Professional Documents
Culture Documents
Spillovers of Stock Return Volatility To Asian Equity Markets From Japan and The US
Spillovers of Stock Return Volatility To Asian Equity Markets From Japan and The US
Spillovers of Stock Return Volatility To Asian Equity Markets From Japan and The US
Abstract
This paper examines the magnitude of return and volatility spillovers from Japan and the
US to seven Asian equity markets. I construct a volatility spillover model that deals with the
US shock as an exogenous variable in a bivariate EGARCH for Japan and Asian markets.
First, only the influence of the US is important for Asian market returns; there is no influence
from Japan. Second, the volatility of the Asian market is influenced more by the Japanese
market than by the US. Third, there exists an adverse influence of volatility from the Asian
market to the Japanese market.
# 2003 Elsevier Science B.V. All rights reserved.
Keywords: Asian markets; Regional and world market factor; Spillover; Volatility
1. Introduction
1042-4431/03/$ - see front matter # 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S1042-4431(03)00015-5
384 T. Miyakoshi / Int. Fin. Markets, Inst. and Money 13 (2003) 383 /399
Most existing studies on volatility spillovers, such as Base and Karolyi (1994),
Karolyi (1995) and Karolyi and Stulz (1996), focus on how an American ‘‘world’’
market influences other markets but do not distinguish between influences from
regional and world markets. However, Ng (2000) constructs a volatility spillover
model by assuming that there are three shocks including local, regional and world,
and that each shock is significant.
Ng (2000) uses more detailed research than previous papers and clarifies the role
of a regional market. Her major contribution is her analyzes on return volatility in
Asia. As expected, she finds that for Asian markets the world shock of the US is
stronger than the regional one of Japan. Thus, Ng (2000) analyzes a regional and
world shock as exogenous variables in the univariate volatility spillover model for
the Asian markets. She uses weekly data from January 1975 to December 19961.
Recently, a large increase of portfolio and direct investment from Japan to the
Asian countries in the latter half of 1990s is shown by the Bank of Japan (2001b) and
Ministry of Finance (2000) official data, except during the period of Asian crisis in
1997. Also, Fornari and Levy (2000) point out that Japanese monetary expansion
primarily translates into capital outflows (rather than domestic demand) toward
Asian countries. Thus, Japanese international investors have held a large amount of
Asian assets and hence the Asian market will affect the Japanese market through the
portfolio management of assets of both countries. The analysis for the latter half of
1990s will have to consider the repercussion of the Asian markets, though Japan is a
regional market. This strong tie between Japan and Asian markets is worthy of
notice.
This paper considers a different volatility spillover model from Ng (2000), noting
the strong relationship between Japan and Asian markets. For example a world
shock of the US is only an exogenous variable in a bivariate volatility spillover model
for Japan and an Asian country. In contrast to Ng (2000), this paper insists on the
endogenity of the Japanese market shock though Japan may be a regional market
against the Asian local markets. Omitting the period during the Asian crisis when
stock data displayed the abnormal behavior, but obtaining the sufficient number of
data, this paper uses recent daily data from January 1998 to April 2000.
This paper draws a major result that a regional factor is stronger than a world one
for Asian market volatility, and also considers why the results are different from Ng
(2000). In addition, the paper finds the endogenity of the Japanese variables and
supports our setting different from Ng. Finally, the paper derives the different policy
implication by asking the following question. Can Asian governments and
international investors obtain some profits by using signals from a market that
opens earlier?
1
Ng (2000) also shows that liberalization events (such as capital market reform and country fund
launching), exchange-rate changes, the number of American Depositary Receipt listings, sizes of trade,
and a country’s fund premiums affect the relative importance of the world and regional market factors
over time.
T. Miyakoshi / Int. Fin. Markets, Inst. and Money 13 (2003) 383 /399 385
2
EGARCH model overcomes the main drawbacks in GARCH for two reasons. First, the
nonnegativity constraints on the coefficients are imposed to ensure that variance remains nonnegative
for all t with probability one. Second, only the magnitude (and not the positivity or negativity of
unanticipated excess returns) determines feature variance.
