Contagion and integration of emerging markets during the global financial crisis

11 

Loading....

Loading....

Loading....

Loading....

Loading....

Full text

(1)

AUSTRALIAN JOURNAL OF BASIC AND

APPLIED SCIENCES

ISSN:1991-8178 EISSN: 2309-8414 Journal home page: www.ajbasweb.com

Open Access Journal

Published BY AENSI Publication

© 2017 AENSI Publisher All rights reserved

This work is licensed under the Creative Commons Attribution International License (CC BY).

http://creativecommons.org/licenses/by/4.0/

To Cite This Article: Lamia SEBAI and Siwar ELLOUZ., Contagion and integration of emerging markets during the global financial crisis. Aust. J. Basic & Appl. Sci., 11(6): 1-11, 2017

Contagion and integration of emerging markets during the global financial

crisis

1Lamia SEBAI and 2Siwar ELLOUZ

1Phd student in finance at Faculty of Economics and Management, University of Sfax 2Professor of finance at higher business school of Sfax, University of Sfax

Address For Correspondence:

Siwar ELLOUZ, Professor of finance at higher business school of Sfax, University of Sfax E-mail: ellousiw@yahoo.fr

A R T I C L E I N F O A B S T R A C T

Article history:

Received 18 January 2017 Accepted 28 March 2017

Available online 1 May 2017

Keywords:

Emerging markets; Financial crises; Financial integration; ICAMP

The aim of this paper is to study the varying nature of integration and contagion during the global financial crisis. We consider three types of risk coming from the local, regional and world markets. These risks require the use of three types of models: a world capital asset pricing model (CAPM), a CAPM with the U.S. equity return as the benchmark asset, and a regional CAPM with a regional portfolio as the benchmark. We investigate theses models with stock returns in four different regions: Latin America, Europe, Asia and Africa / Middle East. Our empirical results show that there are significant contagion effects from the U.S. to the Middle East markets; Asian and Latin during the global financial crisis 2007-2009 periods.

INTRODUCTION

Those years are accompanied by a long series of crises: the crisis of European debt, the global financial crisis, the Asian and Russian crisis, the Mexican crisis and the crisis in Argentina. These crises have affected emerging and developed economies. During the period of crisis, there are very strong interdependence between emerging stock markets which makes us ask questions about the degree of financial integration and the nature of the contagion. Therefore, previous studies have examined the impact of crises on the degree of integration and contagion since every crisis brings a potential effect on the nature of co-movements. The 2007-2009 financial crises were the first global crisis in 1929. The crisis has its origin in the United States. In fact, the crisis affected the global stock markets. It is important to know the sources of "contagion" in the emerging stock markets.

We employ the methodological approach adopted by Bekaert, Harvey, and Ng (2005). Our model has two factors with varying betas in time measure of the degree of financial integration. We apply this model to stock returns in four different regions: Latin America, Europe, Asia and Africa / Middle East. Furthermore, we examine the financial contagion and volatility driven by global, local and regional factors during crisis periods. Bekaert et al. (2005) define contagion as ‘‘correlation over and above what one would expect from economic

(2)

increases in intra-regional co-movements during the crisis period. We assume, as Bekaert et al. (2005-2011); Arouri et al (2007) and Cho et al. (2015) that the US market portfolio represents a good proxy of the global market. In addition, we consider three types of risk from the regional, local and world markets. These risks require the use of three types of models: a world capital asset pricing model (CAPM), a CAPM with the U.S. equity return as the benchmark asset, and a regional CAPM with a regional portfolio as the benchmark.

This paper contributes to the literature in different levels; in fact, we make a number of contributions to the literature, we extend the approach and analysis in Bekaert et al. (2005), the first contribution is to determine the test for global and regional market integration. The second contribution is to know the proportion of the local, regional and global proportions and factors of variance in the period of crisis subprime or non-crisis. The third contribution is to investigate to what extent contagion during the subprime crisis by examining shock correlations of the model idiosyncratic shocks or unexpected returns. The remainder of this paper is organized as follows. Section 1 presents an overview of the existing literature. Section 2 develops the empirical model with time-varying factor exposures and describes the contagion test. Section 3 describes the data. The empirical results are reported in section 4. Section 5 concludes.

Literature review :

Baele (2005) investigates contagion of the US market to a number of local equity markets in Europe in periods of high volatility in global markets, using a swatch of 13 Western European countries. Using a switching regime model, this model allows the shock sensitivity to change over time. He finds regime switches to being both statistically and economically important. The intensity of the shock effect of the European Union and the United States has increased dramatically over the 80’s and 90’s.

In addition the shock intensities have increased more strongly during the second half of the year. They concluded that increased trade integration, the development of stock markets and low inflation contributes to the increase in the intensity of the impact of EU shocks. More evidence is found to the contagion of the US market to a number of local equity markets in Europe in periods of high volatility in global markets.

Bekaert, Harvey and Ng (2005) use a two-factor model or beta variables various types of markets. They apply their model to stock returns on three regions: Europe, Southeast Asia, Latin America and more. They examine the contagion during the crisis periods. Their results suggest that there is no evidence of further contagion caused by Mexican crisis. Moreover, they find a residual correlation increase, particularly in Asia, during the Asian crisis.

Similarly, Baele and Inghelbrecht (2010) developed a model that correctly characterizes fundamental market linkages in an environment of time-varying market integration and investigates to what extent contagion test results depend on the particular choice and complexity of the chosen dynamic factor model. They subordinate the exposure to the global and regional market to a latent variable speed and three structural instruments of 14 European countries over the last 35 years.

Baele and Inghelbrecht (2010) showed that the existence of contagion in Europe for a number of crises, including the beginnings of the crisis and the structural increase exposures and correlations with the world market (regional) and suggested that market integration has increased significantly over their sample period and that contagion of their preferred specification is absent.

