Stock market interdependence, contagion, and the U.S. financial crisis: The case of emerging and frontier markets

Int. Fin. Markets, Inst. and Money 21 (2011) 724–742

Contents lists available at ScienceDirect

Journal of International Financial Markets, Institutions & Money j ou rn al ho me pa ge : w w w . e l s e v i e r . c o m / l o c a t e / i n t f i n

Stock market interdependence, contagion, and the U.S. financial crisis: The case of emerging and frontier markets夽 Lalith P. Samarakoon ∗ Department of Finance, Opus College of Business, University of St. Thomas, 2115 Summit Ave., St. Paul, MN 55105, United States

a r t i c l e

i n f o

Article history: Received 23 December 2010 Accepted 6 May 2011 Available online 14 June 2011 JEL classification: F3 F30 F36 G01 G15 Keywords: Interdependence Contagion U.S. financial crisis Emerging markets Shock models

a b s t r a c t This paper examines transmission of shocks between the U.S. and foreign markets to delineate interdependence from contagion of the U.S. financial crisis by constructing shock models for partially overlapping and non-overlapping markets. There exists important bi-directional, yet asymmetric, interdependence and contagion in emerging markets, with important regional variations. Interdependence is driven more by U.S. shocks, while contagion is driven more by emerging market shocks. Frontier markets also exhibit interdependence and contagion to U.S. shocks. Except for Latin America, there is no contagion from U.S. to emerging markets. But there is contagion from emerging markets to the U.S. © 2011 Elsevier B.V. All rights reserved.

1. Introduction The recent U.S. financial crisis, particularly the severity with which it gripped the markets and economies around the world, was one of the most unanticipated and tumultuous economic events in the recent history. The decline in the U.S. stock market began in late 2007, which was quickly followed by declines in both emerging and frontier markets. During the most turbulent episode of the meltdown

夽 I wish to thank the participants of the 2011 Global Finance Conference in Bangkok, Thailand, an anonymous referee, and Thadavillil Jithendranathan, Dobrina Georgieva, Jeffrey Oxman, and Mufaddal Baxamusa for useful comments. ∗ Tel.: +1 651 962 5444; fax: +1 651 962 5093.

E-mail address: [email protected] 1042-4431/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.intfin.2011.05.001

L.P. Samarakoon / Int. Fin. Markets, Inst. and Money 21 (2011) 724–742

725

that lasted for about 6 months from September 2008 to early March 2009, the U.S. stock market plummeted by 43%, the emerging markets by 50%, and frontier markets by 60%. Do these declines in stock markets around the world during the U.S. financial crisis provide evidence of contagion? If contagion exists during the U.S. crisis, then what is the magnitude of such contagion? How is the contagion during this crisis period different from the transmission of shocks during relatively tranquil periods? The motivation of this paper is to answer these questions by investigating the propagation of return shocks between the U.S. and emerging and frontier stock markets. To answer these questions, this paper develops a framework for estimating the impact of shocks during normal times, i.e. interdependence, and the impact during crisis, i.e. contagion, and implements the models using a comprehensive sample of emerging and frontier markets. The methodology of this paper proceeds as follows. First, unexpected returns or return shocks are calculated by specifying an autoregressive model of returns allowing for time-variation of expected returns for the U.S. market and for each emerging and frontier market studied. Second, U.S. return shocks are related to return shocks in another market, and vice versa, employing the vector auto regressions (VAR) framework. To accommodate differences in trading hours across markets, stock markets are classified into two groups as partially overlapping and non-overlapping markets. Accordingly, two separate shock models, called partially overlapping shock model and non-overlapping shock model, are developed for each market type. In the two models, the interdependence is captured by the coefficient on shocks, and the contagion is measured by the interaction of the U.S. crisis dummy with return shocks. The models are estimated using daily index returns data for 62 stock markets for period from 2000 to 2009. This paper contributes to the literature in a number of important ways. The first major contribution is to develop a straightforward framework for distinguishing between cross-market interdependence and contagion. Unlike most previous studies, cross-market interdependence and contagion is framed on the basis of time-varying return shocks rather than correlation or volatility. The relation between return shocks of one market with another is formulated using the VAR methodology with particular consideration of the differences in trading hours across markets and the need to distinguish between interdependence and contagion. The resulting two models – partially overlapping and nonoverlapping shock models – allow for a clear separation of the transmission of shocks during times of stability vs. crisis, enabling a complete understanding of the propagation of shocks in international markets. The second major contribution of this paper is to provide empirical evidence on the degree of interdependence and contagion between the U.S. and emerging and frontier markets during the U.S. financial crisis. Different from most previous work, interdependence and contagion due to shocks from the U.S. to emerging and frontier markets as well as contagion due to shocks from emerging to the U.S. are investigated. There is very little published work that examines the issue of interdependence and contagion associated return shocks generated during the U.S. financial crisis in international markets. Particularly important is the study of emerging and frontier markets, which have become an increasingly important asset class for investors in international portfolio diversification. The rest of this paper is organized as follows. Section 2 provides an overview of related literature. Section 3 presents data and summary statistics. Section 4 outlines the methodology. The empirical results are discussed in Section 5. Section 6 provides the summary and conclusions.

2. Overview of related literature One approach used in the previous literature to study contagion is to estimate cross-market correlations between stable vs. crisis periods. An increase in correlation during a crisis relative to a stable period is interpreted as evidence of contagion (for example, King and Wadhwani, 1990; Lee and Kim, 1993). These studies find considerable evidence of increases in cross-market correlations during relatively more volatile periods, suggesting contagion. However, some argue that the heteroscedasticity problem caused by an increase in market volatility during crisis periods biases estimated correlations (Forbes and Rigobon, 2002), contagion must involve a dynamic increase in return correlations (Pesaran and Pick, 2007), and that there exists an omitted variable problem in the estimation of cross-country

