Financial contagion in the BRIC countries during the 2007 global financial crisis :

Evidence from Dynamic Conditional Correlation-EGARCH and nonlinear causality

approaches.

Abstract

We examine possible financial contagion versus only interdependence between the

United Stated stock market (ground-zero country) and the BRIC (Brazil, Russia, India
and China) stock markets during the 2007 subprime crisis. Three empirical approaches
have been employed : (1) the traditional adjusted correlation approach of Forbes and
Rigobon (2002), (2) the Dynamic Conditional Correlation (DCC)-EGARCH approach of
Engle (2002) and (3) the nonlinear causality approach of Hiemstra and Jones (1994).
We found strong evidence for the presence of contagion phenomena from the USA stock
market to the India stock market when we use Forbes and Rigobon approach and to the
Brazil stock market when we use the DCC Engle approach. Using nonlinear causality
method, we detect a contagion phenomenon for the Brazil and China stock markets. For
the Russia country there is no evidence for the presence of contagion phenomena using
all methods.

Keywords : BRIC countries, stock return volatility, subprime crises, co-movement,

correlation, DCC-EGARCH, nonlinear causality.

JEL classification : C22, G15.

1
1. Introduction

In the last few decades many emerging and developing countries have accelerated their

financial integration on international markets in order to, at least, benefited from an increasing

of economic growth and employment. Despite these benefits, many others effects of

integration remains ambiguous. For instance, financial market liberalization increases the

vulnerability of domestic’s economies to international factors and news, particularly to

reversals in international capital movements. Subsequently, large co-movements between

assets prices across international stock markets increase portfolio volatility for domestic

investment, see Gagnon and Karolyi (2006), and Karolyi and Stulz (1996). Several others

studies have showed that integration generally increases the interdependence between

economies and then transmit shocks across borders, see for example Bekaert and Harvey,

(1997), Kaminsky and Reinhart (2000), Longin and Solnik, (1995, 2001) and Loretan and

English, (2000).

From previous points, it appears that the measurement of cross-markets linkages and the

assessment of changes in their interdependencies before, during and after crises may be

crucial for decision makers such as portfolio managers, central bankers and regulatory

authorities. As a consequence, it is important to distinguish between interdependence and

contagion in stock markets in period of financial crises. While the first, interdependence term,

is defined as the relationship that exists between stock market returns on average over the

sample period, the second, contagion, is defined as a significant change in the transmission

mechanism between stock markets in crisis times.

In fact, since 1990, the word financial market has characterized by the presence of several

crises originated from one country and extended to a wide range of markets and countries in a

way that was hard to explain that phenomena on the basis of only fundamentals changes. For

example, the US subprime crisis of 2007 have been transmitted to several stock markets in

2
other part of world and have leads that economies towards a declined period. During this

period, the US financial system has suffered from an important recession caused by the

subprime crisis and which has been transmitted to many countries via different channels like

financial markets and trade.

Thus, testing between interdependence and contagion in financial markets remains one of the

most important debates in empirical finance. Many theoretical and empirical works have

examined this problem for different financial crises. Moreover, a multitude of statistical and

econometrics tools have been used. Empirical results are also very mixed (King and

Wadhwani 1990; Eichengreen et al. 1996; Forbes and Rigobon, 2002; Favero and Giavazzi,

2005; Syriopoulos, 2007; Gilmore et al., 2008; Morana and Beltratti, 2008). Until now, no

consensus has been reached about this question. Testing for interdependence and contagion

remains a challenging task for several reasons. First, it appears that results depend on the used

definition of the contagion concept. Second, traditional methods based on testing significant

increases in correlation across-markets are not convincing and suffer from many limitations,

see Calvo and Reinhart (1996) and Baig and Goldfajn (1998). Corsetti et al. (2005). Third,

recent empirical literature that use nonlinear models to investigate contagion phenomena

found also mixed results, see for instance Hamao et al. (1990), Edward and Susmel (2001)

and Wang et al., (2006).

In this paper, we contribute to the literature in this field by investigating the question of

interdependence versus contagion in the BRIC stock markets during the 2007 subprime crisis.

First of all, we use a K-states of Markov switching model in mean and variance to date

exactly the subprime crisis. Then, we propose to use three empirical methodologies to test for

contagion versus only interdependence in the BRIC’s countries during the Subprime crisis.

For instance, in addition to the widely used traditional correlation approach and its adjusted

version of Forbes and Rigobon (2002), we use the recent DCC-EGARCH approach of Engle

3
(2002) and the nonlinear causality approach of Hiemstra and Jones (1994). Empirically, we

apply these three methods to both tranquil and crisis periods. Then, for the cases of Forbes

and Rigobon (2002) and Engle (2002) approaches we test for significant increase in

correlation between the two sub-periods. For the case of nonlinear causality, we distinguish

between four possible cases two of them are for particular interests (the interdependence and

the contagion cases). Our empirical finding shows evidence on the presence of contagion

phenomena from the USA stock market to the India stock market when we use the two

empirical methods. Using the nonlinear causality method we found evidence of contagion for

the Brazil and China countries. For the Russia country there is no evidence for the presence of

contagion phenomena.

The rest of the paper is organized as follows. Section 2 discusses the two concepts of

contagion and interdependency. Section 3 presents the BRIC's countries and the Sub-prime

crises. Section 4 describes the data, the traditional correlation approach, the DCC-EGARCH

approach and the nonlinear causality approach. Section 5 discusses the empirical results and

test for the presence of contagion or only interdependence. Finally, section 6 concludes.

2. Contagion versus interdependence literature review

The empirical literature on contagious effects of financial crises continue to be the topic of

academic research because of its important consequences for the global economy in relation

to monetary policy, optimal asset allocation, risk measurement, capital adequacy, and asset

pricing. Modeling the comovements of stock market returns is, however, a challenging task.

Various empirical approaches have been used to investigate contagion versus interdependence

during financial crisis. These different approaches can be classified in the following

4
categories. the conventional measure of market interdependence, known as the Pearson

correlation coefficient, the adjusted correlation of Forbes and Rigobon, the dynamic

conditional correlation, the regime switching models, the copulas analysis…etc.

The first methodology used to test the financial contagion is the cross-market correlation

coefficients. This approach tests and compares the cross-market correlation during the pre-

identified crisis period relative to the tranquil period. The contagion phenomenon occurs

when the cross-market correlation during a crisis compared to a tranquil period increases.

Empirical results about the existence of contagion based on correlation approach are not

conclusive. For example, King and Wadhwani (1990) find a significant increase in the cross-

country correlation coefficients of stock returns during the 1987 U.S. market crash among

three markets of U.S., the U.K., and Japan. Similarly, Bertero and Majer (1990) and Lee and

Kim (1993) find evidence of significant increase in correlation and conclude to the presence

of contagion phenomena in their investigations of the 1987 U.S. stock market crash. Calvo

and Reinhart (1996) find that correlations increased across weekly equity and Brady bond

returns for emerging markets in Latin America during the turbulence period of the 1994

Mexican crisis. The contagion effect has been investigated during the 1997 Asian crisis by

Baig and Goldfajn (1999), Khan et al. (2005), and Khan and Park (2009). They found

evidence of increased cross-market correlations. For example, Baig and Goldfajn (1999) by

using the correlation analysis test the presence of contagion in the equity, currency and money

markets in emerging economies during the Asian financial crisis. Their result show that

correlations in currency and sovereign spreads increased significantly during the crisis period,

whereas equity market correlations offered mixed evidence.