386 T. Miyakoshi / Int. Fin. Markets, Inst. and Money 13 (2003) 383 /399
o t Ft Zt ; Zt iidN(0; S) (4)
where
!
1 d h1=2 0
S ; ½d½ B1 and Ft jaja;t 1=2 : (5)
d 1 0 hAsAs;t
Under this formulation for error terms, we can derive the covariance matrix Ht ,
!
hjaja;t d(hjaja;t hAsAs;t )1=2
Ht Var(o t ½Ct1 ) Ft SFt (6)
d(hjaja;t hAsAs;t )1=2 hAsAs;t
where
log hiit gi0 gi1 log hiit1 gj2 log hjjt1 g(Zit1 )fii log R2us;t1
It should follow that the maximum likelihood estimate û for u will be asymptotically
normal and consistent with a covariance matrix equal to the inverse of Fisher’s
information matrix. As Nelson (1991) and Braun et al. (1995) do, we assume this
T. Miyakoshi / Int. Fin. Markets, Inst. and Money 13 (2003) 383 /399 387
All data are the daily data purchased from Nomura Research Institute Japan. In
the database, the daily return rt consisted of daily stock closing price Pt , which is
measured in local currency3. Our measurements include Korea Composite Stock
Price Index in Korea, Taiwan Stock Exchange Capitalization Weighted Index in
Taiwan, Strait Times Index in Singapore, Stock Exchange of Thailand Index in
Thailand, The JSX Composite Index in Indonesia, KLSE Composite Index in
Malaysia, Hang Seng Index in Hong Kong, TOPIX in Japan, and S&P500 in the US.
The sample period ranges from 1 January 1998 to 30 April 2000.
The number of observations is approximately 500 for each country4. The data for
the whole period are illustrated in the Appendix A. The data of stock price exhibit
large fluctuations during the whole period.
The graphs for all countries show comovement of Japanese and US stock indexes.
Ng (2000) uses the bivariate GARCH model between Japan (as a regional market)
and the USA (as a global market), picks out the idiosyncratic shocks of both
countries and analyzes the exogenous idiosyncratic effects of the Japanese and the
US return and volatility on each Asian stock market expressed by an univariate
GARCH model. Ng (2000), (p. 219) also suggests other possible models. A
straightforward approach is to estimate a model for one of the Asian market
3
I measure the stock returns in local currency just as Base and Karolyi (1994) and Karolyi (1995) do in
their studies. On the other hand, the stock returns in Karolyi (1995) and Ng (2000) is denominated in US
dollars. Note that when market returns are denominated in US dollars, international investors are
assumed to be unhedged against foreign exchange risk. However, Dumas and Solnik (1995) and De Santis
and Gérard (1998) insist the importance of currency risk on stock markets. Thus, we assume that the
investors are hedged against it.
4
Market holidays differ depending on the country. The same date data in other countries
corresponding to a holiday in one country is deleted from the data set, in each model for Japan, one
Asian country, and the US. The data for return is simply made of the remained data, as seen in Section 2.
The holidays related to the ‘‘old new year’’ holiday (from the end of January to the early February) over 5
days are 10 days for Taiwan in 1998 and 1999 and 7 days for Malaysia in 1998.
388 T. Miyakoshi / Int. Fin. Markets, Inst. and Money 13 (2003) 383 /399
together with Japan and the US in a trivariate GARCH setting. An even more
desirable method is to estimate all Asian markets together in a multivariate GARCH
setting. In this case, the bivariate model with the endogenous Japanese variables will
be more desirable than the univariate model. The model of this paper will be a
bivariate EGARCH between Japan and an Asian market.
The graphs in the Appendix A establish the comovement of the Japanese and the
Asian stock indexes. The comovement suggests the repercussion of the Asian stock
on the Japanese stock and hence the endogenity of Japanese stock variables. The
model I propose here is the bivariate EGARCH model between Japan (as a regional
market) and one of the Asian markets (as a local market). Thus, the paper analyzes
the exogenous effects of US (as a world market) return and volatility on Japan and
Asian countries. Refer to Section 2 for more specifics on return volatility in the US.
