 provides a multiple structural breaks analysis on the country beta, using
the novel Bai and Perron (2003) model, in which the breakpoints are unknown a
priori. The method of Bai and Perron
(2003) has become very useful for
202
further investigation of multiple structural breaks in many time series analyses, but
no other work has yet been done on the examination of structural breaks of the
country beta series. The chapter finds that the country beta of MENA equity markets
have coped with significant structural breaks that coincide with various crises such as
the Asian, Russian, and Turkish financial crises and the September 11, 2001 attacks.
The empirical evidence suggests that these events destabilized the MENA equity
markets and increased their market exposure. In addition, due to the unique
paramatization of the Bai and Perron (2003) model, the results reveal that more
countries have experienced structural breaks due to the war on Iraq compared to the
results found from the Kalman filter; and that the markets of Egypt, Israel, Kuwait,
and Turkey are the most receptive to financial and geopolitical crises.
Chapter 5 examines the determinants of country beta in the MENA region, using a
panel data regression analysis. The dramatic changes in the economic variables of
MENA countries indicate positive and negative effect on the equity markets and their
country betas. Money supply and inflation seem to be major concerns in the MENA
region. The chapter also finds that the MENA region has benefited from high foreign
currency reserves, which tend to alleviate the aggravation of country beta over time.
Chapter 6 examines the integration of the MENA equity markets with the developed
markets of Germany, Japan, UK, and USA. The chapter finds that the MENA
markets are integrated with the developed markets. When the cointegration analysis
is applied to the stock market relationship since the global financial crisis, the results
suggest that integration has significantly increased. The important implication of
increased market integration is that it offers less opportunity for international
investment, since higher market exposure to global capital market tends to shrink
203
opportunities for international portfolio diversification. The chapter also examines
the regional integration of MENA markets without reference to developed markets.
The results show signs of regional financial integration; this level of integration has
also increased significantly since the global financial crisis.
7.3. Implications of the results
This thesis uncovers significant implications for local, regional, and international
investors, and for the development and performance of MENA markets. In Chapters
3 and 4, the empirical evidence shows that MENA equity markets have responded to
many global financial and geopolitical crises, and this will have substantial negative
outcomes on their stock market performance in the future. Further deteriorations in
the global economy or fluctuations in the global capital market, or further
geopolitical and regional unrest, could make a significant contribution to amplifying
systematic risk in the MENA region. This evidence refutes the supposed immunity of
MENA equity markets to financial or geopolitical risks
even during the early stages
of their stock market liberalization. For example, the combination of the results of
Chapters 3 and 4 for the war on Iraq show that further political destabilization in the
region could adversely affect the regional stock markets and cause further
aggravation to the systematic risk perceived by local and international investors. In
addition, further regional or global financial instability could have profound effects
on the pricing of assets or allocative strategies by regional and international investors
since the MENA stock markets are not completely isolated (totally segmented) from
the global financial system. The effects will vary across the MENA markets
depending on their degree of openness to the global economy and their market
exposure to the global financial system.
204
In Chapter 5, monetary policies in the MENA region are revealed to shown
significant signs of weaknesses due to the different monetary systems and exchange
rate regimes. The outcomes show that money supply and inflation are aggravating
the country betas, which is not a good sign for local, regional or international
investors looking for opportunities in the region. High levels of foreign currency
reserves seem to alleviate the country betas of the MENA equity markets. These
results have important implications for local and international investors and for
policy makers in the region. First, investors will be able to single out the major
sources of their investment risk in the region and update their expectations based on
the aggregated economic conditions of the MENA region. Second, investors will be
able to hedge against specific fundamental risks when considering future investment
in these markets. On the other hand, from a policy perspective, policy makers such as
central banks or governments will have a better understanding of the impact of such
variables on the country beta, which will enable them to devise macroeconomic
policies that will lower market risk exposure and improve the financial stability in
regional equity markets. For example, MENA countries could adopt structural
policies like inflation targeting to curb inflation in order to minimize investment
uncertainties. This will attract international investment and help the equity markets
move toward regional integration. Further structural and solid monetary policies and
coordination of MENA countries will reap the benefit of risk reduction in the region.
Coordination across the region will help ameliorate the effects of any financial or
currency crises that might adversely impact their country beta in the future.
In Chapter 6, the evidence shows that the MENA equity markets are integrated with
the major developed stock markets, and that the degree of co-movement has
205
intensified, especially during times of global financial upset. The increased linkages
have profound implications for international investors who seek diversification
benefits in emerging markets. The more MENA equity markets show signs of
increased integration into the global capital market, the more they become sensitive
to global capital market volatility, and hence, offer fewer chances for further gains
and capital growth to international investors. In addition, as MENA equity markets
show signs of increased integration, they become more susceptible to global market
instability, and international investors will revise expectations of asset pricing of
investments in the region.
7.4. Limitations and weaknesses of the thesis
Although the chapters in this thesis have contributed to the literature and offer new
insights into the MENA stock market exposure to global unrest, there are number of
limitations that could undermine the results obtained, and therefore the conclusions
based on the results.
First, the thesis has focused on the vulnerability of market risk exposure to major
external shocks like geopolitical risk and international financial crises. A major
problem arising from this perspective is that there are many incidents of financial and
geopolitical unrest that have not been included in the study. These range from local
events such as the assassination of Lebanese president Rafiq Al Hariri, to regional
events such as Israeli-Palestinian conflicts and the Middle East peace talks, to global
events such as terrorist attacks in England, Spain, Morocco, Egypt, Jordan, Turkey,
and Indonesia.
Second, the use of the single ICAPM in Chapter 3 under the assumption of market
206
integration may undervalue the results concerning stock market exposure. However,
the use of this model has been justified to the best of the author s ability, based on
the literature presented. In addition, the choice of the Kalman filter to estimate the
time-varying country beta has not been based on solid comparison among various
modelling techniques. As the literature shows, the comparisons between these
models do not provide equal utility, and the choice of which model to recommend for
the estimation of beta or country beta has spurred heated debate. The author of this
thesis acknowledges that the choice of the state-space model and the Kalman filter
approach is a personal preference.
Third, the multiple structural breaks analysis in Chapter 4 could have been
constrained by the time span, which affects the number of structural breakpoints.
However, based on recent studies published on the structural breaks analysis, the
time span was carefully chosen based on the length of the sample data and its
frequency.
Fourth, the cointegration technique in Chapter 6 could be sensitive to a large number
of time series. Other cointegration techniques with structural breaks, such as the
Gregory and Hansen s (1996), are even more sensitive to the number of regressors.
In addition, the impact of the global financial crisis on MENA stock markets could
vary significantly from one country to another, if these were studied separately,
instead of drawing conclusions based on the full sample of countries.
Fifth, the financial and economic data on the MENA region suffers from a lack of
important variables, especially for Bahrain, Lebanon, Morocco, Qatar, and UAE. In
addition, other factors such as political, financial, and economic risks ratings would
207
have improved the results, especially in Chapter 5. The data frequency (weekly or
monthly) for every chapter was chosen to suit the model used, although the choice
was justified to the best of the author s ability by the precedents in a large number of
studies. Another limitation is the use of proxies in the thesis, especially for the
interest rate (Chapter 5) and the international risk free rate (Chapter 3).32
7.5. Areas of future research
Based on the findings of this thesis and its limitations, useful areas for future
research have emerged. One of these is to examine the qualitative measure of country
risk ratings in the MENA region based on the data supplied by credit risk rating
agencies. It would also be a great contribution to the literature to examine these
measures based on local and international events such as terrorist attacks and the
ongoing political unrest in the MENA area, especially of the recent uprisings in the
Arab region or what is called the Arab spring . Third, as further data for the MENA
markets emerge, further comprehensive tests of integration, their degree of openness
and vulnerability to the global economic and financial system, can be tested. Another
area of future research is to replicate other major studies on country beta using a
large and more comprehensive sample data, especially using monthly frequency data.