5
However, testing for significance increase in correlation before and during crisis periods using

traditional correlation approach suffers from several limitations. First, there is a problem of

heteroskedasticity when high frequency data are used. More precisely, the estimated

correlation coefficients during the crisis period are in general upwardly biased, and hence a

test based on the biased correlation would imply spurious contagion. To resolve this issue,

Forbes and Rigobon (2002) suggest an adjustment for the correlation coefficient during the

turmoil period. Using adjusted correlation from heteroscedasticity, the authors found no

increase in correlation coefficients during the East Asian crisis, Mexican crisis, and 1987 U.S.

market crash among 29 nations including 9 in Southeast Asia, 4 in Central and South

America, 12 in OECD, and 4 other new nations. Instead, they found a continued high level of

correlation in more tranquil periods and thus concluded that these crises are not the result of

contagion but rather of interdependence. Moreover, Corsetti et al. (2005) stress that the

significant increase in adjusted correlation is not explained by the behavior of the common

factors and the country-specific factor. This implies the generation of new temporary channels

of shocks propagation, in addition to the permanent channels, that characterizes the

interdependence between economies. Second, there is a problem of omitted variables such as

fundamentals variables. Third, contagion must involve evidence of a dynamic increment in

the regressions, affecting at least in the second moments correlations and covariance.

In the empirical analysis of contagion the conventional econometric techniques including

cointegration, causality tests and univariate ARCH and GARCH models has been also used.

The empirical result shows strong evidence in favor of cross-market volatility spillover and in

particular from the crisis country to other economies (Hamao et al. 1990, Chakrabarti and

Roll 2002, Diebold and Yilmaz 2009). For example, Hamao, Masulis, and Ng (1990)

investigate the correlation between three markets volatilities during the 1987 US stock market

6
crisis. They apply the conditional variance estimated under the GARCH model and found that

the spillover effects from New York to London and Tokyo and from London to Tokyo were

observed among the stock markets in New York, London, and Tokyo. Additionally, Diebold

and Yilmaz (2009) find evidence of divergent behavior in the dynamics of return spillovers

vs. volatility spillovers from the early 1990s to the 2000s. Several empirical studies have used

causality test to investigate the interdependence and the contagion of financial crises. For

example, Gomez-Puig and Sosvilla-Rivero (2011) apply the Granger causality test for

European Monetary Union (EMU) during different period since. Their results show the

presence of contagion phenomenon around the first year of EMU in 1999, the introduction of

euro coins and banknotes in 2002, and the global financial crisis in the late-2000s.

Furthermore, they detect a contagion between EMU countries caused by the crises in

sovereign debt markets from 2009.

In the same vein, Gelos and Sahay (2001) using Granger causality tests studied contagion

effects in the economies of Central and Eastern Europe, Russia and the Baltic since 1993. The

results suggest that after the Russian crisis of 1998, the movements of the European emergent

markets were similar to the movements observed in many Asian and Latin American markets

during the Asian crisis. The shocks originating from the Russian shares market caused the

movements in the markets of the Czech Republic, Hungary and Poland in a Grangerian sense.

The authors rejected the hypothesis that there was contagion originating from the markets of

the Czech Republic, Asia and Russia, in the direction of the European financial markets. The

causality test has been used by Masih and Masih (1999) to investigate the contagion between

4 stock markets of Southeast Asia and 4 industrialized markets. The results show the

existence of contagion in Southeast Asia countries in particular the importance of the role

played by Hon Kong in this contagion. Similarly, Khalid and Kawai (2003) applied the

Granger causality test to identify the existence of the contagion phenomena or

7
interdependence during the Asian crisis for 9 Asian countries. Their results were not a support

for the contagion.

This first generation of analysis has been followed by other methodologies, such as the

dynamic conditional correlation, the regime switching models, the copulas analysis. This new

generation offers additional efficient tools in testing contagion and/or interdependence

between stock markets.

In the area of Markov Switching ARCH model (SWARCH) of Hamilton and Susmel (1994),

and copula with extreme value theory several empirical studies tested the evidence of

contagion during the financial crisis. For the case of SWARCH model the intensity of co-

movement between stock market varies under high and low volatility regimes. This model

resolves the problem encountered when employing GARCH models, which are highly

sensitive to regime changes. The issue in the GARCH models is that the results obtained

might not be consistent during periods of low/high volatility. By employing the SWARCH

model, Edwards and Susmel (2001) found evidence of volatility co-movement across Latin

American markets during the crisis in the 1990s, but no volatility dependence between Hong

Kong and Latin American markets using both univariate and multivariate techniques.

Moreover, Boyer et al. (2006) demonstrated that there is greater co-movement during high

volatility periods for numerous accessible and inaccessible stock indices using both regime

switching models and extreme value theory. Canarella and Pollard (2007) found that each

high volatility episode appears to be associated with either a local or an international financial

crisis by applying the SWARCH model to some Latin American countries. By using a

SWARCH-L model for four Latin American stock markets (Argentina, Brazil, Chile, and

Mexico) Diamandis (2008) founded the existence of multiple volatility regimes and a

significant increase in volatility during the Mexican, Brazilian, and Asian crises. Moreover, in

order to test the contagion effects between the U.S stock market and some MENA stock

8
markets Khallouli and Sandretto (2012) use the Markov-Switching EGARCH. The results of

their estimations show that the financial crisis in USA has been transmitted to MENA stock

markets. They interpret this situation as evidence of mean and volatility contagion.

Particularly, they have found mean and volatility contagion in the Bahrain and Egypt stock

markets. In the case of Morocco and Turkey the authors reveal a mean contagion, while the

contagion to Oman and Dubai is explained only by the US volatility.

Regarding the copulas method, Patton (2006) initiated the analysis of time varying copulas for

modeling asymmetric exchange rate dependence. This method has been applied in finance to

test the evidence of contagion effects (see, Cherubini 2004). Employing extreme value copula

functions, De Melo Mendes (2005) examine the dependence of returns for seven emerging

countries markets. His result show the existence of asymmetry in the joint co-exceedances for

the most 21 pairs of markets considered and a strong dependence in cross market tail during

bear market. Similarly, Caillaut and Guegan (2005) use the copulas approach to check the

dependence between markets of Thailand, Indonesia and Malaysia from July 1987 to

December 2002. The analysis of daily data shows a symmetric dependence for Thailand-

Malaysia pair but an asymmetric dependence for Indonesia-Thailand and Malaysia-Indonesia

pairs. Samitas and Tsakalos (2013) investigate the relationships between the Greek stock

market and seven European stock markets and use copula functions to measure financial

contagion. The results of their study provide support to the contagion phenomenon despite the

lower than expected impact. By employing a multivariate copula approach Aloui et al. (2011)

use daily return data from BRIC markets to study its links with the US market during the

period of the global financial crisis. They find that dependency on the U.S. is higher and more

persistent for Brazil–Russia than for China–India.

9
A recent empirical approach not have been widely used in the empirical literature is the

dynamic conditional correlation approach of Engle (2002). In the empirical literature, only

few studies have used the DCC-GARCH approach to investigate possible financial contagion

see for instance, Chiang et al. (2007), Kenourgios et al. (2007) and Yiu et al. (2010), Fidrmuc

and Korhonen (2010), Naoui et al. (2010), Celik (2012), Jones and Olson (2013), Antonakakis

et al. (2013), Papavassiliou (2014). The advantage of this method compared to the previous

estimation technics can be summarized in three points. First, DCC-GARCH model tackle

directly the Heteroscedasticity problem contrary to the Forbes and Rigobon (2002) approach,

as the DCC is estimated by the standardized residual of GARCH model. Second, the DCC

model allows the implementation of other explanatory variables in the estimated equation to

measure the common factor. Third, it is the multivariate form of GARCH model which can be

used to examine multiple asset returns without adding too many parameters (Chiang et al.

2007).

Chiang et al. (2007) checked the presence of contagion by using the DCC approach. The

contagion effects are examined for nine Asian stock markets over the period from 1990 to

2003. Their result show that the correlation between market increases during the financial

crisis. They explained this result as an evidence of contagion. By using the same

methodology, Kenourgios et al. (2007) and Yiu et al. (2010) confirm the presence of

contagion, among BRIC emerging markets and US and UK markets and from the US to 11

Asian markets during the 2007 crisis, respectively. Their empirical results show the absence

of contagion evidence between US and Asian markets during the period of the Asian crisis.

