financial vulnerability in emerging markets evidence from the middle east and north africa

8/13/2019 Financial Vulnerability in Emerging Markets Evidence From the Middle East and North Africa

1/28

Financial Vulnerability in Emerging Markets.Evidence from the Middle East and North Africa

Thomas Lagoarde-Segot 1

Brian M. Lucey 2

Abstract

The purpose of this paper is to test the idea that vulnerability to financial contagion increases as anemerging market reaches a certain threshold in size and integration to the world. Picking up a set of heterogeneous emerging markets located the Middle East and North Africa (MENA), we first conducta series of contagion test for each stock market during seven recent episodes of financial crises. Thesetests include the Fry & Baur (2006), Forbes & Rigobon (2002), Corsetti (2002) and Favero & Giavazzi(2002) approaches. Aggregating the results into a vulnerability index allows us to rank countriesaccording to their sensitivity to international financial crisis. Turkey appears the sample’s mostvulnerable market, followed by Israel, Jordan, Tunisia, Lebanon, Morocco and Egypt. Finally, we poolthe vulnerability indices along with stock market development and integration indicators into anordered logit model. Odds ratios and significance levels suggest that the size, liquidity and integrationof these emerging markets have some explanatory power in determining their vulnerability to financialcontagion.

JEL classification : G11;G12;G15Keywords : Contagion, Emerging Markets, Middle East and North Africa.

1 Corresponding Author: [email protected] . Institute For International Integration Studies, School of Business,

Trinity College, Dublin & CEFI, Universite Aix Marseille II.2 PhD supervisor. School of Business Studies and Institute for International Integration Studies, Trinity CollegeDublin.


2/28


3/28

country or to a group of countries, resulting in a common negative trend (Forbes & Rigobon, 1999).

Contagion thus occurs due to a shift in in cross-market linkages. This definition has the two

advantages of being restrictive and the most appropriate for statistical analysis (Bruner et.al, 2005).

Very close to this definition is the concept of financial vulnerability , which can be defined at the

individual country level as the probability of being affected by shift-contagion in the event of an

external financial turmoil (Dungey et.al, 2005). Whereas shift-contagion deals with country-to-country

crisis transmission within the framework of a specific crisis, vulnerability considers the broader

financial interactions as well as a longer time period. An underlying idea is that financial, economic

and institutional variables can endogenously determine an emerging market’s vulnerability.

However, these theoretical channels are still discussed (Sakho, 2003; Kodres&Pritsker, 2005), and are

reminiscent of the other definitions of contagion. For instance, some authors attribute vulnerability to

real economic integration. In this context, strong interdependences across economies can degenerate

fundamental spill-over into financial contagion when a country is hitten by a crisis (Dornbusch, Park

and Claessens, 2000). Other authors stress the importance of macro-economic destabilization, which

can trigger the crisis when contagion is channelled by macro-economic risk hedging

(Kodres&Pritsker, 2005). Regulatory weaknesses in corporate control and information asymmetries

may also lead to contagion by blurring the information flow to market participants and reinforcing

herding behaviours (Sakho, 2003).

As an alternative to these studies, our paper test for the simple idea that financial vulnerability

increases as an emerging market reaches a certain threshold in size and integration to the world. Our

intuition is that the spread of the crisis may also depend on the degree of development of the afflicted

market and its integration to the world (Bekaert et al. 2005; Biekpens & Collins, 2002). First, in thinly

traded markets, where capitalization and value traded are small, we might expect trading decisions are

to reflect local economic conditions rather than international fluctuations. In addition, the adjustment

of prices to international informations may also be shrinked by an insufficient level of market liquidity

affecting the ability of market participants to accomodate order flows. Second, it is reasonable to

expect that investors need to some extent to be involved in a market in order to trigger a crisis Large

and globally integrated markets are by definition more sensible to shocks and volatility transmissions


4/28

through the inclusion of a world beta into the CAPM equation. By contrast, thinly traded, segmented

stock markets may be immune to such variations, as their systemic risk primarily depends on domestic

factors. In confirmed, such an intuition would be of interest for policy makers. By highlighting the

presence of increasing contagion costs as a stock market develop and become integrated into global

finance, it would constitute a case for gradual, risk-averse financial integration strategies.

We pick up seven emerging markets located in the Middle East and North Africa (MENA) region:

Turkey, Israel, Jordan, Lebanon, Tunisia, Morocco and Egypt. These markets constitute an appropriate

sample for the purpose of our investigation. Due to economic reforms, they are rapidly expanding and

have on average overcome Latin American and Eastern European markets in terms of market

capitalization, value traded and listed firms (table 1). However, the MENA markets are very

heterogeneous of various sizes and maturity, from the largely capitalized stock markets of Turkey,

Israel and Jordan, to the more thinly traded markets of Morocco, Tunisia and Lebanon. These markets

hence provide an appropriate sample for investigating the relationship between contagion and stock

market development. An assessment of vulnerability to financial contagion in the MENA markets

might thus clarify the extent of risks for international investors willing to enter this economic area.