726

L.P. Samarakoon / Int. Fin. Markets, Inst. and Money 21 (2011) 724–742

correlations (Chiang et al., 2007). As a result, some have used alternative techniques to overcome some limitations associated with correlation analysis.1 Instead of simply using cross-market correlation of returns, this paper develops a vector autoregressions (VAR) model of time-varying unexpected returns with a crisis dummy and an interaction of the crisis dummy with return shocks to delineate effects of interdependence from contagion. The model follows the spirit of Bekaert et al. (2005) which also uses a similar model based on idiosyncratic shocks and estimates contagion as the incremental impact of idiosyncratic shocks of one market on the other. In fact, the current paper uses time-varying shocks, which are computed from rolling regressions of a conditional expected return model with autoregressive parameters. When time-varying shocks are combined with the crisis dummy to estimate the incremental impact of the crisis on return shocks, the contagion coefficient captures the dynamic increment in comovements of return shocks across markets. In addition to estimating return shocks after allowing for lagged effects of own market returns, the model also allows for any potential autocorrelations of a market’s own return shocks. This provides sufficient flexibility for effects of all market-related events including changes in fundamental economic variables and investor risk perceptions to be embedded in the expected returns or return shocks, thus minimizing potential omitted variable concerns. Previous studies of cross-market transmission of shocks based on VAR and ARCH-type models find mixed evidence. For example, Eun and Shim (1989) find that return innovations in the U.S. affect major developed markets, and not vice versa. King and Wadhwani (1990) find that contagion among New York, London, and Tokyo rose during and immediately after the October 1987 U.S. market crash. Hamao et al. (1990) show evidence of volatility spillovers from New York to Tokyo, London to Tokyo, and New York to London, but not in other directions. They also find that the effect of a volatility surprise becomes less pervasive across markets when they exclude the post-October 1987 period from the sample, suggesting that volatility spillovers are more pronounced during the market crisis. Lin et al. (1994) report a bi-directional relation between unexpected daytime returns and overnight returns between Tokyo and New York. Karolyi and Stulz (1996) find that large shocks to Japanese and U.S. stock markets positively impact both the magnitude and the persistence of return correlations. Baele (2005) also finds evidence of contagion from the U.S. to a number of European equity markets during periods of high world market volatility. Bekaert et al. (2005) find no contagion between the U.S. and countries in Europe, Asia and Latin America caused by the Mexican crisis. As far as studies that examine the U.S. financial crisis, Dooley and Hutchison (2009) show that lagged U.S. returns are related to Mexican returns over the full period that includes the period before the U.S. financial crisis and the period of crisis, and find some evidence of this relation being influenced by the crisis period. Longstaff (2010) finds that subprime asset-backed collateralized debt obligations had a contagion effect on stock returns, Treasury yields, corporate bond spreads, and in the VIX volatility index in 2007. Guo et al. (2011) show that stock market shock and oil price shock were the main drivers behind credit default and stock market variations. Aloui et al. (2011) find evidence of time-varying dependence between Brazil, Russia, India, China and the U.S. during the global financial crisis. Chan et al. (2011) finds a tranquil regime and a crisis regime across financial, commodities, and real estate asset classes. Pesaran and Pesaran (2010) show that changes to asset return volatilities are shared across assets and markets in the 2008 global financial crisis. 3. Data and summary statistics This study uses price indices for the U.S., emerging, and frontier markets, all measured in U.S. dollars terms. All market indices are obtained from Bloomberg, except for Ecuador (from Datastream) and Indonesia and Vietnam (from MSCI). The index for the U.S. market is the Standard & Poor’s 500 Index. The sample and summary statistics are shown in Table 1. Accordingly, this study includes daily

1 See Kenourgios et al. (2011) for a review of advanced methods to measure contagion. For example, papers have used coincidence of extreme return shocks (Bae et al., 2003), ARCH models (Hamao et al., 1990), switching models (Ramchand and Susmel, 1998; Guo et al., 2011), dynamic correlations (Chiang et al., 2007), Markov switching model (Chan et al., 2011; Okimoto, 2008), multivariate Gaussian dynamic conditional correlation model (Pesaran and Pesaran, 2010), and copula approach (Rodriquez, 2007; Okimoto, 2008; Aloui et al., 2011).

Table 1 The sample and summary statistics. Market

Time zone

N

Total market capitalization (USD Mil.)

Before crisis

During crisis

2000/01–2009/03

2515

10,606,275

17

51

1.00

1.00

−39

2000/01–2009/03 2000/01–2009/03 2000/01–2009/03 2000/01–2009/03 2000/01–2009/03 2000/01–2009/03 2000/01–2009/03 2000/01–2009/03

1971 2057 2155 2016 2028 2021 2053 2013

1,775,591 637,281 95,911 186,323 48,543 484,028 386,688 99,005

26 25 23 16 21 29 25 23

44 51 44 21 40 52 38 42

−0.01 0.07 0.04 0.02 0.01 0.10 0.08 0.05

0.07 0.49 0.16 0.14 0.05 0.26 0.15 0.38

−13 −39 −41 −19 −30 −28 −35 −37

2000/01–2009/03

2038

66,082

25

53

−0.05

0.21

−57

2000/01–2009/03 2002/01–2009/03 2000/01–2009/03

2018 1514 2055

100,397 63,219 242,942

19 13 19

44 23 45

0.22 −0.01 0.24

0.25 0.15 0.41

−34 −20 −33

2000/01–2009/03 2000/01–2009/03 2000/01–2009/03 2000/01–2009/03 2000/01–2009/03

2062 2055 2062 2044 2053

44,483 18,939 92,658 265,217 117,025

20 22 20 34 42

61 56 39 96 51

0.20 0.23 0.19 0.14 0.11

0.35 0.47 0.41 0.24 0.46

−57 −51 −46 −51 −40

2000/01–2009/03 2000/01–2009/03 2000/01–2009/03 2000/01–2009/03 2000/01–2009/03

2028 2035 2049 1961 2048

352,258 588,478 130,046 12,634 44,770 266,024

33 28 15 25 18 24

60 67 34 13 53 47

0.27 0.48 0.40 0.03 0.18 0.14

0.62 0.76 0.67 −0.07 0.57 0.33

−43 −31 −15 −8 −50 −35

2000/01–2009/03 2000/07–2009/03

1730 1871

15,400 21,447

24 48

24 80

0.01 −0.07

0.03 0.06

−8 −69

Before crisis

During crisis L.P. Samarakoon / Int. Fin. Markets, Inst. and Money 21 (2011) 724–742

United States Panel A: emerging markets Asia NO China NO India Indonesia NO Malaysia NO Philippines NO NO South Korea Taiwan NO Thailand NO Africa Egypt NO Middle East Israel PO Morocco PO South Africa PO Europe Czech Republic PO PO Hungary Poland PO PO Russia Turkey PO Latin America PO Argentina PO Brazil PO Chile PO Venezuela PO Peru Average Panel B: frontier markets Asia NO Bangladesh NO Kazakhstan

Cumulative return during crisis (%)

Correlation coefficient with the U.S.

Annualized standard deviation (%)

Data period

727

728

Table 1 (Continued) Market

Cumulative return during crisis (%)

Data period

N

Total market capitalization (USD Mil.)

Before crisis

During crisis

Before crisis

During crisis

NO NO NO NO

2000/01–2009/03 2000/01–2009/03 2000/01–2009/03 2006/12–2009/03

1956 2026 1938 352

407 23,334 4315 12,991

88 25 20 22

21 24 27 37

0.00 0.02 −0.02 −0.04

−0.07 0.00 −0.07 0.01

−44 −38 −30 −54

NO NO NO NO NO NO NO

2000/01–2009/03 2000/01–2009/03 2000/01–2009/03 2000/01–2009/03 2003/12–2009/03 2000/01–2009/03 2000/01–2009/03

1482 1405 1949 1990 1044 2328 1885

3463 2477 10,253 4612 80,375 43,816 5373

13 18 13 12 23 14 9

9 4 22 26 53 16 15

0.05 −0.05 −0.04 0.01 0.20 −0.02 −0.03

0.01 −0.01 0.12 0.02 0.36 −0.06 −0.07

−16 −9 −47 −44 −50 −51 −8

NO NO NO NO NO NO NO NO NO

2000/09–2009/03 2004/07–2009/03 2003/12–2009/03 2000/01–2009/03 2000/01–2009/03 2000/01–2009/03 2000/01–2009/03 2000/01–2009/03 2000/01–2009/03

1762 895 1177 1990 1731 1982 2032 2065 2373

68,812 18,952 63,146 34,245 105,216 8327 14,979 69,055 244,115

17 9 29 16 14 21 13 25 22

36 16 56 29 23 25 39 53 54

−0.03 0.02 0.00 0.01 −0.04 −0.01 −0.03 0.01 0.02

0.19 −0.06 0.07 −0.06 −0.08 0.11 −0.09 0.02 0.46

−46 −41 −67 −40 −55 −41 −49 −58 −50

PO NO PO NO NO NO NO PO NO PO NO

2006/02–2009/03 2000/10–2009/03 2002/06–2009/03 2000/01–2009/03 2000/01–2009/03 2001/12–2009/05 2000/01–2009/03 2004/06–2009/03 2000/01–2009/03 2004/10–2009/03 2000/01–2009/03