Fidrmuc and Korhonen (2010) applied the DCC method and examine the spread of the

subprime crisis to China and India. Their result show that the 2007-2008 crisis effects

business cycle and economic development in both countries.Naoui et al; (2010) used the DCC

analysis to test the contagion effects after the American subprime crisis. To this end they use

10
the stock indices of daily returns of six developed countries and ten emerging countries over

the period 2006-2010. They concluded that during the subprime crisis, contagion is strong

between the US and the developed and emerging countries. More specifically, they noted that

returns conditional correlations of the S&P 500 stock index and the five developed markets

(France, Germany, Italy, Netherlands, and United Kingdom) considerably increased during

the crisis period with values sometimes exceeding 80%. In the case of emerging markets, the

results show that conditional correlations vary according to three types of groups. The first

group, including Brazil, Mexico and Argentina, is characterized by a high dynamic

conditional correlation with the US market. The second group, composed of India, Malaysia

and Singapore, presents correlations variable in time and do not exceed 50%. The third group,

composed of China, Hong Kong, Korea and Tunisia, records weak dynamic conditional

correlations with the US market and seems unaffected by the subprime crisis.

Celik (2012) employed the DCC-MGARCH analysis to investigate the contagion effects. He

found the evidence of contagion across foreign exchange markets of several emerging and

developed countries during the financial crisis. By using the same methodology Jones and

Olson (2013) examine the time varying correlations among macroeconomic uncertainty,

inflation, and output. Their result shows a change in the sign of correlation between

macroeconomic uncertainty and inflation and an unchanged negative sign of correlation

between uncertainty and output. In recent study employing the DCC-MGARCH model,

Antonakakis et al. (2013) investigate the dynamic correlations of stock market returns,

implied volatility, and policy uncertainty. Their result shows that the correlations are time

dependent and sensitive to global recessions. By employing two different methods, the F-G

cross-market correlation coefficient and the DCC approach, Papavassiliou (2014) explores the

correlation intensity between Greek stocks and sovereign bonds during the Greek sovereign

11
debt crisis period. His result show an increase of the correlation between stock and sovereign

bond returns during the Greek debt crisis period and contagion has occurred.

3. BRIC Economies and Subprime crisis

3.1. BRIC countries in globalization process

BRIC countries represent actually a particular interest for financial investors and economists.

These economies are expected to record for the thirty next year a highly potential economic

growth, and become an economic and financial power. Recently, the BRIC have largely

contributed to the world GDP growth. According to various economists' projections, it is only

a matter of time before China becomes the biggest economy in the world - sometime between

2030 and 2050 seems the consensus. In fact, Goldman Sachs believes that by 2050 these will

be the most important economies, relegating the US to fifth place. By 2020, all of the BRIC

countries should be in the top 10 largest economies of the world. The undisputed

heavyweight, though, will be China, also the largest the creditor in the world.

The graph 1 below, describes the GDP growth of BRIC countries over the crises period from

2002 to 2008. For example, the GDP growth in China is equal to 10% per annum. Similarly,

this variable grew in average by 8% annually in India. But the crisis started in USA has

affected the economic performance of these four countries as a new large economic power in

the world. Thus, the crisis has affected Russian economy and the country see its growth

contracted by 8% in 2008, Brazilian – by 0,3%, losing more than 5 p. p since 2007, Chinese

output growth has lost over 5% since the beginning of the crisis, and India has shown almost

the same results as China.

Graphics 3 and 4 show the trend in the inflow of foreign direct investment in these countries

over the same period. Due to various reasons this trend is reversed. In 2008, Russia lost 12

12
billion USD of FDI, while Brazil watched 20 billion USD make a quick exit from their

economy. The corresponding figure for India was a little under 1 billion USD.In these

countries the rapid economic growth is in part explained by a steady rise of FDI. Brazil,

Russia, India and China have seen a steady rise in FDI in a wide range of sectors. The onset of

the global crisis meant the growth rate of FDI plummeted. According to the estimations of the

World Bank, Russia, in particular, will be able to reach the pre-crisis level only in 2014.

The period of crisis has also affected the global trade of BRIC countries. The recession of the

trade as a whole has damaged the economies of export-oriented China and Brazil. For the

countries domestic-demand oriented, like India, the global recession did not hit it so much.

The first countries, China, Russia and Brazil taken a decision to promote exports of goods and

commodities and increases its share, other countries the share of export in GDP was quite

low. The crisis in this case will contract the global economy and consequently the economies

of China, Russia and Brazil. India, on the other hand, weathered the crisis. Of course, as

already mentioned, the contraction of the global economy was not the only factor, but one of

the most significant ones.

Tarzi (2000, 2005) studied the flow of foreign portfolio equity investments and foreign direct

investment to emerging markets between 1986 and 1995 and reported that stock market

capitalization in emerging countries grew from $171 billion to 1.9 trillion and the market

share held in capitalization increased from 4% to 11%, mostly attributed to the BRICs.

Apart from their growth characteristics, the BRIC grouping was not based on economic

similarities. However, for international business and trade purposes, the four countries are

vastly different. India and China are peasant economies with relatively closed, state-

controlled, regulated capital markets; Brazil and Russia are primarily natural resource– based

13
economies that are open to foreign trade and financial flows, and have a mixture of state and

private sector control of capital markets. The first subgroup (China and India) has guided its

exchange rate (more in China than in India), while the second subgroup has more flexible

exchange rates. India and China practice development strategies based on domestic

industrialization (manufacturing and services) for export, while Brazil and Russia follow

export strategies in directing productive structures that are guided by international

comparative advantage. While this latter subgroup has experienced exchange rate and

financial crises that were usually accompanied by high inflation, the former subgroup has

not.

3.2. Subprime crises

The last episode of turbulence triggered by the subprime mortgage crises in the United States

could have affected stock volatility. This crisis transmitted rapidly to other financial market is

due to asymmetric information resulting from the complexity of the structured mortgage

products and, subsequently, as a result of a more widespread reprising of risk which may have

taken the form of a decrease in global investors risk appetite (see Gonzalez-Hermosillo

(2008)). These events resulted in the most by a collapse of the banking industry, stock market

crashes and a large decrease in liquidity on the credit market. In the early stages of the crisis,

the securities backed with subprime mortgages held by many financial institutions rapidly lost

most of their market value because of a dramatic rise in these mortgages’ delinquencies and

foreclosures in the United States. This led to the reorganizations, liquidations, and

government bailouts of major U.S. financial institutions (e.g., Bear Stearns, Lehman Brothers,

and the American International Group) because their capital largely vanished.

14
The major feature of the U.S. subprime crisis is its rapidly propagation, spilling over into not

only other sectors of the economy but also other countries. This propagation of crisis takes the

form of a dramatic decline in the value of equities markets and commodities worldwide. The

successive failures of bank in US, such as Lehman Bankruptcy, triggered this transmission of

such impact to other markets.

Some research confirms this observation like the increase of volatility of Turkey price indice

(ISE-100) after the bankruptcy of Lehman Brothers, see Celikkol et al. (2010). Longstaff

(2010) find that there was financial contagion spillover across to other financial markets as

the Subprime crisis developed. Ramlal (2010) find that volatility clustering had increase and

interpret this result by the existence of transmission effect of subprime crisis to other

emerging countries. Moreover, he shows that leverage effects are higher in crises period

compare to pre-crisis period in most of stock markets.

4. Empirical methodology

In this paper, we employ three different econometrics and statistical approaches to examine

the existence of stock market contagion during the subprime crisis from the U.S stock market

to the four BRICs countries. In contrast to other previous study and to date exactly the break

date, we use a K-states Markov switching model with both changes in mean and in variance.

In addition, as a starting point of our empirical approach, we use the adjusted correlation

approach proposed by Forbes and Rigobon (2002, 2003), the multivariate DCC-EGARCH

model as in Chiang et al. (2007) and Syllignakis and Kouretas (2001) and the for nonlinear

causality between the U.S stock market and each BRIC stock market. For all three

approaches, we investigate the behavior of the statistic of interest, the adjusted correlation or

the nonlinear causality test, for both tranquil and crisis periods.