Bekaert et al. (2011) analyzed the transmission of crises to country-industry equity portfolios in 55 countries during the 2007-2009 financial crises. They developed an international model with three factors the US market factor the global market factor and the domestic one. They identified three types of contagion: global contagion, US contagion and domestic contagion. Also, they identify the channels of influence through which contagion spreads. They examined the period of the global financial crisis to investigate how and why that crisis transmitted so readily and extremely across the world. Their model allows for whether the global crisis 2007-2009 mainly reflects a global financial shock, specific shock to the US economy, which, then spreads worldwide: they also they to know to what extent the crisis spreads the troughs countries. They found statistically significant evidence of contagion during the 2007-2009 financial crises in the US markets and the global financial sector but the effects are economically weak. However, they concluded that there is evidence of significant US and global contagion. In addition, the overall effect is economically small.

(3)

financial markets during the recent global financial crisis. The contagion effects are stronger for emerging markets than the developed markets and it is shown that financial markets have a high incidence of contagion for American markets to other markets despite that the financial sector recorded a low contagion.

Bekaert et al, (2014) identified the mechanisms of transmission the 2007-2009 financial crises in fifty five stock markets and ten sectors markets. They find an evidence of contagion from the United States and the global financial sector that the effects are small. In addition, there was a significant contagion domestic market to individual national portfolios, with a severity related quality of economic fundamentals of the countries.

Cho, Hyde and Nguyen (2015) examined the variable nature of integration and contagion modes of asset portfolios in five recent periods of crisis, including the global financial crisis and the crisis of European debt. They follow the methodological approach adopted by Bekaert, Harvey and Ng (2005) and Baele and Inghelbrecht (2010) using a model of regime change GARCH with regional-local CAPM to measure the level of integration and co-movement. About financial analysis of contagion, the contagion test pushed Bekaert et al

(2011) to consider whether the contagion affects portfolio feature beyond the standard levels of co- movement during periods of economic and financial crisis. Their results show that there are significant changes in the level of comovement with global and regional factors for regional style portfolios between low and high volatility. In addition, there are signs of contagion at a regional and a global scale. They also found evidence of contagion in the financial crisis in relation to previous literature on market returns: Baur (2012); Bekaert et al. (2011); Dungey and Gajurel (2013).

However, the effects of Mexico and Asian crisis are limited to regional effects. Cho, Hyde and Nguyen (2015) treat the extent of contagion between the BRICS and US stocks. These markets were influenced by the global financial crisis 2007-2009. During the global financial crisis, the authors show that there are significant spillovers from the United States to market BRICS stocks. But Dooley and Hutchison (2009) as well as Demetriou et al. (2013) examine the transmission of the 2007-2009 global financial crisis in the emerging markets and BRICS, respectively. These authors do not suggest contagion for emerging stocks and BRICS In mid - 2008; rather, they examine relationship to the US market back in early 2009.

Bekaert et al. (2014) analyzed the transmission of the 2007 to 2009 financial crisis to 415 country industry equity portfolios. They use a factor model to predict crisis returns, defining unexplained increases in factor loadings and residual correlations as indicative of contagion. They see signs of contagion from the US and the global financial sector, the effects are low. They find significant evidence of contagion during the period 2007-2009. In addition, they are statistically of an economically weak evidence of contagion of the US markets and the global financial sector. There is also strong evidence of domestic contagion, with factor loadings. Aloui et al. (2011) examined the extent of the 2007-2009 global financial crisis and the contagion effects using empirical investigation of the extreme financial interdependencies of some emerging markets with selected United States. They show strong evidence of time-varying dependence between the US and selected emerging markets.

Long et al, (2014) examined the conditional time-varying currency betas from five developed markets and six emerging financial ones with contagion and spillover effects. They find empirical evidence of contagion and spillover between the stock market and the foreign exchange market during the recent global financial crisis. The effect is stronger in the emerging markets than in the developed markets.

An important paper by Syriopoulos et al. (2015) that examined the dynamic risk-return BRICS capital markets and potential time model variation correlations and volatility spillovers effects on the US stock market. They conclude that BRICS equity markets were severely affected by the global financial crisis of 2008. Guyot et al. (2014) studied the impact of foreign financial shocks of emerging countries of MENA. Using a set of MMCD RBAP and models, they found that external shocks can cost of equity in mature emerging markets. Maghyereh et al. (2015) studied the stock returns and volatility between the United States and a group of stock markets in the Middle East and North Africa before and after the global financial crisis in 2008. The empirical findings suggest that the pre-crisis relationship with the United States was weak and insignificant, which jumped to a high level after the crisis.

Model:

We follow the framework of Bekaert, Harvey and Ng (2005). First, we develop an international version of the conditional CAPM who characterizes fundamental linkages between markets. We provide a very general specification for the time-varying factor exposures. Then, we discuss the estimation and the model selection strategy. Finally, we briefly describe the contagion test.

Let Ri,t: represent the excess return of country𝑖, our model can be written as follows:

Ri,t= δiZi,t−1+ βi,t−1us eus,t−1+ βi,t−1 reg

ereg,t−1+ βi,t−1us eus,t+ βi,t−1 reg

ereg,t+ ei,t (1)

(4)

shock ereg,t. Ω𝑡−1 : contains all the information available at the time(𝑡 − 1). Zi,t−1 : The vector of information on the market.

