Download Market Integration and Contagion in Asian Emerging Stock and

Market Integration and Contagion in Asian Emerging Stock and Foreign
Exchange Markets
Chu-Sheng Tai*
Department of Economics and Finance
College of Business Administration
Texas A&M University-Kingsville
MSC 186, 700 University Blvd.
Kingsville, TX 78363-8203, USA
Abstract
This paper examines whether Asian emerging stock markets have become integrated into world
capital markets since their official liberalization dates by estimating and testing a dynamic
international asset pricing model (ICAPM) in the absence of purchasing power parity (PPP)
using an asymmetric multivariate GARCH-in-Mean (MGARCH-M) approach. Also examined
in this paper is whether there are pure contagion effects between stock and foreign exchange
markets for each Asian country during the 1997-98 Asian crisis. The empirical results show that
first, both currency and world market risks are priced, suggesting that omitting currency risk in
pricing international assets might give rise to model misspecification. Second, the stock markets
for India, Korea, Malaysia, Philippines, and Thailand are segmented from the world capital
markets before their liberalization dates, but have become fully integrated since then. Third, the
market liberalization has reduced the cost of capital and price volatility for most of the markets.
Finally, as for the contagion effects, data from Korea, Malaysia, and Philippines indicate strong
feedback relations between stock prices and exchange rates. This feedback relation is consistent
with the predictions of both the stock- and flow-oriented exchange rate models. Data from India
and Taiwan show that exchange rates led stock prices, whereas that of Thailand fails to reveal
any recognizable pattern during the crisis.
JEL Classifications: C32, G12, G15
Key Words: Market Integration; Contagion; Currency Risk; Multivariate GARCH-M
Market Integration and Contagion in Asian Emerging Stock and Foreign
Exchange Markets
Abstract
This paper examines whether Asian emerging stock markets have become integrated into world
capital markets since their official liberalization dates by estimating and testing a dynamic
international asset pricing model (ICAPM) in the absence of purchasing power parity (PPP)
using an asymmetric multivariate GARCH-in-Mean (MGARCH-M) approach. Also examined
in this paper is whether there are pure contagion effects between stock and foreign exchange
markets for each Asian country during the 1997-98 Asian crisis. The empirical results show that
first, both currency and world market risks are priced, suggesting that omitting currency risk in
pricing international assets might give rise to model misspecification. Second, the stock markets
for India, Korea, Malaysia, Philippines, and Thailand are segmented from the world capital
markets before their liberalization dates, but have become fully integrated since then. Third, the
market liberalization has reduced the cost of capital and price volatility for most of the markets.
Finally, as for the contagion effects, data from Korea, Malaysia, and Philippines indicate strong
feedback relations between stock prices and exchange rates. This feedback relation is consistent
with the predictions of both the stock- and flow-oriented exchange rate models. Data from India
and Taiwan show that exchange rates led stock prices, whereas that of Thailand fails to reveal
any recognizable pattern during the crisis.
JEL Classifications: C32, G12, G15
Key Words: Market Integration; Contagion; Currency Risk; Multivariate GARCH-M
1
1. Introduction
A large number of Asian emerging markets have embarked on a series of reforms in
recent years, including liberalization of their national stock markets.
As a result of these
developments and the important implications of market integration on international capital
budgeting and investment, market integration has emerged as an important body of literature.
Two recent examples of this literature that study the impact of market liberalization on market
integration for Asian emerging markets are Bekaert and Harvey (1995) and De Santis and
Imrohoroglu (1997).1 Bekaert and Harvey (1995) propose a one-factor asset pricing model that
allows the conditional expected returns of a country to be affected by their covariance with a
world benchmark portfolio and by the variance of the country returns. They use a conditional
regime-switching model to account for periods when national markets were segmented from
world capital markets and when they became integrated later in the sample. In contrast to
general perceptions that markets are becoming more integrated, their results suggest that some
countries are becoming less integrated into the world market. However, based on specification
tests, their model is rejected in most countries. They point out that one possible extension of
their study is to consider currency risk as another potential priced factor.
Instead of using the
conditional regime-switching methodology, De Santis and Imrohoroglu (1997) utilize the
multivariate GARCH-in-Mean (MGARCH-M) approach. They introduce a dynamic integration
version of the classic CAPM framework that assumes full market segmentation until the official
1
Earlier theoretical and empirical papers on market integration include Solnik (1974), Stehle (1977), Stulz (1981),
Errunza and Losq (1985), Eun and Janakiramanan (1986), Jorion and Schwartz (1986), Wheatley (1988), and
Errunza, Losq and Padmandaban (1992). More recent papers include Hardouvelis, Malliaropulos, and Priestley
(2000), Bekaert, Harvey and Lumsdaine (2001), and Carrieri, Errunza and Hogan (2001). However, these papers
either focus on developed markets or do not consider how market liberalization affects market integration. Instead
of focusing on how market liberalization affects market integration, Obstfeld (1994), Bekaert, Harvey and
2
liberalization date of each market, and full integration thereafter to capture the fact that the
analyzed markets were legally segmented for part of the sample period. Their empirical results
show that neither the country-specific risk, nor the world market risk is priced and thus no
conclusion can be made regarding the impact of market liberalization on market integration.
One possible common cause for the weak findings of De Santis and Imrohoroglu (1997) and the
rejection of Bekaert and Harvey’s model is that both assume purchasing power parity (PPP) and
thus ignore currency risk, which motivates the current research.
Due to a series of financial crises in 1990s including the Exchange rate Mechanism
(ERM) attacks of 1992, the Mexican peso collapse of 1994, the Asian crisis of 1997, the Russian
collapse of 1998, and the Brazilian devaluation of 1999, the study of the transmission of
financial shocks/crisis across markets/countries has also emerged as one of the most intensive
research topics in international financial literature in recent years. Previous papers on this topic
have failed to take into account an important distinction between the two concepts of
interdependence and contagion except Forbes and Rigobon (2002).2
Masson (1998) argues that
there are three main channels that financial markets turbulence can spread from one country to
another: monsoonal effects, spillovers and pure contagion effects.
‘Monsoonal’ effects, or
‘contagions from common causes’ tend to occur when affected countries have similar economic
fundamentals or face common external shocks. The second type of financial market interlinkages arises from spillover effects, which may be due to trade linkages or financial
Lundblad (2000, 2001) and Henry (2000a) study the impact of market liberalization on the economic development
of the underlying countries.
2
Forbes and Rigobon (2002) define contagion as a significant increase in cross-market linkages after a shock to one
country or group of countries, and find that there was virtually no increase in unconditional correlation coefficients
during the 1997 Asian crisis and thus conclude that there was no contagion but interdependence. However, they
also point out that their definition of contagion is not universally accepted, and therefore it warrants another
examination of whether contagion did occur during 1997 Asian crisis.
3
interdependence. The first two channels of financial crises can be categorized as fundamentalsdriven crises since the affected countries share some macroeconomic fundamentals, which
implies that the transmission of financial crises is due to the interdependence among those
countries and not necessarily due to contagion. The third transmission channel is the pure
contagion effect. Contagion here refers to the cases where crisis in one country/market triggers a
crisis elsewhere for reasons unexplained by macroeconomic fundamentals. For instance, a crisis
in one country may lead creditors and investors to pull out from other countries over which they
have a poor understanding resulting from information asymmetries.
Given the facts that both market integration and contagion have important implications in
international finance and that previous studies in these two topics are inconclusive and thus
debatable, in this paper I attempt to provide new empirical evidence on theses two issues.
Specifically, I develop a dynamic version of international capital asset pricing model (ICAPM)
in the absence of PPP, and then test the model using data from six Asian emerging countries.
This study contributes our understanding in the return dynamics of emerging markets with
respect to market integration and contagion in several ways. First of all, unlike previous studies
on market integration (e.g., Bekaert and Harvey (1995) and De Santis and Imrohoroglu (1997)),
PPP is not assumed when testing market integration/segmentation hypothesis. Many empirical
studies have documented that PPP does not hold, especially in short horizons.3 In the absence of
PPP, international investors will face different real returns when holding the same assets. In this
case, currency risk will emerge as another potential priced factor. Secondly, previous studies
have failed to control for the economic fundamentals when testing contagion. In this paper, I
rely on the developed ICAPM, which provides me a theoretical basis in selecting the economic
4
fundamentals. The economic fundamentals under ICAPM are the world market and currency
risks, so the evidence of contagion is based on testing whether idiosyncratic risks - the part that
cannot be explained by the world market and currency risks, are significant in describing the
return dynamics of both stock and foreign exchange markets for each emerging country during
the 1997-98 Asian crisis. Finally, a parsimonious parameterization of asymmetric trivariate
GARCH-M process is employed to model the conditional covariance matrix of asset returns,
which is very important in testing contagion.4 The advantage of this multivariate approach is that
it utilizes the information in the entire variance-covariance matrix of the errors, which, in turn,
leads to more precise estimates of the parameters of the model.