Previous investigations of international linkages of these stock markets have mostly focused on spill-

over sensitivity and long-run co-movements. For instance, Neaime (2002) considered a mix of MENA

and Gulf Cooperation Council countries over the 1990-2000 periods. He found that financial

integration of the MENA markets seemed to go along with a strong sensitivity to unidirectional shocks

flowing from the US and the UK. Erdal and Gundunz (2001) investigated the interdependence of the

Istanbul Stock Exchange with the G-7 equity markets and with the stock markets of Israel, Jordan,

Egypt and Morocco, before and after the Asian crisis. Based on Granger causality tests, they rejected

the hypothesis of significant linkages among the MENA markets. They also found one co-integrating

vector between the Istanbul Stock Exchange and the G-7 markets, but no lead-lag relationship.

Another similar study was carried by Gundunz and Omran (2001), in which the hypothesis of a

common stochastic trend between the markets of Turkey, Israel, Egypt, Morocco and Jordan was

rejected over the period 1997-2000. Girard, Omran and Zaher (2004) implemented a state-dependent

multivariate GARCH methodology and found, in contrast with previous studies, that the MENA


5/28


6/28

The preliminary step to the investigation of contagion is the accurate identification of the crisis

interval. For each crisis, we divide the dataset into a stable and a turmoil period. Our starting dates are

based on the literature, and the length of the turmoil is chosen to be one or two months depending on

crisis development. Following Rigobon(2001), we assume that the breakout of the East Asian crisis

can be identified with the dramatic increase of short term interest rates in Hong Kong on October 23,

1997. The dates for the Russian crisis and its Brazilian sequel are based on the results from

Rigobon(2001) and Baig and Goldfajn(2001). According to this timeline, the initial shock to the

Russian bond market took place on August 6, 1998. The stock market reacted one week later and the

turmoil persisted until the end of September. The Brazilian crisis, which was often associated with

contagion from the Russian crisis, lasted from the end of November 1998 to January 1999. Following

Mishkin and White (2002), the starting dates of the two American market crashes are taken from daily

newspaper. Terrorists acts in New York and Washington took place on September 11, 2001, and

WorldCom revealed its accounting fraud on June 25, 2002. Dates for the Turkish crisis were selected

following Alper(2001) and Yeldan(2002), and the duration of the Argentinean crisis is identified

following Serwa and Bohl(2004).

We used daily dollar stock market returns for Morocco, Egypt, Tunisia, Lebanon, Jordan, Turkey and

Israel, as well as for each of the crisis markets. We also use a MENA, a composite emerging market

and a world benchmark. Data are taken from the S&P/IFCG emerging markets database. For the US

market we used the MSCI database. The time series ranged from September 1997 to September 2002.

All series were smoothed using a two-day moving average filter in order to neutralize the possible

impact of different trading days. Turning to stock market development indicators, we selected factors

including the market capitalization to GDP ratio, the number of listed firms, value traded in dollars

and liquidity as measured by turnover ratio. These were taken from the World Bank WDI Database,

and averaged over the study period. Turning to measures of global integration, we observe the average

percentage change of the country S&P investable index. We also include the Akdogan index of global

financial integration for the MENA countries, as taken from Lagoarde-Segot & Lucey (2006). This


7/28

index is based on a market-capitalization adjusted ratio of the variance of the weight of an individual

market into the reference benchmark systematic risk. Details are given in annex.

2. Methodology

We adopt a three steps methodology. As a preliminary investigation, we ask whether the MENA

markets are subject to joint vulnerability to common exogenous shocks, using a fixed effect panel

approach. Second, we investigate financial vulnerability at a country level using a country

vulnerability index based on a battery of bi-variate tests for shift-contagion. Third, we analyze the

impact of stock market development and integration indicators on the latter based on an ordered logit

model.

2.1 Joint vulnerability

Baur & Fry (2006) developed a multivariate test of contagion based on a panel data model which

controls for common vulnerabilities through the inclusion of a world and emerging equity market

index. The framework is a basic regression model of the form:

t it emergingit globalit it i y ,,2,1, ε τ β τ β γ α ++++= (1)

Where t i y , is the return of country i at time t , and t global ,τ and t emerging ,τ are global and emerging

markets factors, respectively. The model contains a constant, iα , for each country return vector i y and

a fixed time effect t γ which is defined for a period a K days through time across all countries. The

fixed time effect is interpreted in comparison to a base period and capture contagion in this model. The

error terms are given by t i ,ε and are assumed to be independent and independently distributed with

zero mean and unit variance.