The sensitivity of equity market i to the foreign news factors is measured by the parameters βi,t−1us and βi,t−1 reg

. Following: Bekaert & Harvey, 1997; Ng, 2000); Bekaert, Harvey and Ng (2005); Baele and Inghelbrecht (2009, 2010) and Cho, Hyde and Nguyen (2015) the betas depend on both structural information as follows:

𝛽𝑖,𝑡−1𝑢𝑠 = 𝑝1,𝑖′ 𝑋𝑖,𝑡−1𝑢𝑠 + 𝑞𝑖′𝑋𝑖,𝑡−1𝑤 . 𝑤𝑢𝑠,𝑡−1 (2)

𝛽𝑖,𝑡−1𝑟𝑒𝑔 = 𝑝2,𝑖′ 𝑋𝑖,𝑡−1 𝑟𝑒𝑔

+ 𝑞𝑖′𝑋𝑖,𝑡−1𝑤 . (1 − 𝑤𝑢𝑠,𝑡−1) (3)

Wherewus,t−1: denotes the market capitalization of the United States, relative to the total world market capitalization, at time(t − 1). 𝑋𝑖,𝑡−1𝑢𝑠 ; 𝑋𝑖,𝑡−1

𝑟𝑒𝑔

are the trade of word and trade of region. 𝑋𝑖,𝑡−1𝑢𝑠 :is the trade with the rest of the word data.

The idiosyncratic shock ei,t And the conditional variance of the global, regional-specific and country-specific shocks are estimated by a single state asymmetric GARCH (1,1) with a Normal distribution.

𝑒𝑖,𝑡/Ω𝑡−1 ~𝑁(𝑂, 𝜎𝑖,𝑡2) (4)

𝜎𝑖,𝑡2 = 𝑎𝑖+ 𝑏𝑖𝜎𝑖,𝑡−12 + 𝑐𝑖𝑒𝑖,𝑡−12 + 𝑑𝑖𝜂𝑖,𝑡−12 (5)

Before we can estimate equation (1), we first need to identify the global and regional market shocks:

𝜀𝑖,𝑡= 𝛽𝑖,𝑡−1𝑢𝑠 𝑒𝑢𝑠,𝑡+ 𝛽𝑖,𝑡−1 𝑟𝑒𝑔

𝑒𝑟𝑒𝑔,𝑡+ 𝑒𝑖,𝑡 (6)

Whereεi,t: denotes the return residual of market i.

This decomposition guarantees that global market shocks are orthogonal to the region-specific shocks. Concerning the analysis of variance, we follow the same Bekaert, Harvey and Ng (2005). We also try to calculate the share of the conditional correlation and the share of the world market conditional variance and regional following a crisis. Indeed, the U.S. and regional correlations are given by:

ρi,us,t=

βi,t−1US σus,t

√hi,t (7)

ρi,reg,t=

βi,t−1US βreg,t−1us σus,t2 +βi,t−1reg σreg,t2

√hi,threg,t

(8)

In addition, the variance ratios are follows:

𝑉𝑅𝑖,𝑡𝑢𝑠=

i,t−1US )2σus,t2

𝑖𝑡 (9)

𝑉𝑅𝑖,𝑡𝑟𝑒𝑔=(βi,t−1

reg )2σ reg,t 2

𝑖𝑡 (10)

Integration tests :

We examine the integration and segmentation of stock markets using the model (1) to (5). First, if the profitability of a stock market is explained by global risk factors local variables have no impact on the market. In the case, 𝛿𝑖= 0; the segmentation hypothesis is rejected. Therefore, we interpret that integration can be global or regional. Furthermore, if 𝛿𝑖= 0 and 𝛽𝑖,𝑡−1

𝑟𝑒𝑔

= 0 our model is reflected in the traditional and the global integration ICAPM hypothesis is accepted. Finally, if 𝛿𝑖= 0 and 𝛽𝑖,𝑡−1𝑢𝑠 = 0 regional integration and CAPM hypothesis is accepted.

Contagion tests:

In this section, we examine the contagion by measuring the correlations of the model of idiosyncratic shocks or unexpected returns. We adopt the approach advocated by Bekaert, Harvey and Ng (2005). The contagion test is specified as follows:

𝑒̂𝑖,𝑡= 𝑤𝑖+ 𝑣𝑖,𝑡𝑒̂𝑧,𝑡+ 𝑢𝑖,𝑡 (11)

(5)

Where êi,t and êz,t are the estimated idiosyncratic return shocks of country i and benchmark market z (i.e. Global or regional market), and Di,t Is a dummy variable that takes on a value of 1 in a particular period and zero. Our tests determine whether v0 and v1Are jointly equal to zero (overall contagion), and v1 is significantly different from zero (contribution of particular periods to contagion). Di,t : Variable that represents three sample periods: the subprime crisis (2007: 7, 2009: 8); Post-crisis subprime period (2010: 01-2014: 04) and before the crisis Deleted (1994: 04- 2006: 12).

Estimation procedure :

We use a three-step procedure to keep estimation feasible. First, we estimate the global and regional market shocks. Second, we relate country shock to both global and regional shocks obtained in the first step. Third, we estimate the specifications of country by country on the U.S. and regional market model estimates. Givien that, the regional shocks are more independent from global market shocks than the country-specific shocks are independent from both region-specific and global market shocks. All estimates are obtained by maximum likelihood.

Data description:

The data comprise monthly Morgan Stanley Capital Index (MSCI) aggregate prices for U.S. and 20 countries classified into four regions: Latin America, Europe, Asia and Africa / Middle during a period from April1994 to April 2014. Real exchange rates represent the value of the local currency against the U.S. dollar and they are extracted from the IMF’s International Financial Statistics (IFS) and the U.S. Federal Reserve databases. The real effective exchange rate index is the geometric average of bilateral real exchange rates among the countries under consideration. The expected country return is modeled as a linear function of lagged values of the US and the regional market return, the US short rate, dividend yield, term spread, the default spread, the local interest rate and the local return. The data for trade with the U.S, region and word are calculated as the ratio of imports plus exports over GDP more detailed view as Bekaert, Harvey, and Ng (2005).

Empirical results:

Table 1 summarizes the averages of betas, correlations and variance ratios for all the countries with respect to the US and regional markets. The averages of beta, correlations and variances are variable over time as a function of macroeconomic variables.