The use of the US Treasury bill rate as a proxy for the international risk free rate may be
constrained by the recent decision S&P to downgrade the credit rating of the US government from
AAA to AA+.
32
208
Appendix
The state-space form of the time-varying market model in equation (3.4) can be
expressed as (Wells, 1996, p.76-79, Yao and Jao, 2004):
Yt
t
t
t
t
~
( 0, )
The state or transition equation takes the form:
t
t 1
t
t
~
(0, )
Where Yt is the country s market returns, and
Zt
(1, Rmt ) is a vector. The matrices
the error terms
t
and
t
,
t
and
the unobserved state vector, and
are assumed to be known, and
are serially independent from each other. By setting the
to the identity matrix, the model evolves as a random-walk
transition matrix
model. The Kalman filter consists of two important steps:
Prediction:
Treating period t 1 as the initial period, the estimate of the state and its covariance
at time t , conditional on information available at t 1 , is:
t |t 1
t 1| t 1
t |t 1
t 1| t 1
,
When the new observation and corresponding
prediction error
t
t
are available, the one-step-ahead
and its variance, ft can be obtained by:
209
Rt Rt |t
t
ft
t |t 1
t
Rt
1
t
t
2
B t |t 1 ,
Updating:
Given the new information about
t
, the state and its covariance can now be
updated, conditional on the information available at time t :
Bt |t
Bt |t
Kt
1
where
t |t 1
,
is the weight assigned to the new information about
contained in the
prediction error. The covariance of time t can be obtained:
t |t
t |t 1
where K t
K t Zt
t \t 1
t |t 1
t
to information about
f t |t 1 1 , is the Kalman gain which determines the weight assigned
t
contained in the prediction errors. The Kalman filter enables
the likelihood function to be estimated by providing the minimum mean squared
errors estimates of the unobserved variable. The likelihood function can be given as:
L
T
2
T
t
1
log ( f t |t 1 )
2
1
T
t 1
2
t |t 1
f t |t
1
The Berndt, Hall, Hall, and Hausman (1974) iterative algorithm can be used to
maximize the log-likelihood function.
210
References
Abell, J. & Krueger, T. (1989). Macroeconomic Influence on Beta. Journal of
Economics and Business, 41, 185-93.
Aggarwal, R. & Kyaw, N. (2005). Equity Market Integration in the NAFTA Region:
Evidence from Unit root and Cointegration Tests. International Review of
Financial Analysis, 14, 393-406.
Aguiar, M. & Bruner, F. (2006). Determining Underlying Macroeconomic
Fundamentals during Emerging Market Crises: Are Conditions as Bad as they
Seem? Journal of Monetary Economics, 53, 699 724.
Agus, A. & Erbil, C. (2010). The Effects of Instability on Bilateral Trade with Iraq.
Turkish Policy Quarterly, 4, 169 77.
Ahlgren, N. & Antell, N. (2002). Testing for Cointegration between International
Stock Prices. Applied Financial Economics, 12, 851 61.
Akyuz, Y. & Boratav, K. (2003). The Making of the Turkish Financial Crisis. World
Development, 31, 1549 66.
Alexander, G. & Benson, P. (1982). More on Beta as a Random Coefficient. Journal
of Financial and Quantitative Analysis, 17, 27 36.
Al-Otaibi, B. & Slywester, K. (2007). Oil Price Movements and the GDP Growth in
GCC Countries: Is the Relationship Asymmetric? Indian Journal of Economics
and Business, 6, 299 309.
Al Refai, H. (2011). The Impact of the Iraq War on the Country Beta of MENA
Markets. International Journal of Economics and Finance, 3, 12 8.
Alper, E. & Yilaz, K. (2004). Volatility and Contagion: Evidence from the Istanbul
Stock Exchange. Economic Systems, 28, 353 67.
Amihud, Y. & Wohl, A. (2004). Political News and Stock Prices: The Case of the
Saddam Hussein Contracts. Journal of Banking and Finance, 28, 1185 2000.
Andern, N. & Kjellsson, M. (2005). Regional and Global Stock Market Integration in
the EU. European Financial Management Association. Working Paper, Online:
http://www.efmaefm.org/efma2005/papers/10-andren_paper.pdf.
Anderson, T., Bollerslev, T., Diebold, F. & Wu, G. (2005). A Framework of
Exploring the Macroeconomic Determinants of Systematic Risk. American
Economic Review, 95, 398 404.
Andrade, J. & Teles, V. (2006). An Empirical Model of the Brazilian Country risk
An Extension of the Beta Country Risk Model. Applied Economics, 38, 1271
78.
211
Andrade, J. & Teles, V. (2008). Monetary Policy and Country Risk. Applied
Economics, 40, 2017 24.
Arestisa, P., Caporaleb, M., Cipollinic, A. & Spagnolo, N. (2005). Testing for
Financial Contagion between Developed and Emerging Markets during the
1997 East Asian Crisis. International Journal of Finance and Economics, 10,
259 67.
Arouri, M. (2004). The Impact of Increasing Stock Market Integration on Expected
Gains from International Portfolio Diversification: Evidence from a
Multivariate Approach with Time Varying Risk. Economic Bulletins, 6, 1 13.
Arouri, M. (2006). International Asset Pricing and World Market Integration:
Evidence from a Partially Integrated ICAPM with Asymmetric Effects.
Frontiers in Finance and Economics, 3, 70 94.
Arouri, M., Bellalah, M. & Nguyen, D. (2009). Comovements in International Stock
Markets: New Evidence from Latin American Emerging Countries. Applied
Economics Letters, 17, 1323 28.
Athanassiou, E., Kollias, C. & Syriopoulos, T. (2006). Dynamic Volatility and
External Security Related Shocks: The Case of the Athens Stock Exchange.
Journal of International Financial Markets, Institutions and Money, 16, 411
24.
Attia, A. (2004). The Oil Stock Market Reaction to the Gulf War. Working Paper,
accessed from: http://www.sbaer.uca.edu/research/sribr/1998/pdf/19.pdf.
Bai, J. & Perron, P. (1998). Estimating and Testing Linear Models with Multiple
Structural Changes. Econometrica, 66, 47 78.
Bai, J. & Perron, P. (2003). Computation and Analysis of Multiple Structural Change
Models. Journal of Applied Econometrics, 18, 1 22.
Basher, S. & Sadorsky, P. (2006). Oil Price Risk and Emerging Stock Markets.
Global Finance Journal, 17, 224 51.
Basse, T., Friedrich, M. & Vazquez Bea, E. (2009). REITs and the Financial Crisis:
Empirical Evidence from the US International Journal of Business and
Management, 4, 3 10.
Beirne, J., Caporale, G., Schulze-Ghattas, M. & Spagnolo, N. (2009). Volatility
Spillovers and Contagion from Mature to Emerging Stock Markets. The
European Central Bank, Working Paper No. 1113 / November.
Bekaert, G., Erb, C., Harvey, C. & Viskanta, T. (1996a). The Behaviour of Emerging
Market Returns, in The Future of Emerging Capital Flows (Eds) E. Altman, R.
Levich and J. Mei, Kluwer Academic Publishers, Boston, pp. 107 73.