8/28

The model differentiates between common vulnerabilities and contagion through the relative

importance of the regressors compared to the fixed time effects. It is assumed that vulnerabilities exist

in both the benchmark and crisis period and capture the systematic relationship between the equity

markets of each country, emerging markets and the world. In this framework, the fixed time effect

captures time-varying joint positive and negative movements across markets that are unexplained by

the loading factors over the period of study. The idea is then that contagion occurs wherever these

fixed time effects reach a certain threshold, highlighting the fact that asset prices are determined by a

large unexplained common factor. The threshold is reached if the t-statistic of an estimate of the fixed

effect is significant at the 5% level. The advantage of this approach is that the model can

endogenously determine contagion and hence avoid the sample selection bias discussed in Pesaran and

Pick (2004). The panel model is multivariate, and therefore gives evidence of joint contagion through

an estimation of global interdependencies.

2.1.2 Bi-variate contagion

There is now a reasonably large body of empirical work testing for the existence of contagion during

financial crisis. The seminal methodology used to analyze simultaneously falling stock markets over

breakdown periods was to compare correlation coefficient with a benchmark (Longin&Solnik, 1995;

Karolyi & Stulz, 1996). The presence of heteroscedasticity in the studied markets poses another

problem to coefficient analysis, since heteroscedasticity is a typical feature of crisis periods, the latter

generally corresponding to an increase in volatility (Forbes&Rigobon, 2002). We also compute the

Forbes and Rigobon (2002) adjustment of heteroscedasticity in correlation coefficients; as well as the

Corsetti (2003) correction of the latter, based on the inclusion of a common factor variance effect.

However simple tests based on changes of coefficient can have low power as they are based on an

exogenous definition of the crisis period (Dungey & Zhumabekova, 2001). We deal with these

difficulties by implementing an outlier based structural model as in Favero and Giavazzi (2002). We

finally aggregate results from these approaches into a vulnerability index.

2.1.2.1 Adjusted correlation coefficient


9/28

In a seminal paper, Forbes&Rigobon (2002) pointed out that the traditional comparison of correlation

coefficient is biased due to heteroscedasticity in market returns during crisis periods. They

subsequently proposed a methodology to correct for that bias. Consider the basic conditional

correlation coefficient between country 1 and 2:

21

2,1

σ σ σ

= p(5)

An adjustment can be done using the following transformation:

( )[ ]2*

11 p

p p

−+=

δ

(6)

Where 111

11 −= lh

σ σ

δ measures the change in high period volatility against the low period volatility in

the crisis country. The null hypothesis of no contagion is then tested as: 0:0 =− lh p p H .

(b) The common factor model

However, this approach has been criticized by Corsetti et.al(2002) on the basis that it is built on

arbitrary and unrealistic restrictions on the variance of country-specific shocks. Whereas the

Forbes&Rigobon (2002) methodology identifies tranquil and crisis periods by different levels of asset

return volatility, a change in variance might actually be driven by an increase in the variance of a

common factor, which then causes unusual volatility in other markets. In this case, the event of a

significant change in the magnitude of co-movement between markets does not necessarily require a

rise in correlation between these markets; and contagion can be defined as the presence of co-

movements in significant excess from what could be expected from an unchanged transmission

mechanism. Accordingly, the methodology proposed by Corsetti et.al (2002) consists of testing for

structural breaks in the international transmission mechanism. The model first creates data-generating

process in country 1 and country 2, where country 2 is the country where the crisis occur:


10/28

++=++=

2222

1211

r

r

ε γ α ε γ α

f

f (7)

Where s'α are constants, 1γ and 2γ are country-specific factor loading, f is a common factor, iε and

jε are country-specific factors. Correlation coefficients are defined as:

( )

+++=

+++=

2

)()(

1)(

)(1

1

)(1

)|()(

1

1

21

22

22

1

21

1

21

22

22

1

21

1

f Var

Var

f Var

Var p

f Var

Var

C f Var

Var p

t

c

γ ε

γ ε

γ ε

γ ε

(8)

Where c p and t p are coefficients for the crisis and tranquil period, respectively. If the transmission

mechanism is left unchanged between the tranquil and crisis period, 1γ , 2γ ,

( )1ε Var and ( )21ε ε Cov will be constant and the correlation coefficient between asset returns becomes:

( )( ) ( )

21

22

22

2

2

221

11)11

(11

111

,,,

+−++++

+

++≡

λ λ λ

δ

δ λ λ

δ λ λ φ

C

C C

p

p p

(9)

Where( )

)(222

2 f Var

Var

γ

ε λ = and ( )

)|(

|

2

2

2 C f Var

C Var C

C

γ

ε λ = .