The emerging countries of Latin America have very high betas with respect to the US market. Betas vary between 0.35 in Chile and 0.95 in Brazil. The regional betas are high. This ranges from 0.27 in Chile to 0.42 in Brazil. Correlations are also small. The highest value of US betas and regional betas is 0.14 for Brazil and 0.32 for Peru.

The variance share explained by the US market is equal to 13.84%. The shares explained by the regional market are also lower and average 4.62%. Except for Colombia, the US market accounts for considerably larger shares of the variability in the returns of the Latin American markets than the regional market.

For the emerging countries in Europe the beta with respect to the US market are very high. The highest value is 0.95 in Hungary followed by Greece 0.94. The regional betas are also lower. It ranges from 0.25 in Hungry to 0.37 in the Czech Republic. Correlations with the US and regional markets are also low. The highest value in the US market is 39% in Hungry. While the highest correlation value with the regional market is 42% in Hungary. The variance explained by the US market is 19.16%. The shares explained by the regional market are also lower than 6.3%. The US market also accounts for considerably larger shares of the variability of European market returns than the regional market.

For the emerging countries of the Middle East the betas with respect to the US market are very high. The highest value is (0.92) in Turkey followed by Egypt (0.85). The regional beats are average. It exceeds 0.5 for all countries. Correlations with the US and regional markets are also low. The highest correlation value with the US market is 23% in South Africa and Turkey. Then the highest value with the regional market is (43%) in South Africa. The ratio variance of the US market is equal to 2.63%. The ratio variance of the regional market is also lower than 3.97%.

(6)

Table 1: Implicit statistics of the country specific model

Market βi,tus βi,t

reg

ρi,us ρi,reg VRusi,t VRi,t

reg

Brazil (0.95

0.76) (

0.42

0.61) (

0.14

0.20) (

0.24

0.26) (

0.22

0.75) (

0.026

0.003)

Chile (0.35

0.29) (

0.27

0.65) (

0.041

0.47) (

0.028

0.50) (

0.26

0.99) (

0.018

0.064)

Colombia (0.73

0.63) (

0.42

0.91) (

0.011

0.13) (

0.05

0.61) (

0.021

0.076) (

0.030

0.20)

Mexico (0.88

0.67) (

0.35

0.59) (

0.078

0.081) (

0.148

0.27) (

0.031

0.058) (

0.027

0.16)

Peru (0.93

0.80) (

0.37

0.66) (

0.092

0.14) (

0.32

0.60) (

0.16

0.57) (

0.13 0.84) Czech Republic (0.48

0.41) (

0.37

0.66) (

0.24

0.36) (

0.28

0.57) (

0.20

0.71) (

0.022

0.15)

Greece (0.94

0.81) (

0.29

0.61) (

0.128

0.192) (

0.233

0.51) (

0.125

0.44) (

0.037

0.26)

Hungary (0.95

0.81) (

0.25

0.52) (

0.39

0.59) (

0.42

0.86) (

0.25

0.87) (

0.13 0.94)

Egypt (0.85

0.73) (

0.77

0.92) (

0.11

0.17) (

0.33

0.44) (

0.0145

0.51 ) (

0.071

0.13)

Jordan (0.87

0.74) (

0.63

0.59) (

0.22

0.34) (

0.26

0.29) (

0.053

0.18) (

0.028

0.056)

South Africa (0.66

0.57) (

0.50

0.60) (

0.23

0.35) (

0.43

0.55) (

0.031

0.059) (

0.030

0.11)

Turkey (0.92

0.80) (

0.52

0.58) (

0.23

0.35) (

0.34

0.41) (

0.007

0.025) (

0.030

0.057)

China (0.82

0.70) (

0.46

0.47) (

0.178

0.26) (

0.34

0.42) (

0.04

0.15) (

0.15 0.29) India (0.70

0.59) (

0.60

0.58) (

0.26

0.40) (

0.50

0.57) (

0.030

0.12) (

0.030

0.05)

Indonesia (0.62

0.53) (

0.6

0.58) (

0.21

0.32) (

0.42

0.44) (

0.06

0.20) (

0.020

0.034)

Koera (0.82

0.80) (

0.33

0.32) (

0.174

0.28) (

0.33

0.36) (

0.08

0.42) (

0.04 0.21)

Malaysia (0.6

0.52) (

0.40

0.41) (

0.21

0.32) (

0.42

0.44) (

0.027

0.16) (

0.015

0.039)

Philippines (0.76

0.65) (

0.30

0.32) (

0.23

0.35) (

0.46

0.57) (

0.079

0.29) (

0.066

0.14)

Taiwan (0.82

0.70) (

0.31

0.32) (

0.20

0.30) (

0.44

0.25) (

0.095

0.32) (

0.077

0.19)

Thailand (0.94

0.80) (

0.49

0.42) (

0.23

0.35) (

0.30

0.33) (

0.14

0.49) (

0.014

0.009)

Hypothesis test of global and regional financial integration:

Table 2 summarizes the results of the univariate model in equation (1) to (5) country by country. The first two columns report the beta estimated. We estimate our model using the quasi-maximum likelihood (QMV) method of Bollerslev and Wooldridge (1992). The estimator (QMV) follows the normal distribution. Our estimation is performed by the algorithm (BHHH) developed by Berndt et al. (1974).

The emerging countries of Latin America, the American beta is statically significant for all the markets except Mexico and Peru. The regional beta is statically significant except for Colombia.

Concerning Emerging countries in the Middle East region, the US index beta, is statically significant for all countries except Egypt. However, the regional beta is not significant except for Turkey. For the emerging countries of Europe region, the American beta and the regional beta are statically significant except for the Czech Republic. For Hungary and Greece the regional and global beta is not significant.