15
The rest of this section is organized as follow. First, in subsection 4.1., we present the Markov

switching model with two states in variance. Second, in subsection 4.2., we present the

adjusted correlation approach as developed by Forbes and Rigobon (2002, 2003). Then, in

subsection 4.3., we describe the DCC-GARCH approach. Finally, in subsection 4.4. we

present how we adapt the nonlinear causality approach proposed by Hiemstra and Jones

(1994) to test for contagion versus interdependence.

Insert Figures 1 to 8 Here

4.1. K-states Markov Switching model

To date the subprime crisis we use a K states Markov switching model with changes in means

and variances.

The model is given by,

r t =α S + σ S ε t
t t

Where ε t N ( 0 , 1 ).

Where r t is the U.S stock market return and α 1, α 2, α 3, …, α m are the estimated coefficients of

the mean equation. σ S is the standard deviation which depend on the state of the stock
t

markets. We estimate here all Markov switching models until the loglikelihood ratio test show

that the alternative model is rejected.

The mean and variance equations are defined by,

16
Mean-equation :

α S =α 1 S1 t + α 2 S 2t +…+ α m S mt
t

Variance-equation :
2 2 2
σ S =σ 1 S 1t +σ 2 S 2 t +…+ ¿ σ 2m S mt
t

Skt =1 if St =k , and Skt =0 otherwise; k =1 ,2 , 3 ,4,…,m

The of regime is governed by a first order three states Markov chain given by,

k
pij =P [ S t = j∨S t−1=i ] for i , j=1 ,2 , 3 , … , k and ∑ pij =1
j=1

In addition, we suppose that σ 21< …<σ 2m. For example for the case of three states Markov

switching model, σ 1 corresponds to the low volatility state, σ 2 to the medium volatility state

and finally σ 3 to the high volatility state.

4.2. Correlation analysis

Forbes and Rigobon (2002) and Rigobon (2003) define shift contagion by the significant rise

in cross-market interdependencies. In order to explain the phenomenon of contagion, this

paper builds from, Forbes and Rigobon (2002) model given by,

US
r ¿ =μ+ β r t +ε t

(1)

17
Where r ¿ is the returns of country i=( , Brazil, Russia, India, China ), r US
t is the returns ground

zero country which is the U.S in our study. μ and β are the coefficients to estimate of the

model. ε t is the error term.

The adjusted correlation coefficient proposed by Forbes and Rigobon (2002) is given by,

ρ
ρ¿ =
√1+ δ ( 1−ρ ) 2

(2)

Where ρ is the unadjusted correlation coefficient supposed to vary with period of low

volatility (before crisis period) and period of medium-high volatility (during crisis period).

This correlation coefficient is given by,

[ ]
−1/ 2
cov (r ¿ , r US
t ) Var ( r ¿ )
ρ= = 1+ for i=( , Brazil, Russia, India, China )
√ Var ( r ) Var (r
¿
US
t )
2
β Var (r t )
US

(3)

And δ is a measure of the relative increase in the observed variances of the stock

US crisis
US Var ( r t )
market returns variable r . The quantity δ is defined by δ =
t US tranquil
−1, where
Var (r )t

US crisis
Var ( r t ) US tranquil
and Var (r t ) are the variances, during the crisis and the tranquil

periods, respectively.

18
To test for the presence of contagion between the country source of crises, the USA here, and

each of the BRIC countries, we use the student test based on the differences between

correlation coefficients.

Hence, a statistical analysis of contagion versus interdependence can be performed using

the two hypotheses:

{ H 0 : ρ¿tranquil= ρ¿crisis ( interdependence )

H 1 : ρ¿tranquil ≠ ρ¿crisis ( pure contagion )

¿
Where, ρtranquil is the correlation coefficient between returns ( r ¿) of country i and returns (

US
r t ) during the tranquil period.
¿
ρcrisis is the correlation coefficient between returns ( r ¿) of country i and returns (r US
t )

during the crisis period.

To test for significance increase in adjusted correlations, we use a Student test as suggested by

Collins and Biekpe (2002). This test is given by,

t=( ρ crisis −ρtranquil )

¿ ¿

√ ntranquil +n crisis−2
1−( ρcrisis − ρtranquil )
¿ ¿

This t statistic follows under the null hypothesis of interdependence a student

distribution with ( n tranquil +ncrisis−2 ) degrees of freedom, where ntranquil and n crisis are the

number of observations under the tranquil and crisis periods, respectively.

19
 ¿
The decision rule is as follows. If there is no a significant difference between ρtranquil and
¿
ρcrisis , then there is no evidence for the presence of contagion. Conversely, if an opposite
¿ ¿
result is obtained when testing for the significant difference between ρtranquil and ρcrisis ,

then we conclude that there is a pure contagion between the country origin of crises

(USA here) and the corresponding BRIC country. This adjusted correlation test will be

applied for both the original series as in Forbes and Rigobon (2002) and also for the

estimated correlations using the DCC-EGARCH model which will presented in the

following section.

4.3. The DCC-EGARCH model

The multivariate GARCH class of models have been widely used in the recent empirical

finance literature for modeling stock markets comovement and testing for possible stock

markets contagion, (Chiang et al., 2007; Syllignakis and Kouretas, 2001; Lahiani and

Nguyen, 2013; and Hemche et al., 2015). Since the work of Bollerslev (1990) which

introduce the Constant Conditional Correlation (CCC)-GARCH models, different other

multivariate extension have been developed such as the BEKK-GARCH models of Engle

and Kroner (1995), the Dynamic Conditional Correlation (DCC-GARCH models of

Engle (2002), the Asymmetric DCC models of Cappiollo et al. (2006) and the dynamic

equicorrelation (DECO) models of Engle and Kelly (2012) among many others.

The DCC and ADCC models present many advantages compared to the other models.

For instance, compared to the CCC-GARCH models which suppose constant correlation

over time and to the full vec model and the BEKK model that are time consuming as the

number of parameters to estimate increases, the DCC-GARCH model overcomes these

problems and presents some others advantages. First, as the DCC-GARCH model use

standardized residuals to estimate correlation coefficients, then it overcomes the main

20
drawback of the standard correlation approach (heteroscedasticity). Second, the DCC-

GARCH model provides a time series of estimated correlation coefficients which allow

for testing for significant increases in the correlation coefficient before and during the

subprime crisis (Lahiani and Nguyen, 2013). Third, the DCC multivariate model allow

to include multiple stock markets returns without adding too many parameters (see

Engle, 2002; and Chiang et al., 2007). This will allow to estimate several pair-wise

correlations coefficients series.

In contrast to previous empirical study, in this part we use the DCC-EGARCH model.

This model have proven superiority to the DCC-GJR and DCC-GARCH models.

Moreover, we use also Bai and Perron (1998, 2003) to test for possible breaks in the

estimated conditional correlation series.

We extend the DCC-GARCH framework of Chiang et al. (2007) by including an

exponential part in the variance equation to cope with the evidence of asymmetric

volatility reaction to positive and negative return innovations pointed out by many

authors (Christie, 1982; and Nelson. 1991). The estimation procedure of this model is

performed in two-stage as in the standard DCC-GARCH of Engle (2002) where in the

first stage an univariate volatility models are fitted for each stock market returns. Then,

in the second stage, the standard residuals from the first stage are used to estimate the

DCC equation.

The return specification used in this paper is the same as in Chiang et al. (2007),

US
r t =μ+φ 1 r t −1+ φ2 r t−1 +ε t

21
Where r t =(r USA ,t ,r BRA ,t ,r RUS ,t , r IND ,t , r CHI ,t )' where USA, BRA, RUS, IND and CHI indices

refer to the U.S, Brazil, Russia, India and China countries.

ε t=(ε USA ,t , ε BR A ,t , ε RUS , t , ε IND, t , ε CHI ,t ) is the vector of the errors terms, ε t= √ H t z t .
'

z t N (0 , 1)

Where H t is a ( k × k ) matrix, where k =5 is the number of stocks markets returns. This

H t is decomposed as,

H t =Dt Rt Dt

Where Dt is a ( k × k ) diagonal matrix of time-varying standard deviations and Rt is ( k × k )

time-varying correlation matrix.