Testing the null hypothesis of interdependence versus contagion amounts to measuring whether C p is

significantly higher than φ , which represents the theoretical measure of interdependence:

φ ≤C p H :0


11/28

In implementing the correlation-based methodology, we draw on two test-statistics to measure the

significance of the difference between coefficients. Following Forbes & Rigobon (2002) we begin

with a test based on the Fisher transformation. However, this approach makes the assumption of

normality, and might suffer from a lack of robustness in the case of skewed stock market returns. In an

attempt to improve the finite sample properties of the statistic we therefore complement the analysis

with an exact t-test based on actual sample correlation coefficients (as suggested in Collins&Biekpe,

2003) 3:

( ) ( )221

1

4

y x

y x p p

nn p pt

−−−+−=

(10)

Where ( )4,05.0 21 −+ nnt .

(c) The structural modeling approach

Favero & Giavazzi (2002) have proposed a methodology allowing to endogenously defining contagion

by identifying many short lived crisis periods associated with extreme returns. The idea is to

implement a VAR to control for the interdependence between asset returns, and subsequently used the

heteroscedasticity and non-normalities of the residuals from that VAR to identify unexpected shocks

transmitted across countries, which are considered as contagion. The first step is to estimate a simple

VAR and to consider the distribution of the residuals. Crisis observations are then defined through a

set of dummies associated with extreme residuals for each country. Consider the following VAR

model:

t t t v z z += −1φ (2)

3 Corsetti et al. (2002) also suggest calculating the test based on threshold values derived from the variance

ratios. However, this framework requires that studied market display high correlation levels (>0.32) during thecrisis period, otherwise threshold values tend to infinity and the null hypothesis cannot be rejected at all. Resultsare available on request.


12/28

Where t z are pooled asset returns across the sample period, φ contains the N N × VAR parameters,

and t v are the reduced-form disturbance with zero means and constant covariance matrix with

variances given by 22ii

E σ ν =

. The dummy variables are then defined as:

>=otherwise

d it it k i:0

3:1 2,,,

σ ν (3)

Where we define one single dummy variable per observation. These dummy variables are then

included in the following structural model:

t t t t t t d d z z z ,1,1,22,1,1,11,11,11,22,1,1 ηγ γ θ α ++++= −

t t t t t t d d z z z ,2,2,22,1,1,11,21,22,11,2,2 ηγ γ θ α ++++= −

(4)

Where 1θ and 2θ are the parameters on own lags and t i ,η are the structural disturbances. In order to

correct for simultaneity bias, this model is implemented using an FIML variable estimator where

instruments are the dummy variables and each country’s own lagged returns. Finally, contagion from

country 1 to country 2 is tested by checking the significance of the shock in asset returns in the second

country on asset returns in the first country:

0: 2,10 =γ H

(d) The vulnerability index

The objective is to build an individual index reflecting each country’s sensitivity to financial contagion

as derived from our underlying tests. For each country, we first create one series of contagion

dummies per test, taking the value of 1 if contagion is found. Averaging the latter over the number of

crisis and methodologies is a first way to capture vulnerability to contagion. However, we need to

discriminate between countries where contagion was found the same number of times. Using the linear

transformation, we calculate the p-values for each test. We then add to the index a component

reflecting the inverted average p-value among all tests and crisis. This component can be interpreted as


13/28

an indicator of overall sensitivity to financial contagion. It is included to facilitate comparisons, but

should not substitute for the finding of contagion at the 5% level of significance. We therefore give the

p-values a smaller weight in the index. Weights were bootstrapped using 10000 random numbers,

contagion dummy weights being constrained to be greater than the p-values’ . We then calculated our

index for each of these weights, and selected the value corresponding to a 50% cumulative distribution

function.

The final index can be described as follows:

( )

C M

p

C M

CONTAGION INDEX

C

c

M

mi

C

c

M

mi

i .

1

.1 11 1

∑∑∑∑= == =

−+= β α

(2)

Where C is the number of crises, M the number of methodologies, iCONTAGION is the contagion

dummies for country i, and i p the test p-value. α and β are weights for each component of the index.

2.1.3 Results

We begin the analysis with an investigation of joint vulnerability to financial crisis. As shown in table

3, results from the fixed-effect panel regression suggest that the world index is significant in

explaining co-movements between the MENA markets. By contrast, the emerging market index is

insignificant. This reflects both the weak share of the MENA markets in emerging markets total

capitalization; and the fact that most economic interaction of these countries takes place with

developed countries rather than with each other (FEMISE, 2004).