The setting proposed by this paper displays one possibility by using daily data (see
Fig. 1). Japanese and the Asian markets operate in a broadly synchronized manner.
Due to the time flow, the Japanese variables are endogenous for the Asian countries
and the USA variables are exogenous. The time difference in terms of Tokyo trading
hours is at most 2 h 30 min for the opening hour and at most 3 h 30 min for the
closing hours between Japan and other Asian countries. The US market opens 4 h
after the Thai market closes, and the Japanese market opens 5 h after the US market
closes. The paper uses the daily data depending on such a time flow5.
Moreover, as mentioned in Section 1, the Japanese official data and Fornari and
Levy (2000) point out that Japanese investors hold a large amount of Asian assets.
Therefore, the Asian market can affect the Japanese market through the portfolio
management of assets of both countries. Hence the Japanese stock variables may be
endogenous.
We check the statistical features of the data. Table 1 lists the country name of
sample stocks, the number of samples and the nonnormality of the unconditional
distribution of daily returns. The table reports the mean, standard deviation of
return, the Kendall /Stuart skewness, the excess kurtosis, and their tests. The Ljung-
Box Q-statistic Q(20) and Q2(20) are reported under the null hypothesis of nonserial
correlation tests in daily return and squared returns, respectively. At significance
levels of the 5%, the null hypotheses (skewness /0 or excess kurtosis/0) and of
nonserial correlation are generally rejected, respectively. These time series have the
typical features of stock returns as fat tail, spiked peak, and the persistence of
variance6. Therefore, the ARCH-type model including such features is appropriate
for analyzing these series.
5
Since there are no overlapping trading hours between the US and any of the Asian markets and the
US return and volatility are exogenous, it is appropriate to use daily open-to-close returns than the daily
closing returns. However, today’s opening prices on Asian markets are already decided, reflecting the
information from the yesterday in the US. Thus, the open-to-close returns lose a part of yesterday’s
information from the US. This criticism will be considered in future work.
6
All computations in the paper have been performed with the computer package WINRATS-32 Version
4.30.
T. Miyakoshi / Int. Fin. Markets, Inst. and Money 13 (2003) 383 /399
Table 1
Summary statistics for daily returns in Japanese and Asian markets: 1 January 1998 /30 April 2000
Country Sample Mean t -Value (mean/0)a Standard deviation Skewnessa Ex-Kurta Q(20)b Q2(20)b
389
390 T. Miyakoshi / Int. Fin. Markets, Inst. and Money 13 (2003) 383 /399
We explain the test of hypotheses. If the parameters lja , lAs , fjaja , and fAsAs are
significantly nonzero, it shows that the exogenous effects of US market return and
return volatility are transmitted to Japan and the Asian markets, respectively.
Similarly, the significant bAs 1 and gja 2 (from Japan to Asian returns and return
volatility), and bja 2 and gAs 2 (from the Asian to the Japanese returns and return
volatility) show how much the endogenous effects explain. Finally, the significant
bja 1, gja 1, bAs 2, and gAs 1 show Japanese and the Asian country’s own effects on the
return and return volatility. These tests are for the null hypotheses of zero
coefficients. However, the tests for the null hypotheses of non-difference between
some coefficients are not implemented.
4. Empirical results
Ng (2000) deals with a regional and a world shock as exogenous variables in the
univariate GARCH model for the Asian markets. However, this paper deals only
with a world shock of the US as exogenous variable in a bivariate EGARCH model
for Japan and each of the Asian countries Korea, Taiwan, Thailand, Singapore,
Malaysia, Indonesia, and Hong Kong.
Table 2 reports the estimation results of endogenous relations between Japan and
the Asian countries, their own effects, and US exogenous effects7. (All d are
significant and satisfied with jd jB/1.) The estimated coefficients show that in all
case,
7
Actually arbitrary initial values for the Maximum Likelihood algorithm could affect the results of
estimated parameters. In Table 2, the initial values are estimated parameters by OLS for parameters in Eq.