The third column of Table 2 summarizes the result of the Wald test of the hypothesis:𝐻0: 𝐾𝑖= 𝛽𝑖,𝑡𝑢𝑠= 0. Our model refers to the regional CAPM using as a benchmark portfolio the regional market index. Estimations show that the regional beta is significant for all Latin American countries except Chile; Two European countries (the Czech Republic and Greece); all countries in the Middle East except Turkey and all Asian countries except Malaysia. The significant regional beta is a perfect regional integration phenomenon. The Wald test is rejected for countries with a significant regional beta. However, for Chile, Hungary, Turkey and Malaysia the regional beta is not significant and the regional integration hypothesis is rejected.

The fourth column in Table 2 summarizes the result of the Wald test of the hypothesis: 𝐻0: 𝐾𝑖= 𝛽𝑖,𝑡 𝑟𝑒𝑔

= 0.

Our model refers to MEDAF International using the benchmark portfolio as the global market index. The global beta is significant for all countries except Hungary, Turkey and three Asian countries (China, Malaysia and Taiwan). The trade with the rest of the word is important for these countries.

(7)

Table 2: Global and regional integration

Market βi,tus βi,t

reg

Ki= βi,tus= 0 Ki= βi,t

reg = 0

Brazil (0.97∗

0.38) (

0.24 0.28)

−0.237 [0.000]

0.15 [0.000] Chile (0.19∗∗∗

0.007) (

−0.02

0.86)

0.54 [0.97]

0.31 [0.821] Colombia (1.22∗∗∗

0.003) (

0.49∗∗∗

0.01 )

4.02 [0.000]

9.1 [0.000]

Mexico (0.35

0.25) (

0.058

0.84)

2.5∗∗∗ [0.000]

0.153 [0.000]

Peru (0.139

0.24) (

0.023

0.99)

0.349 [0.000]

0.023 [0.000] Czech Republic (0.73∗∗∗

0.0004) (

2.18∗∗∗

0.002)

131 [0.000]

0.73 [0.000]

Greece (0.63

0.71) (

0.23 0.55)

1.83 [0.000]

102 [0.000]

Hungary (0.91

0.63) (

0.029

0.25)

0.311 [0.857]

0.091 [0.725]

Egypt (0.031

0.056) (

0.29 0.37)

0.86 [0.155]

0.608 [0.099] Jordan (1.25∗∗∗

0.013) (

0.14∗∗∗

0.003)

−0.054 [0.000]

0.23 [0.079] South Africa (0.29∗

0.148) (

0.20 0.22)

0.36 [0.000]

3.89 [0.000] Turkey (0.22

0.47) (

0.26 0.57)

0.63 [0.884]

0.38 [0.318] China (0.49

1.18) (

0.75∗∗∗

0.18 )

0.76 [0.255]

0.022 [0.785] India (0.59∗∗∗

0.0001) (

1.28∗∗∗

0.0002)

1.2 [0.000]

8.04 [0.000]

Indonesia (0.24

0.19) (

0.50∗∗∗

0.13 )

4.47 [0.004]

0.246 [0.000]

Koera (0.76

0.03) (

0.29 0.37)

43.25 [0.000]

0.093 [0.000]

Malaysia (0.004

0.47) (

0.47 0.86)

0.79 [0.282]

0.47 [0.833] Philippines (−0.05

0.26) (

6.79∗∗∗

0.87 )

7.67 [0.000]

−0.14 [0.000] Taiwan (−0.108

0.098) (

0.65 0.90)

−21.30 [0.000]

0.528 [0.403]

Thailand (0.52∗

0.25) (

0.047

0.13)

91.70 [0.000]

0.080 [0.039] ***,** , * Indicate that the coefficients are significant at the 1%, 5%, and 10% levels, respectively.

We study patterns in global and regional integration by examining how the estimated betas, correlations and ratio of variance change during particular periods. We consider four sub-periods. Dt: an indicator variable marking the five sub-periods. We have divided the total period (1994: 5 - 2014: 4) as; the subprime crisis (2007: 7- 2009: 8), before the subprime crisis (1994: 5-2006: 12) and Post-crisis subprime period (2010: 1-2014: 4). The indicator variable takes the value 1 during the indicated period and 0 other. We estimate our model for four regions: Latin America; Europe; Middle East and Asia.

Panel A in Table 3 summarizes the results of estimations of betas, correlations, and variance ratios for the total period. For most countries, betas, correlations and variance ratios for the US market are higher than the regional market. This increase can be explained by the increasing links between the different countries.

Panel B in Table 3 reports the results of beta estimations, correlations and variance ratios during the subprime crisis. During this period, the betas, correlations and variance ratio with respect to the US market increase by more than in the regional countries for the different regions. The movements in US market increase due to global integration during the subprime crisis.

Panel C in Table 3 summarizes the results of estimations of betas, correlations, and variance ratios before the subprime crisis. The betas, correlations and variance ratios with respect to the US increase significantly before the subprime crisis period for three regions: Latin America, Asia and the Middle East. In Europe, The regional beta is higher than the global beta. In addition, our model does not suggest any change in correlation, variance ratio and beta before the deletion crisis. We explain the increase in American market movements by global integration during this period with the different regions.

(8)

market are low due to the regional integration of the Middle East region before the subprime crisis. Consequently, the regional effects are evident in Latin America.