INSERT TABLE 3 ABOUT HERE

Turning to the analysis of joint contagious shocks, the time series of the fixed time effect over the

whole sample period, including the seven investigated crisis is presented in figure 1. The first panel of

the figure presents coefficients estimates and the second panel presents the t-values associated with

critical values at the 5% significance level. Inspection of this figure shows the absence of joint


14/28

contagion over the period of study, which suggests that taken as a group, the MENA financial markets

are not sensitive to regional re-allocation of international portfolios in the event of an international

financial crisis. This finding is not surprising considering both the weak share of the region in

international portfolio investment (World Bank, 2004) and previous studies highlighting small co-

movements among the MENA markets (Lagoarde-Segot & Lucey, 2005, Girard, 2004). However,

does not inform on the potentially divergent reactions of individual stock markets to exogenous

shocks. The MENA markets being rather heterogeneous in size, liquidity and integration, and the

region being weakly integrated in financial and economic terms; we expect to find different outcomes

for the country level analysis.

INSERT FIGURE 1 ABOUT HERE

Our bi-variates analyses suggest contagion for every single MENA market in at least one out of the

seven crisis episodes. The coincidence of results using different approaches gives a stronger

robustness to the contagious hypothesis. As shown in table 4, the most significant evidence in favor of

contagion are found in the case of Israel during the Turkish crisis, Jordan during the WTC breakdown,

Tunisia during the Brazilian crisis, and Turkey during the Enron crisis.

Taking different methodologies altogether, there is suspicion for contagion for every single MENA

market in at least one out of the seven crisis episodes. However, results are contrasted among

countries. Israel and Turkey are the only two markets that we can suspect to have endured contagion

during the Asian crisis. Considering that they are the oldest, largest and most developed markets in the

MENA and that the Asian crisis occurred early in our timeline, this constitutes preliminary evidence

that contagion requires a high participation of international investors in the afflicted markets.


Moreover, Turkey seems to be the sample’s most vulnerable market, as it seems to have endured

contagion three times, during the Asian crisis, the WTC breakdown and the Enron crisis. It is followed


15/28

by Israel during the Asian and Turkish crisis, Jordan during the two American crisis, and Morocco and

Lebanon during the Turkish and Enron crisis. Egypt seems to be the region’s less vulnerable market,

as it seems to have been affected only during the Russian crisis. Two reasons may be put forward to

explain the absence of contagion during the Argentinean crisis. First, the nature of the crisis itself,

being primarily as currency crisis. Second, the relative small size of the MENA markets and the

weakness of trade linkages might also explain the absence of contagion, as it suggests that the sample

countries were immune both from balance of payment deficits and from the capital flights that were

implied by the restructuring of international portfolios.

Finally, another striking fact is that evidence of contagion in the MENA seems to increase over time:

looking at the number of contagion relationships per crisis, we yield two relationships during the 1997

Asian crisis, four during the 2001 Turkish crisis, and our results culminate with five relationships

during the 2002 Enron crisis. The decade of study being a period of significant developments in the

MENA markets, the increase in contagion relationships and the appearance of new recipient markets

as we move trough time suggest that along with improved resource allocation benefits, the risks of

contagion tend to increase as emerging markets reach higher levels liquidity and capitalization. These

findings are summarized through the vulnerability index.


As shown in table 5, we find that Turkey is the sample’s most vulnerable market. It is followed by

Israel and Jordan. They are followed by Tunisia, Morocco, Lebanon and Egypt. These results are

intuitively appealing as Turkey, Israel and Jordan are the three largest markets of the sample (see table

1). By contrast, Tunisia, Morocco, Lebanon and Egypt were on average smaller over the period of

study. We now rely on a specific model in order to make robust conclusions about the relationship

between vulnerability and stock market development.

2.1.3 Ordered logit regression


16/28

Taking the vulnerability index as dependent variables, we use an ordered logistic regression including

indicators of stock market development and financial integration as regressors. Results from the

regression are shown in table 6.


We first observe from the LR test statistic that our model appears reasonably well-specified at the 10%

level of significance. Second, our variables seems to have explanatory power in determining our

financial vulnerability index. Market capitalization, value traded, liquidity as measured by the turnover

ratio, and the two measures of integration are significant at the 5% level. The number of firms listed is

significant at the 10% level. This constitutes a validation of our intuition that an increase in market

size, liquidity and integration to the world increases the probability of being affected by contagion.