The empirical results show that first, both currency and world market risks are priced,
suggesting that omitting currency risk in pricing international assets might give rise to model
misspecification. Second, the stock markets for India, Korea, Malaysia, Philippines, and
Thailand are segmented from the world capital markets before their liberalization dates, but have
become fully integrated since then. Third, consistent with Bekaert and Harvey (1997), Stulz
(1999) and Henry (2000b), the market liberalization has reduced the cost of equity capital and
price volatility for most of countries. Finally, as for the contagion effects, data from Korea,
Malaysia, and Philippines indicate strong feedback relations between stock prices and exchange
3
4
Rogoff (1996) provides a detailed discussion on the issue of PPP.
According to Forbes and Rigobon (1999), Dornbusch, Park and Claessens (1999), and Kaminsky and Reinhart
(2000), previous empirical studies on contagion can be categorized by methodology into four groups: (1) the
testing of significant increases in correlation (Calvo and Reinhart (1996), Baig and Goldfajn (1999), Forbes and
Rigobon (1999, 2002) and Park and Song (1999)); (2) the testing of significance in innovation correlation (Baig
and Goldfajn (1999)); (3) the testing of significant volatility spillover (Edwards (1998), Edwards and Susmel
(1999)); (4) crisis prediction regression (Bae, Karolyi, and Stulz (2000), Eichengreen, Ross, and Wyplosz (1996),
Kaminsky and Reinhart (2000), Van Rijckeghem and Weder (1999), Sachs, Tornell, and Velasco (1996)). None
of the contagion studies mentioned above explicitly takes the time dependencies in the second moment into
account. A recent paper by Bekaert, Harvey, and Ng (2002) applies three-stage univariate GARCH model to study
contagion in equity markets by testing whether there is evidence of significant increase in cross market residual
5
rates. Data from India and Taiwan show that exchange rates led stock prices, whereas that of
Thailand fails to reveal any recognizable pattern during the crisis.
The remainder of this paper is organized as follows. Section 2 develops the dynamic
version of ICAPM in the absence of PPP. Section 3 presents the econometric methodologies
used to estimate and test the model. Section 4 discusses the data. Section 5 reports the empirical
results. Summary and concluding remarks are offered in Section 6.
2. The dynamic ICAPM in the absence of PPP
Since one of the purposes of this paper is to investigate whether Asian emerging stock
markets have become fully integrated after their liberalization, it would be appropriate to
consider a dynamic version of ICAPM. De Santis and Imrohoroglu (1997) introduce a structure
that allows in a single model a situation of full market segmentation until the official
liberalization date, and full integration thereafter. I modify their model by incorporating the
currency risk.5 The dynamic ICAPM with currency risk applied to each emerging stock market
index can be written as follows.
correlation during a crisis. Although they model conditional second moments, they can not answer whether return
shocks emanating from one market will significantly affect the other markets during the crisis.
5
Unlike the regime-switching model of Bekaert and Harvey (1995), De Santis and Imrohoroglu (1997) assume that
the integration process is irreversible and recognize that their assumption has a drawback to detect the presence of
the two regimes in the data if the official liberalization date is a poor indicator of when that market actually became
accessible to foreign investors and allowed residents to invest abroad. This simplistic specification for a test on the
relevance of liberalization would imply that an abrupt change on the pricing of risk happened on the day of the
event, and thus is subject to criticism. Although interest rate differentials could be used as integration proxy to
allow for time-varying integration (e.g., Hardouvelis, Malliaropulos, and Priestley (2000)), the interest rate data for
most of Asian emerging countries are not available until 1990s, which is the period when most of these countries
have liberalized their markets, thus making this kind of approach unfeasible.
6
ri ,t = (1 − LDi ,t )λ w,t −1 hiw ,t + λ c ,t −1hic ,t + LDi ,t λl ,t −1 hi ,t + ε i ,t
(1)
where ri ,t is the excess return from date t − 1 to date t on local stock market i ; LDi ,t is a
liberalization dummy variable, which is equal to one before the opening date of stock market i
to foreign investors and zero otherwise.6 λ w,t −1 , λ c ,t −1 and λl ,t −1 are time-varying prices of world
market, currency, and local market risks, respectively.
hiw ,t is the conditional covariance
between the excess returns on the local market index and world market portfolio; hic ,t is the
conditional covariance between the excess returns on the local market index and the bilateral
U.S. dollar exchange rate changes, and hi ,t is the conditional variance of the excess returns on
the local market index. Finally, ε i,t denotes the country-specific, idiosyncratic shock for market
i . The dynamic ICAPM in the absence of PPP specified in equation (1) indicates that if a capital
market is completely segmented, the expected returns on the local market index is only
associated with the country-specific risk, which is proxied by the conditional variance of its
stock index returns. However, if capital market is fully integrated, then the source of risk of the
expected returns is affected by its covariance with the world market portfolio returns.
Equation (1) requires not only the estimation of the conditional variance of each local
stock market, but also the estimation of the conditional covariances between the local stock
market and world market portfolio returns, and between the local stock market returns and
exchange rate changes. Therefore, I need to generalize the process for the conditional second
6
Table 1 reports the International Finance Corporation (IFC) official liberalization dates for six Asian emerging
countries under study. The IFC date is based on the Investibility Index, which represents the ratio of the market
capitalization of stocks that foreigners can legally hold to total market capitalization.
7
moments to a trivariate framework to include equations representing the world market portfolio
returns and exchange rate changes, which are estimated as follows:
rc ,t = λ w,t −1 hcw,t + λc ,t −1 hc ,t + ε c ,t
(2)
rw,t = λ w,t −1 hw,t + λc ,t −1 hcw ,t + ε w,t
(3)
where hcw,t is the conditional covariance between exchange rate changes and world market
portfolio returns. hw,t and hc ,t are conditional variances of world market portfolio returns and
exchange rate changes, respectively. ε w,t is the unexpected return on the world market portfolio
conditional on expectations based on the information available at t − 1 , and similarly ε c,t is the
unexpected change on the bilateral exchange rate conditional on expectations based on the
information available at t − 1 .
To model the time-varying prices of world market and local market risk, their dynamics
are chosen according to the theoretical asset-pricing model developed by Merton (1980). In his
model, both the prices of world market and local market risk are the coefficients of risk aversion
of risk-averse investors, and thus should be positive. Consequently, similar to Bekaert and
Harvey (1995), De Santis and Gerard (1997, 1998), an exponential function is used to model
both the dynamics of λ w,t −1 and λl ,t −1 . For the dynamic of λ c ,t −1 , a linear specification is adopted
because the model does not restrict the price of currency risk to be positive.7
7
As pointed out by De Santis and Gerad (1997), the conditional ICAPM is only a partial equilibrium model and the
theory does not help identify the state variables that affect the prices of world market, currency and local market
risks, so inevitably any parameterization of the dynamics of λ w, t − 1 , λ c , t − 1 , and λl ,t −1 can be criticized for being
ad hoc.
8
λ w,t −1 = exp(ϕ w' z t −1 )
(4)
λ c ,t −1 = ϕ c' z t −1
(5)
λl ,t −1 = exp(ϕ l' z l ,t −1 )
(6)
where Z t −1 and Z l ,t −1 are vectors of information variables observed at the end of time t − 1 and
ϕ ’s are time-invariant vectors of weights. Thus, the price of currency risk is assumed to be a
linear function of the information variables in Z t −1 , and the price of world market risk is assumed
to be an exponential function of information variables in Z t −1 . Similarly, the price of local
market risk is also an exponential function of local information variables in Z l ,t −1 . Given the
dynamics of price of risk, I can then test whether the prices of world market and currency risks
are significantly priced and change over time by testing whether the information variables in
Z t −1 are significant in addition to significant GARCH parameters. Similarly, the time-varying
price of each local market risk can be tested whether the information variables in Z l ,t −1 are
significant.