531 1834 1970 2081 2107 2088 2036 921 2001 872 1945

5643 6143 26,647 1961 1623 3603 3423 2863 12,672 12,165 6350

26 22 19 16 22 14 13 22 28 13 20

33 47 51 35 42 39 13 54 56 26 20

−0.07 −0.04 0.09 0.06 −0.01 −0.01 −0.02 0.02 0.02 −0.01 0.04

0.11 −0.01 0.53 0.17 0.12 0.15 −0.02 0.06 0.36 0.11 0.06

−57 −74 −60 −51 −58 −62 −26 −37 −65 −62 −25

L.P. Samarakoon / Int. Fin. Markets, Inst. and Money 21 (2011) 724–742

Mongolia Pakistan Sri Lanka Vietnam Africa Botswana Ghana Kenya Mauritius Namibia Nigeria Tunisia Middle East Abu Dhabi Bahrain Dubai Jordan Kuwait Lebanon Oman Qatar Saudi Arabia Europe Bosnia Bulgaria Croatia Estonia Latvia Lithuania Malta Montenegro Romania Serbia Slovakia

Correlation coefficient with the U.S.

Annualized standard deviation (%)

Time zone

Market

Time zone

Data period

N

Total market capitalization (USD Mil.)

Before crisis

During crisis

Slovenia Ukraine Central America Bermuda Costa Rica Jamaica Trinidad and Tobago Latin America Ecuador Average Overall average

NO PO

2000/01–2009/03 2000/01–2009/03

2008 1986

10,904 42,537

12 28

39 54

0.03 0.01

0.19 0.25

−51 −60

PO PO PO PO

2000/01–2009/03 2000/01–2009/03 2000/01–2009/03 2000/01–2009/03

1931 1970 1997 1163

1666 1905 3727 7904

16 19 13 6

16 36 20 8

0.02 −0.03 0.06 0.04

−0.16 −0.08 0.08 −0.05

−36 −31 −23 −26

PO

2000/01–2009/03

1936

4562 25,285 110,708

28 21 22

8 32 37

0.03 0.07 0.16

1 −43 −40

Annualized standard deviation (%)

Correlation coefficient with the U.S. Before crisis

0.01 0.005 0.05

Cumulative return during crisis (%)

During crisis

This table shows the time-zone, data period, number of daily return observations, the total market capitalization (as of year-end 2008 unless noted otherwise), annualized standard deviation before and during the crisis, correlation of daily returns of each market with the U.S. daily returns before and during the crisis, and the cumulative return during the crisis for the U.S. market and for each emerging and frontier market in the sample. Time zones represent partially overlapping (PO) and non-overlapping (NO) markets with respect to the U.S. market. The U.S. market is represented by the S&P 500 Index, and foreign markets are represented by market index in U.S.$ in each market. Data sources are Bloomberg, Datastream and MSCI. n.a. means not available.

L.P. Samarakoon / Int. Fin. Markets, Inst. and Money 21 (2011) 724–742

Table 1 (Continued)

729

730

L.P. Samarakoon / Int. Fin. Markets, Inst. and Money 21 (2011) 724–742

market indices for 62 markets comprising of 22 emerging markets and 40 frontier markets, during the sample period from 01/04/2000 to 03/09/2009. All returns are continuously compounded. The breakdown of markets as emerging vs. frontier is based on the Standard and Poor’s (S&P) classification as of May 2009. The S&P classifies a country as an emerging market if it has a low- or middle-income economy as defined by the World Bank. In addition, the country’s investable market capitalization must be low relative to its most recent GDP figures, and its equity market must exhibit substantial features of emerging markets. The frontier markets are defined as relatively small and illiquid markets, even by emerging market standards and include all the markets that are not classified as emerging markets. For the purpose of the study, markets are classified into two time zones as partially overlapping markets and non-overlapping markets, based on the time each market opens and closes for trading relative to the U.S. market. The partially overlapping markets are the ones which open either before or concurrently with the opening of the U.S. market, partially overlap trading hours with the U.S. market, and close before the closing of the U.S. market. There are 23 markets that partially overlap the trading with the U.S. market, which are mostly located in the European and Latin American regions. Non-overlapping markets open and close before the U.S. market opens for trading, and hence, have no overlapping trading times with the U.S. market. These markets are mainly located in Asia, Africa, the Middle-East, and the Eastern Europe, and the sample includes 39 of such non-overlapping markets.2 4. Methodology 4.1. Defining interdependence and contagion In order to address the research questions, it is necessary to separate the impact of return shocks during normal periods, i.e. interdependence, from incremental effect associated with the crisis period, i.e. contagion. The propagation of shocks across markets is a continuous phenomenon that occurs constantly. During periods of crisis, shocks become larger, and their effects across markets is likely to be different from the effects during relatively stable time periods. Forbes and Rigobon (2002) define co-movements during stable periods as “interdependence” driven by strong linkages among markets. In line with this definition, this paper defines the co-movement of shocks during stable time periods as interdependence. The term interdependence refers to normal co-movements of shocks and provides a baseline to compare the excessive or incremental impact, if any, that shocks might have during a crisis. Edwards (2000) defines contagion as a situation where the extent and the magnitude of the international transmission of shocks exceed what was expected by market participants. Forbes and Rigobon (2002) define contagion as a significant increase in cross-market linkages after a shock to one market. Bekaert et al. (2005) define contagion as excess correlation, which is the correlation over and above what is expected. Consistent with this literature, the present study defines contagion as excessive impact of shocks of one market on another during a period of crisis. Contagion represents the transmission of shocks attributable to the crisis and can be evaluated against interdependence to ascertain the particular impact of shocks of one market on another during the financial crisis.3 4.2. Estimating return shocks The first step is to estimate return shocks or unexpected returns. A return shock is defined as the difference between the actual return and the conditional expected return, based on information available at time t − 1. The use of return shocks rather than raw returns closely follows the spirit of Bekaert et al. (2005), which uses unexpected returns in studying the transmission of shocks during normal and crisis time periods. The conditional expected return is modeled as a function of lagged

2 The markets that trade completely concurrent with the U.S. (Mexico, and Puerto Rico) as well as markets that open before the U.S. and close after the U.S. (Colombia) are excluded from the sample because of the simultaneity bias. 3 See Claessens et al. (2001), Forbes and Rigobon (2001), Forbes and Rigobon (2002), Karolyi (2002), Rigobon (2002), Bae et al. (2003), and Bekaert et al. (2005) for discussion of the issue of interdependence and contagion.