Odds ratios give further information. Other factors being held constant, an increase of one unit in the

logarithm of average market capitalization would make the vulnerability 1.24 times more likely to

move up one category than to diminish. These are similar for the number of listed firms (1.53) and

value traded (1.33). This suggests that emerging markets need to reach a critical size before

experiencing vulnerability to external crises: in thinly traded markets, price movements may tend to

solely reflect variations in local economic factors. This reasoning seems appealing when we consider

the impact of financial integration measured as the Akdogan score (1.13) and changes in the S&P

IFCI index (1.15). These results highlight that segmentation diminishes financial vulnerability, an

intuition that can be traced from a capital asset pricing model. Also, changes in the turnover ratios

seem to have the strongest impact on the financial vulnerability index (2.81). This seems to confirm

the intuition that stock market liquidity is a necessary condition for financial vulnerability. By

allowing market participants to quickly accommodate order flows, liquidity significantly increases the

probability of incorporating the dynamics of an external financial crisis into domestic price

movements.

4. Conclusion


17/28

The objective of this paper was to test the idea that vulnerability to financial contagion increases as an

emerging market reaches a certain threshold in size and integration to the world. Picking up a set of

heterogeneous emerging markets located the Middle East and North Africa (MENA), we first conduct

a series of contagion test for each stock market during seven recent episodes of financial crises. These

tests included the Fry & Baur (2006), Forbes & Rigobon (2002), Corsetti (2002) and Favero &

Giavazzi (2002) approaches. Aggregating the results into a vulnerability index allowed us to rank

countries according to their sensitivity to international financial crisis. Turkey appears the sample’s

most vulnerable market, followed by Israel, Jordan, Tunisia, Lebanon, Morocco and Egypt. Finally,

we pooled the vulnerability indices along with stock market development and integration indicators

into an ordered logit model. Odds ratios and significance levels suggested that the size, liquidity and

integration of these emerging markets have some explanatory power in determining their vulnerability

to financial contagion. At a policy level, this implies that stock market development policies should

acknowledge and adress the costs of increased financial vulnerability. Future research could try to

improve the vulnerability index by extending the number of underlying tests. We could also check

whether additional factors help explain the degree of financial vulnerability. Finally, we could also

extend our dataset by including more emerging markets in order to make international comparisons.

References


18/28

Alper, C.E., 2001. The liquidity crisis of 2000: what went wrong? Russian East Europe Finance Trade

37, 54–75.

Bae, K.-H., Karolyi, G.A., Stulz, R.M., 2003. A new approach to measuring financial contagion.

Reviw of Financial Studies 16, 717–763.

Baig, T., Goldfajn, I., 1999. Financial market contagion in the Asian crisis. IMF Staff Paper 46, 167–

195.

Baele, L. Volatility spillover effects in European equity markets. J. Financial Quantitative Analysis, in

press.

Baur, D., Fry, R. 2005 Endogenous contagion – A panel analysis . CAMA and EC Working Paper.

Bekaert, G., Harvey, C.R., 2002. Research in emerging markets finance: looking to the future.

Emerging Markets Review 3, 429–448.

Bekaert, G., Harvey, C.R., 2003. Emerging markets finance. Journal of Empirical Finance 10, 3–55.

Bekaert, G., Harvey, C.R., Ng, A., 2005. Market integration and contagion. Journal of Business 78,

39–70.

Bekaert, G., Harvey, C.,&Lumsdaine, R.,2002). Dating the integration of world equity markets.

Journal of Financial Economics, 65(2), 203–249.

Billio, M., Pelizzon, L., 2003. Contagion and Interdependence in Stock Markets: Have they been

misdiagnosed? Journal of Economics and Business 55, 405–426.

Bordo, M., Eichengreen, B., Klingebiel, D., & Martinez-Peria, M, 2001, Spring. Is the crisis problem

growing more severe? Economic Policy, 32, 51–75.

Boyer, B.H., Gibson, M.S., Loretan, M., 1999. Pitfalls in tests for changes in correlations.

International Finance Discussion Paper No. 597R, Federal Reserve Board, Washington, DC.

Calvo, G.A., Goldstein, M., Hochreiter, E. (Eds.), 1996. Private Capital Flows to Emerging Markets

after the Mexican Crisis. Institute for International Economics, Washington, DC.

Calvo, S., & Reinhart, C., 1996. Capital flows to Latin America: Is there evidence of contagion

effects? In G. Calvo, M. Goldstein, & E. Hochreiter (Eds.), Private capital flows to emerging markets

after the Mexican crisis. Washington, D.C.: Institute for International Economics.


19/28

Chan-Lau, J.A., Mathieson, D.J., Yao, J.Y., 2004. Extreme contagion in equity markets. IMF Staff

Paper 51, 386– 408.

Chen, G., Firth, M., Rui, O.M., 2002. Stock market linkages: evidence from Latin America. Journal of

Banking and Finance 26, 1113–1141.