Bekaert, G., Erb, C., Harvey, C. & Viskanta, T. (1996b). The Cross-Sectional
Determinants of Emerging Equity Market Returns. Working Paper:
http://ssrn.com/abstract=826454.
212
Bekaert, G., Erb, C., Harvey, C. & Viskanta, T. (1998). Distributional Characteristics
of Emerging Markets Returns and Asset Allocation. Journal of Portfolio
Management, 24, 102 16.
Bekaert, G. & Harvey, C. (1995). Time-Varying World Market Integration. Journal
of Finance, 1, 403 44.
Bekaert, G. & Harvey, C. (1997). Emerging Equity Market Volatility. Journal of
Financial Economics, 43, 29 78.
Bekaert, G. & Harvey, C. (2000). Foreign Speculators and Emerging Equity Markets.
Journal of Finance, 55, 564 614.
Bekaert, G. & Harvey, C. (2002). Research in Emerging Markets Finance: Looking
to the Future. Emerging Markets Review, 3, 429 48.
Bekaert, G. & Harvey, C. (2003). Emerging Markets Finance. Journal of Empirical
Finance, 10, 3 56.
Bekaert, G., Harvey, C. & Ng, A. (2005). Market Integration and Contagion. Journal
of Business, 78, 39 70.
Bely, J. (2007). How Homogeneous are the Stock Markets of the Middle East and
North Africa? Quarterly Journal of Business & Economics, 46, 3 26.
Ben Naceur, S., Ghazouani, S. & Omran, M. (2008). Does Stock Market
Liberalization Spur Financial and Economic Development in the MENA
Region? Journal of Comparative Economics, 36, 673 93.
Berglund, T. & Knif, J. (1999). Accounting for the Accuracy of Beta Estimates in
CAPM Tests on Assets with Time-Varying Risks. European Financial
Management, 5, 29 42.
Bernadette, L. & O Callaghan, A. (2009). The Current Global Financial Crisis: Do
Asian Stock Markets Show Contagious or Interdependency Effects? XVth
Euro-Asia International Research Seminar, 10th 11th of September, PauFrance.
Berndt, E., Hall, B., Hall, R. & Hausman, J. (1974). Estimation and Inference in
Nonlinear Structural Models. Annals of Economic and Social Measurement, 3,
653 65.
Bilson, C., Brailsford, T. & Hooper, V. (2001). Selecting Macroeconomic Variables
as Explanatory Factors of Emerging Stock Market Returns. Pacific-Basin
Finance Journal, 9, 401 26.
Blass, O., Peled, O. & Yafeh, Y. (2004). The Determinants of Israel s Cost of
Capital: Globalization, Reforms and Politics. Israel Economic Review, 2, 29
54.
Blume, M. (1975). Betas and their Regression Tendencies. Journal of Finance, 30,
785 95.
213
Bodurtha, J., Chinhyung, C. & Senbet, L. (1989). Economic Forces and the Stock
Market: An International Perspective. Global Finance Journal, 1, 21 46.
Bodurtha, J. & Mark, J. (1991). Testing the CAPM with Time-varying Risks and
Returns. Journal of Finance, 46, 1485 1505.
Bos, T. & Newbold, P. (1984). An Empirical Investigation of the Possibility of
Stochastic Systematic Risk in the Market Model. Journal of Business, 57, 35
41.
Boyd, J., Levine, R. & Smith, B. (2001). The Impact of Inflation on Financial Sector
Performance. Journal of Monetary Economics, 47, 221 48.
Breitung, J. (2000). The Local Power of some Unit Root Tests for Panel Data.
Advances in Econometrics, 15, 161 77.
Brooks, C., Faff, R. & McKenzie, M. (2002). Time-Varying Country Risk: An
Assessment of Alternative Modelling Techniques. The European Journal of
Finance, 8, 249 74.
Brooks, R. (2003). Time Varying Betas on China s A and B Share Indices.
Proceedings of the 15th Annual Conference of the Association for Chinese
Economics Studies, Australia (ACESA), Melbourne.
Brooks, R. & Del Negro, M. (2002). International Stock Returns and market
Integration: A Regional Perspective. IMF Working Paper No. 02/202.
Brooks, R., Faff, R. & Lee, J. (1992). The Form of Time Variation of Systematic
Risk: Some Australian Evidence. Applied Financial Economics, 2, 191 98.
Brooks, R., Faff, R. & McKenzie, M. (1998). Time Varying Beta Risk of Australian
Industry Portfolios: A Comparison of Modelling Techniques. Australian
Journal of Management, 23, 1 22.
Brown, R., Durbin, J. & Evans, J. (1975). Techniques for Testing the Constancy of
Regression Relationships over Time. Journal of Royal Statistician Association,
37, 145 64.
Brown, S. & Warner, J. (1980). Measuring Security Price Performance. Journal of
Financial Economics, 8, 205 58.
Aguiar, M. & Bruner, F. (2006). Determining Underlying Macroeconomic
Fundamentals during Emerging Market Crises: Are Conditions as Bad as they
Seem? Journal of Monetary Economics, 53, 699 724.
Bruner, R., Li, W., Kritzman, M., Myrgren, S. & Page, S. (2008). Market Integration
in Developed and Emerging Markets: Evidence from the CAPM. Emerging
Markets Review, 9, 89 103.
Cai, C., Faff, R., Hillier, D. & McKenzie, M. (2006). Modelling Return and
Conditional Volatility Exposures in Global Stock Markets. Review of
Quantitative
Finance
and
Accounting, 27, 125 42.
214
Carrera, L. & Musso, I. (2009). Geopolitical Risk and Market Behaviour: Do
Terrorist Attacks Affect Market Returns? Watson Wyatt Worldwide, Working
Paper, online: http://www.watsonwyatt.com/repecarticle/geopolitical_risk.pdf.
Cashin, P. & McDermott, J. (1998). Testing the Consumption-CAPM in Developing
Equity Markets. International Journal of Finance and Economics, 3, 127 41.
Chan, K., Karolyi, A. & Stulz, R. (1992). Global Financial Markets and the Risk
Premium on US Equity. Journal of Financial Economics, 32, 137 67.
Chen, A. & Siems, T. (2004). The Effects of Terrorism on Global Capital Markets.
European Journal of Political Economy, 20, 249 66.
Chen, G., Firth, M. & Rui, O. (2002). Stock Market Linkages: Evidence from Latin
America. Journal of Banking and Finance, 26, 1113 41.
Chen, N., Roll, R. & Ross, S. (1986). Economic Forces and Stock Market. Journal of
Business, 59, 383 403.
Cheung, Y. & Ng, L. (1998). International Evidence on the Stock Market and
Aggregate Economic Activity. Journal of Empirical Finance, 5, 281 96.
Choudhry, T. (2001). The Long Memory of Time-Varying Beta: Examination of
Three Emerging Asian Stock Markets. Managerial finance, 27, 5 23.
Choudhry, T. (2002). The Stochastic Structure of Time-Varying Beta: Evidence from
the UK Companies. The Manchester School, 70, 768 91.
Choudhry, T. (2005a). September 11 and Time-Varying Beta of United States
Companies. Applied Financial Economics, 15, 1227 42.
Choudhry, T. (2005b). Time-Varying Beta and the Asian Financial Crisis: Evidence
from Malaysian and Taiwanese Firms. Pacific-Basin Finance Journal, 22, 93
118.
Chung, H. (2005). The Contagious Effects of the Asian Financial Crisis: Some
Evidence from ADR and Country Funds. Journal of Multinational Financial
Management, 15, 67 84.