(1), 0.1 for error terms in Eq. (1), the sample variances of estimated error in Eq. (1) by OLS for variances in
Eq. (7), and 0.05 for the other parameters in Eq. (7) except for d (/0.0) and gi 0(/the logarithm of the
sample variances). I examine the robustness of the estimation results in Table 2. In particular, the
following initial value set affected the estimation results to some degree. Yet different initial values are all
0.1 for variances in Eq. (7) and the others are the same. However, the results for Eqs. (9) /(12) are not
changed greatly.
T. Miyakoshi / Int. Fin. Markets, Inst. and Money 13 (2003) 383 /399
Table 2
Parameter estimations of bivariate EGARCH (1.1)
bja 2 bAs 1 gAs 2 gja 2 bja 1 bAs 2 gja 1 gAs 1 lja lAs fjaja fAsAs
Korea 0.287** (6.92) /0.022 (/1.18) 0.132 (1.31) /0.303* (/2.31) 0.155 (1.79) 0.052 (1.27) 0.054 (1.01) 1.017** (29.13) 0.367 (1.82) 0.349** (9.00) 0.478** (4.50) 0.017 (1.15) 0.033 (1.21)
Taiwan 0.261** (6.33) /0.017 (/0.48) 0.073 (1.35) 0.002 (0.03) /0.146* (/1.96) 0.042 (0.79) 0.080 (1.45) 0.723** (6.92) 0.779** (16.72) 0.376** (9.14) 0.262** (4.60) 0.014 (1.02) 0.053** (2.47)
Thailand 0.303** (7.19) /0.011 (/0.44) 0.010 (0.12) /0.098** (/4.10) 0.412** (3.51) 0.019 (0.46) 0.072 (1.40) 1.067** (59.08) 0.582** (7.68) 0.335** (7.62) 0.295** (3.39) 0.013* (2.12) /0.027 (/0.92)
Singapore 0.275** (6.49) 0.006 (0.23) /0.008 (/0.14) 0.123** (2.79) 1.232** (4.15) 0.005 (0.12) 0.065 (1.52) 0.675** (9.05) 0.300** (4.75) 0.374** (10.21) 0.432** (6.52) 0.013 (1.22) 0.038 (1.39)
Malaysia 0.281** (6.42) 0.036 (1.87) /0.056 (/0.85) /0.036 (/1.43) 0.094 (1.70) 0.018 (0.43) 0.245** (6.57) 0.876** (17.99) 0.892** (48.72) 0.300** (7.32) 0.297** (5.31) 0.034 (1.70) 0.010 (0.64)
Indonesia 0.161** (3.49) 0.004 (0.13) /0.102 (/1.36) /0.223** (/4.64) 0.945** (4.41) 0.031 (0.79) 0.111** (2.65) 1.232** (60.10) 0.169* (2.91) 0.288** (6.94) 0.370** (4.20) 0.016* (2.05) 0.035 (1.37)
Hong Kong 0.336** (8.56) 0.008 (0.30) /0.049 (/0.73) 0.008 (0.06) 0.596** (3.16) 0.031 (0.68) /0.004 (/0.08) 0.525** (3.85) /0.221 (/1.74) 0.384** (9.54) 0.768** (10.97) 0.068* (2.65) 0.021 (0.90)
Asymptotic t -statistics for the estimated parameters appear in parentheses. **, * Statistically significant at 1 and 5% assuming conditional normality. The
estimated parameters in the table correspond to the following bivariate EGARCH of Japan and each of the Asian countries, abbreviating constant terms and
the g(.).
Rja;t bja1 Rja;t1 bja2 RAs;t1 lja Rus;t1 log hjaja;t gja1 log hjaja;t1 gAs2 log hAsAs;t1 fjaja log R2us;t1
/ : /
RAs;t bAs1 Rja;t1 bAs2 RAs;t1 lAs Rus;t1 log hAsAs;t gAs1 log hAsAs;t1 gja2 log hjaja;t1 fAsAs log R2us;t1
391
392 T. Miyakoshi / Int. Fin. Markets, Inst. and Money 13 (2003) 383 /399
with all significant gja 2 except for Korea and Malaysia and one significant fAsAs out
of seven. The reason for nonsignificant fAsAs could be that their volatility are mainly
explained by the Asian own and the Japanese volatility, gja 2 and gAs 1. The reason
that Malaysian volatility did not affect (or was not affected by) foreign countries
may come from the capital control adopted by Malaysian government during these
periods. The result is significantly different than those of Ng (2000). The different
results depend on the structure of the model and the sample span. More
consideration is given later in this Section 4.