Notice that the parameterization of the dynamic ICAPM differs from that of Bekaert and
Harvey (1995). Bekaert and Harvey (1995) apply a two-step estimation procedure, in which they
use an univariate GARCH-in-Mean approach to estimate the price of world market risk in the
first step, and then in the second step the estimated price of world market risk is treated as data to
estimate the country-specific risk associated with each local stock market returns. However, my
parameterization not only allows a country’s stock returns to include country-specific risk ( λl ,t −1 )
9
and world market risk ( λ w,t −1 ), but also requires the estimation of the price of currency risk
( λ c ,t −1 ). More importantly, all the prices of risk are allowed to change through time and are
estimated simultaneously to overcome the potential efficiency loss in Bekaert and Harvey
(1995).
Another main purpose of the paper is to test whether there are any pure contagion effects
between a country’s stock market and its foreign exchange market during the 1997-98 Asia
crisis.
Consequently, I modify equations (1) and (2) to allow the local stock market’s
idiosyncratic shocks at t − 1 to affect the return of its foreign exchange market at t , and vice
versa. Specifically, equations (1) and (2) can now be written as:
ri ,t = (1 − LDi,t )λw,t −1hiw,t + λc,t −1hic,t + LDi ,t λl ,t −1hi ,t + (δ ii ε i ,t −1 + δ ciε c,t −1 ) + CDt (ωii ε i ,t −1 + ω ciε c,t −1 ) + ε i ,t (7)
rc ,t = λ w,t −1 hcw,t + λ c ,t −1 hc ,t + (δ cc ε c ,t −1 + δ ic ε i ,t −1 ) + CDt (ω cc ε c ,t −1 + ω ic ε i ,t −1 ) + ε c ,t
(8)
where " CDt " is a crisis dummy variable, which is equal to one during the Asian crisis and zero
otherwise.8
In testing the pure contagion effects, I allow the past market-specific shocks to
affect current asset returns in the entire sample period (i.e., mean spillovers), and then test
whether there are any incremental influences of past return shocks on these returns during the
crisis period (i.e., contagion). Thus, the hypothesis of no contagion between stock and foreign
exchange markets for each Asian emerging country can be examined by testing whether the
contagion coefficients, ω ic or ω ci , are individually or jointly significant after the systematic risks
and any mean spillovers have been accounted for.
10
3. Econometric Methodology
To model the conditional second moments, several multivariate GARCH models have
been proposed such as the diagonal VECH model of Bollerslev, Engle, and Wooldridge (1988),
the constant correlation (CCORR) model of Bollerslev (1990), the factor ARCH (FARCH)
model of Engle, Ng, and Rothschild (1990), and the BEKK model of Engle and Kroner (1995).
Among these four popular MGARCH models, the diagonal BEKK model is selected and is
modified to accommodate the asymmetric volatility effects in variances and covariances, which
has been documented in recent papers by, among others, Kroner and Ng (1998) and Bekaert and
Wu (2000). The diagonal BEKK model is preferred because it not only yields a positive definite
covariance matrix for all values of ε t −1 , but also economizes on parameters relative to other
MGARCH processes.9 Specifically, the dynamic process for the conditional variance-covariance
matrix of the asset returns is specified as:
H t = C ' C + A ' ⋅ H t −1 ⋅ A + B ' ⋅ ε t −1ε t' −1 ⋅ B + D ' ⋅ η t −1η t' −1 ⋅ D
(9)
where H t is 3× 3 time-varying variance-covariance matrix of asset returns; C is restricted to be
a 3× 3 upper triangular matrix and A , B , and D are 3× 3 diagonal matrices. η t −1 is 3× 1
vector such η i ,t −1 = ε i ,t −1 if ε i ,t −1 < 0 , 0 otherwise which captures the asymmetric impact that the
vector of past negative shocks has on the conditional covariance matrix in a manner similar to
that of Glosten et al. (1993). In this model the conditional variance and covariance of each
8
9
I assume that Asian crisis began in July 1997 and ended in June 1998 in the sample.
De Santis and Gerard (1997) find strong support for this parameterization of matrices
11
A and B .
return are related to the past squared residuals and cross residuals, past squared asymmetric
shocks and cross asymmetric shocks of all returns while they are only related to their own past
conditional variance and covariance. Even with this diagonal BEKK parameterization, it still
requires the estimation of 35 parameters in the conditional mean and variance equations.
Under the assumption of conditional normality, the log-likelihood to be maximized can
be written as:
ln L(ϖ ) = −
TN
1 T
1 T
ln 2π − ∑ ln | H t (ϖ ) | − ∑ ε t (ϖ ) ' H t (ϖ ) −1 ε t (ϖ )
2
2 t =1
2 t =1
(10)
where ϖ is the vector of unknown parameters in the model. Since the normality assumption is
often violated in financial time series, I use quasi-maximum likelihood estimation (QML)
proposed by Bollerslev and Wooldridge (1992) which allows inference in the presence of
departures from conditional normality. Under standard regularity conditions, the QML estimator
is consistent and asymptotically normal and statistical inferences can be carried out by
computing robust Wald statistics. The QML estimates can be obtained by maximizing equation
(10), and calculating a robust estimate of the covariance of the parameter estimates using the
matrix of second derivatives and the average of the period-by-period outer products of the
gradient. Optimization is performed using the Broyden, Fletcher, Goldfarb, and Shanno (BFGS)
algorithm.10
10
I have also tried the estimation by fitting a t-distribution on the vector of errors, with v degrees of freedom as
recommended by Engle and Bollerslev (1986) and Bollerslev (1987). However, the t-distribution did not give
good results. As noted by Bera and Higgings (1993) this could be due to the fact that although conditional tdistribution allows kurtosis to exceed 3, it assumes that it is constant since the estimated degrees of freedom v are
time invariant. Therefore, I proceed by focusing on QML estimates.
12
4. Data and summary statistics
The sample of Asian countries examined in the paper includes India, Korea, Malaysia,
Philippines, Taiwan and Thailand. The data consists of end-of-month observations of stock
market total (i.e. dividend adjusted) return indices, and of local bilateral spot exchange rates
expressed as units of U.S. dollar against one unit of each Asian currency.11 The sample covers
monthly observations for the period 1980:01 to 2002:03 for India, Korea and Thailand, and
1984:12 to 2001:03 for Malaysia and Taiwan, and 1986:01 to 2001:03 for Philippines. The
MSCI world market total return index is used to proxy the world market portfolio. The USdollar denominated excess stock return is computed as: ri ,t = ln(
pt +1
1
) − ln(1 + Rtus $ ) where pt
pt
12
is either the IFC emerging market total return index or MSCI world market total return index at
time t , and RtUS $ is annualized one-month Eurodollar deposit rate. The currency return is
computed as: rc ,t = ln(S i ,t ) − ln(S i ,t −1 ) where S i ,t is the spot exchange rate at time t expressed as
U.S. dollar price of one unit of local currency of country i .
The information variables selected in this paper to model the dynamics of λ w,t −1 and λ c ,t −1
are based on previous studies. Harvey (1991) shows that U.S. information variables are useful in
predicting foreign stock returns. Giovannini and Jorion (1987) find that nominal interest rates
have explanatory power for the time variation of currency returns. Thus, several information
variables are chosen to be included in Z t −1 . They are dividend yield on the world market index
11
The International Finance Corporation (IFC) total return indices are used to study the market integration for six
Asian emerging countries. They are India ( IFCIND ), Korea ( IFCKOR ), Malaysia ( IFCMAL ), Philippines
( IFCPHI ), Taiwan ( IFCTAI ), and Thailand ( IFCTHA ).
13
in excess of the one-month Eurodollar deposit rate ( DIV ), the change in the U.S. term premium,
measured by the yield difference between 10-year Treasury constant maturity rate and 1-month
Eurodollar rate ( ∆USTP ), the U.S. default premium, measured by the yield difference between
Moody’s Baa-rated and Aaa-rated U.S. corporate bond ( USDP ), the lagged world excess returns
( WORLD ), and a constant ( CON ). Given the limited local information variables available for
each Asian emerging country, the information variables contained in Z l ,t −1 only include lagged
local market returns ( LOCAL ) and a constant ( CON ).
All the data are extracted from
Datastream.