Clark, E. & Kassimatis, K. (2006). Constructing and Testing the World Market
Portfolio for Dollar Based Investors. Working Paper, EFA Zurich Meetings.
Cordesman, A. & Al-Rodhan, K. (2005). The Changing Risks in Global Oil Supply
and Demand: Crisis or Evolving Solutions? Centre for Strategic and
International Studies Publications, online: http://csis.org/publication/changingrisks-global-oil-supply-and-demand-crisis-or-evolving-solutions.
Corrado, C. (1989). A Nonparametric Test for Abnormal Security-Price Performance
in Event Studies. Journal of Financial Economics, 23, 285 95.
Cowan, A. (1992). Nonparametric Event Study Tests. Review of Quantitative
Finance and Accounting, 2, 343 58.
215
Cuyvers, L., Plasmans, J., Soeng, R. & Van Den Bulcke, D. (2008). Determinants of
Foreign Direct Investment in Cambodia: Country-Specific Factor Differentials.
University of Antwerp, Faculty of Applied Economics, Working Paper No.
2008/003.
Damar, E. (2007). The Effect of the Iraq War on Foreign Bank Lending to the
MENA Region. Emerging Markets Finance and Trade, 43, 20 36.
Das, D. (2010). Contours of Deepening Financial Globalization in the Emerging
Market Economies. Global Journal of Emerging Market Economies, 2, 45 67.
Dhakal, D., Kandil, M. & Sharma, S. (1993). Causality between the Money Supply
and Share Prices: a VAR Investigation. Quarterly Journal of Business and
Economics, 32, 52 74.
Dickey, D. & Fuller, W. (1981). Likelihood Ratio Statistics for Autoregressive Time
Series with Unit Root Test. Econometrica, 49, 1057 72.
Dickinson, D. (2000). Stock Market Integration and Macroeconomic Fundamentals:
An Empirical Analysis, 1980 95. Applied Financial Economics, 10, 261 76.
Dominguez, J. (2007). Cost of Equity Capital and Country Risk: An econometric
Analysis of the Expected Rate of Return for Four Latin American Countries.
Economia, 23, 63 90.
Dotan, A. & Ofer, A. (1984). Variable versus Stationary Beta in the Market Model:
A Comparison Analysis. Journal of Banking and Finance 8, 525 34.
Dowling, J. & Verbiest, J. (2002). Is Volatility Built into Today s World Economy?
Asian Development Bank. ERD Policy Brief No. 4.
Drakos, K. (2004). Terrorism-Induced Structural Shifts in Financial Risk: Airline
Stocks in the Aftermath of the September 11th Terror Attacks. European
Journal of Political Economy, 20, 435 45.
Dungey, M., Fry, R., Gonzalez-Hermosillo, B. & Martin, V. (2003). Empirical
Modeling of Contagion: A Review of Methodologies. IMF Working Paper No.
04/78.
Ebner, M. & Neumann, T. (2005). Time Varying Betas of German Stock Returns.
Financial Markets and Portfolio Management, 19, 29 46.
Ender, W. & Sandler, T. (2008). After 9/11 is it all different Now? Journal of
Conflict Resolution, 49, 259 77.
Engle, C. & Rodrigues, A. (1993). Test of Mean-Variance Efficiency of International
Equity Markets. Oxford Economic Papers, 45, 403 21.
Engle, R. (1982). Autoregressive Conditional Heteroscedasticity with Estimates of
the Variance of U.K. Inflation. Econometrica, 50, 987 1008.
Engle, R., Lilen, D. & Robins, R. (1987). Estimating Time Varying Risk Premia in
216
the Term Structure: The ARCH-M Model. Econometrica, 55, 391 407.
Episcopos, A. (1996). Stock Return Volatility and Time Varying Betas in the
Toronto Stock Exchange. Quarterly Journal of Business and Economics, 35,
28 38.
Erb, C., Harvey, C. & Viskanta, T. (1995). Country Credit Risk and Global Portfolio
Selection. Journal of Portfolio Management, Winter, 74 83.
Erb, C., Harvey, C. & Viskanta, T. (1996a). Expected Returns and Volatility in 135
Countries. Journal of Portfolio Management, 22, 46 58.
Erb, C., Harvey, C. & Viskanta, T. (1996b). Political Risk, Financial Risk and
Economic Risk. Financial Analysts Journal, 52, 28 46.
Erdal, F. & Gunduz, L. (2001). An Empirical Investigation of the Interdependence of
Istanbul Stock Exchange with Selected Stock Markets. Conference paper,
Global Business and Technology Association International Conference
Proceedings.
Erdem, C., Arslan, C. & Erdem, M. (2005). Effects of Macroeconomic Variables on
Istanbul Stock Exchange Indices. Applied Financial Economics, 15, 987 94.
Fabozzi, F. & Francis, J. (1978). Beta as a Random Coefficient. Journal of Financial
and Quantitative Analysis, 13, 106 16.
Faff, R. & Heaney, R. (1999). An Examination of the Relationship between
Australian Industry Equity Returns and Expected Inflation. Applied Economics,
31, 915 33.
Faff, R. W., Hillier, D. & Hillier, J. (2000). Time Varying Beta Risk: An Analysis of
Alternative Modelling Techniques. Journal of Business Finance and
Accounting, 27, 523 54.
Fama, E. (1981). Stock Returns, Real Activity, Inflation, and Money. American
Economic Review, 71, 545 65.
Fama, E. & French, K. (1992). The Cross Section of Expected Returns. Journal of
Finance, 47, 427 65.
Fama, E. & French, K. (1998). Value versus Growth: The International Evidence.
Journal of Finance, 53, 1975 99.
Faris, H. (2003). The Socio-Economic and Political Impact of the War on the Arab
Region. United Nations, Beirut, Working Paper No. 03 0774.
Fedorova, E. & Vaihekoski, M. (2009). Global and Local Sources of Risk in Eastern
European Emerging Stock Markets. Journal of Economics and Finance, 59, 2
19.
Fernandez, V. (2006). The Impact of Major Global Events on Volatility Shifts:
Evidence from the Asian Crisis and 9/11. Economic Systems, 30, 79 97.
217
Fernandez, V. (2007). Stock Market Turmoil: Worldwide Effects of Middle East
Conflicts. Emerging Market Finance and Trade, 43, 58 102.
Fernandez, V. (2008). The War on Terror and its Impact on the Long-Term Volatility
of Financial Markets. International Review of Financial Analysis, 17, 1 26.
Fernandez-Serranoa, J. & Sosvilla-Riverob, S. (2001). Modelling Evolving LongRun Relationships: The Linkages between Stock Markets in Asia. Japan and
the World Economy, 13, 145 60.
Ferson, W. & Harvey, C. (1991). The Variation of Economic Risk Premiums.
Journal of Political Economy, 99, 285 315.
Ferson, W. & Harvey, C. (1993). The Risk and Predictability of International Equity
Returns. Review of Financial Studies, 6, 527 66.
Ferson, W. & Harvey, C. (1995). Predictability and Time-Varying Risk in World
Equity Markets. Research in Finance, 13, 25 88.
Ferson, W. & Harvey, C. (1999a). Conditional Variables and the Cross Section of
Expected Returns. Journal of Finance, 54, 1325 60.
Ferson, W. & Harvey, C. (1999b). Economic, Financial and Fundamental Global
Risk in and out of the EMU. Swedish Economic Policy Review, 6, 123 84.