In addition, new findings appear, which cannot be obtained in Ng (2000) model
where the Japanese shock is an exogenous variable. Since, in all cases except for
Korea,
0B½gAs2 ½ B½gja2 ½ (11)
with four significant gAs 2 out of seven, the Asian return volatility adversely affects
the Japan volatility. However, Japanese influences are stronger than those of the
Asian country. This may be due to a strong economic relationship between Japan
and the Asian countries through large amounts of portfolio investment during the
latter half of 1990s. More detailed research is necessary for future study of the
subject.
The result of Ng (2000) is supplemented by the results of my paper. Though the
Asian market returns and volatility are influenced by the US market, their own
effects of the previous day on the present day remain smaller for return and larger
for volatility than the US. The following equation shows:
½lAs ½ ½bAs2 ½; ½fAsAs ½ B½gAs1 ½ (12)
with all nonsignificant bAs 2 except for Malaysia and Indonesia, and all significant
gAs 1 except for Korea and Hong Kong. The significant bAs 2 for Malaysia and
Indonesia may mean that the US return did not fully explain the Asian.
The results of Eq. (11) suggest a strong support for endogenity of the Japanese
variables, in addition to the explanation in Section 3. In fact, both significant gja 2
T. Miyakoshi / Int. Fin. Markets, Inst. and Money 13 (2003) 383 /399 393
and gAs 2 indicate the interaction between the Japanese and the Asian volatility. The
Japanese volatility will not be exogenous. Bekaert and Harvey (1997) make some
explicit assumptions so as to break down the joint likelihood function of a
multivariate model into a set of univariate likelihood. Ng (2000) follows their
method. However, in our model, one of Bekaert and Harvey (1997) assumptions,
‘‘the joint density of the Japanese and the US returns depends only on the
parameters in the bivariate distribution and not on any parameters in the univariate
GARCH model for the Asian markets’’(pp. 40 /41), seems left unsatisfied.
For the careful reader, the issue of endogeneity of Japanese market and of
exogeneity of the US market is still an issue. If we treat the US market as an
endogenous variable, the results will appear different from those of our model. We
have to test the specification of our model. The arguments for the exogeneity of the
variables are not developed on the multivariate ARCH-type model, as opposed to
the arguments of Engle et al. (1983) on the VAR model. Therefore, I estimate the US
market as endogenous and the Japan market as exogenous. Next, I check the
exogeneity or endogeneity of the US market and consider the validity of the model
specification. In Table 3, the bAs 1 and gUS 2 (from the US to the Asian returns and
return volatility) and bUS 2 and gAs 2 (from the Asian to the US returns and return
volatility) show how much the effects explain and the d shows the magnitude of
covariance. These parameters correspond to those in Table 2, changing the Japanese
and the US variables. Except for Taiwan and Indonesia, the US market return and
return volatility (bAs 1, gUS 2) significantly affect those of the Asian in one way. There
is no adverse effect from the Asian to the US. In addition, for Indonesia, because of
d /0, the covariance at the same time between the US and the Asian markets is zero.
In these senses, the model specification with the US market being endogenous is not
appropriate. I rely on the results from the model with the US market being
exogenous and the Japan market endogenous in Table 2.