Table 3: Regional Integration

Panel A : Total period

Coun try Group

βi,t−1us βi,t−1

reg

βi,t−1reg − βi,t−1us

ρi,us ρi,reg ρi,reg

− ρi,us

VRi,tus VRi,t

reg

VRi,treg − VRusi,t Midd

le East ( 0.001

0.57) (

0.003

0.06) (

0.0004

0.35 ) (

0.013

0.005) (

0.057

0.98) (

0.007

0.098) (

0.013

0.011) (

0.018

0.003) (

0.138

0.114)

Latin America (

0.006

0.067) (

0.003

0.06) (

0.055

0.12) (

0.038

0.018) (

0.032

0.51) (

0.001

0.051) (

0.045

0.04) (

0.005

0.024) (

0.040

0.047)

Euro

pe (

0.068

0.23) (

0.063

0.22) (

0.017

0.12) (

0.34

0.14) (

0.05

0.84) (

0.087

0.25) (

0.086

0.51) (

0.011

0.11) (

−0.07

0.50)

Asia (0.052

0.049) (

0.0002

0.03 ) (

0.051

0.049) (

0.071

0.031) (

0.005

0.43) (

−0.066

0.43 ) (

0.085

0.070) (

0.016

0.02) (

0.069

0.123)

Panel B : Subprime Crise

Coun try Group

βi,t−1us βi,t−1

reg

βi,t−1reg − βi,t−1us

ρi,us ρi,reg ρi,reg

− ρi,us

VRi,tus VRi,t

reg

VRi,treg − VRusi,t Midd

le East ( 0.15

0.125) (

0.031

0.33) (

−0.120

0.32 ) (

0.41

0.070) (

0.0004

0.24 ) (

−0.042

0.33 ) (

0.23

0.20) (

0.11

0.21) (

−0.11

0.33)

Latin America (

0.32

0.29) (

0.11

0.107) (

−0.21

0.19) (

0.69

0.23) (

0.29

0.40) (

−0.203

0.25 ) (

0.11

0.08) (

0.014

0.058) (

0.053

0.049)

Euro

pe (

0.107

0.56) (

0.069

0.197) (

−0.038

0.37 ) (

0.26

0.79) (

0.11

0.53) (

−0.151

0.33 ) (

0.16

0.72) (

0.003

0.012) (

−0.017

0.72 )

Asia (0.24

0.22) (

0.03

0.04) (

−0.24

0.21) (

0.38

0.16) (

0.015

0.66) (

−0.038

0.70 ) (

0.43

0.36) (

0.053

0.091) (

−0.037

0.40 )

Panel C : Before Subprime Crise

Coun try Group

βi,t−1us βi,t−1

reg

βi,t−1reg − βi,t−1us

ρi,us ρi,reg ρi,reg

− ρi,us

VRi,tus VRi,t

reg

VRregi,t − VRi,tus Midd

le East ( 0.03

0.037) (

0.0002

0.0002) (

−0.03

0.037) (

0.17

0.07) (

0.001

0.017) (

−0.169

0.073) (

0.033

0.025) (

0.003

0.010) (

−0.03

0.028)

Latin America (

0.060

0.064) (

0.006

0.009) (

−0.054

0.069) (

0.191

0.095) (

−0.047

0.65 ) (

−0.238

0.65 ) (

0.045

0.041) (

0.0036

0.021) (

−0.041

0.044)

Euro

pe (

0.002

0.002) (

0.043

0.051) (

0.041

0.049) (

0.117

0.305) (

0.004

0.408) (

−0.11

0.427) (

0.139

0.758) (

0.003

0.042) (

−0.135

0.075)

Asia (0.056

0.051) (

0.023

0.021) (

−0.032

0.030) (

0.258

0.104) (

0.074

0.429) (

−0.183

0.457) (

0.194

0.144) (

0.010

0.029) (

−0.183

0.152)

Panel C : Post-crisis subprime period

Coun try Group

βi,t−1us βi,t−1

reg

βi,t−1reg − βi,t−1us

ρi,us ρi,reg ρi,reg

− ρi,us

VRi,tus VRi,t

reg

VRregi,t − VRi,tus Midd

le East ( 0.043

0.034) (

0.76

0.068) (

0.032

0.68) (

0.09

0.04) (

0.103

0.48) (

0.014

0.50) (

0.048

0.040) (

0.066

0.26) (

0.017

0.27)

ASI

A (

0.033

0.029) (

0.014

0.012) (

−0.019

0.017) (

0.28

0.124) (

−0.03

0.45) (

−0.28

0.45) (

0.230

0.194) (

0.051

0.198) (

−0.178

0.313)

Euro

pe (

0.079

0.064) (

0.036

0.030) (

−0.042

0.036) (

0.022

0.042) (

0.029

0.6 ) (

0.007

0.59) (

0.077

0.28) (

0.001

0.008) (

−0.015

0.055)

Latin America (

0.083

0.11) (

0.085

0.109) (

0.002

0.015) (

0.192

0.088) (

−0.038

0.33 ) (

−0.23

0.034) (

0.44

0.38) (

0.006

0.015) (

−0.437

0.38 )

America; Asia and Europe during before the subprime crisis, but they are smaller that are dominated by the increase in movements with the US market.

Hypothesis test of financial contagion:

Table 4 summarizes the correlation between the country’s idiosyncratic shocks with the U.S. residuals, the regional residuals and country’s idiosyncratic residuals.

All Latin American countries have excess correlation with the U.S residual. There are signs of American contagion in these markets. Similarly, the Brazil and Mexico have excess correlation with the regional residual. There is evidence of regional market contagion on the countries of Latin America.

The Europe Greek countries have excess correlation with the U.S residual but the Czech Republic is the only that the market has excess correlation with the regional residual. There is regional contagion to its Czech Republic market.

(9)

Middle East (Egypt and Jordan) and three countries in Asia (Indonesia, Taiwan and Thailand) have excess correlation with the regional residual. In other words, the residual correlation of these countries with the regional market is statically significant at the 10% threshold.