Filis, G., Degiannakis, S. and Floros, C. (2011). Dynamic Correlation between Stock
Market and Oil Prices: The Case of Oil-Importing and Oil-Exporting
Countries. International Review of Financial Analysis, 20, 152 164
Flavin, T., Panopoulou, E. & Unalmis, D. (2008). On the Stability of Domestic
Financial Market Linkages in the Presence of Time-Varying volatility.
Emerging Markets Review, 9, 280 301.
Fontaine, T. (2005). Currency Crises in Developed and Emerging Market
Economies: A Comparative Empirical Treatment. IMF Working Paper No.
05/13.
Forbes, K. & Rigobon, R. (2002). No Contagion, Only Interdependence: Measuring
Stock Market Comovements. Journal of Finance, 57, 2223 61.
Gallup, J., Sachs, J. & Mellinger, A. (1998). Geography and Economic Growth.
Paper prepared for the Annual Bank Conference on Development Economics,
Washington, D. C., April 20 21.
Gangemi, M., Brooks, R. & Faff, R. (1999). Mean Reversion and the Forecasting of
Country Betas: A Note. Global Finance Journal, 10, 231 45.
Gangemi, M., Brooks, R. & Faff, R. (2000). Modeling Australia s Country Risk: A
Country Beta Approach. Journal of Economics and Business, 52, 259 76.
Gelos, G. & Sahay, R. (2001). Financial Market Spillovers in Transition Economies.
Economics of Transition, 9, 53 86.
218
Gerard, B., Hillion, P., De Roon, F. & Eiling, E. (2007). International Portfolio
Diversification: Currency, Industry and Country Effects Revisited. EFA
Working Paper, Berlin.
Ghanem, D. (2010). Fixed Exchange Rate Regimes and Price Stability: Evidence
from the MENA Countries. Laboratoire Montpellierain d Economie Theorique
et Appliquee (LAMETA), Working Paper No. 16.
Ghosh, J. (2009). The Global Financial Crisis, Developing Countries and India.
Online: http://www.networkideas.org/ideasact/jan09/PDF/Jayati.pdf.
Giannopoulos, K. (1995). Estimating the Time Varying Components of International
Stock Markets Risk. The European Journal of Finance, 1, 129 64.
Gilmore, C. & McManus, G. (2002). International Portfolio Diversification: US and
Central European Equity Markets. Emerging Markets Review, 3, 69 83.
Girard, E. & Ferreira, S. (2003). On the Evolution of Inter and Intraregional
Linkages to Middle East and North African Capital Markets. Quarterly Journal
of Business and Economics, 43, 21 43.
Gjerde, S. & Saettem, F. (1999). Causal Relations among Stock Returns and
Macroeconomic Variables in a Small, Open Economy. Journal of International
Financial Markets, Institutions and Money, 9, 61 74.
Goldberg, J. & Delgado, F. (2001). Financial Integration of Emerging Markets: An
Analysis of Latin America versus South Asia Using Individual Stocks.
Multinational Finance Journal, 5, 259 301.
Goldberg, J. & Veitch, M. (2002a). Country Risk and the Currency Contagion Effect.
Journal of Management Research, 2, 13 21.
Goldberg, J. & Veitch, M. (2002b). Mexico s Tequila Crisis-Hangover or Hair of the
Dog? Country Risk and Exchange Rate Regimes. Multinational Business
Review, 10, 82 90.
Goldberg, J. & Veitch, M. (2010). Country Risk and Financial Integration: A Case
Study of South Africa. Research in International Business and Finance, 24,
138 45.
Gregory, A. & Hansen, B. (1996). Residual-Based Tests for Cointegration in Models
with Regime Shifts. Journal of Econometrics, 70, 99 126.
Grieb, T. & Reyes, M. (2001). Time Varying Betas in an Emerging Stock Market:
Case of Brazil. American Business Review, 19, 118 25.
Griffin, J. (2002). Are Fama and French Factors Global or Country Specific? Review
of Financial Studies, 15, 783 803.
Groenewold, N. & Fraser, P. (1997). Share Prices and Macroeconomic Factors.
Journal of Business Finance and Accounting, 24, 1367 83.
219
Groenewold, G. & Fraser, P. (1999). Time-Varying Estimates of CAPM Betas.
Mathematics and Computers in Simulation, 48, 531 39.
Groenewold, N. & Fraser, P. (2000). Forecasting Beta: How Well Does the FiveYear Rule of Thumb Do? Journal of Business Finance and Accounting, 27,
953 82.
Gunasinghe, W. (2005). Investigation of Behaviour of Stock Markets in South Asia:
An Econometric Investigation. South Asia Economic Journal, 6, 165 91.
Gunduz, L. & Hatemi-J, A. (2004). The Effect of Asian Financial Crisis on the
Causal Relationship between Stock Prices and Exchange Rates: Evidence from
MENA Region. Finance Letters, 2, 1 8.
Gupta, R. and Guidi, F. (2012). Cointegration Relationship and Time Varying Comovements among Indian and Asian Developed Stock Markets. International
Review of Financial Analysis, 21 , 10 22.
Haddad, M. & Hakim, S. (2008). The Impact of War and Terrorism on Sovereign
Risk in the Middle East. Journal of Derivatives and Hedge Funds, 14, 237 50.
Hammoudeh, S. & Li, H. (2008). Sudden Changes in Volatility in Emerging
Markets: The Case of the Gulf Arab Stock Markets. International Review of
Financial Analysis, 17, 47 63.
Harvey, C. (1991). The World Price of Covariance Risk. Journal of Finance, 46,
111 57.
Harvey, C. (1995a). Predictable Risk and Returns in Emerging Markets. Review of
Financial Studies, 8, 773 816.
Harvey, C. (1995b). The Risk Exposure of Emerging Equity Markets. The World
Bank Economic Review, 9, 19 50.
Harvey, C. (2000). The Drivers of Expected Returns in International Emerging
Markets. Emerging Markets Quarterly, 4, 32 48.
Harvey, C. & Siddique, A. (2000). Conditional Skewness in Asset Pricing Tests.
Journal of Finance, 55, 1263 95.
Harvey, C., Solnik, B. & Zhou, G. (2002). What Determines Expected International
Asset Returns? Annals of Economics and Finance, 3, 249 98.
Harvey, C. & Zhou, G. (1993). International Asset Pricing with Alternative
Distributional Specifications. Journal of Empirical Finance, 1, 107 31.
Hassan, K., Maroney, N., El-Sady, H. & Telfah, A. (2003). Country Risk and Stock
Market Volatility, Predictability, and Diversification in the Middle East and
Africa. Economic Systems, 27, 63 82.
Hatemi-J, A., Eduardo, R. & Fang, T. (2005). US Equity Market Spillover and
Contagion Effects on Selected Asian Markets Vis-à-vis September 11.
220
International Economics, 58, 449 70.
Hatemi-J, A. & Hacker, R. (2009). Can the LR Test be Helpful in Choosing the
Optimal Lag Order in the VAR Model when Information Criteria Suggest
Different Lag Orders? Applied Economics, 41, 1121 25.
Heaney, R. & Hooper, V. (1999). World, Regional, and Political Risk Influences
upon Asia Pacific Equity Market Returns. Australian Journal of Management,
24, 131 42.
Hodrick, R., Tat-Cheeng, D. & Sengmueller, P. (1999). An International Dynamic
Asset Pricing Model. International Tax and Public Finance, 6, 597 620.
Hong, P., Nsimba, E., Gray, C. & Diallo, O. (2007). The Impact of Higher Oil Prices
on the Global Economy A Tale of Two Different Cases. Available at SSRN:
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=632367.
Hooker, M. (2004). Macroeconomic Factors and Emerging Market Equity Returns:
A Bayesian Model Selection Approach. Emerging Market Review, 5, 379 87.