The paper gives an economic justification about the results on volatility between
Japan and Asian countries. As seen in Fig. 2, the Japanese direct investment (a large
Table 3
The exogeneity of the US market
Korea 0.099* (2.14) 0.028 (1.59) 0.544** (5.17) /0.243 (/1.44) 0.461* (2.41)
Taiwan 0.109* (2.42) /0.052* (/2.14) 0.272** (4.63) 0.048** (2.72) 0.079* (2.43)
Thailand 0.151** (3.23) 0.030 (1.51) 0.316** (3.27) /0.035 (/0.93) 0.072 (1.17)
Singapore 0.177** (3.75) /0.006 (/0.24) 0.452** (6.41) 0.052 (1.23) 0.633** (5.34)
Malaysia 0.147** (3.10) 0.020 (0.93) 0.315** (4.50) /0.000 (/0.02) 0.284** (3.63)
Indonesia 0.041 (0.79) 0.038* (2.11) 0.379** (3.93) /0.123** (/3.54) 0.176** (3.01)
Hong Kong 0.184** (4.34) /0.016 (/0.60) 0.745** (10.15) 0.043 (0.55) 0.677** (3.72)
Asymptotic t -statistics for the estimated parameters appear in parentheses. **, * Statistically significant
at 1, 5% assuming conditional normality. The estimated parameters in the table correspond to the
bivariate EGARCH of the US and each of the Asian countries.
394 T. Miyakoshi / Int. Fin. Markets, Inst. and Money 13 (2003) 383 /399
Fig. 2. The Japanese direct and portfolio investments per year and the credits at the end of the year
against other Asian countries.
8
In Fig. 2, the Ministry of Finance (2000) reported the amount of investment in US dollar before 1994
and in the Japanese Yen after 1995. The amounts after 1995 are converted into the US dollars by using the
exchange rates at the end of the year (International Financial Statistics, line ae, in Japan). The ‘‘Asian’’
means Korea, Singapore, Thailand, Indonesia, Malaysia, and Philippines.
9
In Fig. 2, the ‘‘Credit’’ covers all balance-sheet position vis-à-vis non-residents in any currencies. All
banks authorized to conduct business in the Japan Offshore Market report on their institutions located in
Japan (i.e. including foreign-owned bank affiliates in Japan, but excluding Japanese bank affiliates in
foreign countries) in principle. Positions in Yen are converted into US dollars at the exchange rate
prevailing on the reporting date. The ‘‘Asia’’ means the category of the Asian and Pacific. This is
consistent with the geographical breakdown in the BIS (Bank of International Settlement) statistics.
T. Miyakoshi / Int. Fin. Markets, Inst. and Money 13 (2003) 383 /399 395
Table 4
The portfolio investment and direct investment positions at the end of the years (in a 100 million Japanese
Yen)
Panel A: the Japanese portfolio investment position against the Asian (equity securities )
1996 354 144 12 962 3370 1490 667 5231
1997 272 91 5319 1539 288 219 1386
1998 278 157 4712 870 430 122 518
1999 574 671 7468 2173 696 165 399
2000 471 334 5429 1782 417 85 317
Aa 390 279 7180 1947 665 252 1570
A/JA (%)b 0.01 0.00 0.21 0.56 0.02 0.00 0.05
A/AS (%)b 0.12 0.17 1.29 0.82 1.18 0.47 0.88
Panel B: the Asian portfolio investment position against Japan (equity securities )
1996 235 60 9584 9091 0 3 0
1997 171 0 7407 6436 0 2 1
1998 81 0 4322 5002 0 0 0
1999 1596 6 11 199 4506 4 2 9
2000 1061 190 13 422 12 732 4 1 3
Ba 629 51 9187 7553 2 2 3
B/JA (%)b 0.02 0.00 0.27 0.22 0.00 0.00 0.00
B/AS (%)b 0.19 0.03 1.65 3.19 0.00 0.00 0.00
Panel C: the Japanese direct investment position against the Asian (equity capital on affiliated enterprises )
1996 3176 2606 6441 6273 5308 16 917 3692
1997 4743 10 566 6474 8283 4434 7445 3393
1998 3115 3567 6707 7093 5144 11 376 3157
1999 2852 2174 3679 5273 4140 2937 2340
2000 2865 3321 5205 6313 4958 3731 2785
Ca 3350 4447 5701 6647 4797 8481 3073
C/JA (%)b 0.10 0.13 0.17 0.19 0.14 0.25 0.09
C/AS (%)b 1.01 2.67 1.02 2.81 8.52 15.65 1.72
Source; Nomura Research Institute for the data of JA and AS, Bank of Japan (2001a) for the other
data.
a
A, B, C denotes the average number during 5 years.
b
JA and AS denote the average number during 5 years of the total values of the Japanese and the Asian
stock markets, respectively. A/JA (%) and A/AS (%) indicate the ratio of A to JA and AS in percentage;
the same meaning for B/ and C/. The positions in each panel are, respectively, categorized according to the
nationality of security issuers and the nationality of the final investor.