In the DCC model, the variances are modeled as univariate GARCH models, whereas the

covariances are modeled as nonlinear functions of the variances.

hij ,t =ρij ,t √ h ii, t h jj ,t

The DCC GARCH model proposed by Engle (2002) defines the condition correlation as

follows
−1/ 2 −1/ 2
Rt =diag (Qt ) Qt diag (Qt )

diag(Qt )−1/ 2=Diag

(√ 1
q11 ,t
,…,
1
√ q kk ,t )
The evolution of the correlation in the DCC model is given by :

22
 '
Qt =( 1−α−β ) Q+α u t−1 u t−1 + β Qt −1,

ε i ,t
where ut = . Qt =( q ij, t ) is the ( k × k ) time varying covariance matrix of ut and
√ hii, t
Q=E [ ut u't ] is an ( k × k ) unconditional variance matrix of ut .

The correlation matrix Rt have one in the diagonal and a values equal to ρij ,t =q ij, t / √ q ii, t q jj, t

for i , j=1 ,2 , 3 , … , k and i≠ j.

in this paper we consider that hii ,t evolves according to a univariate EGARCH model of the

form

log ( h ii, t )=ωi + β log ( hii ,t −1 ) + ( 1+ αL ) g ( z ii ,t −1 )

Where,

√
g ( z ii, t ) =θ1 z ii ,t +θ 2 [| zii ,t|−E|z ii ,t|], for our case E|z ii ,t| is equal to 2 .
π

For more details about the estimation procedure of the DCC-EGARCH model we refer

to Engle (2002). Once estimated, the DCC model yields an N x N matrix of pairwise

correlations at each time t. This rich conditional correlation parameterization will allow to

study the contagion versus interdependence of one market with each other market.

To test for significance increase in the correlation coefficients estimated before and

during the Subprime crisis we employ the Adjusted correlation coefficients of Forbers

and Rigobon (2002,2003) presented above.

4.4. Nonlinear causality tests

23
The third method considered in this paper to test for the presence of contagion or

interdependence phenomena between the USA and the BRIC countries is nonlinear

causality tests. This method is proposed by Baek and Brock (1992) which offer a

nonparametric statistical method to detect nonlinear causal relations. This method basically

relies on the assumption that the variables are mutually independent and identically

distributed. However, this assumption seems to be quite restrictive as it eliminates the time

dependence of variables and does not consider the nature and range of the dependence.

Hiemstra and Jones (1994) modify the Baek and Brock (1992)'s test to allow the testing

variables to exhibit short-term temporal dependence. By defining the m-length lead vector of

Y t noted Y mt , and the Ly-length and le-length lag vectors of Y t and X t , respectively, by Y tL−L y

Le
and X t −L , we obtain the following representations (Hiemstra and Jones, 1994),
e

m
Y t =( Y t , Y t +1 , … , Y t +m−1 ) , m=1 ,2 , … , t=1 ,2 , …

Ly
Y t −L =( Y t− L ,Y t −L +1 ,… , Y t −1) , L y =1 , 2 , … ,t=L y +1 , L y +2 ,…
y y y

Le
X t −L =( X t− L , X t− L +1 , … , X t −1) , Le =1 , 2, … , t=Le + 1, Le + 2 ,…
e e e

The definition of nonlinear Granger noncausality is then given by

Pr (‖Y t −Y s ‖<ϵ ‖Y t −L −Y s−L ‖<ϵ ‖X t− L −X s−L ‖<ϵ )=Pr (‖Y t −Y s ‖<ϵ ‖Y t−L −Y s− L ‖< ϵ ) ,
m m Ly Ly Le Le m m Ly Ly
y y e e y y

(1)

where Pr{.} is probability and ∥∥ is the maximum norm. If Eq. (1) holds for given values of

m, Ly and Le ≥ 1 and for ϵ > 0, then { X t } does not strictly Granger cause { Y t }. the Hiemstra-

Jones test consists of choosing a value of ϵ whose typical values are between 0.5 and 1.5 after

normalizing the series to obtain unit variance, and to test subsequently Eq. (1) by estimating

the conditional probabilities as ratios of unconditional probabilities.

24
Hiemstra and Jones (1994) show that under Granger noncausality the null hypothesis

formulated by Eq. (1), the following statistic follows an asymptotic normal distribution as:

C 1 ( m+ L y , Le , ϵ , n ) C 3 ( m+ L y , ϵ , n )
√ n( −
2
) AN (0 , σ ( m , L y , Le , ϵ ) )
C 2 ( m+ L y , ϵ , n ) C 4 ( Ly , ϵ , n)

(2)

where n = T +1 − m − max(Ly, le), C1(m + Ly, Le, ϵ, n), C2(m + Ly, ϵ, n), C3(m + Ly, ϵ, n),

and C4(Ly, ϵ, n) are correlation-integral estimators of the point probabilities corresponding to

the left hand side and right hand side of Eq. (1). The test statistic in Eq. (2) is applied to the

estimated residual series from the bivariate VAR model. The VAR models remove any linear

predictive power and therefore the remaining incremental predictive power of the one series

for another is an indication of the nonlinear predictive power.

The hypotheses of interest are,

H0 : there is no Granger causality from USA to country i.

H1 : presence of Granger causality from USA to country i.

Where i takes respectively the Brazil, Russia, India and China countries. To test for the

presence of interdependence or contagion we should compare between the nonlinear test

results between the tranquil period and the crises period. Four cases can arise,

Table (1) : Decision table

Tranquil period Crises period Decision

Case 1 Presence of nonlinear causality Presence of nonlinear causality Interdependence

25
Case 2 Presence of nonlinear causality Absence of nonlinear causality -

Case3 Absence of nonlinear causality Presence of nonlinear causality Contagion

Case4 Absence nonlinear causality Absence of nonlinear causality No dependence

In practice, only the first, third and fourth cases are of interest and can hold. The second case

is not of particular interest and rarely hold in practice. This can be due to the low probability

of occurrence of nonlinear causality in tranquil periods and its absence in crises periods.

The two cases of interest are the first and the third cases. The first case correspondents to the

case of a simultaneous presence of nonlinear causality in the tranquil and crises period (H 1

holds for the two sub-periods). This case is characterized by a significant correlation between

stocks markets before and during the subprime crises. In general, the correlation increases in

crises period compared to tranquil periods but the increases is not significantly different from

zero. This result can be interpreted as presence of interdependence. The third case

correspondents to the case where the tranquil period is characterized by the absence of

nonlinear causality (H0 holds) and the presence of nonlinear causality in the crises period (H 1

holds). This third case is considered as the contagion case.

Finally, the fourth case is characterized by the absence of a simultaneous nonlinear causality.

This case correspondents to the case of absence of dependence between the USA and the

BRIC stocks markets.

5. Empirical results

5.1. Descriptive analysis

The data used in this paper consists on the daily indices denominated in US dollars, for four

BRIC’s countries named the Brazil, China, India, Russia, and USA countries. The sample

26
covers the period from February 10, 2005 through April 14, 2009. The data were taken from

the Morgan Stanley Capital International data set. We limit our sample to this period in order

to account for only the period of the subprime crisis and the period after the subprime crisis.

An important step before testing for the presence of significance increase in correlations

between the tranquil and crises periods is to date exactly the tranquil and crises periods.

A misspecification on determining these periods lead to a bias in the results. To this end, we

use a two states Markov switching specification with only changes in variance where the high

volatility state corresponds to the crisis period and low volatility state corresponds to the

tranquil period. According to the selected specification, the tranquil period range from

February 10, 2005 to July 13, 2007 and the crisis period is from July 14, 2007 to 14 April,

2009. The break point that separates between the tranquil period and the crises period is

obtained using the smoothed probabilities of the Markov switching specification.1

Descriptive statistics and unit roots tests results are reported in Table (1) and (2). Table (1)

presents information on the mean, standard deviation, skewness coefficient, Kurtosis

coefficient, the Jarque–Bera Normality test (JB), and Ljung–Box test (LB) for the level and

squared series for the tranquil, crises and total periods.