L.P. Samarakoon / Int. Fin. Markets, Inst. and Money 21 (2011) 724–742

731

returns of that market. This specification of expected returns with lags is particularly important to account for infrequent trading that is prevalent in most small markets in the sample. To compute the expected returns, the return series in each market is specified as an autoregressive process up to order three, as shown below.4 rf,t = ˛f +

3 

ˇf,j (rf,t−j ) + εf,t

(1)

j=1

rus,t = ˛us +

3 

ˇus,j (rus,t−j ) + εus,t

(2)

j=1

rf is the foreign market return while rus is the U.S. market return. The subscript t represents the trading day rather than the calendar day. To allow for the time variation in the autoregressive coefficients, the above two equations are estimated using rolling regressions with 240 daily observations per regression. Using the estimated coefficients, return shocks are specified as given in Eqs. (3) and (4) below:

⎡

ef,t = rf,t − ⎣˛ ˆf +

⎡

3 

⎤

ˆ f,j (rf,t−j )⎦ ˇ

j=1

eus,t = rus,t − ⎣˛ ˆ us +

3 

(3)

⎤ ˆ us,j (rus,t−j )⎦ ˇ

(4)

j=1

ef is the return shock in a given foreign market and eus is the return shock in the U.S. market. 4.3. Shock models The second step is to estimate the relation between return shocks in the U.S. and foreign markets and vice versa. This paper accomplishes this by developing two shock models based on the VAR framework. Because of the timing difference in the relation between the U.S. market and foreign markets, two separate models are specified as the partially overlapping shock model and the non-overlapping shock model. Partially overlapping markets are the markets that open either before or concurrently with the opening of the U.S. market, partially overlap trading hours with the U.S., and close before the closing of the U.S. Since these markets close before the U.S. – close, the full impact of shocks originated in the U.S. on a given trading day is likely to be incorporated into stock prices in these foreign markets over a period of two days – the same day and the following day. The transmission of U.S. shocks during the overlap of trading is captured through concurrent interdependence and contagion coefficients, and the potential lagged effect of U.S. shocks is captured through lagged interdependence and contagion coefficients. The model for partially overlapping markets is specified Eqs. (5) and (6) below: ef,t = Af + Bf,1 (ef,t−1 ) + Bf,2 (ef,t−2 ) + Df,t (CDt ) + Cf,t (eus,t ) + Cf,t−1 (eus,t−1 ) + Ff,t (eus,t × CDt ) + Ff,t−1 (eus,t−1 × CDt−1 ) + Vf,t

eus,t = af + bus,1 (eus,t−1 ) + bus,2 (eus,t−2 ) + df,t (CDt ) + cf,t (ef,t ) + ff,t (ef,t × CDt ) + vf,t

(5)

(6)

Eq. (5) estimates the impact of U.S. return shocks on foreign markets after controlling for the possible lagged effects of own-market shocks.5 Cf,t and Cf,t−1 , called interdependence coefficients, 4 5

An AR(3) process was found to be optimal in describing the return series. The number of lags is based on the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC).

732

L.P. Samarakoon / Int. Fin. Markets, Inst. and Money 21 (2011) 724–742

estimate of the concurrent and lagged impact of U.S. shocks on a given partially overlapping foreign market during normal times. As such, these coefficients measure the magnitude of interdependence between the U.S. and any partially overlapping foreign market due to normal linkages between them and capture the baseline level of spillover of shocks from the U.S. to the particular foreign market. The excessive impact of shocks or the contagion that is uniquely attributable to the U.S. financial crisis is captured by the interaction of U.S. return shocks with the crisis dummy (CD). The crisis dummy takes the value of one during the period from 9/01/2008 to 03/09/2009 and zero otherwise. Accordingly, Ff,t and Ff,t−1 represent the magnitude of concurrent and lagged contagion of U.S. shocks to a partially overlapping foreign market. Eq. (6) estimates the effect of shocks from a partially overlapping foreign market on the U.S. market after controlling for the influence of the lagged U.S. return shocks. Since partially overlapping foreign markets close before the close of the U.S., the full information of the foreign shock is known to the U.S. market on the same day. Consequently, such return shocks should be incorporated into U.S. prices on the same day. Therefore, only the concurrent interdependence (cf,t ) and concurrent contagion (ff,t ) of foreign return shocks are incorporated into the model. Next, let us turn to non-overlapping markets. These markets open and close before the U.S. market opens for trading, and hence, have no overlapping trading times with the U.S. Therefore, the transmission of shocks between the U.S. and non-overlapping foreign markets lags by a complete trading day, and there is no contemporaneous channel for the spillovers of the crisis. The model for non-overlapping markets is given in Eqs. (7) and (8) below: ef,t = Af + Bf,1 (ef,t−1 ) + Bf,2 (ef,t−2 ) + Df,t−1 (CDt−1 ) + Cf,t−1 (eus,t−1 ) + Ff,t−1 (eus,t−1 × CDt−1 ) + Vf,t

(7)

eus,t = af + bus,1 (eus,t−1 ) + bus,2 (eus,t−2 ) + df,t−1 (CDt−1 ) + cf,t−1 (ef,t−1 ) + ff,t−1 (ef,t−1 × CDt−1 ) + vf,t

(8)

Eq. (7) describes the effect of U.S. shocks on non-overlapping foreign markets while Eq. (8) shows how non-overlapping foreign markets shocks impact the U.S. market. These equations for non-overlapping markets differ from the ones for partially overlapping markets due to the timing of the interdependence and contagion effects. Since there is no contemporaneous relation between non-overlapping foreign markets and the U.S., the effect of U.S. return shocks should be felt in non-overlapping markets only on the following day of trading in such markets. Similarly, the impact of foreign market shocks should be incorporated into U.S. prices on the following day of trading in the U.S. To capture this lagged effect, Eq. (7) specifies Cf,t−1 , and Ff,t−1 to capture lagged interdependence and lagged contagion with respect to U.S. shocks, and cf,t−1 , and ff,t−1 to measure lagged interdependence and lagged contagion with respect to foreign market shocks. The two models are estimated for each foreign market, each of the two groups classified by trading time and for each region within each group.6 The VAR system is estimated using the Ordinary Least Squares, and standard errors are adjusted for both heteroskedasticity and autocorrelation using the Newey–West technique. 5. Empirical results 5.1. Impact of U.S. shocks on emerging markets Table 2 (Panel A) provides the results relating to the effect of U.S. shocks on partially overlapping emerging markets.7 The concurrent and lagged interdependence coefficients are generally large and statistically significant, providing strong evidence that all partially overlapping emerging markets are influenced by U.S. return shocks during relatively stable times, reflecting long-term linkages. Significant lagged interdependence coefficients, except for Argentina, confirm that U.S. shocks impact these markets on the following trading day as well. The Latin American markets, such as Argentina, Brazil, and Chile, that have a longer overlap of trading with the U.S. compared to partially overlapping markets, exhibit larger concurrent interdependence coefficients. In contrast, the markets with shorter

6 The estimations for multiple markets, called composites, are based on cross-sectional and time-series regressions with fixed effects. 7 For parsimony, the tables show estimates of interdependence and contagion coefficients only.

Region

Market

Interdependence Concurrent

Panel A: partially overlapping markets Morocco Africa South Africa Composite Middle East Israel Europe Czech Republic Hungary Poland Russia Turkey Composite Argentina Latin America Brazil Chile Peru Venezuela Composite Composite

Contagion Lagged

R

Concurrent

2

DW

Lagged

Ct

t-Stat

Ct−1

t-Stat

Ft

t-Stat

Ft−1

t-Stat

−0.05 0.37 0.25 0.27 0.23 0.24 0.33 0.33 0.59 0.32 0.56 0.98 0.44 0.21 0.05 0.46 0.37

(−1.24) (6.37)*** (7.87)*** (6.34)*** (4.87)*** (3.62)*** (7.59)*** (4.90)*** (5.18)*** (12.91)*** (8.46)*** (11.23)*** (11.48)*** (5.88)*** (0.99) (18.82)*** (23.67)***

0.09 0.53 0.36 0.31 0.39 0.43 0.33 0.48 0.70 0.40 −0.02 0.20 0.15 0.18 0.08 0.09 0.27

(2.24)** (10.45)*** (10.77)*** (7.76)*** (8.56)*** (8.14)*** (7.87)*** (9.66)*** (6.42)*** (16.35)*** (−0.20) (2.02)** (3.67)*** (4.87)*** (1.70)* (4.03)*** (18.01)***