Table 2 presents summary statistics of the continuously compounded stock and currency
returns for each emerging country.12 As can be seen from Panel A, IFCMAL has the highest
mean returns (1.153%) with a standard deviation of 11.191%, while IFCTHA has the lowest
mean returns (0.385%) with a standard deviation of 11.112%. Comparing the performance of six
currency returns, Taiwan dollar ( TAIWDUS ) is the best one and the only one with a positive
mean return of 0.098%, and a standard deviation of 1.376%. This is not surprising since Taiwan
is one of the few countries that was not seriously affected by the Asian currency crisis.
Table 2 also reports Bera-Jarque and Ljung-Box statistics. In eleven out of twelve cases,
the Bera-Jarque test statistic strongly rejects the hypothesis of normally distributed stock and
currency returns. The Ljung-Box test statistic for raw returns, LB (20) , is significant in six cases
(three stock returns: IFCMAL , IFCPHI , and IFCTHA , and three currency returns: SKORWUS ,
TAIWDUS , and THAIBUS ), implying linear dependencies in these returns. For squared returns,
14
LB 2 (20) is significant for all stock returns except IFCPHI , and for all currency returns except
INDNRUS and SKORWUS , indicating strong nonlinear dependencies in these returns. This is
consistent with the volatility clustering observed in most stock and foreign exchange markets:
Large (small) changes in prices tend to be followed by large (small) changes of either sign. The
GARCH models used in this study are well known to capture this property.
5. Empirical results
The quasi-maximum likelihood estimation of the dynamic ICAPM in the absence of PPP
(equations (3)-(8)) is reported in Table 3. The parameter estimates of the time-varying prices of
risk are shown in Panel A. The estimates of mean spillovers and contagion parameters are
depicted in Panel B. Panel C reports the parameter estimates for the conditional variance
process. The diagnostic test statistics for the standardized residuals are shown in Panel D.
Finally, The hypothesis tests concerning the time-varying prices of risk and the summary
statistics about the estimated risk premia and conditional volatility are shown in Panels E and F,
respectively.
5.1. The dynamic ICAPM with currency risk
Panel A of Table 2 presents the estimation results for the prices of risk for each emerging
country. First considering the price of country-specific risk before liberalization, it is significant
12
Given the different sample size for each country, I do not report the descriptive statistics for the selected
information variables and MSCI world market index in Table 2, but they are available upon request.
15
and changes over time in three cases (India, Korea, and Malaysia) since the lagged local stock
market returns are significant in describing the price dynamic of the country-specific risk. For
Philippines and Thailand, their prices of country-specific risk are significant but time-invariant.
However, it is insignificant for Taiwan. These results can be confirmed by the hypothesis tests
(#7) reported in Panel E of Table 3. As can be seen, the null hypothesis of zero price of countryspecific risk is significantly rejected at least at the 5% level in all cases except Taiwan. Next,
considering the prices of world market and currency risk, the hypothesis tests reported in Panel E
show that the price of world market risk not only is significant (#3) but also changes over time
(#4) in all cases. As for the currency risk, it is also significantly priced (#5) and changes over
time (#6) in all cases except India. These empirical results imply that India, Korea, Malaysia,
Philippines, and Thailand were fully segmented before their liberalization dates, but have
become fully integrated with the world capital markets thereafter. This result is in sharp contrast
to De Santis and Imrohoroglu (1997), who find neither the country-specific risk nor the world
market risk is significantly priced for the six Asian emerging countries examined in this paper,
and consequently they can not draw any conclusion regarding how market liberalization affects
the market integration for each country studied. The significant currency risk found in this paper
indicates that an international asset pricing model under PPP such as the one used by De Santis
and Imrohoroglu (1997) would be misspecified.
5.2 Evidence of Mean Spillover and Contagion
After controlling the systematic market and currency risks, I can then test pure contagion
effects between stock and foreign exchange markets for each country. However, before that, I
16
need to control for the overall mean spillovers in the entire sample period, so any incremental
mean spillover effects can be tested during the crisis period. It can be seen from Panel B of
Table 3 that the parameter estimates for mean spillovers are significant in three cases: India,
Korea, and Malaysia. For example, the past return shocks originating from Indian currency
market have significant and positive ( δ ci = 0.027) impact on its domestic stock market,
suggesting that exchange rates lead stock prices. On the other hand, the past return shocks
emanating from Malaysian stock market have significant and positive ( δ ic = 1.226) impact on the
value of its domestic currency, implying that stock prices lead exchange rates. Finally, the past
return shocks from Korean stock market also have significant and positive ( δ ic = 0.035) impact
on domestic currency value and vice versa ( δ ci = 0.006), indicating a feedback relation between
Korean stock and foreign exchange markets. The finding of significant positive impact of past
return shocks from the stock market (foreign exchange market) on foreign exchange market
(stock market) is consistent with Ajayj and Mougoue (1996) where they conclude that currency
appreciation has a positive effect on domestic stock market. This finding can be explained based
on stock-oriented model of exchange rates (Frankel (1983)). An increase in stock prices causes
an increase in the wealth of domestic investors, which in turn leads to a higher demand for
money with ensuing higher interest rates. The higher interest rates encourage capital inflows
ceteris paribus, which in turn is the cause of currency appreciation.
Since significant systematic risk premia have been founded and the overall mean
spillovers have been controlled for the entire sample period, I can now test whether there are any
pure contagion effects between domestic stock and foreign exchange markets and examine the
dynamic relation between these two markets during the crisis. As shown in Panel B, the
17
contagion effects are significant in five cases for the stock markets, and three cases for the
foreign exchange markets. Among the five cases where contagion effects are due to the shocks
from the foreign exchange market to the stock market, two have the negative effects: India ( ω ci =
-0.054) and Malaysia ( ω ci = -0.146), and the other three have positive effects: Korea ( ω ci =
0.132), Philippines ( ω ci = 1.047), and Taiwan ( ω ci = 0.100). The significantly negative impact of
past foreign exchange shocks on the domestic stock market (India and Malaysia) can be
explained by the flow-oriented exchange rate model (Dornbusch and Fischer (1980)). The model
focuses on the current account or trade balance, which posit that currency movements affect
international competitiveness and the trade balance, thereby influencing real income and output.
As a country’s currency appreciates, it decreases her international competitiveness in good
markets, which has a negative effect on a firm’s future cash flow. Consequently, returns on
domestic stock market decrease. On the other hand, the significantly positive impact of past
foreign exchange shocks on the domestic stock market (Korea, Philippines, and Taiwan) can be
explained by portfolio balance approach. According to this approach, a rise in the value of
domestic currency against the U.S. dollar raises the returns on domestic assets. Investors quickly
shift funds from dollar assets to domestic assets such as stock due to higher returns. This shift of
portfolio composition in favor of domestic stocks and against dollar assets results in decreases in
stock supply and increases in stock demand, which then raises domestic stock prices and their
returns. The portfolio balance model, thus, implies that currency appreciation tends to have a
positive effect on the local stock market, which is the opposite of the flow-oriented exchange rate
model discussed earlier.
18
As for the three cases where the contagion effects are due to the shocks from the stock
market to foreign exchange market, the effects are negative for Philippines ( ω ic = -0.709), but
positive for Korea ( ω ic = 1.034) and Malaysia ( ω ic = 2.170). The significantly negative impact of
past stock return shocks on current spot exchange rates for Philippines can be explained by the
stock market’s providing a barometer for the health of an economy (Solnik (1987)). During
economic expansion, investors are bullish on stock market, and this tends to fuel inflation
expectations, which exerts downward pressure on the value of domestic currency. On the other
hand, the significantly positive impact of past stock return shocks on current spot exchange rates
for Korea and Malaysia can again be explained by stock-oriented exchange rate model (Frankel
(1983)) discussed earlier.
Overall data from Korea, Malaysia, and Philippines indicate strong feedback relations
between stock prices and exchange rates during the crisis. Data from India and Taiwan show
that exchange rates led stock prices, whereas that of Thailand fails to reveal any recognizable
pattern during the crisis. 13
5.3 Conditional Variance Process and Residual Diagnostics
13
Notice that although the dynamic ICAPM in the absence of PPP with trivariate GARCH-M parameterization may
enable one to uncover some economic and/or statistical relations, it is sometimes difficult to interpret the
underlying fundamental economic relation based on those results. It is likely that the results may be generated
from other structure relations, i.e., via interest rate parity condition or IS-LM-related policies. For example, some
recessionary shocks or unfavorable information on the country will cause a stock price decrease and an exchange
rate depreciation. In this case the time relation between the stock price and foreign exchange rate will be
generated from the relative efficiency of the stock market and foreign exchange market. Consequently, both
stock- and flow-oriented exchange rate models may not play any role in generating the feedback relation found
here.