Table 4: Correlations of Market Residuals

Market 𝑒𝑖,𝑡 and 𝑒𝑟𝑒𝑔,𝑡 𝑒𝑖,𝑡 and 𝑒𝑢𝑠,𝑡

Brazil 0.186∗∗∗ −0.398∗∗∗

Chile 0.104 0.234∗∗∗

Colombia 0.068 0.209∗∗

Mexico 0.159∗ 0.171∗∗

Peru 0.074 0.223∗∗∗

Czech Republic 0.144∗ −0.008

Greece 0.046 0.157∗

Hungary 0.245 −0.067

Egypt 0.231∗∗∗ 0.175∗∗

Jordan 0.191∗∗∗ 0.312∗∗∗

South Africa −0.008 −0.336∗∗∗

Turkey 0.164 0.226∗∗∗

China 0.148 0.389∗∗∗

India 0.221 0.356∗∗∗

Indonesia 0.188∗∗∗ 0.566∗∗∗

Koera 0.136 0.306∗∗∗

Malaysia 0.028 0.356∗∗∗

Philippines 0.199 0.259∗∗∗

Taiwan 0.267∗∗∗ 0.399∗∗∗

Thailand 0.155∗ 0.589∗∗∗

***,** , * Indicate that the coefficients are significant at the 1%, 5%, and 10% levels, respectively.

Table 5 presents the results of tests of the contagion hypothesis. In fact, we test the significance of v0 and v1. Test the contagion during the four periods studied: The Subprime crisis (2007: 7- 2009: 8); After the Subprime crisis of Delete (2010: 01-2014: 04) and Before the Subprime crisis (1994: 04- 2006: 12). The coefficient v1 measures the increase in residual correlations of different zones studied with the benchmark (world market) and with the regional market during the sub-periods studied. The nullity of the coefficients v0 and v1[H0: v0= v1= 0] is a sign of contagion.

Panel A in Table 5 summarizes the estimations the contagion assumption during the Subprime crisis. During this period, the coefficient 𝑣1 is positive in all. By using global or regional index v1 is positive, this implies the idiosyncratic residuals are more correlated in the Subprime crisis of the sample.

In addition, the correlations with the respect to the regional index is significantly higher in the Middle East markets; Asian and Latin American markets. However, the hypothesis of the absence of a contagion test during the Subprime crisis [𝐻0: 𝑣0= 𝑣1= 0] is rejected for all regions studied when using the global index. The hypothesis of the absence of contagion is also rejected in the regions of the Middle East; Asians and Latin America when considering the residues of the regional indices. Contagion effects during the global financial crisis are detected in almost all regions.These findings are consistent with those of the earlier literature that appropriately capture integration results that vary over time and show signs of contagion during the Subprime crisis, Baur (2012); Bekaert et al. (2011); Al. (2011), Baur (2012) and Dungey A Gajurel (2013), Bekaert et al. (2014); Long, Albert, Tsui and Zhang (2014); Syriopoulos et al. (2015).

Table 5: Cross-section Analysis of Market Residuals

U. S. Residual Returns (êus,t) Regional Residual Returns (êreg,t)

𝑣0 𝑣1

𝑇𝑒𝑠𝑡 𝑑𝑒 𝑊𝑎𝑙𝑑

𝑣0 𝑣1

𝑇𝑒𝑠𝑡 𝑑𝑒 𝑊𝑎𝑙𝑑

𝜋𝑖= 0 𝑝 = 𝑞 = 0 𝜋𝑖= 0 𝑝 = 𝑞 = 0

Panel A : Subprime crise Middle East −1.37∗∗∗

(0.06)

0.77∗∗∗ (0.20)

0.61 [0.433]

504.2 [0.000]

−0.22∗∗∗

(0.059)

0.33∗∗∗ (0.09)

0.28 [0.596]

16.82 [0.000]

Asia −2.23∗∗∗

(0.008)

0.23∗ (0.13)

6.9 [0.008]

763.9 [0.000]

−2.68∗∗∗

(0.063)

1.28∗∗∗ (0.78)

2.8 [0.998]

182 [0.000] Latin America 1.57∗∗∗

(0.072)

0.37∗ (0.22)

0.081 [0.774]

474.5 [0.000]

−0.0001∗∗∗

(0.0001)

0.004∗∗∗ (0.0008)

134.6 [0.000]

80.9 [0.000]

Europe −1.31∗∗∗

(0.058)

0.71∗∗∗ (0.035)

0.27 [0.603]

837.7 [0.000]

−0.039 (0.067)

0.025 (0.2)

0.0001 [0991]

0.342 [0.842] Panel B : Before Subprime crise

Middle East −1.88∗∗∗ (0.05)

1.88∗ (0.929)

0.03 [0.861]

116 [0.000]

−1.04∗∗∗

(0.038)

0.419 (1.25)

0.052 [0.818]

0.44 [0130]

Asia −2.57∗∗∗

(0.033)

0.196∗∗∗ (0.029)

0.457 [0.499]

661 [0.000]

−2.85∗∗∗

(0.052)

0.82∗∗∗ (0.091)

0.015 [0.900]

296 [0.000] Latin America 0.20∗∗∗

(0.0004)

0.320∗∗∗ (0.019)

0.049 [0.823]

260 [0.000]

0.88∗∗∗ (0.009)

−0.069∗

(0.041)

0.649 [0.420]

(10)

Europe −2.82∗∗∗ (0.027)

0.146 (0.198)

0.650 [0.420]

112 [0.230]

−0.04 (0.04)

0.04 (0.97)

0.0006 [0.979]

1.39 [0.498]

Panel C : After Subprime crise Middle East −1.28∗∗∗

(0.053)

0.95 (0.27)

0.253 [0.614]

574 [0.000]

−0.276∗∗

(0.002)

0.78∗∗ (0.26)

4.12 [0.423]

166 [0.000]

Asia −0.005

(0.032)

0.041 (0.28)

2.26 [0.996]

0.24 [0.988]

−2.85∗∗∗

(0.052)

0.82∗∗∗ (0.091)

0.015 [0.901]

296 [0.000] Latin America 1.59∗∗∗

(0.074)

0.102∗∗ (0.031)

0.061 [0.804]

467 [0.000]

4.01∗∗∗ (0.023)

−0.309∗

(0.129)

0.094 [0.758]

313 [0.000]

Europe −1.59∗∗∗

(0.058)

1.03 (1.34)

0.357 [0.55]

742 [0.200]

−0.027 (0.055)

0.026 (1.87)

3.40 [0.995]

0.256 [0.879] ***,** , * Indicate that the coefficients are significant at the 1%, 5%, and 10% levels, respectively.