0.14 0.25 0.15 0.08 0.36 0.56 0.21 0.15 0.05 0.32 0.25 0.34 0.06 0.58 −0.08 0.23 0.25

(1.57) (1.31) (1.81) (0.75) (2.10)** (2.68)*** (1.27) (0.50) (0.23) (6.19)*** (1.62) (1.35) (0.58) (4.01)*** (−1.26) (2.56)** (5.44)***

0.07 0.23 0.06 −0.01 0.17 0.20 0.16 0.03 −0.33 0.07 0.28 0.27 0.12 0.11 −0.01 0.08 0.08

(1.15) (1.42) (0.62) (−0.05) (1.12) (1.03) (1.04) (0.15) (−1.51) (1.20) (2.53)** (1.99)** (1.91)** (0.83) (−0.14) (1.98)** (2.06)**

0.02 0.28 0.17 0.13 0.21 0.20 0.17 0.12 0.12 0.14 0.13 0.29 0.24 0.19 0.01 0.15 0.14

1.94 1.99 1.98 2.01 1.97 2.06 1.90 2.00 1.90 1.98 1.91 1.92 2.07 2.06 1.83 2.00 1.99

L.P. Samarakoon / Int. Fin. Markets, Inst. and Money 21 (2011) 724–742

Table 2 Impact of U.S. shocks on emerging markets.

733

734

Table 2 (Continued) Region

Market

Interdependence Concurrent

Lagged t-Stat

Ct−1 0.10 0.43 0.38 0.29 0.49 0.64 0.50 0.30 0.36 0.18 0.35

R

Concurrent t-Stat

Ft

2

DW

Lagged t-Stat

(2.27)** (7.71)*** (5.42)*** (10.02)*** (9.07)*** (13.77)*** (12.26)*** (8.01)*** (20.43)*** (3.97)*** (20.89)***

Ft−1

t-Stat

0.17 −0.33 0.03 −0.01 0.07 −0.13 −0.14 −0.14 −0.04 0.35 −0.02

(1.48) (−3.01)*** (0.27) (−0.16) (0.79) (−0.86) (−2.05)** (−1.23) (−0.91) (2.58)** (−0.65)

0.03 0.08 0.06 0.17 0.22 0.14 0.15 0.06 0.07 0.08 0.07

2.04 2.06 2.02 2.09 1.85 2.01 1.97 1.98 2.00 2.05 2.00

This table reports the results relating to the impact of U.S. shocks on emerging market shocks using the partially overlapping shock model (Eq. (5)), and the non-overlapping shock model (Eq. (7)). Partially overlapping shock model: ef,t = Af + Bf,1 (ef,t−1 ) + Bf,2 (ef,t−2 ) + Df,t (CDt ) + Cf,t (eus,t ) + Cf,t−1 (eus,t−1 ) + Ff,t (eus,t × CDt ) + Ff,t−1 (eus,t−1 × CDt−1 ) + Vf,t

(5)

Non-overlapping shock model: ef,t = Af + Bf,1 (ef,t−1 ) + Bf,2 (ef,t−2 ) + Df,t−1 (CDt−1 ) + Cf,t−1 (eus,t−1 ) + Ff,t−1 (eus,t−1 × CDt−1 ) + Vf,t

(7)

The table shows the coefficient estimates and t-statistics for concurrent interdependence, lagged interdependence, concurrent contagion, and lagged contagion. Composite represents all the markets in each panel. The last two columns show the adjusted R2 and the Durbin–Watson statistic. The total sample period is from 01/04/2000 to 03/09/2009, and the crisis period for measuring contagion is from 09/01/2008 to 03/09/2009. * ** ***

Significant at 10%. Significant at 5%. Significant at 1%.

L.P. Samarakoon / Int. Fin. Markets, Inst. and Money 21 (2011) 724–742

Ct Panel B: non-overlapping markets Asia China India Indonesia Malaysia Philippines South Korea Taiwan Thailand Composite Africa Egypt Composite

Contagion

L.P. Samarakoon / Int. Fin. Markets, Inst. and Money 21 (2011) 724–742

735

overlap of trading with the U.S., which are in Africa, Middle East, and Europe, have larger lagged interdependence coefficients. Clearly, consistent with intuition, these results suggest that markets with a longer overlap of trading with the U.S. are more sensitive to concurrent shocks, while markets with a shorter overlap of trading with the U.S. are more sensitive to lagged U.S. shocks. When both the concurrent and lagged effects are combined, the European emerging markets exhibit the highest degree of dependence with respect to U.S. shocks. There is no pervasive evidence of either concurrent or lagged contagion of U.S. shocks to partially overlapping markets in Africa, Middle East, and Europe. However, except for Venezuela, all the Latin American markets exhibit strong evidence of contagion, particularly lagged contagion, of U.S. shocks. Turning to non-overlapping emerging markets (Panel B), there are large and significant effects of U.S. shocks on all of them during normal times. The lagged interdependence coefficient ranges from 0.10 for China to 0.64 for South Korea with an overall coefficient of 0.36 for Asian markets. Surprisingly, China exhibits the lowest degree of dependence with respect to U.S. return shocks during tranquil periods.8 As for contagion, there is contagion from the U.S. to India, Taiwan, and Egypt, although the coefficients are negative for India and Taiwan, suggesting that negative U.S. shocks during the crisis led to positive shocks in these two markets. There is no evidence of contagion to China, and this finding is consistent with the fact that the Chinese stock market declined only by 13%, compared with a 39% drop in the U.S. market, during the financial crisis, and that the increase in the correlation of the Chinese market with the U.S. during the crisis was small (see Table 1). Overall, there is no evidence of contagion from the U.S. to non-overlapping emerging markets. The evidence of interdependence is consistent with the existing literature (for example, Forbes and Rigobon, 2002), which finds broad support for the existence of long-term co-movements or linkages among markets. Unlike Forbes and Rigobon (2002) that show no evidence of contagion in emerging markets, and Bekaert et al. (2005) that find contagion in emerging markets during periods of crisis, the present study shows that contagion from U.S. is important in Latin America, but not in Europe, Asia, Africa, and Middle East. The bottom line is that, in response to U.S. shocks, Latin American emerging markets show strong evidence of both interdependence and contagion, while all other regions show strong evidence of interdependence but not contagion. The main implication of these results is that, except for Latin America, the decline in stock prices globally was induced primarily by the interdependence of foreign markets with the U.S. due to normal linkages and not due to any crisis-induced contagion. The declines in stock markets in Latin America were further exacerbated through contagion effects. Emerging markets have fairly large correlations and normal sensitivity to U.S. shocks, and large declines in stock prices in these markets reflected these dependencies. The essential message is that portfolio diversification into emerging markets does not provide a hedge against U.S. stock market shocks. 5.2. Impact of U.S. Shocks on frontier markets Table 3 shows the results on the impact of U.S. shocks on frontier markets. Return shocks originating from the U.S. have no influence on partially overlapping frontier markets during normal times (Panel A), with the exception of Croatia and Ukraine. While there is evidence of contagion of the U.S. financial crisis to these frontier markets, such evidence is limited to Croatia, Serbia, Ukraine, Bermuda, and Jamaica only. Thus, contagion of U.S. shocks is not large or pervasive across partially overlapping frontier markets. In contrast, there is strong evidence of widespread influence of U.S. shocks on nonoverlapping frontier markets (Panel B). These are the markets that respond to U.S. shocks with a one-day lag. The lagged interdependence coefficients are significant for 16 of the 30 such markets, and such evidence of interdependence is particularly strong and pervasive across the European frontier markets, which exhibit an overall interdependence coefficient of 0.16. This is also consistent with the earlier observation that, among the emerging markets, interdependence is more pronounced in European emerging markets. Further, some Asian, African and Middle Eastern frontier markets also

8 The results are based on the Shanghai Stock Exchange Composite Index. The Shanghai and Shenzhen A Shares Index produced similar results.