Hong Kong and the Singapore stocks, respectively. These large ratios suggest that
the Hong Kong, Singapore, and Thai markets will be rocked by such Japanese
investors holding the Asian stocks and by such Asian investors holding Japanese
stocks through the arbitrage in the Japanese markets. Finally, the C/AS% in Panel C
denotes the ratio (in percent) of each Asian equity capital (C) on affiliated enterprises
held by the Japanese to the total market values of each Asian country’s stock.
Japanese investors hold 2.67, 2.81, 8.52, and 15.65% of the total market values of
the Korea, Singapore, Thai, and Indonesia stocks, respectively. The stock prices of
the head quarters in Japan will affect the stock price of affiliated enterprises, which
396 T. Miyakoshi / Int. Fin. Markets, Inst. and Money 13 (2003) 383 /399
will affect each Asian market: Korea will be an exception because it has the strongest
and widest restrictions for stock trading among Asian countries. On the other hand,
these countries become a symbol of expanding business oversea for Japanese
investors. Therefore, these markets will affect the Japanese market. These facts will
justify the results of significant gAs 2 and gja 2. Finally, the result for Malaysia
provides an exception worthy of note. That there is no relation between Japan and
the Malaysia markets is due to the strong capital control by the Malaysian
government after the crisis in 1997. However, the effect of Japanese volatility on
the Taiwanese volatility cannot be so easily explained and could be a task for future
study.
5. Concluding remarks
This paper investigates the magnitude of the return and volatility spillover from
Japan and the US to the Asian markets. The major empirical findings are 3-fold.
First, only the influence of the US is important on the Asian market returns; there is
no influence of Japan. Second, Asian market volatility is influenced more by the
Japanese market than by the US. Third, there exists an adverse influence of volatility
from the Asian to the Japanese market. In contrast to Ng (2000) study which
represents the primary previous research in this field, the results of this study suggest
a greater regional influence from Japan on Asian volatility than the world influence
from the US and find a new adverse influence from Asia to Japan.
The difference of the results between this paper and Ng (2000) stem from the two
factors. First, there is a strong economic relationship between Japan and Asian
countries through the large amount of portfolio investment during the latter half of
1990s. Therefore, the Japanese variables will be endogenous in our model, as
explained in Section 4. Second, there is a structural difference between the models.
Roughly speaking, Ng separates the effect of idiosyncratic shocks y2Ast1, y2,jat1 and
y2ust1 of Asia, Japan and the US on Asian market volatility hAsAst ? at period t:
?
hAsAst ?
/hAsAst (y2Ast1, y2,jat1 and y2ust1). Ng compares the impact of the US and the
Japanese shocks on Asian volatility. However, our paper compares the impacts of
US market volatility (consisting only of the US idiosyncratic shock y2ust1) and
Japanese market volatility hjajat1 (including the previous US shock y2ust2) on Asian
volatility: hAsAst /hAsAst (hAsAst1(y2ust2), hjajat1(y2ust2), yust1
2
). Our model does
not incorporate the idiosyncratic shocks of the Japanese and the Asian markets.
Thus, our paper purely examines the spillover effects of the market volatility, thereby
differing from Ng (2000). Therefore, the results of our paper should be carefully
compared with the results of Ng. More research for the difference will be needed in
future.
However, our finding implies that Asian governments and international investors
can obtain some profits using signals from a market that opens earlier. At period t in
Asian countries, the earliest market in time flow is the US market at period t/1.