Table (1) shows also that all series are non-normally distributed, as suggested by the two

statistics, skewness and Kurtosis, and summarized by the JB test results. This behavior is

largely observed on financial time series. Moreover, the high value of the kurtosis coefficient

is also typical of high frequency financial time series, and is behind the rejection of normality.

In addition, the Ljung–Box LB statistics for the level and squared series suggest significant

autocorrelation except for the china return series in the level. The high dependence in the

squared returns series indicates the presence of ARCH effects.

1
Results of the two states Markov switching specification can be obtained upon request from the authors.

27
Comparing the descriptive statistics of the three periods, we remark that the mean return is

positive during the tranquil period and negative during the crises period. Moreover, the

variance is more pronounced in the crises period compared to the tranquil and total periods.

This is expected because during crises periods investors generally sell stocks in order to

reduce losses, then price fall and returns on these stocks markets will be negative. Crises

period is also characterized by a higher level of variance due to the instability that

characterize this period.

To test the stationary of the stock markets indexes, we have carried out the ADF, PP and

KPSS unit roots tests. Empirical results for all the BRIC countries are reported in table (2)

below. In our cases neither the trend nor the intercept are significant. Following this table, the

stock markets indices series are not stationary in level but they are in first difference.

Insert Tables 1 and 2 here

Markov switching results

The results of the estimation of the linear, two states, three states and four states Markov

switching models are reported in Table 3 below. The likelihood ratio test is used to select the

more appropriate model. At this level it is very important to mention that this test does not has

the usual khi-squared distribution. To overcome this problem, we adopt the Davies (1977)

bound and the Garcia (1998) critical values in addition to the usual information criteria show

that the 3 state Markov Switching specification is more appropriate than the linear, 2 states

and also the 4 states specifications. This results is confirmed by the AIC and SC information

criteria’s which indicate that the 3 states specification describe better the evolution of the US

stock market evolution.

28
 Insert Table 3 here

Probability smoothing results show that low volatility regime last two years (509 days), the

medium volatility regime last for 419 days and the high volatility regime for 161 days.

As in this paper we consider two periods to investigate significant increase or presence of

nonlinear causality between the tranquil and crisis period, thus we consider that the low

volatility period corresponds to the period before crisis and the two others periods (medium

and high volatility) correspond to the period of subprime crisis.

Based on this result, the probabilities smoothing show that the date of beginning of the

subprime crisis corresponds to 12/07/2007. This date will be used all along the paper when

testing of significance increase of the adjusted correlations before and during the subprime

crisis.

Insert Figure 9 here

5.2. Correlation approach results

In order to test for significant changes in correlation coefficients between the tranquil and

turmoil periods we follow the same line as in Collins and Biekpe (2003), Corsetti et al. (2005)

and Forbes and Rigobon (2002). As suggested by Forbes and Rigobon (2002) the estimation

of cross-market correlation coefficients is biased because of heteroskedasticity in market

returns. In other words, correlation coefficients tend to increase in crisis period due to the

increase in volatility. To take into account heteroskedasticity, we adjust correlation

coefficients as proposed by Forbes and Rigobon (2002).

29
It is evident that during the two periods there is co-movement between all stock markets of

BRIC countries and the US market. The correlation shows that the link between BRIC

countries market with US market has strengthened in the crisis period since July 13, 2007 as

compared with the earlier period (February 2005-July 2007). The most important result is that

the China market has an edge over the major other market such as Brazil and Russia in terms

of the sharp increase in return correlation between the two periods, 2005-2007 and 2007-

2009. Illustratively, it is evident that the increase in correlation between stock market in the

China market and the US market during the second period as compared to the earlier period

was 707 percent, the highest among other painting of regional market with the US market.

Nevertheless, the stock return correlation of the China market with global market is lower

than that of other countries markets with U.S markets. However, the results show that there

was increase correlation coefficient during the U.S. subprime crisis. The null hypothesis of no

increase in correlation is rejected by all cases during the U.S. subprime crisis.

Also, one can realize that correlation analysis is not enough to investigate financial contagion

during the U.S. subprime crisis. In the next section, we discuss the econometric methodology

that enables us to investigate financial contagion during the U.S. subprime crisis.

Insert Table 4 here

The adjusted correlation coefficients suggest that there is no evidence of contagion for any

BRIC market. In this case, we witness an interdependence phenomenon between the BRIC

markets and the US market and not a pure contagion after the US sub-prime mortgages crises

in 2007. Our results are similar to previous researches by Omri et Frikha (2011) and Hsien-Yi

Lee (2012). This empirical finding is consistent with previous empirical based correlation

approach studies suggesting that correlations of stock returns have been increased in recent

30
periods as a result of increasing financial integration across national stock markets. This

increasing of correlation leads to lower diversification benefits especially in the long-run

horizons.

5.3. DCC-EGARCH(1,1) results

The results of estimation of the DCC-EGARCH(1,1) model are reported in Table 5 below.

This Table reports the results of the first step estimation of an univariate EGARCH models for

the five stock markets in Panel A. It reports the results of the variance equation in Panel B and

the results of the multivariate DCC equation in Panel C. It reports also the results of the

residuals diagnostics in terms of Ljung-Box for the residuals and squared residuals series.

From this Table different important results can be drawn. For the mean equations results, it

appears that the lagged coefficient of each dependent variable become more significant in the

crises period compared to the tranquil period. For instance, during the tranquil period no one

of the lagged coefficient is significant at conventional significance level. In contrast, during

the crises period, the only the Russia and India lagged coefficients remain non-significant.

Concerning the U.S stock market coefficient, it is significant for all the BRIC’s stock markets

and for the two periods.

Turning now to the variance equations. In contrast to previous empirical studies, we consider

in this paper an EGARCH(1,1) specification for modeling the variance equations. Empirical

results not reported here and can be obtained upon request from the authors show that this

specification better perform the standard GARCH(1,1) and GJR(1,1) specifications. For the

GARCH and ARCH coefficients, the empirical results show that the α coefficients are not

significant for both periods in contrast to the β coefficients that are significant at the 1% level

significance and are more significant under the second period. For the EGARCH coefficients

31
the empirical results show that these coefficients are significant at conventional level except

for the θ1 coefficent for the China stock market for both periods and the θ2 coefficient for

Russia and China stock markets for the first and second period respectively.

Insert Table 5 here

Finally, the results of the third Panel, Panel C, show that the coefficients of the multivariate

DCC equation are all significant and for both periods. The α and β are both significant under

the two periods except for the α coefficent under the tranquil period. Moreover, the results

show that the sum of these two coefficients is close to one especially under the crises period.

For the correlations coefficients, coefficients of interests in this paper, the results show that

there are an increase of these coefficients between the tranquil period and crises period. The

four correlation of interest have increased from 0.425, 0.198, 0.202 and 0.310 to 0.677, 0.207,

0.282 and 0.383 for the U.S-Brazil, U.S-Russia , U.S-India and U.S-China respectively. The

question that need an answer at this level is whether that increase statistically significant or

not. To answer this question, we employ the Forbes and Rigobon (2002) approach on the

estimated correlation coefficients before and during the financial crises. The empirical results

are reported in Table 6 below.

Insert Table 6 here

From Table 6, we can conclude that only for the US-Brazil couple that the increase in the

DCC-EGARCH correlation is significant as shown by the t-statistic of Forbes and Rigobon

(2002) reported in the last column of Table 6. This result indicates that we found only a

contagion phenomena from the U.S stock market to the Brazil stock market. For the other

stock markets, there is no evidence of contagion.

32
5.4. Nonlinear causality results

To examine whether there is contagion or only interdependence from the USA to the BRIC

countries, we start by performing the nonlinear causality tests to residuals extracted from the

estimated VAR models for each couples of times series, (USA, Brazil), (USA, Russia),

(USA, India) and (USA, China). Before applying these tests we need to determine the values

of the lead length, m, the lag lengths, Lx and Ly, and the scale parameter, ϵ . Based on the

empirical results of the Hiemstra and Jones (1993) Monte Carlo simulations, we set the

parameter values of the nonlinear causality tests equals to m 1, Lx 1 to 8, Ly 1 to 8 , and

ϵ 1.5 , where 1.