Region

Market

736

Table 3 Impact of U.S. shocks on frontier markets. Interdependence Concurrent

Lagged

R

Concurrent

2

DW

Lagged

Ct

t-Stat

Ct−1

t-Stat

Ft

t-Stat

−0.04 0.05 0.01 −0.56 0.04 0.01 0.03 0.01 0.04 −0.01 0.03 0.01 0.01

(−0.43) (0.82) (0.08) (−0.89) (0.76) (0.36) (0.68) (0.30) (1.47) (−0.26) (0.92) (0.11) (0.36)

0.09 0.33 0.01 0.07 0.15 0.16 0.05 −0.02 −0.03 0.01 0.03 0.01 0.06

(1.03) (4.16)*** (0.01) (1.21) (2.64)*** (5.85)*** (1.44) (−0.70) (−0.91) (0.18) (0.81) (0.65) (4.68)***

0.08 0.48 0.10 0.23 0.51 0.31 −0.05 0.07 0.01 −0.01 −0.09 0.01 0.17

(0.57) (3.16)*** (0.88) (2.52)** (2.07)** (5.43)*** (−0.93) (1.13) (0.26) (−0.05) (−1.56) (0.13) (4.64)***

0.07 0.19 0.20 0.08 0.05 0.05 0.08 0.17 −0.02 0.03 0.05 0.81 −0.02 0.06 0.07 0.09 −0.05 0.19 0.04 0.05 0.02 0.02

(1.33) (1.90)* (1.37) (2.02)** (1.10) (0.46) (2.39)** (5.21)*** (−0.63) (0.65) (1.79)* (7.56)*** (−0.33) (2.34)** (4.25)*** (2.13)** (−2.11)** (2.51)** (1.80)* (1.48) (0.57) (1.28)

Ft−1

t-Stat

0.17 −0.06 0.16 0.21 0.06 0.15 −0.09 0.06 0.15 0.01 −0.03 0.04 0.12

(1.56) (−0.45) (1.37) (2.02)** (0.29) (2.99)*** (2.14)** (0.83) (2.79)*** (0.14) (0.76) (1.63) (3.83)***

0.05 0.14 0.01 0.10 0.04 0.07 0.01 0.00 0.02 0.00 0.01 0.01 0.03

1.97 2.06 2.02 2.16 1.98 1.99 2.01 2.11 2.02 2.24 1.85 2.01 2.00

−0.11 −0.09 −0.12 −0.10 0.27 0.30 0.08 0.02 0.01 0.20 0.17 −0.02 0.04 0.09 0.16 0.23 0.15 0.28 0.28 0.07 0.10 0.37

(−1.50) (−0.40) (−0.82) (−1.79) (3.69)*** (2.21)** (1.60) (0.39) (0.26) (2.98)*** (1.89)* (−0.13) (0.63) (1.69) (3.59)*** (2.74)** (3.27)** (2.45)** (6.31)*** (1.24) (1.58) (4.01)***

0.01 0.01 0.00 0.00 0.04 0.12 0.01 0.05 0.04 0.02 0.04 0.20 0.00 0.05 0.03 0.09 0.04 0.09 0.10 0.02 0.02 0.14

1.90 2.03 2.03 1.92 2.13 1.95 2.00 2.04 2.09 2.07 2.06 1.92 2.03 2.06 1.99 1.83 1.92 1.91 1.80 2.03 1.96 1.82

L.P. Samarakoon / Int. Fin. Markets, Inst. and Money 21 (2011) 724–742

Panel A: partially overlapping markets Bosnia Europe Croatia Montenegro Serbia Ukraine Composite Latin America Bermuda Costa Rica Jamaica Trinidad and Tobago Ecuador Composite Composite Panel B: non-overlapping markets Bangladesh Asia Kazakhstan Mongolia Pakistan Sri Lanka Vietnam Composite Botswana Africa Ghana Kenya Mauritius Namibia Nigeria Tunisia Composite Middle East Abu Dhabi Bahrain Dubai Jordan Kuwait Lebanon Oman

Contagion

Table 3 (Continued) Region

Market

Interdependence Concurrent Ct

Composite

Lagged t-Stat

Ct−1 0.05 0.05 0.04 0.10 0.32 0.11 0.21 0.15 0.24 0.02 0.19 0.16 0.10

R

Concurrent t-Stat

Ft

2

DW

Lagged t-Stat

(2.03)** (1.53) (2.72)*** (1.51) (9.46)*** (2.92)*** (6.46)*** (0.38) (4.11)*** (0.51) (6.09)*** (11.81)*** (10.32)***

Ft−1

t-Stat

0.49 0.10 0.18 0.28 0.11 0.13 015 0.02 0.13 0.11 0.34 0.21 0.16

(5.03)*** (0.78) (5.11)*** (2.04)** (1.35) (1.29) (2.11)** (0.44) (0.85) (1.47) (4.25)*** (5.51)*** (6.87)***

0.14 0.01 0.04 0.06 0.16 0.03 0.10 0.02 0.05 0.01 0.18 0.06 0.02

1.94 2.07 1.95 1.88 2.00 1.93 2.02 2.00 1.87 2.00 1.92 1.95 1.99

This table reports the results relating to the impact of U.S. shocks on frontier market shocks using the partially overlapping shock model (Eq. (5)), and the non-overlapping shock model (Eq. (7)). Partially overlapping shock model: ef,t = Af + Bf,1 (ef,t−1 ) + Bf,2 (ef,t−2 ) + Df,t (CDt ) + Cf,t (eus,t ) + Cf,t−1 (eus,t−1 ) + Ff,t (eus,t × CDt ) + Ff,t−1 (eus,t−1 × CDt−1 ) + Vf,t

(5)

Non-overlapping shock model: ef,t = Af + Bf,1 (ef,t−1 ) + Bf,2 (ef,t−2 ) + Df,t−1 (CDt−1 ) + Cf,t−1 (eus,t−1 ) + Ff,t−1 (eus,t−1 × CDt−1 ) + Vf,t

(7)

The table shows the coefficient estimates and t-statistics for concurrent interdependence, lagged interdependence, concurrent contagion, and lagged contagion. Composite represents all the markets in each panel. The last two columns show the adjusted R2 and the Durbin–Watson statistic. The total sample period is from 01/04/2000 to 03/09/2009, and the crisis period for measuring contagion is from 09/01/2008 to 03/09/2009. * ** ***

L.P. Samarakoon / Int. Fin. Markets, Inst. and Money 21 (2011) 724–742

Europe

Qatar Saudi Arabia Composite Bulgaria Estonia Latvia Lithuania Malta Romania Slovakia Slovenia Composite

Contagion

Significant at 10%. Significant at 5%. Significant at 1%.