Horta, P., Mendes, C. & Vieira, I. (2008). Contagion Effects of the US Subprime
Crisis on Developed Countries. CEFAGE-UE, Working Paper No. 2008/08.
Hoti, S. (2005). Modelling Country Spillover Effects in Country Risk Ratings.
Emerging Markets Review, 6, 324 45.
Hoti, S. & McAleer, M. (2004). An Empirical Assessment of Country Risk Ratings
and Associated Models. Journal of Economic Surveys, 18, 539 88.
Hsiao, C. (1981). Autoregressive Modeling and Money-Income Causality Detection.
Journal of Monetary Economics, 9, 85 106.
Hwang, S. & Pedersen, C. (2004). Asymmetric Risk Measures when
ModelingEmerging Markets Equities: Evidence for Regional and Timing
Effects. Emerging Markets Review, 5, 109 28.
Ibrahim, M. (2004). Time Varying Beta and Volatility in the Kuala Lumpur Stock
Exchange. International Journal of Business, 6, 117 30.
Im, K., Pesaran, M. & Shin, Y. (2003). Testing for Unit Roots in Heterogeneous
Panels. Journal of Econometrics, 115, 53 74.
International Monetary Fund (2004). Classification of Exchange Rate Arrangements
and Monetary Policy Frameworks. Accessed from: http://www.imf.org/
external/np/mfd/er/2004/eng/0604.htm#fn1.
Jayasuriya, S., Shambora, W. & Rossiter, R. (2009). Asymmetric Volatility in
Emerging and Mature Markets. Journal of Emerging Market Finance, 8, 25
43.
Jeon, B. (2001). Macroeconomic influences on Country Risk. Working Paper, Rice
University, Texas.
221
Johansen, S. (1988). Statistical Analysis of Cointegration Vectors. Journal of
Economic Dynamics and Control, 12, 231 54.
Johansen, S. & Juselius, K. (1990). Maximum Likelihood Estimation and Inference
on Cointegration with Application for the Demand for Money. Oxford Bulletin
of Economics and Statistics, 52, 169 210.
Johansson, A. (2009a). Stochastic Volatility and Time-Varying Country Risk in
Emerging Markets. The European Journal of Finance, 15, 337 63.
Johansson, A. (2009b). An Analysis of Dynamic Risk in the Greater China Equity
Markets. Journal of Chinese Economic and Business Studies, 7, 299 320.
Johnson, L. & Sakolis, G. (2008). Maximizing Equity Market Sector Predictability in
a Bayesian Time Varying Parameter Model. Computational Statistics and Data
Analysis, 52, 3083 106.
Johnson, R. & Soenen, L. (2002). Asian Economic Integration and Stock Market
Comovement. Journal of Financial Research, 7, 141 57.
Jones, D. (2002). Unlocking the Secrets of the Fed, Hoboken, New Jersey, John
Wiley & Sons, Inc.
Jones, D., Leiby, P. & Paik, I. (2004). Oil Price Shocks and the Macroeconomy:
What has been learned since 1996? Energy Journal, 25, 1 32.
Jorion, P. (1991). The Pricing of Exchange Rate Risk in the Stock Market. Journal of
Financial and Quantitative Analysis, 26, 363 76.
Juselius, K. (2011). Time to Reject the Privileging of Economic Theory over
Empirical Evidence? A Reply to Lawson. Cambridge Journal of Economics,
35, 423 36.
Kamin, S., Schinder, J. & Samuel., S. (2001). The Contribution of Domestic and
External Factors to Emerging Market Devaluation Crises: An Early Warning
Systems Approach, Board of Governors of the Federal Reserve System,
International Finance Discussion Papers, No.711.
Kaminsky, G. & Schmukler, S. (2002). Emerging Market Instability: Do Sovereign
Ratings Affect Country Risk and Stock Returns? The World Bank Economic
Review, 16, 171 95.
Kanas, A. (1998). The Linkages between the US and European Equity Markets:
Further Evidence from cointegration Tests. Applied Financial Economics, 8,
607 14.
Karakatsani, N. & Bunn, D. (2004). Modeling Stochastic Volatility in HighFrequency Spot Electricity Prices. Second Energy Management Seminar,
Working Paper, London Business School.
Kasibhatla, K., Stewart, D., Sen, S. & Malindretos, J. (2006). Are Daily Stock Price
Indices in the Major European
Equity Markets Cointegrated? Test and
222
Evidence. The American Economist, 5, 47 57.
Kaul, G. (1990). Monetary Regimes and the Relation between Stock Returns and
Inflationary Expectations. Journal of Financial and Quantitative Analysis, 15,
307 31.
Kenourgios, D., and Padhi, P. (2012). Emerging Markets and Financial Crises:
Regional, Global or Isolated Shocks? Journal of Multinational Financial
Management, 22, 24-38.
Khallouli, W. & Sandretto, R. (2010). Testing for Contagion of the Subprime Crisis
on the Middle East and North African Stock Markets: A Markov Switching
EGARCH Approach. Available at SSRN: http://ssrn.com/abstract=1687528.
Khoury, S. (2003). Country Risk and International Portfolio Diversification for
Individual Investor. Financial Services Review, 12, 73 93.
Kim, H. & Gu, Z. (2004). Impact of 9/11 Terrorist Attacks on the Return and Risk of
Airline Stocks. Tourism and Hospitality Research, 5, 150 63.
Klinz, W. (2008). The International Financial Crisis: Its Causes and what to do about
it? In: Liberals and Democrats Workshops, F. T. (Ed.). Available online at:
http://www.alde.eu/fileadmin/webdocs/key_docs/Finance-book_EN.pdf.
Koh, W. (2007). Terrorism and its Impact on Economic Growth and Technological
Innovation. Technological Forecasting and Social Change, 74, 129 38.
Koutmos, G., Lee, U. & Theodossiou, P. (1994). Time Varying Betas and Volatility
Persistence in International Stock Markets. Journal of Economics and
Business, 46, 101 12.
Kwiatkowski, D., Phillips, P., Schmidt, P. & Shin, Y. (1992). Testing the Null
Hypothesis of Stationarity against the Alternative of a Unit Root. Journal of
Econometrics, 54, 159 78.
Lagoarde-Segot, T. & Lucey, B. (2006). Financial Contagion in Emerging Markets:
Evidence from the Middle East and North Africa, Trinity College, Dublin. IIIS
Discussion Paper No. 114.
Lagoarde-Segot, T. & Lucey, B. (2009). Shift-Contagion Vulnerability in the MENA
Stock Markets. The World Economy, 32, 1478 97.
Lahreche-Revil, A. & Milgram, J. (2006). Exchange-Rate Policies and Trade in the
MENA Countries. Department of Economic Theory and Economic History of
the University of Granada, Working Paper No. 06/07.
Leippold, M. & Lohre, H. (2010). The Dispersion Effect in International Stock
Returns. Available at SSRN: http://ssrn.com/abstract=1139412.
Leong, S. & Felmingham, B. (2003). The Interdependence of Share Markets in the
Developed Economies of East Asia. Pacific-Basin Finance Journal, 11, 219
37.
223
Lie, F. & Faff, R. (2003). Global Industry Betas. Applied Economics Letters, 10, 21
26.
Lim, K., Brooks, R. & Kim, J. (2008). Financial Crisis and Stock Market Efficiency:
Empirical Evidence from Asian Countries. International Review of Financial
Analysis, 17, 571 91.