Claessens, S., Forbes, K.J. (Eds.), 2001. International Financial Contagion: How it Spreads and How it

Can be Stopped. Kluwer Academic Publishers, Boston, MA.

Climent, F.J., Meneu, V., 2003. Has 1997 Asian crisis increased information flows between

international markets? International Review of Economics and Finance 12, 111–134.

Collins, D., Biekpe, N. 2003. Contagion: A fear for African equity markets? Journal of Economics and

Business Vol 55, pp. 285-297.

Corsetti, G., Pericoli, M., Sbracia, M. Some contagion, some interdependence, more pitfalls in tests of

financial contagion. Journal of International Money and Finance, in press.

Darvas, Z., Szapa´ry, G., 2000. Financial contagion in five small open economies: does the exchange

rate regime really matter? International Finance 3, 25–51.

Dungey, M., Zhumabekova, D., 2001. Testing for contagion using correlations: Some words of

caution. Pacific Basin Working Paper PB01-09, Federal Reserve Bank of San Francisco, San

Francisco, CA.

Edwards, S., 2000. Interest rates, contagion and capital controls. NBER Working Paper 7801,

Cambridge, MA.

Eun, C.S., Shim, S., 1989. International transmission of stock market movements. Journal of Financial

Quantitative Analysis 24, 241–256.

Favero, C.A., Giavazzi, F, 2002. Is the international propagation of shocks non linear? Evidence from

the ERM. Journal of Financial Economics, Vol. 57(1), 231-246.

FEMISE Annual Report on the Euro-Mediterranean Partneship, 2005, European Commission.

Forbes, K.J., Rigobon, R., 2001. Measuring contagion: conceptual and empirical issues. In: Claessens,

Stijn, Forbes, Kristin J. (Eds.), International Financial Contagion: How it Spreads and How it Can be

Stopped. Kluwer Academic Publishers, Dordrecht, pp. 43–66.


20/28

Forbes, K.J., Rigobon, R., 2002. No contagion, only interdependence: measuring stock market co-

movements. Journal of Finance 57, 2223–2261.

Patro, D.K., Wald, J.K.2002. Firm characteristics and the impact of emerging market liberalizations.

Working Paper, Reutgers Business School


21/28

Table 1 Comparative Indicators for emerging markets (2003)

Area Market Capitalisation /GDP Liquidity Listed Companies

Asia

India 46.80% 31.97% 5644

China 25.50% 71.08% 780Malaysia 156.00% 32.45% 902

Hong-Kong 456.10% 41.44% 1037

Korea 48.50% 156.20% 684

Philippines 29.20% 11.52% 236

Taiwan 132.53% 156.10% 674

Average 127.80% 71.50% 1422

Latin America

Argentina 27.00% 8.80% 110

Brazil 45.90% 29.35% 391

Mexico 19.50% 21.11% 237

Chile 11.97% 7.70% 240

Colombia 18.10% 5.65% 108

Peru 19.90% 10.00% 227

Average 23.70% 13.80% 218

MENA

Egypt 33.79% 15.61% 967

Morocco 29.32% 18.72% 52

Tunisia 10.03% 7.73% 45

Jordan 110.73% 23.78% 161

Lebanon 7.91% 8.72% 14

Israel 67.23% 27.74% 577Turkey 29.36% 143.55% 285

Average 41.20% 35.12% 300

Source: Federation Internationale des Bourses de Valeur, 2005Note: Market Capitalization/GDP is the market capitalization at the end of each year dividedby GDP for the year Liquidity corresponds to total value traded for the year divided by market capitalization Listed Companies are the number of listed companies at the end of the year


22/28

Table 2: Crisis Timeline

Table 3: Estimation Results of equation (1): regional and global vulnerabilities

Indep. Variable Coefficients t P>TWorld_benchmark 0,09 2,8* 0,005

Emerging_benchmark 0,00 0,11 0,911_cons 0,00 -0,40 0,69

R square 0,21 F(3,9783) 37,98

Note: (*) indicates significance at the 5% level.