737

738

L.P. Samarakoon / Int. Fin. Markets, Inst. and Money 21 (2011) 724–742

exhibit significant interdependence with respect to U.S. shocks. As for the contagion of U.S. shocks, 13 of the 30 non-overlapping frontier markets produce significant lagged contagion coefficients. Five of them (Sri Lanka, Vietnam, Kenya, Lebanon and Bulgaria) are markets that do not show evidence of interdependence, suggesting that markets that are generally immune from U.S. shocks in relatively stable periods become vulnerable to shock spillovers during the crisis. Among the regions, contagion from U.S. is most pronounced in the Middle-East. Overall, this evidence of contagion of U.S. shocks to non-overlapping frontier markets sharply contrasts with the earlier finding of no contagion for non-overlapping emerging markets. Overall, there is pervasive evidence of interdependence and contagion with respect to U.S. shocks in frontier markets. When compared with emerging markets, however, the degree of interdependence is lower and the contagion is larger across frontier markets, suggesting that the U.S. financial crisis has had a more contagious effect on frontier markets than on emerging markets. Frontier markets display very low correlation with the U.S. market (see Table 1), and this is interpreted in the literature as evidence of more portfolio diversification opportunities in frontier markets relative to emerging markets. The results of the present paper, however, point to an important caveat to this general proposition. During the financial crisis, the frontier market correlations increased proportionately more than the emerging market correlations (see Table 1), and some frontier markets in Asia, Africa, Middle East, and Europe as well as those that do not have a strong linkage with the U.S. during relatively stable times, became more susceptible to contagion than emerging markets during the crisis. The implication of these findings is that, even those international portfolios that are diversified across markets with low correlations with the U.S. become more exposed to contagion-induced losses in asset prices during crisis periods.

5.3. Impact of emerging market shocks on the U.S. market The results concerning the impact of emerging market return shocks on the U.S. are presented in Table 4. During tranquil periods, there exists a strong and positive influence of return shocks from partially overlapping emerging markets on the U.S. market, except for shocks originating in Morocco and Venezuela, suggesting strong interdependence. The markets with a larger impact on the U.S. during normal times include Chile (0.39), Israel (0.22), Brazil (0.21), South Africa (0.20), Poland (0.19), and Peru (0.19). There is also important contagion from partially overlapping emerging markets to the U.S. with the most prominent among them being Argentina (0.49), Peru (0.42), and Chile (0.31). Among the regions, the Latin American markets have the most pervasive and large interdependence as well as contagion on the U.S. The results also show that the U.S. market is not influenced by shocks emanating from non-overlapping emerging markets, except those from India, South Korea, and Taiwan, which have small and significant interdependence coefficients. This lack of importance of shocks from non-overlapping emerging markets along with the importance of shocks from partially overlapping markets suggest that return shocks that are generated in markets that are relatively closer to the trading hours in the U.S. market, i.e. primarily Latin American and European, have a more important impact on U.S. than those generated in the previous trading day in the Asian markets. Compared with this evidence of interdependence, the magnitude of contagion from nonoverlapping emerging markets to the U.S. is much larger. Such contagion is pervasive and clearly evident from all Asian markets except Philippines. Particularly larger contagion effects originate from Malaysia (0.51), Thailand (0.50), India (0.31), Taiwan (0.33), Indonesia (0.32), India (0.31), and China (0.28). These results also indicate that, although return shocks from China, Indonesia, Malaysia, and Thailand do not have an economically important effect on the U.S. market during normal times, they indeed have a larger contagious impact during the financial crisis. On average, the contagion from non-overlapping emerging markets is larger than that from partially overlapping emerging markets. These non-overlapping markets are all Asian markets, providing powerful evidence of contagion from Asian emerging markets to the U.S. The spillover of shocks from emerging markets possibly reflect adverse economic implications of the U.S. crisis on these markets which, in turn, affect the U.S. market due to negative effects on factors such as U. S. exports and withdrawal of foreign portfolio investments from the U.S.

Region

Market

Interdependence Concurrent

Panel A: partially overlapping markets Morocco Africa South Africa Composite Middle East Israel Europe Czech Republic Hungary Poland Russia Turkey Composite Latin America Argentina Brazil Chile Peru Venezuela Composite Composite

Contagion Lagged

ct

t-Stat

−0.01 0.20 0.16 0.22 0.15 0.12 0.19 0.11 0.07 0.11 0.10 0.21 0.39 0.19 0.01 0.16 0.14

(−0.44) (7.19)*** (8.81)*** (5.91)*** (5.23)*** (4.25)*** (8.03)*** (5.13)*** (5.15)*** (13.24)*** (3.68)*** (12.92)*** (12.78)*** (6.25)*** (0.52) (12.65)*** (20.50)***

ct−1

R

Concurrent t-Stat

2

DW

Lagged

ft

t-Stat

0.36 0.12 0.24 0.28 0.11 0.18 0.14 0.06 0.24 0.21 0.49 0.12 0.31 0.42 −0.48 0.37 0.27

(1.19) (1.77)* (4.37)*** (2.13)** (1.87)* (2.35)** (1.75)* (0.99) (3.71)*** (6.95)*** (8.03)*** (2.03)** (2.51)*** (7.21)*** (−1.10) (10.20)*** (11.07)***

ft−1

t-Stat 0.03 0.13 0.10 0.13 0.09 0.11 0.12 0.07 0.12 0.11 0.21 0.28 0.25 0.20 0.02 0.20 0.13

2.05 1.91 2.04 2.20 1.94 1.96 1.95 1.91 2.05 2.04 1.95 1.94 1.97 1.98 1.89 2.00 2.04

L.P. Samarakoon / Int. Fin. Markets, Inst. and Money 21 (2011) 724–742

Table 4 Impact of emerging market shocks on the U.S. market.

739

740

Table 4 (Continued) Region

Market

Interdependence Concurrent

Lagged t-Stat

R

Concurrent

ct−1

t-Stat

−0.03 0.07 0.02 0.05 0.03 0.07 0.06 0.04 0.06 −0.03 0.05

(−1.49) (2.80)*** (0.94) (1.13) (1.24) (2.87)*** (2.37)** (1.15) (8.07)*** (−1.01) (7.45)***

ft

2

DW

Lagged t-Stat

ft−1

t-Stat

0.28 0.31 0.32 0.51 0.22 0.14 0.33 0.50 0.29 0.17 0.30

(2.45)** (3.74)*** (3.21)*** (2.62)*** (1.43) (2.27)** (2.47)** (2.54)** (8.89)*** (0.79) (9.62)***

0.03 0.10 0.06 0.04 0.03 0.08 0.06 0.08 0.05 0.03 0.05

1.94 1.90 1.95 1.96 1.96 1.96 1.98 1.98 2.01 2.07 2.03

This table reports the results relating to the impact of emerging market shocks on the U.S. market shocks using the partially overlapping shock model (Eq. (6)), and the non-overlapping shock model (Eq. (8)). Partially overlapping shock model: eus,t = af + bus,1 (eus,t−1 ) + bus,2 (eus,t−2 ) + df,t (CDt ) + cf,t (ef,t ) + ff,t (ef,t × CDt ) + vf,t

(6)

Non-overlapping shock model: eus,t = af + bus,1 (eus,t−1 ) + bus,2 (eus,t−2 ) + df,t−1 (CDt−1 ) + cf,t−1 (ef,t−1 ) + ff,t−1 (ef,t−1 × CDt−1 ) + vf,t

(8)

The table shows the coefficient estimates and t-statistics for concurrent interdependence, and concurrent contagion. Composite represents all the markets in each panel. The last two columns show the adjusted R2 and the Durbin–Watson statistic. The total sample period is from 01/04/2000 to 03/09/2009, and the crisis period for measuring contagion is from 09/01/2008 to 03/09/2009. * ** ***

Significant at 10%. Significant at 5%. Significant at 1%.