19
Panel C presents the parameter estimates for the conditional variance process. As can be
seen, most of elements in the vectors a and b are statistically significant at 1% level, implying
that strong GARCH effect is present for all the return series. In addition, the estimates satisfy
the stationarity conditions for all the variance and covariance processes.14 As for the asymmetric
volatility parameter, d , it is much stronger in the stock market (four cases) than in the foreign
exchange markets (two cases). To access the fit of the dynamic ICAPM in the absence of PPP
with trivariate GARCH-M specification, Panel C reports the Ljung-Box statistics for 20th-order
serial correlation in the level ( LB (20) ) and squared standardized residuals ( LB 2 (20) ). Under
the multivariate framework, the standardized residuals at time t is computed as Z t = H t−1 / 2 ε t ,
where H t−1 / 2 is the inverse of the Cholesky factor of the estimated variance-covariance matrix.
Panel D shows that none of the LB (20) statistics is significant except in three cases. Similarly,
none of the LB 2 (20) statistics is significant except in one case. Overall, the dynamic ICAPM in
the absence of PPP with trivariate GARCH-M parameterization effectively eliminates most of
the linear and nonlinear dependencies found in the raw data. As for B − J test statistics, most of
them are smaller than those found in the raw data but still significant in most cases, indicating
departure from the normality, which justifies the use of robust standard errors computed from
using the quasi-maximum likelihood method of Bollerslev and Wooldridge (1992).
5.4 The Cost of Capital and Conditional Volatility
14
For the process in
H t to be covariance stationary, the condition a i a j + bi b j < 1 ∀i, j has to be satisfied
(Bollerslev, 1986; De Santis and Gerard, 1997, 1998).
20
One important implication of market liberalization is the issue of cost of capital. The cost
of capital is the discount rate used to discount the expected cash flows of a project that a firm
intends to take on. This discount rate is the rate of return required by the firm’s shareholders and
is determined by the CAPM. If a market is fully segmented from other capital markets, this
required rate of return is equal to a risk-free rate plus a risk premium equal to the firm’s beta
times a local market’s risk premium. If a country is opened up to foreign investors and lets its
residents invest abroad, then the residents of the country no longer have to bear all the risks
associated with the domestic economic activities because foreign investors will bear some of
these risk when they invest in the country. On the other hand, the domestic investors will bear
some foreign risks when they buy foreign securities. As a result, the domestic investors will
benefit from the market liberalization through the process of diversification since some of the
domestic and foreign risks will offset each other (e.g., Stulz (1999) and Henry (2000b)).
Therefore, the required rate of return for domestic investors will be smaller due to the lower risk
they are bearing. Consequently, one would expect that market liberalization would reduce the
cost of capital for the country. Panel E of Table 3 confirms this conjecture. For example, the
predicted average monthly total risk premium has decreased from 0.177% in the preliberalization sub-period to -0.166% in the post-liberalization sub period for India. This is also
true for the other countries except Malaysia.
To examine whether these decreases of cost of
capital are really statistically significant, an OLS was run to see if the slope coefficient for the
liberalization dummy variable for each emerging market is significantly negative. Panel A of
Table 4 shows that the slope coefficients are negative in all cases except Malaysia, and are
significant in three cases: Malaysia, Philippines, and Taiwan, which confirms the conclusion that
the costs of capital for these markets have indeed decreased because of their liberalization.
21
Another important implication of market liberalization is the issue of excess price
volatility induced by foreign capital movements. Government officials in emerging markets
were concerned whether their stock markets would become more volatile when they decided to
open up their markets to foreign investors. Panel E of Table 3 presents the conditional volatility
calculated from before and after the liberalization date for each market. As can be seen, the
conditional volatility has decreased from 9.705% to 8.907% for Malaysia, from 11.268% to
11.011% for Philippines, and from 14.345% to 11.465% for Taiwan since they were opened up
to the foreign investors. For the other three markets, it appears that their conditional volatilities
have increased somewhat, but these increases are due to the market turmoil during the 1997-98
Asian crisis. If the conditional volatility is calculated from the liberalization date up to the
outbreak of the crisis in middle-1997 for each market, then we would expect to see a decrease in
the conditional volatility for most of the markets. To officially test this hypothesis, an OLS was
run on the predicted conditional volatility over the pre-crisis sample for each market. Panel B of
Table 4 shows that the conditional volatility has indeed diminished in four of six markets (Korea,
Malaysia, Philippines, and Taiwan) since their liberalization dates because the slope coefficients
are statistically significant and negative for these markets. This result is consistent with Bekaert
and Harvey (1997), and Henry (2000b), who also find that the volatilities for emerging markets
have diminished through the process of liberalization. One explanation for this finding is that the
liberalization process has expanded the base of the domestic stock market due to the influx of
foreign institutional investors who make their investment decisions based more on a rational
investment analysis focusing on fundamental values.
22
6. Summary and Concluding Remarks
In this paper, I have developed a dynamic ICAPM in the absence of PPP in an attempt to
examine market integration and contagion using data from six Asian emerging countries. In
testing market integration, the dynamic ICAPM allows full market segmentation until the official
liberalization date and full market integration thereafter. To test pure contagion effect between
foreign exchange and stock markets, I first control for the economic fundamentals or systematic
risks and overall mean spillovers. I then allow the past return shocks from one market to affect
the current returns of the other market during the 1997-98 Asian crisis. This study contributes
our understanding in the dynamics of asset returns in emerging markets in several ways. First of
all, unlike previous studies on market integration (e.g., Bekaert and Harvey (1995) and De Santis
and Imrohoroglu (1997)), PPP is not assumed when testing market integration/segmentation
hypothesis. Secondly, previous studies have failed to control for the economic fundamentals
when testing contagion. In this paper, I rely on the developed ICAPM, which provides me a
theoretical basis in selecting the economic fundamentals. The economic fundamentals under
ICAPM are the world market and currency risks, so the evidence of contagion is based on testing
whether idiosyncratic risks - the part that cannot be explained by the world market and currency
risks, are significant in describing the return dynamics of foreign exchange and stock markets for
each emerging country during the crisis. Finally, a parsimonious parameterization of asymmetric
trivariate GARCH-M process is employed to model the conditional covariance matrix of asset
returns.
Since the ICAPM is fully parameterized, several interesting statistics including
estimated risk premia and conditional volatility can be recovered. The empirical results can be
summarized as follows.
23
First, both currency and world market risks are priced, suggesting that omitting currency
risk in pricing international assets might give rise to model misspecification as I have shown to
be the case for De Santis and Imrohoroglu (1997). Since currency risk is priced, investors are
compensated for bearing such risk, and thus should not be discouraged by more flexible
exchange rate regimes from investing in emerging markets. Second, I find that the stock markets
for India, Korea, Malaysia, Philippines, and Thailand are segmented from the world capital
markets before their liberalization dates, but have become fully integrated since then. Moreover,
the market liberalization has reduced the cost of capital and price volatility for most of countries.
Finally, as for the contagion effects, data from Korea, Malaysia, and Philippines indicate strong
feedback relations between stock prices and exchange rates. This feedback relation is consistent
with the predictions of both the stock- and flow-oriented exchange rate models. Data from India
and Taiwan show that exchange rates led stock prices, whereas that of Thailand fails to reveal
any recognizable pattern during the crisis.
24
References
Adler, M., Dumas, G., 1983. International portfolio choice and corporate finance: A synthesis.
Journal of Finance 38, 925-984.
Ajayi, R.A., Mougoue, M., 1996. On the dynamic relation between stock prices and exchange
rates. The Journal of Financial Research XIX(2), 193-207.
Bae, K.-H., Karolyi, G. A., Stulz, R., 2000. A New approach to measuring financial contagion.
NBER Working Paper, No.7913.
Baig, T., Goldfajn, I., 1999. Financial market contagion in the Asian Crisis. IMF Staff Papers,
46, 167-195.
Bekaert, G., Hodrick, R.J., 1992. Characterizing predictable components in excess returns on
equity and foreign exchange markets. Journal of Finance 47, 467-508.
Bekaert, G., Harvey, C.R., 1995. Time-varying world market integration. Journal of Finance 50,
403-444.
Bekaert, G., Harvey, C.R., 1997. Emerging equity market volatility. Journal of Financial
Economics 43, 29-77.