Lintner, J. (1965). The Valuation of Risk Assets and the Selection of Risky
Investments in Stock Portfolios and Capital Budgets. Review of Economics and
Statistics, 47, 13 37.
Madhusoodanan, P. & Kumar, H. (2008). An Empirical Verification of Cointegration
and Causality in Indian Stock Markets. A Case of BSE and NSE. South Asia
Economic Journal, 9, 159 72.
Maghyereh, A. (2006). Regional Integration of Stock Markets in MENA Countries.
Journal of Emerging Market Finance, 5, 59 94.
Malliaris, A. & Urrutia, J. (1995). The Impact of the Persian Gulf Crisis on National
Equity Markets. In Advances in International Banking and Finance, 1, 43 65.
Mallik, G. (2006). Has the Stock Market Integration between the Asian and OECD
Countries Improved after the Asian Crisis? Frontiers in Finance and
Economics, 3, 55 69.
Marashdeh, H. (2005). Stock Market Integration in the MENA Region: An
Application of the ARDL Bounds Testing Approach. Department of
Economics, University of Wollongong, Working Paper No. 05 27.
Mariano, G. & Roman, M. (2005). Study of Country-Risk for Non-Developed
Countries in 1980 2000. Applied Econometrics and International
Development, 5, 65 92.
Marshall, A., Maulana, T. & Tang, L. (2009). The Estimation and Determinants of
Emerging Market Country Risk and the Dynamic Conditional Correlation
GARCH Model. International Review of Financial Analysis, 18, 250 9.
Maysami, R. & Koh, T. (2000). A Vector Error Correction Model of the Singapore
Stock Market. International Review of Economics and Finance, 9, 79 96.
McAleer, M., Da Veiga, B. & Hoti, S. (2009). Value-at-Risk for Country Risk
Ratings. Center for Advanced Research in Finance, Faculty of Economics,
University of Tokyo, Working Paper No. 169.
McKenzie, M., Brooks, R. & Faff, R. (2000). The Use of Domestic and World
Market Indices in the Estimation of Time-Varying Betas. Journal of
Multinational Financial Management, 10, 91 106.
McMillan, D. (2005). Time Variation in the Cointegrating Relationship between
Stock Prices and Economic Activity. International Review of Applied
Economics, 19, 259 68.
224
McSweeny, E. & Worthington, A. (2007). A Comparative Analysis of Oil Prices as a
Risk Factor in Australian Industry Stock Returns, 1980 2006. University of
Wollongong, Working Paper No. 07/07.
Mergner, S. & Bulla, J. (2008). Time-Varying Beta Risk of Pan-European Sectors: A
Comparison of Alternative Modelling Techniques. The European Journal of
Finance, 14, 771 802.
Mie, J. & Gau, L. (2004). Political Uncertainty, Financial Crisis and Market
Volatility. European Financial Management, 10, 639 57.
Miller, S. (2006). Global Equity Systematic Risk Hedging and the Asian Crisis.
Conference Paper, 11th Finsia-Melbourne Centre for Financial Studies
Banking and Finance Conference, Melbourne.
Ministry of Economics and Trade, Lebanon. Accessed from: www.economy.gov.lb.
Mohanty, M. & Turner, P. (2006). Foreign Exchange Reserve Accumulation in
Emerging Markets: what are the Domestic Implications? BIS Quarterly
Review, September: 39 52.
Mookerjee, R. & Yu, Q. (1997). Macroeconomic Variables and Stock Prices in a
Small Open Economy: The Case of Singapore. Pacific-Basin Finance Journal,
5, 377 88.
Moonis, M. & Shah, J. (2003). Testing for Time-Variation in Beta in India. Journal
of Emerging Market Finance, 2, 163 83.
Mukherjee, T. & Naka, A. (1995). Dynamic Relations between Macroeconomic
Variables and the Japanese Stock Market: an Application of a Vector Error
Correction Model. Journal of Financial Research, 18, 233 37.
Nandha, M. & Faff, R. (2007). Does Oil Move Equity Prices? A Global View.
Energy Economics, 30, 986 97.
Narayan, R. & Smyth, R. (2005). Cointegration of Stock Markets between New
Zealand, Australia and the G7 Economies: Searching for Co Movement under
Structural Change. Australian economic papers, 44, 231 47.
Neaime, S. (2002). Liberalization and Financial Integration of MENA Stock
Markets. the 9th Economic Research Forum meeting. American University in
Sharja, UAE.
Neaime, S. (2005). Financial Market Integration and Macroeconomic Volatility in
the MENA Region: An Empirical Investigation. Review of Middle East
Economics and Finance, 3, 231 55.
Nell, E. & Semmler, W. (2007). The Iraq War and the World Oil Economy.
Constellations, 14, 557 85.
Newey, W. & West, K. (1987). A Simple Positive-Definite Heteroscedasticity and
225
Autocorrelation Consistent Covariance Matrix. Econometrica, 55, 703 8.
Nguyen, D. & Bellalah, M. (2007). Testing for Structural Breaks and Dynamic
Changes in Emerging Market Volatility. DEPOCEN, Working Paper No.
2007/02.
Nikkinen, J., Omran, M., P., Sahlstrom, P. & Aijo, J. (2008). Stock Returns and
Volatility Following the September 11 Attacks: Evidence from 53 Equity
Markets. International Review of Financial Analysis, 17, 27 46.
Nummenlin, K. & Vaihekoski, M. (2002). World Capital Markets and Finnish Stock
Returns. The European Journal of Finance, 8, 322 43.
Onour, I. (2007). Impact of Oil Price Volatility on Gulf Cooperation Council Stock
Markets Return. OPEC Review, 31, 171 89.
Page, J. (1999). The Impact of Lower Oil Prices on the Economies of Gulf States.
Middle East Policy, 6, 59 67.
Paleari, S., Redondi, R. & Vismara, S. (2005). Stock Market Interdependence during
the Iraq War. Investment Management and Financial Innovations, 3, 25 38.
Panetta, F. (2002). The Stability of the Relation between the Stock Market and
Macroeconomic Forces. Economic Notes, 31, 471 50.
Park, K. & Kim, M. (2007). The Industrial Relationships in Time-Varying Beta
Coefficients between Korea and United States. Applied Economics, 41, 1929
38.
Patell, J. (1976). Corporate Forecasts of Earnings per Share and Stock Price
Behavior: Empirical Tests. Journal of Accounting Research, 14, 246 76.
Patro, D., Wald, J. & Wu, Y. (2002). The Impact of Macroeconomic and Financial
Variables on Market Risk: Evidence from International Equity Returns.
European Financial Management, 8, 421 47.
Public Broadcasting Service (PBS), www.pbs.org.
Phillip, P. & Perron, P. (1988). Testing for Unit Root in Time Series Regression.
Biometrica, 75, 599 607.
Phylaktis, K. & Ravazzolo, F. (2002). Measuring Financial and Economic
Integration with Equity Prices in Emerging Markets. Journal of International
Money and Finance, 21, 879 903.
Pownall, R. & Koedijk, K. (1999). Capturing Downside Risk in Financial Markets:
The Case of the Asian Crisis. Journal of International Money and Finance, 18,
853 70.
Razzak, W. (2007). In the Middle of the Heat: The GCC Countries between Rising
Oil Prices and Sliding Greenback. MPRA, Working Paper No. 6591.
226
Reyes, M. (1999). Size, Time Varying Beta and Conditional Heteroscedasticity in
UK Stock Returns. Review of Financial Economics, 8, 1 10.
Richardson, D. (2003). Economics of War with Iraq. Economics, Commerce and
Industrial Relations Group. Current Issues Brief, No. 20/2002 03.