L.P. Samarakoon / Int. Fin. Markets, Inst. and Money 21 (2011) 724–742

ct Panel B: non-overlapping markets Asia China India Indonesia Malaysia Philippines South Korea Taiwan Thailand Composite Africa Egypt Composite

Contagion

L.P. Samarakoon / Int. Fin. Markets, Inst. and Money 21 (2011) 724–742

741

Finally, the results of this paper also shed some light into the decoupling–recoupling hypothesis. In general, the evidence of bi-directional interdependence and contagion indicates that emerging markets are coupled with the U.S., i.e. move in tandem, in response to both U.S. and emerging market shocks. However, there are two important exceptions. The U.S. seems decoupled from the Asian emerging markets during tranquil periods in regard to shocks emanating from such markets. More importantly, the lack of contagious effect of U.S. shocks points out that emerging markets, except for Latin American, are decoupled from the U.S. with respect to additional impact of U.S. shocks during the crisis. Consistent with the findings of Forbes and Rigobon (2002) in other crises episodes, the widespread and large decline in stock markets in emerging markets during the U.S. financial crisis was largely due to normal, interdependence and not due to contagion of U.S. crisis. The results are robust to alternative specifications of the model and variables.9 6. Summary and conclusions This paper examines the transmission of shocks between the U.S. and foreign stock markets to delineate the effects of interdependence from contagion of the U.S. financial crisis using two shock models representing partially overlapping and non-overlapping emerging and frontier markets. There exits important bi-directional, yet asymmetric, interdependence and contagion between the U.S. and emerging markets, with important regional variations. The interdependence is driven more by U.S. shocks than by emerging market shocks, whereas contagion is driven more by emerging market shocks than by U.S. shocks. While Asian emerging markets do not impact the U.S. in tranquil periods, they have a strong contagious effect on the U.S. during the crisis. Except for Latin America, there is no contagion of U.S. crisis to emerging markets in other regions, whereas there is strong contagion from emerging markets in all regions to the U.S. There is evidence of interdependence and contagion, although small in magnitude, in frontier markets with respect to U.S. shocks. Frontier markets are influenced by U.S. shocks more during crisis than during normal times, and the U.S. financial crisis had a more contagious effect on frontier markets than on emerging markets. These findings enrich the existing literature on transmission of shocks in international capital markets. The central message from these findings is that emerging markets have large normal sensitivities to U.S. shocks, and large declines in stock prices in these markets reflected these dependencies rather than contagion. Portfolio diversification into emerging markets or even less-correlated frontier markets does not provide an effective hedge against U.S. shocks in crisis times. Further research, using advanced approaches and exploring different asset classes such as bonds, real estate, and commodities, can further enrich our understanding of the impact of the crisis and its lessons for asset pricing and international portfolio diversification. References Aloui, R., Alissa, M.S.B., Nguyen, D.K., 2011. Global financial crisis, extreme interdependences, and contagion effects: the role of economic structure? Journal of Banking & Finance 35, 130–141. Bae, K.H., Karolyi, G.A., Stulz, R.M., 2003. A new approach to measuring financial contagion. Review of Financial Studies 16, 717–763. Baele, L., 2005. Volatility spillover effects in European equity markets. Journal of Financial and Quantitative Analysis 40, 373–401. Bekaert, G., Harvey, C.R., Ng, A., 2005. Market integration and contagion. Journal of Business 78, 39–69. Chan, K.F., Treepongkaruna, S., Brooks, R., Gray, S., 2011. Asset market linkages: evidence from financial, commodity and real estate assets. Journal of Banking and Finance 35, 1415–1426. Chiang, T.C, Jeon, B.N., Li, H., 2007. Dynamic correlation analysis of financial contagion: evidence from the Asian Markets. Journal of International Money and Finance 26, 1206–1228.

9 The robustness checks included (1) 120 days vs. 240 days in rolling regressions to compute return shocks, (2) lower or higher number of lags of own returns in rolling regressions to compute returns shocks, (3) additional lags of own return shocks in the shocks models, (4) the Dow Jones Industrial Average as the U.S. market proxy, (5) use of local currency-denominated returns rather the U.S. dollar-denominated returns, (6) alternative definitions of the U.S. crisis period including (a) 06/02/2008 to 03/09/2009 to include the summer of 2008 and (b) 01/02/2008 to 03/09/2009 to expand the crisis period to the beginning of 2008, (7) crisis dummies to control for other crisis episodes that occurred during the estimation period, i.e., the 9/11 terrorist attacks and the Argentina economic and debt crisis in late 2001 to early 2002. The results and conclusions are remarkably robust to these alternative specifications.

742

L.P. Samarakoon / Int. Fin. Markets, Inst. and Money 21 (2011) 724–742

Claessens, S., Dornbusch, R., Park, Y.C., 2001. Contagion: how it spreads and how it can be stopped? In: Claessens, S., Forbs, K. (Eds.), International Financial Contagion. Kluwer Academic Publishers. Dooley, M., Hutchison, M., 2009. Transmission of the U.S. subprime crisis to emerging markets: Evidence on the decouplingrecoupling hypothesis. Journal of International Money and Finance 28, 1331–1349. Edwards, S., 2000. Contagion. Working Paper. NBER. Eun, C.S., Shim, S., 1989. International transmission of stock market movements. Journal of Financial and Quantitative Analysis 24, 241–256. Forbes, K., Rigobon, R., 2001. Contagion in Latin America: definitions, measurement, and policy implications. Economica 1, 1–46. Forbes, K., Rigobon, R., 2002. No contagion, only interdependence: measuring stock market comovements. Journal of Finance 57, 2223–2261. Guo, F., Chen, C.R., Huang, Y.S., 2011. Markets contagion during financial crisis: a regime-switching approach. International Review of Economics and Finance 20, 95–109. Hamao, Y., Masulis, R.W., Ng, V., 1990. Correlations in price changes and volatility across international stock markets. Review of Financial Studies 3, 281–308. Karolyi, G.A., 2002. Does International Financial Contagion Really Exist? Working Paper. The Ohio State University. Karolyi, G.A., Stulz, R.M., 1996. Why do markets move together? An investigation of U.S.–Japan stock return comovements. Journal of Finance 51, 951–986. Kenourgios, D., Samitas, A., 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., Wadhwani, S., 1990. Transmission of volatility between stock markets. Review of Financial Studies 3, 5–33. Lee, S.B., Kim, K.J., 1993. Does the October 1987 crash strengthen the co-movements among national stock markets? Review of Financial Economics 3, 89–102. Lin, W.L., Engle, R.F., Ito, T., 1994. Do bulls and bears move across borders? International transmission of stock returns and volatility. Review of Financial Studies 7, 507–538. Longstaff, F.A., 2010. The subprime credit crisis and contagion in financial markets. Journal of Financial Economics 97, 436–450. Okimoto, T., 2008. New evidence of asymmetric dependence structures in international equity markets. Journal of Financial and Quantitative Analysis 43, 787–815. Pesaran, B., Pesaran, M.H., 2010. Conditional volatility and correlations of weekly returns and the VaR analysis of 2008 stock market crash. CESIFO Working Paper No. 3023. Pesaran, M.H., Pick, A., 2007. Econometric issues in the analysis of contagion. Journal of Economic Dynamics & Control 31, 1245–1277. Ramchand, L., Susmel, R., 1998. Volatility and cross-correlation across major stock markets. Journal of Empirical Finance 5, 397–416. Rigobon, R., 2002. Contagion: how to measure it? In: Edwards, S., Frankel, J. (Eds.), Preventing Currency Crises in Emerging Markets. University of Chicago Press. Rodriquez, J.C., 2007. Measuring financial contagion: a copula approach. Journal of Empirical Finance 14, 401–423.