However, Asian volatility is influenced more from Japanese market volatility at
period t/1. Even if the US shock is large (the Japanese government has already
T. Miyakoshi / Int. Fin. Markets, Inst. and Money 13 (2003) 383 /399 397
stabilized its own market volatility at period t/1), Asian market volatility at period
t is small. Symmetrically, the stabilization of the Asian governments for their own
markets is needed for the Japanese market. For international investors to get profits
from the returns of Asian securities, it is necessary to pay attention on the US (as a
world market) directly before. However, implementing global hedging strategies and
asset allocation decisions on Asian markets requires the information concerning
Japanese volatility (as a regional market) on the previous day. These implication will
be different from Ng (2000), since the estimation results are different.
Acknowledgements
I am very grateful to the referees and the editor of this journal for valuable
comments. Also, I would like to thank Toshihiko Hayashi, Katsumi Matsuura,
Mototsugu Fukushige, Colin McKenzie, Wen-Jseng Chu, Y.J.James Goo, Makoto
Ohta, Haruo Onishi, Mamoru Kaneko, Masatoshi Yoshida, and Yoshihiko
Tsukuda for their helpful comments at the February, March, and April 2001
seminars. The research was supported by the Nomura Foundation for Social Science
in 2000 and by the Japan Economic Research Foundation in 1999.
398 T. Miyakoshi / Int. Fin. Markets, Inst. and Money 13 (2003) 383 /399
Appendix A: Graphs of the observed data series: 1 January 1998 /30 April 2000
Notes: the stock price indexes are depicted for Japan, the USA, and Asian
countries: in the case of Taiwan and Hong Kong, the right axis is for Taiwan and
Hong Kong.
T. Miyakoshi / Int. Fin. Markets, Inst. and Money 13 (2003) 383 /399 399
References
Bank of Japan, 2001. Regional direct investment position and regional portfolio investment position.
Statistics by the Bank of Japan.
Bank of Japan, 2001. Locational international banking statistics in Japan. Statistics by the Bank of Japan.
Base, K.H., Karolyi, G.A, 1994. Goodnews, bad news and international spillovers of stock return
volatility between Japan and the US. Pacific-Basin Finance Journal 2, 405 /438.
Bekaert, G., Harvey, C.R., 1997. Emerging equity market volatility. Journal of Financial Economics 43,
29 /77.
Bollerslev, T., 1986. Generalized autoregressive conditional hetroskedasticity. Journal of Econometrics 31,
307 /327.
Bollerslev, T., Engle, R.F., Wooldridge, J.M., 1988. A capital asset pricing model with time-varying
covariances. Journal of Political Economy 96, 116 /131.
Braun, P.A., Daniel, B., Nelson, D.B., Sunier, A.M., 1995. Good news, bad news, volatility, and betas.
The Journal of Finance 50, 1575 /1603.
De Santis, G., Gérard, B., 1998. How big is the premium for currency risk. Journal of Financial
Economics 49, 375 /412.
Dumas, B., Solnik, B., 1995. The world price of foreign exchange risk. The Journal of Finance 50, 445 /
479.
Engle, R.F., Hendry, D.F., Richard, J.F., 1983. Exogeneity. Econometrica 51, 277 /304.
Fornari, F., Levy, A., 2000. Global liquidity in the 1990s: geographical allocation and long-run
determinants. In: International Financial Markets and the Implications for Monetary and Financial
Stability, Bank for International Settlements Conference, Papers No. 8, pp.1 /36.
Karolyi, G.A., 1995. A Multivariate GARCH Model of International Transmissions of Stock Returns and
Volatility: The Case of the United States and Canada, vol. 13. American Statistical Association, pp.
11 /25.
Karolyi, G.A., Stulz, R.M., 1996. Why do markets move together? An investigation of US /Japan stock
return comovements. Journal of Finance 51, 951 /986.
Ministry of Finance, 2000. Regional Balance of Payment. Ministry of Finances Statistics Monthly.
Nelson, D.B., 1991. Conditional hetroskedasticity in asset returns: a new approach. Econometrica 59,
347 /370.
Ng, A., 2000. Volatility spillover effects from Japan and the US to the Pacific-Basin. Journal of
International Monetary and Finance 19, 207 /233.