Crisis name Crisis country Stable periods Crisis periodsAsian “Flu” Hong Kong 1997 :10:1–1997:10:22 1997:10 :23–1997 :11:22Russian “Virus” Russia 1998:6:6–1998:8:5 1998:8:6–1998:10:5Brazilian crisis Brazil 1998:11:1–1998:12:31 1999:1:1–1999:3:1Turkish collapse Turkey 2000:12:5–2001:2:14 2001:2:15–2001:3:13Terrorist acts and economicslowdown

U.S. 2001:6:27–2001:8:26 2001:9:14–2001:10:13

Argentinean crisis Argentina 2001 :10:13–2001:12 :12 2001:12 :27–2002 :2:26Accounting scandals U.S. 2002:4:25–2002:6:24 2002:6:25–2002:7:24


23/28

Figure 1: Estimates of the fixed time effects in equation (1), corresponding t-values and 95% critical

values

Fixed time effects

-0,02

-0,015

-0,01

-0,005

0

0,005

0,01

0,015

07/10/97 07/04/98 07/10/98 07/04/99 07/10/99 07/04/00 07/10/00 07/04/01 07/10/01 07/04/02 07/10/02

t-values

-2,5

-2

-1,5

-1

-0,5

0

0,5

1

1,5

2

2,5

07 /1 0/ 97 07 /0 4/ 98 07 /1 0/ 98 07 /0 4/ 99 07 /1 0/ 99 07 /0 4/ 00 07 /1 0/ 00 07 /0 4/ 01 07 /1 0/ 01 07 /0 4/ 02 07 /1 0/ 02


24/28

Table 4 Shift contagion analysis

Asian Russian Brazil Turkey WTC Argentina Enron

Egypt 0.180.780.72

0.027*0.210.79

0.990.290.24

0.490.720.54

0.600.570.99

0.990.630.99

0.430.570.77

Israel 0.220.51

0.049*

0.920.490.38

0.900.680.16

0.045*0.74

0.031**

0.500.2

0.35

0.990.070.99

0.990.2

0.99Morocco 0.99

0.210.28

0.990.260.23

0.990.120.33

0.048*0.250.12

0.360.960.99

0.790.770.99

0.046*0.960.99

Jordan 0.990.150.99

0.990.430.99

0.990.760.33

0.990.990.35

0.039*0.35

0.042*

0.990.570.61

0.130.35

0.018*Tunisia 0.99

0.850.80

0.990.070.95

0.025*0.001**

0.95

0.990.850.32

0.036*0.9

0.66

0.460.190.99

0.990.9

0.82Lebanon 0.13

0.340.97

0.900.110.99

0.990.480.99

0.340.00**

0.56

0.400.15

0.3431

0.130.380.99

0.003**0.150.32

Turkey 0.990.89

0.006**

0.990.170.99

0.960.230.99

---

0.100.15

0.011*

0.990.560.99

0.026*0.15

0.011*

Note : for each country and crisis, the first coefficient gives the t-statistic for the Forbes-Rigobon analysis. Thesecond row gives the p-value for the Favero-Giavazzi analysis. The third row gives the t-statistic for the Corsettianalysis.


25/28

Table 5 The vulnerability index

Country Average (1-p value) Average contagion dummy Vulnerability Index(CDF=0.5)

Turkey 0.53 0.22 0.42

Israel 0.49 0.17 0.38

Jordan 0.40 0.17 0.32

Tunisia 0.34 0.17 0.28

Lebanon 0.56 0.11 0.26

Morocco 0.45 0.11 0.22

Egypt 0.42 0.06 0.18

Note : The first column shows the first component of the index. The second column shoxs the second component of the index. The third column shows the selected value of the index with booststrapped weights corresponding to the50% cumulative distributive function.


26/28

Table 6 Ordered Logit Regression

Variables Oddsratios

Std. Err. z p value

CAP 1.22 0.133 2.02 0.043**

FIRMS 1.49 0.341 1.91 0.056*

VALUE 2.81 0.178 2.17 0.009**

TURNOVER 2.01 0.689 2.24 0.023**

INTEGRATION 1.13 0.074 2.01 0.044**

S&P INV 1.15 0.069 2.33 0.02**

Number of obs = 42

LR test p-value = 0.094

Note : The vulnerability index is the dependent variable. The firstcolumn lists the regressors. See section 2 for description of thedataset.


27/28

ANNEX 1 : The Akdogan Integration score

Following Akdogan (1996,1997) and Barari (2004), first consider the following international risk

decomposition model:

I gi

R R ε β α ++= (1)

Where R i is the rate of return on the ith country, R g is the global rate of return, b is the beta of the i

th

country with respect to the global index, and ε i is the error term. The variance of the ith country’s

portfolio can then be decomposed into:

)()()( 2 I gi Var RVar RVar ε β += (2)

i

i

i

g

i

i

VarRVar

VarR

VarR

VarRVarR ε β

+=

2 (3)


28/28

ii q p +=1 (4)

In equation (4), p i measures the country’s contribution to worldwide systemic risk and is the proposedmeasure of market integration. Results are available on request. See Akdogan (1996), Barari (2004)

and Lagoarde-Segot & Lucey (2006) for further details.

financial vulnerability in emerging markets evidence from the middle east and north africa

Documents