Tables 7 and 8 report the results of nonlinear Granger causality tests for the tranquil and crises

periods, respectively. From table 7, we report no significant nonlinear causality running from

USA to BRIC countries during tranquil countries except for India at only the fourth lag. This

result is consistent with the fact that under the tranquil period, the five financial markets are

characterized by low volatility and small correlation between stocks markets. As a

consequence, it is expected that no significant nonlinear causality will be detected.

Insert Table 7 here

In contrast, during the crises period, we note a significant nonlinear causality relationship

from the USA to Brazil, to India and to China countries. This finding suggests a contagion

phenomenon from the USA to the three BRIC countries (Brazil, USA and India). This

contagion is more pronounced for the India country. For the Brazil and China countries the

33
test of nonlinear causality is significant only in one lag, eight for the Brazil country and 4 for

the China country.

Insert Table 8 here

6.Implications and summary

This paper investigated empirically whether any of the stock markets of BRICs region have

been affected by the US subprime crisis. More precisely, it investigated financial contagion

versus financial interdependence. To this end a daily indexes denominated in US dollars, for

the Brazil, China, India, Russia, and USA countries have been used to examine whether

contagion occurred across markets during the Global financial crises of 2007. The sample

used covers the period from 10/02/2005 to 14/04/2009. To test for possible contagion versus

interdependence, we use three econometrics approaches to test for changes in the relation

between the USA stock market and other financial markets after the onset of the crisis : (1)

the traditional adjusted correlation approach of Forbes and Rigobon (2002), (2) the DCC-

EGARCH model of Engle (2002) and (3) the nonlinear causality approach of Hiemstra and

Jones (1994).

The results obtained from the application of the three methods are not uniform. We found

strong evidence for the presence of contagion phenomena from the USA stock market to the

India stock market when we use Forbes and Rigobon approach as we found a significant

evidence of an increase in cross-market linkages between the tranquil and turmoil period. The

34
same phenomenon has been observed for the case of the Brazil stock market when the DCC

Engle approach has been employed. Using nonlinear causality method, we detect a contagion

phenomenon for the Brazil and China countries. For the Russia country there is no evidence

for the presence of contagion phenomena using all methods. In general, we can say that for

most BRICs countries there is a menace of financial contagion which spread across markets

as the subprime crisis developed. Crises and contagion seem to be the price that some

countries have to pay to integrate with the international financial system. Since we focus only

on the subprime crisis, it is important to acknowledge that our results are limited to this

specific episode of financial crises.

The results of this empirical analysis seem to support the operational advantages associated

with definition of contagion proposed by Forbes and Rigobon (2001). In fact, the evidence of

increased dependence between countries before and during crisis period should be prudently

considered by portfolio managers as it suggests that a simple strategy of geographical

diversification may not all the time be successful. The results provide a significant implication

for the international investor, multinational corporations and portfolio managers, who all are

involved in reducing and managing their financial risk exposure. Moreover, as the results

shown the benefits from international diversification depend from the country of BRIC

region. Former researcher papers demonstrate that a negative shock to an emerging country

can increase the relation between stock returns across countries, confusing investors’ ability to

differentiate their investments. At such times, investors may want to limit their international

assets to minimize their risk. From policymakers' perspective, this study provides important

information about the directions for possible measures to take in order to protect emerging

markets from contagion during future crises. The challenge for policymakers is to manage the

process as to take full advantage of the opportunities, while minimizing the risks. However

not all conditions are to be met before governments liberalize the financial sector, countries

should guarantee that the financial system is prepared to manage with foreign capital flows

35
and external shocks. More complete policies for risk management are needed to build solid

economies, in particular in terms of directive and administration of the financial system. The

increasing integration of countries gives governments less policy instruments to manage the

shock, so there is a crucial need for international financial policy coordination.

Refrences

Aloui, R., Aissa, M.S.B. and Nguyen, D.K. (2011). “Global financial crisis, extreme

interdependences, and contagion effects: the role of economic structure?” Journal of Banking

and Finance 35, pp. 130–141.

Angkinand A., Chiu, E.M.P. and Willett, T.D. (2009), “Testing the unstable middle and two

corners hypotheses about exchange rate regimes”. Open Economies Review, Vol. 20 No. 1,

pp. 61-83.

Antonakakis, N., Chatziantoniou, I., & Filis, G. (2013). Dynamic co-movements of stock

market returns, implied volatility and policy uncertainty. Economics Letters, 120, 87-92.

http://dx.doi.org/10.1016/j.econlet.2013.04.004

Baek, E. and Brock, W. (1992). “A general test for nonlinear granger causality: bivariate

model”. Working Paper. Iowa State University.

Baig, T. and Goldfajn, I. (1999). "Financial market contagion in the asian crisis." IMF Staff

Papers, 46, pp. 167-195.

36
Bekaert, G. and Campbell H. (1997). “Emerging Equity Market Volatility”. Journal of

Financial Economics, 43, pp. 29-77. (O)

Beltratti, A. and Morana, C. (2008). "International shocks and national house prices," ICER

Working Papers - Applied Mathematics Series 14-2008, ICER - International Centre for

Economic Research.

Bertero, E. and C. Mayer (1990). “Structure and performance: global interdependence of

stock markets around the crash of 1987”. European Economic Review, 34; 1155-1180.

Billio, M. and Pelizzon, L. (2003). "Volatility and shocks spillover before and after EMU in

European stock markets". Journal of Multinational Financial Management, Elsevier, vol.

13(4-5), pp. 323-340, December.

Bonfiglioli, A. and Favero, C. (2005). “Explaining Co-movements Between Stock Markets:

the case of US and Germany”. Journal of International Money and Finance, Vol.78, pp.1299-

1316

Boyer, B.H., Kumagai,T.,Yuan, K., (2006). How do crises spread? Evidence from accessible

in inaccessible stock indices. Journal of Finance 61, 957–1003.

Caillault, C. and Guegan, D. (2005). Empirical estimation of tail dependence using copulas:

application to Asian markets. Quantitative Finance, 5(5):489–501.

Canarella, G., Pollard, S.K., (2007). A switching ARCH (SWARCH) model of stock market

volatility: some evidence from Latin America. International Review of Economics 54, 445–

462.

37
Calvo, S. and Reihart, C. (1996). "Capital flows to Latin America: Is there evidence of

contagion effect?" In. Calvo, G., Goldstein, M., Hochreiter, E., esd. Private capital flows to

emerging markets after the Mexican Crisis, Institute of International Economics.

Celik, S. (2012). The more contagion effect on emerging markets: The evidence of DCC-

GARCH model. Economic Modelling, 29, 1946-1959.

Celikkol, H., Akkoc, S. and Akarim, YD. (2010). “The Impact of Bankruptcy of Lehman

Brothers on the Volatility Structure of ISE-100 price Index”, Journal of Money, Investment

and Banking, Issue 18.

Chakrabarti, R., Roll, R., 2002, East Asia and Europe during the 1997 Asian collapse: A

clinical study of a financial crisis. Journal of Financial Markets 5, 1-30.

Chiang, T.C., Jeon, B.N., Li, H.M., (2007). "Dynamic correlation analysis of financial

contagion: Evidence from asian markets." Journal of International Money and Finance, 26,

1206-1228

Corsetti, G., Pericoli, M. and Sbracia, M. (2005). “Some contagion, some interdependence:

More pitfalls in tests of financial contagion.” Journal of International Money and Finance, 24,

pp. 1177-1199.

Corsetti, G., Dedola, L. and Leduc, S. (2005). “DSGE models of high exchange-rate volatility

and low pass-through”. International Finance Discussion Papers 845, Board of Governors of

the Federal Reserve System (U.S.).

De Melo Mendes, B.V., (2005). Asymmetric extreme interdependence in emerging equity

markets. Applied Stochastic Models in Business and Industry 21, 483–498

38
Diamandis, P.F., (2008). Financial liberalization and changes in the dynamic behaviour of

emerging market volatility: evidence from four Latin American equity markets. Research in

International Business and Finance 22, 362–377.