25
Bekaert, G., Harvey, C.R., Lumsdaine, R., 2000. Dating the integration of world equity markets.
Working paper, Columbia University, Duke University, and Brown University.
Bekaert G., Harvey, C.R., Lundblad, C., 2000, Emerging equity markets and economic
development. NBER Working Paper, No. 7763.
Bekaert G., Harvey, C.R., Lundblad, C., 2001. Does financial liberalization spur growth? NBER
Working Paper, No. 8245.
Bekaert, G., Harvey, C. R., Ng, A., 2002. Market integration and contagion. Working Paper,
Columbia University, Duke University, and Hong Kong University of Science and
Technology.
Bekaert, G., Wu, G., 2000. Asymmetric volatility and risk in equity markets. Review of
Financial Studies, 13, 1-42.
Bera, A.K., Higgins, M.L., 1993. ARCH models: properties, estimation and testing. Jouranl of
Economic Surveys 4, 305-362.
Black, F., 1972. Capital market equilibrium with restricted borrowing. Journal of Business 45,
444-454.
26
Bollerslev, T., 1986. Generalized autoregressive conditional heteroskedasticity. Journal of
Econometrics 31, 307-327.
Bollerslev, T., 1987. A conditional heterosckedastic time series model for speculative prices and
rates of return. Review of Economics and Statistics 69, 542-547.
Bollerslev, T., Engle, R.F., 1986. Modelling the persistence of conditional variances.
Econometric Reviews 5, 1-50.
Bollerslev, T., Engle, R. F., Wooldridge, J. M., 1988. A capital asset pricing model with timevarying covariance. Journal of Political Economy 96, 116-131.
Bollerslev, T., 1990. Modelling the coherence in short-run nominal exchange rates: a
multivariate generalized ARCH approach. Review of Economic and Statistics, 72, 498505.
Bollerslev, T., Wooldridge, J.M., 1992. Quasi-maximum likelihood estimation and inference in
dynamic models with time-varying covariances. Econometric Review 11, 143-172.
Calvo, S,. and Reinhart, C. R., 1996. Capital flows to Latin America: Is there evidence of
contagion Effects?. In Private Capital Flows to Emerging Markets, ed. by G. A. Calvo,
M. Goldstein, and E. Hochreittter. Institution for International Economics: Washington
D.C.
27
Carrieri, F., Errunza, V., Hogan, K., 2001. Characterizing world market integration through time.
Working Paper, McGill University and Barclays Global Investors.
De Santis, G., Imrohoroglu, S., 1997. Stock returns and volatility in emerging financial markets.
Journal of International Money and Finance 16, 561-579.
De Santis, G., Gerard, B., 1997. International asset pricing and portfolio diversification with
time-varying risk. Journal of Finance 52, 1881-1912.
De Santis, G., Gerard, B., 1998. How big is the premium for currency risk?. Journal of Financial
Economics 49, 375-412.
Dornbusch, R., Fischer, S., 1980. Exchange rates and the current account. American Economic
Review 70, 960-971.
Dornbusch, R., Park, C.Y., Claessens, S., 1999. Contagion: How it spreads and how it can be
stopped. Working Paper, MIT.
Dumas, B., 1993. A test of the international CAPM using business cycles indicators as
instrumental variables. In J. Frankel, Eds.: The internationalization of equity markets
(university of Chicago Press, Chicago, IL).
28
Dumas, B., 1994. Partial-equilibrium vs general-equilibrium models of international capital
market equilibrium. NBER Working Paper, No. 4446.
Dumas, B., Solnik, B., 1995. The world price of exchange rate risk. Journal of Finance 50, 445479.
Ebrahim, S., 2000. Volatility transmission between foreign exchange and money markets. Bank
of Canada Working Paper, No.16.
Edwards, S., 1998. Interest rate volatility, capital controls and contagion. NBER Working Paper,
No.6756.
Edwards, S., Susmel, R., 1999. Interest rate volatility in emerging markets: Evidence from the
1990s. Working Paper, UCLA and University of Houston.
Eichengreen, B., Rose, A. K., Wyplosz, C., 1996. Contagious currency crises. NBER Working
Paper, No.5681.
Engle, R. F., Ng, V. K., Rothschild, M., 1990. Asset pricing with a FACTOR-ARCH covariances
structure: Empirical estimates for Treasury bills. Journal of Econometrics, 45, 213-237.
Engle, R. F., Ng, V. K., 1993. Measuring and testing the impact of news on volatility. Journal of
Finance, 48, 1749-1778.
29
Engle, R. F., Kroner, K., 1995. Multivariate simultaneous GARCH. Econometric Theory, 11,
122-150.
Errunza, V., Losq, E., 1985. International asset pricing under mild segmentation: Theory and
test. Journal of Finance 40, 105-124.
Errunza, V.R., Losq, E., Padmanabhan, P., 1992. Tests of integration, mild segmentation and
segmentation hypotheses. Journal of Banking and Finance 16, 949-972.
Forbes, K., Rigobon, R., 1999. Measuring contagion: conceptual and empirical issues. Working
Paper, MIT.
Forbes, K., Rigobon, R., 2002. No contagion, only interdependence: Measuring stock market comovements. Journal of Finance 57, 2223-2261.
Frankel, J. A., 1983. Monetary and portfolio-balance models of exchange rate determination, in
J. S. Bhandari and B. H. Putnam, eds: Economic Interdependence and Flexible Exchange
Rates (MIT Press, Cambridge, MA).
Glosten, L. R., Jagannathan, R., Runkle, D., 1993. On the relation between expected value and
the volatility of the nominal excess return on stocks. Journal of Finance, 48, 1779-1801.
30
Hardouvelis, G., Malliaropulos, D., Priestley, R., 1999. EMU and European stock market
integration. CEPR Discussion Paper No.2124.
Henry, P.B., 2000a. Equity prices, stock market liberalization, and investment. Journal of
Financial Economics 58, 301-334.
Henry, P.B., 2000b. Stock market liberalization, economic reform, and emerging market equity
prices. Journal of Finance 55, 529-564.
Jorion, P., Schwartz, E., 1986. Integration vs. segmentation in the Canadian stock market.
Journal of Finance 41, 603-616.
Kaminsky, G. L., Reinhart, C. M., 2000. On crises, contagion, and confusion. Journal of
International Economics, forthcoming.
Kroner, K. F., Ng, V. K., 1998. Modeling asymmetric comovements of asset returns. Review of
Financial Studies 11, 817-844.
Lintner, J., 1965. The valuation of risk assets and the selection of risk investments in stock
portfolios and capital budgets. Review of Economics and Statistics 47, 13-37.
Masson, P., 1998. Contagion: Monsoonal effects, spillovers, and jumps between multiple
equilibria. IMF Working Paper, WP/98/142.
31
Mossin, J., 1966. Equilibrium in a capital asset market. Econometrica 34, 768-783.
Obstfeld, M., 1994. Risk taking, global diversification and growth. American Economic Review
84, 1310-1329.
Park, Y. C., and Song, C.-Y., 1999. Financial contagion in the East Asian crisis-with special
reference to the Republic of Korea. Working Paper, Korea University.
Rogoff, K., 1996, The purchasing power parity puzzle. Journal of Economic Literature 34, 647668.
Sachs, J., Tornell, A., Velasco, A., 1996. Financial crises in emerging markets: The lessons from
1995. NBER Working Paper, No.5576.
Sharpe, W., 1964. Capital asset prices: A theory of market equilibrium under conditions of risk.
Journal of Finance 19, 425-442.
Solnik, B.H., 1974. An equilibrium model of the international capital market. Journal of
Economic Theory 8, 500-524.
Stehle, R., 1977. An empirical test of alternative hypotheses of national and international pricing
of risky assets. Journal of Finance 32, 493-502.
32
Stulz, R.M., 1981. A model of international asset pricing. Journal of Financial Economics 9,
383-406.
Stulz, R.M., 1999. Globalization of equity markets and the cost of capital. Working Paper, The
Ohio State University.
Wheatley, S., 1988. Some tests of international equity integration. Journal of Financial
Economics 21, 177-212.
Van Rijckeghem, C., Weder, B., 1999. Financial contagion: Spillovers through banking centers.
Working Paper, IMF.