0.0E+00 5.0E-08 1.0E-07 1.5E-07 2.0E-07 2.5E-07 3.0E-07

1994 1996 1998 2000 2002 2004 2006 2008 2010 2012

the conditional variance behavoir over time

0.0E+00

5.0E-08 1.0E-07 1.5E-07 2.0E-07 2.5E-07 3.0E-07

1994 1996 1998 2000 2002 2004 2006 2008 2010 2012

the conditional varaaice behavoir over time

Africa / Middle East Asia

0.0E+00 5.0E-08 1.0E-07 1.5E-07 2.0E-07 2.5E-07 3.0E-07

1994 1996 1998 2000 2002 2004 2006 2008 2010 2012

The conditional variance behavior over time

0.0E+00

5.0E-08 1.0E-07 1.5E-07 2.0E-07 2.5E-07 3.0E-07

1994 1996 1998 2000 2002 2004 2006 2008 2010 2012

The conditional variance behavior over time

Latin America Europe

Fig. 1: The conditional variance behavior over time

Fig 1 plots the evolution of time-varying conditional variance dynamics among the U.S and regional markets. The chart shows strong signs of volatility for the four regions during the course of the 2007-2009 global financial crisis.

Conclusion:

(11)

Czech Republic and Hungary). For the second method we adopted the definition of Bekaert et al. (2005) who defined the contagion as an excess of correlations, not explained by the fundamentals. More precisely, we tried to find out the variation in contagion during the three studied periods: the Subprime crisis (2007: 7- 2009: 8); after the crisis of Subprime (2010: 01-2014: 04) and before the crisis of the Subprime (1994: 04- 2006: 12). The main results do not allow us to reject the hypothesis of contagion during the crisis of Subprime and after the crisis of Subprime.

As recommendations and from policymakers perspective, this study provides important information about the direction for possible undertaking measures in order to protect emerging markets from contagion during future crises.

REFERENCES

Aloui, R., M.S.B. Alissa and D.K. Nguyen, 2011. Global financial crisis, extreme interdependence, and contagion effects: The role of economic structure? Journalof Banking and Finance, 35: 130-141.

Baele, L., 2005. Volatility spillover effects in European equity markets. Journal of Financial and Quantitative Analysis, 40: 373-401.

Baele, L and K. Inghelbrecht, 2010. Time-varying integration, interdependence and contagion. Journal of International Money and Finance, 29: 791-818.

Baur, D., 2012. Financial contagion and the real economy. Journal of Banking and Finance, 36: 2680-2692. Bekaert, Geert, Michael Ehrmann, Marcel Fratzscher and Arnaud Mehl, 2014. The Global Crisis and Equity Market Contagion, Journal of Finance, 69: 2597-2649.

Bekaert, G., M. Ehrmann, M. Fratzscher and A. Mehl, 2011. Global crises and equity market contagion. NBER Working pp: 17121.

Bekaert, G. and C.R. Harvey, 1997. Emerging equity market volatility. Journal of Financial Economics, 43: 29-77.

Bekaert, G., C.R. Harvey and A. Ng, 2005. Market integration and contagion. Journal of Business, 78: 39-69.

Cho, S., S. Hyde and N. Nguyen, 2015. Time-varying regional and global integration and contagion: Evidence from style portfolios. International Review of Financial Analysis, 42: 109-131.

Dungey, M. and D. Gajurel, 2013. Equity market contagion during the global financial crisis: Evidence from the world’s eight largest economies. University of Tasmania School of Economics and Finance Discussion 2013-15.

Dimitriou, D., D. Kenourgios and T. Simos, 2013. Global financial crisis and emerging stock market contagion: A multivariate FIAPARCH–DCC approach. International Review of Financial Analysis, 30: 46-56.

Dooley, M., M. Hutchison, 2009. Transmission of the U.S. subprime crisis to emerging markets: Evidence on the decoupling-recoupling hypothesis. Journal of International Money and Finance, 28: 1331-1349.

Guyot, A., T. Lagoarde-Segot, S. Neaime, 2014. Foreign shocks and international cost of equity destabilization: Evidence from the MENA region. Emerging Markets Review, 18: 101-122.

Long, L., A.K. Tsui and Z. Zhang, 2014. Estimating time-varying currency betas with contagion: New evidence from developed and emerging financial markets. Japan and the World Economy, 30: 10-24.

Maghyereh, A.I., B. Awartani, K. Al Hilu, 2015. Dynamic transmissions between the U.S. and equity markets in the MENA countries: New evidence from pre- and post-global financial crisis. The Quarterly Review of Economics and Finance., 56: 123-138.

Figure

Table 1: Implicit statistics of the country specific model  Market

Table 1.

Implicit statistics of the country specific model Market . View in document p.6
Table 2: Global and regional integration Market

Table 2.

Global and regional integration Market . View in document p.7
Table 3:  Regional Integration                                                                                                  Panel A : Total period

Table 3.

Regional Integration Panel A Total period . View in document p.8
Table 4: Correlations of Market Residuals Market

Table 4.

Correlations of Market Residuals Market . View in document p.9
Table 5: Cross-section Analysis of Market Residuals  U. S. Residual Returns (êus,t)

Table 5.

Cross section Analysis of Market Residuals U S Residual Returns e us t . View in document p.9
Fig. 1: The conditional variance behavior over time

Fig 1.

The conditional variance behavior over time . View in document p.10