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

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

Edwards, S. and Susmel, R. (2001). “Volatility dependence and contagion in emerging equity

markets”. Journal of Devevelopment Economy 66, pp. 505–532. Eic

Eichengreen, B., Rose, A.K. and Wyplosz, C. (1996). “Contagious currency crises”.

Scandinavian Economics Review 98 (4), pp. 463–484.

Engle, R. (2002), Dynamic conditional correlation: A simple class of multivariate generalized

autoregressive conditional heteroskedasticity models, Journal of Business and Economic

Statistics, 20(3), 339-350.

Fidrmuc, J. and Korhonen, I. (2010). “The impact of the global financial crisis on business

cycles in Asian emerging economies”. Journal of Asian Economics, Elsevier, vol. 21(3), pp.

293-303, June.

Forbes, K.J. and Rigobon, R. (2000). “No contagion, only interdependence: measuring stock

market comovements”. Journal of Finance 57, pp. 2223–2261.

Gagnon, L. and Karolyi, G. A., (2006). “Price and Volatility Transmission across Borders.”

Working Paper Series 2006-5, Ohio State University, Charles A. Dice Center for Research in

Financial Economics.

Gelos, R. and Sahay, R. (2001) Financial Market Spillovers in Transition Economies,

Economies of Transition, 9, 53-86.

39
Gilmore, C. K., McCarthy, S. E. and Spelke, E. S. (2007). “Symbolic arithmetic knowledge

without instruction”. Nature, 447, pp. 589–591.

González-Hermosillo, B. (2008). “Investors Risk Appetite and Global Financial Market

Conditions”. IMF Working Papers, Vol., pp. 1-75. Available at SSRN:

http://ssrn.com/abstract=1119425

Gómez-Puig, M., & Sosvilla-Rivero, S. (2013). Granger causality in peripheral EMU debt

markets: A dynamic approach. Journal of Banking and Finance, 37, 4627–4649.

Granger, C.W.J. (1969). “Investigating causal relation by econometric and cross-sectional

method”, Econometrica 37, pp. 424–438.

Hamao, Y.R., Masulis, R.W. and Ng, V.K. (1990). “Correlations in Price Changes and

Volatility Across International Stock Markets”, Review of Financial Studies, 3, pp. 281-307.

Hamilton, James D. and R. Susmel (1994), "Autoregressive Conditional Heteroskedasticity

and Changes in Regime," Journal of Econometrics, 64, 307-333.

Hiemstra, C. and Jones, J.D. (1993). “Some information relating to the finite sample

properties of modified baek, and brock nonlinear granger causality test”. Working paper,

University of Strathclyde and Securites and Exchange Commission.

Hiemstra, C. and Jones, J.D. (1994). “Testing for linear and nonlinear granger causality in the

stock price-volume relation”. Journal of Finance. 49, pp. 1639–1664.

Hsieh, D. (1991). “Chaos and Nonlinear Dynamics: Application to Financial Markets”.

Journal of Finance, 46, pp. 1839-1877.

Hsieh, D. (1989). “Testing for Nonlinearity in Daily Foreign Exchange Rate Changes”.

Journal of Business, 62, pp. 339-368.

40
Jones, P., & Olson, E. (2013). The time-varying correlation between uncertainty, output, and

inflation: Evidence from a DCC-GARCH model. Economics Letters, 118, 33-37.

Khalid, Ahmed M., and Mashiro Kawai. (2003). Was Financial Market Contagion the Source

of Economic Crisis in Asia? Evidence Using a Multivariate VAR Model. Journal of Asian

Economics 14:131–56.

Khallouli, W., Sandretto, R., (2012). Testing for “contagion” of the subprime crisis on the

Middle East and North African stock markets : A Markov Switching EGARCH approach.

Journal of Economic Integration 27, 134’166.

Khan, S., Park, K. W. (2009), Contagion in the stock markets: the Asian financial crisis

revisited. Journal of Asian Economics, 20(5), 561-569

Khan, S., F. Islam, and S. Ahmed (2005), "The Asian Crisis: An Economic Analysis of the

Causes," Journal of Developing Areas, Vol. 39, 1, 169-90.

Karolyi, G. A. and Stulz, R. M. (1996). “Why Do Markets Move Together? An Investigation

of U.S.-Japan Stock Return Comovements”. Journal of Finance, American Finance

Association, vol. 51(3), pp. 951-86, July.

Kenourgios, D., Samitas, A., and Paltalidis, N. (2011). “Financial crises and stock market

contagion in a multivariate time-varying asymmetric framework”. Journal of International

Financial Markets, Institutions & Money 21 (92), 106

King, M. and Wadhwani, S. (1990). “Transmission of volatility between stock markets”.

Review of Financial Studies 3, pp. 5–33.

Lee, S. B., & Kim, K. J. (1993): “Does the October 1987 crash strengthen the co-movements

among

41
national stocks markets?”, Review of Financial Economics, 3, 89-102.

Longin, F. and Solnik, B. (1995). “Is the Correlation in International Equity Returns Constant:

1960-1990? ”. Journal of International Money and Finance, 14, pp. 3-26.

Longin, F. and Solnik, B. (2001). “Extreme Correlation of International Equity Markets,”

Journal of Finance, 56, pp. 651-678.

Longstaff, F.A. (2010), “The Subprime Credit Crisis and Contagion in Financial Markets”.

Journal of Financial Economics, vol. 97, no. 3, pp. 436-450.

Loretan, M., and English, W.B. (2000). “Evaluating Correlation Breakdowns during Periods

of Market Volatilit”. Intemationai finance Discussion Papers, no. 658. Board of Governors of

the Federa! Reser\'e Board.

Murat, Y., (2010). “Auctioning a Discrete Public Good under Incomplete Information”.

Working Papers 2010/14, Bogazici University, Department of Economics.

Masih. A. M., Masih. R. (1999), Are Asian stock market fluctuations due mainly to intra-

regional ontagion effects? Evidence based on Asian emerging stock markets, Pacific-Basin

Finance Journal, vol. 7, pp: 251-282.

Naoui, K., Liouane, N., Brahim, S., (2010). A dynamic conditional correlation analysis of

financial contagion: the case of the subprime credit crisis. International Journal of Economics

and Finance 2 (3), 85–96

Papavassiliou, V. G. (2014). Cross-asset contagion in times of stress. Journal of Economics

and Business, Vol 76, pp. 133–139

Patton, A. J. (2006). Modelling asymmetric exchange rate dependence. International

Economic Review 47, 527–556

42
Pesaran, M.H., Pettenuzzo, D. and Timmermann, A. (2006). “Forecasting time series subject

to multiple structural breaks”. Review of Economic Studies 73, pp. 1057–1084.

Omri, A. and Frikha, M. (2011). "No Contagion, Only Interdependence During the US Sub-

Primes Crisis," Transition Studies Review, Springer, vol. 18(2), pp. 286-298, December.

Ramlall, I. (2010). “Has the US Subprime Crisis Accentuated Volatility Clustering and

Leverage Effects in Major International Stock Markets? ”. International Research Journal of

Finance and Economics, Issue 39, pp. 157-185.

Reinhart, C. and Kaminsky, G. (2000). "Crisis financieras en Asia y Latinoamerica: ahora y

entonces [Financial Crises in Asia and Latin America Different: Then and Now]”. MPRA

Paper 13735, University Library of Munich, Germany.

Rigobón, R. (2003). “On the measurement of the international propagation of shocks: is the

transmission stable? ”. Journal of International Economics 61 (2), pp. 261–283.

Samitas, A., Tsakalos, I., (2013). How can a small country affect the European economy? The

Greek contagion phenomenon. Journal of International Financial Markets, Institutions and

Money 25, 18-32

Yiu, M. S., Ho, W.-Y. A., & Choi, D. F. (2010). Dynamic Correlation Analysis of Financial

Contagion in Asian Markets in Global Financial Turmoil. Applied Financial Economics , 20

(4), 345-354.

43