33
Table 1: Regulation of Asian emerging markets
Country
India
Korea
Malaysia
Philippines
Taiwan
Thailand
IFC official liberalization date
November 1992
January 1992
December 1988
a
June 1991
January 1991
December 1988
Degree of openness
24% of issued share capital
10% of capital of listed companies; 25% after July 1992
30% for banks and institutions; 100% for remaining stocks
100% foreign ownership
Investable up to 10%
Investable up to 49%
a
The IFC official liberalization date for Philippines is October 1989, which is hard to justify according to Bekaert and Harvey (2000), who
argue that a Foreign Investment Act was signed into law in June 1991 by Philippine government. Under this Act, all restrictions on foreign
investments were removed over a period of three years. As a result, I chose June 1991 as the liberalization date for Philippines.
34
Table 2: Summary Statistics
The six US-dollar denominated monthly emerging equity/currency returns are India ( IFCIND / INDNRUS ), Korea
( IFCKOR / SKORWUS ), Malaysia ( IFCMAL / MALAYUS ), Philippines ( IFCPHI / PHILPUS ), Taiwan
( IFCTAI / TAIWDUS ), and Thailand ( IFCTHA / THAIBUS ). 1-month Eurodollar deposit rate is used to calculate
excess equity returns. The log first differences of bilateral exchange rates between the U.S. dollar and each of the
six emerging currencies are used to proxy the currency risk. The Bera-Jarque ( B − J ) tests normality based on
both skewness and excess kurtosis and is distributed χ 2 with two degrees of freedom. LB (20) and LB 2 (20)
denote the Ljung-Box test statistics for up to the 20th order autocorrelation of the raw and squared returns,
respectively. * and ** denote statistical significance at the 5% and 1% level, respectively.
Sample period
1981:02 –
2001:03
Mean
Std. Dev.
Min
Max
B−J
IFCIND
0.643
8.598
-27.948
30.210
4.677
LB (20)
30.000
2
1981:02 –
1985:01 –
1986:02 –
2001:03
2001:03
2001:03
Panel A: Stock returns (%)
IFCKOR
IFCMAL
IFCPHI
0.545
1.153
0.185
10.996
11.191
10.238
-40.888
-34.672
-37.365
53.600
38.467
43.008
107.430**
91.088**
21.180**
20.905
43.525**
33.294*
1985:01 –
2001:03
1981:02 –
2001:03
IFCTAI
1.084
12.924
-43.888
42.749
22.954**
IFCTHA
0.385
11.112
-41.281
38.454
86.193**
18.820
46.458**
58.874**
156.439**
TAIWDUS
0.098
1.376
-6.827
4.581
204.351**
THAIBUS
-0.312
2.865
-21.023
14.161
5921**
LB (20)
50.007**
Mean
Std. Dev.
Min
Max
B−J
INDNRUS
-0.731
1.845
-19.561
5.911
22929**
LB (20)
28.054
90.565**
25.742
20.652
79.009**
38.829**
9.492
15.522
149.830**
34.615*
48.652**
130.942**
2
LB (20)
141.580**
103.919**
14.846
Panel B: Currency returns (%)
SKORWUS
MALAYUS
PHILPUS
-0.268
-0.232
-0.525
2.974
2.376
3.102
-35.889
-15.117
-14.602
8.926
14.483
8.628
80759**
3268.913**
373.171**
35
Table 3: Quasi-Maximum Likelihood estimation of the dynamic ICAPM: Trivariate GARCH(1,1)-M
ri ,t = (1 − LDi ,t )λ w,t −1 hiw ,t + λ c ,t −1 hic ,t + LDi ,t λl ,t −1 hi ,t + (δ ii ε i ,t −1 + δ ci ε c ,t −1 ) + CDt (ω ii ε i ,t −1 + ω ci ε c ,t −1 ) + ε i ,t
rc ,t = λ w,t −1 hcw,t + λ c ,t −1 hc ,t + (δ cc ε c ,t −1 + δ ic ε i ,t −1 ) + CDt (ω cc ε c ,t −1 + ω ic ε i ,t −1 ) + ε c ,t
rw,t = λ w,t −1 hw,t + λc ,t −1 hcw,t + ε w,t
where
λ w,t −1 = exp(φ w' z t −1 ) ; λc ,t −1 = φ c' z t −1 ; z t −1 = {CON , USDIV , ∆USTP, USDP, WORLD}
λl ,t −1 = exp(φ l' z l ,t −1 ) ; z l ,t −1 = {CON , LOCAL} ε t | I t −1 ~ N (0, H t )
H t = C ' C + A ' ⋅ H t −1 ⋅ A + B ' ⋅ ε t −1ε t' −1 ⋅ B + D ' ⋅ η t −1η t' −1 ⋅ D
“ LDi ,t ” is a liberalization dummy variable for emerging market i . “ CDt ” is a dummy variable for Asian crisis period.
Robust t-statistics are given in parentheses. * and ** denote statistical significance at the 5% and 1% level, respectively.
Panel A: Conditional mean process - prices of world market, currency, and local market risks
CON
IFCIND
φw
φc
IFCKOR
φw
φc
IFCMAL
φw
φc
1.926
(3.324)**
-6.563
(-1.628)
21.309
(4.838)**
-0.081
(-0.011)
-0.873
(-0.513)
-0.493
(-0.149)
63.534
(2.339)*
-2.743
(-0.882)
-5.069
(-0.251)
IFCPHI
φw
φc
(-0.474)
-8.400
(-2.029)*
0.052
(1.150)
0.250
(1.886)
-10.365
(-1.795)
2.992
(1.912)
3.845
(0.315)
1.487
(1033.86)**
0.364
(201.16)**
-0.064
(-1.585)
0.220
(9.755)**
1.243
(2.408)*
IFCTAI
φw
φc
(-2.500)*
-1.749
(-2.084)*
1.383
(2.171)*
-4.037
(-1.515)
41.714
(5.487)**
-1.660
(-0.447)
-1.138
(-2.251)*
11.849
(5.629)**
-28.995
(-2.206)
66.746
(2.814)**
59.760
(10.378)**
IFCTHA
φw
φc
-4.812
(-2.340)*
28.909
(3.136)**
-5.054
(-1.995)*
39.822
(7.654)**
-44.728
(-8.154)**
48.461
(3.433)**
-7.216
(-1.307)
(3.030)** 1670.390
(5.375)**
174.055
(4.860)**
LOCAL
(-12.254)**
9.338
(3.358)**
1.052
(1.380)
2.882
(5.439)**
-4.718
(-8.670)**
-1.698
(-3.051)** -34.437 (-10.686)**
LOCAL
0.915
(2.696)**
-0.380
(-0.420)
-2.039
(-1.382)
1.784
(1.831)
-165.663 (-5.277)**
-3.709
-41.400 (-3.025)** 215.528
(-1.291)
LOCAL
-1.407
(-0.667)
2.085
(1.116)
0.290
(0.299)
2.537
(2.585)**
-8.100
(-2.339)*
25.841
(3.220)**
-7.598
(-0.837)
-8.749
(-1.505)
3.130
(1.016)
78.593
(3.273)**
108.974
(1.214)
107.946
(3.241)**
CON
φl
-17.320 (-5.062)**
-9.509
CON
φl
LOCAL
-1.572
CON
φl
LOCAL
-0.775
CON
φl
WORLD
(2.075)*
CON
φl
USDP
0.666
CON
φl
∆USDP
USDIV
-3.577
(-2.929)*
LOCAL
1.896
(0.077)
36
Table 3 (continued)
Panel B: Mean spillover and contagion
Mean spillover
Contagion
δ ii
δ ci
δ cc
δ ic
ω ii
ω ci
ω cc
ω ic
0.007
(0.094)
0.007
(0.586)
0.027
(9.967)**
0.006
(2.300)*
0.078
(0.548)
0.561
(779.581)**
0.275
(0.927)
0.035
(17.066)**
-0.221
(-0.514)
-1.530
(-15.919)**
-0.054
(-3.311)**
0.132
(6.068)**
0.480
(1.940)
-0.654
(-819.963)**
1.140
(0.640)
1.034
(535.050)**
Malaysia
0.007
(0.254)
-0.011
(-1.717)
0.397