Richman, V., Santos, M. & Barkoulas, J. (2005). Short- and Long-Term Effects of
the 9/11 Event: The International Evidence. International Journal of
Theoretical and Applied Finance, 8, 947 58.
Rigobon, R. (2003). On the Measurement of the International Propagation of Shocks:
Is the Transmission Stable? Journal of International Economics, 61, 261 83.
Rigobon, R. & Sack, B. (2005). The Effects of War Risks on the U.S Financial
Markets. Journal of Banking and Finance, 29, 1769 89.
Roll, R. (1977). A Critique of the Asset Pricing Theory s Tests, Part I: On Past and
Potential Testability of the Theory. Journal of Financial Economics, 4, 129
76.
Rouwenhorst, G. (1999). Local Return Factors and Turnover in Emerging Stock
Markets. The Journal of Finance, 54, 1439 64.
Saadi, T. & Williams, O. (2011). Global and Regional Spillovers to GCC Equity
Markets. IMF Working Paper Number 11/138.
Saez, L., Fratzscher, M. & Thimann, C. (2009). The Transmission of Emerging
Market Shocks to Global Equity Markets. Journal of Empirical Finance, 16, 2
17.
Saif, I. & DeBartolo, D. (2007) The Impact of Iraq War on Growth and Inflation in
Jordan. Centre for Strategic Studies, Jordan University. Online, available at:
http://www.jcss.org/uploadeconomic/118.pdf.
Salameh, M. (2008). The Oil Price Rise Factor in the Iraq War: A Macroeconomic
Assessment. The United States Association for Energy Economics (USAEE),
Working Paper No. 08 008.
Saleem, K. & Vaihekoski, M. (2008). Pricing of Global and Local Sources of Risk in
Russian Stock Market. Emerging Market Review, 9, 40 56.
Schneider, G. & Troeger, V. (2006). War and the World Economy: Stock Market
Reactions to International Conflicts. Journal of Conflict Resolution, 50, 623
48.
Schwert, W. (1989). Why does Stock Market Volatility Change over Time? Journal
of Finance, 44, 1115 53.
Schwert, W. & Seguin, P. (1990). Heteroscedasticity in Stock Returns. Journal of
Finance, 45, 1129 55.
Serra, A. (2000). Country and Industry Factors in Returns: Evidence from Emerging
227
Markets Stocks. Emerging Markets Review, 1, 127 51.
Shachmurove, Y. (2003). Financial Markets of the Middle East and North Africa:
The Past and Present. Penn Institute for Economic Research, University of
Pennsylvania, Working Paper No. 03 017.
Shamila, J. & Shambora, W. (2008). The World is Shrinking: Evidence for Stock
Market Convergence. Economics Bulletin, 7, 1 12.
Sharma, S. (2010). The Arab World amidst the Global Financial Crisis of 2008
2009. Contemporary Arab Affairs, 3, 38 52.
Sharpe, W. F. (1964). Capital Asset Prices: A Theory of Market Equilibrium under
Conditions of Risk. Journal of Finance, 19, 425 42.
Siklos, P. & Ng, P. (2001). Integration among Asia-Pacific and International Stock
Markets: Common Stochastic Trends and Regime Shifts. Pacific Economic
Review, 6, 89 110.
Solnik, B. (1974). An Equilibrium Model of the International Capital Market.
Journal of Economic Theory, 8, 500 24.
Soussa, F., Marchand, L. & Salters, R. (2008). Oil Boom Gives Gulf Cooperation
Council Sovereign s Room to Address Economic and Employment
Imbalances. Standard & Poor s Ratings. www.standard&poor s.com.
Soydemir, G. (2000). International Transmission Mechanism of Stock Market
Movements: Evidence from Emerging Equity Markets. Journal of Forecasting,
19, 149 76.
Timeline, http://www.xtimeline.com/timeline/1997-Asian-Financial-Crisis. Accessed
on September, 2010.
Tourani-Rad, A. Winson, B. & Choi, D. (2006). A Time-Varying Beta Approach to
Measuring New Zealand s Country Risk. Managerial Finance, 32, 257 69.
US Energy and Information Administration (EIA), http://www.eia.gov/.
Verma, R. & Soydemir, G. (2006). Modeling Country Risk in Latin America: A
Country Beta Approach. Global Finance Journal, 17, 192 213.
Voronkova, S. (2004). Equity Market Integration in Central European Emerging
Markets: A Cointegration Analysis with Shifting Regimes. International
Review of Financial Analysis, 13, 633 47.
Wang, Y. & Di Iorio, A. (2007). The Cross-Sectional Relationship between Stock
Returns and Domestic and Global Factors in the Chinese A-Share Market.
Review of Quantitative Finance and Accounting, 29, 181 203.
Wassal, K. (2005). Stock Market Growth: An Analysis of Cointegration and
Causality. Economic Issues, 10, 37 58.
228
Wdowinski, P. (2004). Determinants of Country Beta Risk in Poland. CESifo
Working Paper Series No. 1120.
Wells, C. (1994). Variable Betas on the Stockholm Exchange; 1971 1989. Applied
Financial Economics, 4, 75 92.
Wells, C. (1996). The Kalman Filter in Finance. Kluwer Academic Publishers.
Dorfrecht, Netherlands.
Williams, P. (2006). Projections for the Geopolitical Economy of Oil after War in
Iraq. Futures, 38, 1074 88.
Woertz, E. (2008). The Impact of the US Financial Crisis on GCC Countries. Report,
Gulf Research Center (GRC). Dubai, UAE.
Wolfers, J. & Zitzewitz, E. (2006). Using Markets to Inform Policy: The Case of the
Iraq War. Economia, 76, 225 50.
World Economic Forum (2007). A Global Risk Network Briefing. Middle East Risk.
Report.
Yang, J., Hsaio, C., Li, Q. & Wang, Z. (2006). The Emerging Market Crisis and
Stock Market Linkages: Further Evidence. Journal of Applied Econometrics,
21, 727 44.
Yao, J. & Gao, J. (2004). Computer Intensive Time-Varying Model Approach to The
Systematic Risk of Australian Industrial Stock returns. Australian Journal of
Management, 29, 121 46.
Yao, J., Gao, J. & Alles, L. (2005). Dynamic Investigation into the Predictability of
Australian Industrial Stock Returns: Using Financial and Economic
Information. Pacific-Basin Finance Journal, 13, 225 45.
Yesilada, B., Noordijk, P. & Nelson, H. (2004). Lessons Learned from the Financial
Crisis in Turkey: Implications for Economic and Political Stability. Paper
presented at the annual meeting of the American Political Science Association,
Hilton Chicago and the Palmer House Hilton, Chicago, IL.
Zeileis, A. & Kleiber, C. (2005). Validating Multiple Structural Change Models
Case Study. Journal of Applied Econometrics, 20, 685 90.
A
Zeileis, A., Kleiber, C., Kramer, W. & Hornik, K. (2003). Testing and Dating of
Structural Changes in Practice. Computational Statistics and Data Analysis, 44,
109 23.
Zhang, D., Dickinson, D. & Barassi, M. (2006). Structural breaks, Cointegration and
B Share Discount in Chinese Stock Market. International Conference of Policy
Modeling, Hong Kong.
Zhang, X. (2006). Specification Tests of International Asset Pricing Models. Journal
of International Money and Finance, 25, 275 307.
229
Zivot, E. & Andrew, D. (1992). Further Evidence on the Great Crash, the Oil Price
Shock, and the Unit Root Hypothesis. Journal of Business and Economic
Statistics, 10, 251 70.
230