(10.589)**
1.226
(4.159)**
0.057
(0.480)
-0.146
(-9.947)**
1.019
(8.866)**
2.170
(4.928)**
Philippines
0.517
(18.644)**
0.005
(1.523)
0.182
(4.807)**
0.200
(1.226)
0.504
(5.519)**
1.047
(16.086)**
0.098
(1.400)
-0.709
(-4.512)**
Taiwan
-0.001
(-0.013)
0.007
(1.075)
0.662
(5.842)**
0.605
(0.581)
0.836
(5.623)**
0.100
(5.037)**
-0.520
(-4.084)**
-2.092
(-1.526)
Thailand
0.179
(2.801)**
-0.006
(-0.848)
0.273
(1.750)
-0.249
(-0.706)
-0.002
(-0.011)
0.050
(0.946)
-0.393
(-3.486)**
1.358
(1.631)
India
Korea
37
Table 3 (continued)
Panel C: Conditional variance process
IFCIND
INDNRUS
WORLD
ai
0.948
(67.873)**
0.102
(0.723)
0.540
(2.409)*
bi
di
0.219
(3.192)**
1.709
(7.805)**
0.023
(0.352)
-0.048
(-0.254)
-0.677
(-0.676)
1.739
(0.402)
IFCKOR
ai
bi
di
SKORWUS
WORLD
0.921
(170.515)**
-0.543
(-121.048)**
0.335
(8.492)**
0.289
(11.373)**
0.966
(1678.70)**
-0.056
(-1.114)
0.916
(2.087)*
-0.225
(-1.717)
3.183
(0.868)
IFCMAL
MALAYUS
WORLD
ai
0.828
(55.878)**
-0.467
(-9.720)**
0.622
(17.227)**
bi
di
0.420
(14.220)**
0.168
(1.846)
0.125
(1.691)
1.145
(3.292)**
39.982
(22.828)**
2.355
(2.623)**
IFCPHI
ai
bi
di
PHILPUS
WORLD
0.717
(17.808)**
0.112
(1.094)
0.754
(7.036)**
0.274
(4.849)**
1.953
(19.687)**
0.249
(4.646)**
-1.329
(-2.976)**
-3.789
(-1.885)
-1.083
(-1.412)
IFCTAI
TAIWDUS
WORLD
ai
0.866
(61.816)**
-0.015
(-0.123)
0.072
(0.187)
bi
di
-0.212
(-0.985)
0.574
(5.987)**
-0.084
(-0.314)
-1.004
(-2.732)**
-34.604
(-5.701)**
-0.026
(-0.021)
IFCTHA
ai
bi
di
THAIBUS
WORLD
0.913
(38.849)**
-0.228
(-5.470)**
0.909
(48.239)**
-0.395
(-6.710)**
1.065
(8.852)**
-0.231
(-2.928)**
-0.001
(-0.009)
-0.002
(-0.006)
0.001
(0.005)
38
Table 3 (continued)
a
Panel D: Diagnostics for the residuals
IFCIND INDNRUS
WORLD
B−J
5.858
348.691**
61.967**
LB (20)
25.461
13.626
24.905
26.808
31.960*
9.596
IFCKOR
SKORWUS
B−J
WORLD
3.602
118.661**
82.316**
LB (20)
13.542
77.891**
20.219
15.720
20.210
12.157
IFCMAL
MALAYUS
B−J
WORLD
77.810**
340.193**
2.792
LB (20)
16.491
31.485*
23.990
2
LB (20)
2
LB (20)
2
LB (20)
10.014
10.456
15.143
IFCPHI
PHILPUS
B−J
WORLD
10.283**
39.005**
30.808**
LB (20)
24.383
36.637*
16.535
20.490
8.200
10.015
IFCTAI
TAIWDUS
B−J
WORLD
40.738**
3.101
24.483**
LB (20)
16.527
18.079
25.945
17.914
10.401
17.054
IFCTHA
THAIBUS
WORLD
2
LB (20)
2
LB (20)
B−J
LB (20)
2
LB (20)
a
60.123** 44297.219**
28.517**
20.472
11.175
16.228
10.118
0.350
12.833
LB ( 20 ) and LB 2 ( 20) are the Ljung-Box test statistics of order 20 for serial correlation in the standardized residuals and standardized
residuals squared.
respectively.
B−J
is the Bera-Jarque test statistic for normality. * and ** denote statistical significance at the 5% and 1% level,
39
Table 3 (continued)
a
Panel E: Hypothesis testing concerning time-varying prices of world market, currency, and local market risk
IFCIND IFCKOR IFCMAL IFCPHI IFCTAI IFCTHA
Null hypothesis
k
k
260.242 4.13E+9 1296.648 1108.214 363.231 424.336
1. H0: φ = φ = 0
w
c
∀k = CON , USDIV , ∆USTP , USDP, WORLD
2. H0:
φ wk = φ ck = 0
∀k = USDIV , ∆USTP , USDP, WORLD
3. H0:
φ wk = 0
∀k = CON , USDIV , ∆USTP , USDP, WORLD
4. H0:
φ wk = 0
∀k = USDIV , ∆USTP , USDP, WORLD
5. H0:
[0.000]
[0.000]
[0.000]
[0.000]
59.817
7.84E+8
968.011
704.474
72.452
64.795
[0.000]
[0.000]
[0.000]
[0.000]
[0.000]
[0.000]
244.435
248.630
137.139
318.52
141.310 258.324
[0.000]
[0.000]
[0.000]
[0.000]
[0.000]
[0.000]
49.647
14.750
61.209
40.843
17.844
11.591
[0.000]
[0.005]
[0.000]
[0.000]
[0.001]
[0.020]
9.714
2.63E+9
693.274
449.080
60.763
20.028
[0.083]
[0.000]
[0.000]
[0.000]
[0.000]
[0.001]
7.628
1.81E+8
693.064
449.019
45.936
19.600
[0.106]
[0.000]
[0.000]
[0.000]
[0.000]
[0.000]
9.271
25.797
194.711
7.371
2.653
9.143
φ ck = 0
∀k = USDIV , ∆USTP , USDP, WORLD
7. H0:
[0.000]
φ ck = 0
∀k = CON , USDIV , ∆USTP , USDP, WORLD
6. H0:
[0.000]
φ lk = 0
[0.009]
[0.000]
[0.000] [0.025] [0.265] [0.010]
∀k = CON , LOCAL
Panel F: Estimated time-varying risk premia and conditional volatility (%)
Risk premium
IFCIND IFCKOR IFCMAL IFCPHI IFCTAI IFCTHA
Before the liberalization
0.177
0.442
-0.246
3.149
2.352
0.241
After the liberalization
-0.116
-0.745
0.132
0.394
0.340
0.238
Conditional volatility
Before the liberalization
8.276
9.205
9.705
11.268 14.345
6.603
After the liberalization
8.774
10.228
8.907
11.011 11.465 12.007
a
Robust Wald statistics are in the first row, and the corresponding p-values are in the brackets.
40
Table 4: The impact of liberalization on the estimated risk premium and conditional volatility
RPi ,t = β 0 + β 1 LDi ,t + u i ,t ∀i
CVi ,t = β 0 + β 1 LDi ,t + u i ,t ∀i
where “ RP ” and “ CV ” denote estimated risk premium and conditional volatility, respectively; “ LDi ,t ” is a
liberalization dummy variable for each emerging market i . t-statistics are given in parentheses. * and ** denote
statistical significance at the 5% and 1% level, respectively.
Panel A:
IFCIND
IFCKOR
IFCMAL
IFCPHI
IFCTAI
IFCTHA
intercept
LD
intercept
LD
intercept
LD
intercept
LD
intercept
LD
intercept
LD
0.002
-0.002
0.004
-0.012
-0.002
0.004
0.031
-0.028
0.023
-0.020
0.004
-0.002
RP
(2.188)*
(-2.342)*
(0.869)
(-1.585)
(-0.350)
(0.470)
(4.297)**
(-3.035)**
(6.892)**
(-4.695)**
(1.380)
(-0.649)
Panel B:
Full sample
0.083
(100.67)**
0.005
(3.927)**
0.092
(59.073)**
0.010
(4.456)**
0.097
(23.340)**
-0.008
(-1.674)
0.113
(68.665)**
-0.003
(-1.261)
0.143
(38.224)**
-0.029
(-6.110)**
0.078
(17.523)**
0.041
(7.230)**
41
CV
Pre-crisis
0.083
(92.317)**
0.007
(4.097)**
0.092
(98.539)**
-0.004
(-2.786)**
0.097
(30.278)**
-0.019
(-5.017)**
0.113
(70.159)**
-0.006
(-2.629)**
0.143
(33.604)**
-0.027
(-4.660)**
0.078
(28.867)**
0.015
(3.952)**