How Does the Asian Crisis Affect the Interdependencies Between Major Financial Markets in Asia and the US



How Does the Asian Crisis Affect the Interdependencies Between Major Financial Markets in Asia and the US



Description:
Empirical studies support evidence of major financial markets affecting minor markets. During the Asian Crisis, however, major markets, such as the US and Japan, reacted to the problems in smaller markets such as Hong Kong and Singapore

Abstract

Empirical studies support evidence of major financial markets affecting minor markets. During the Asian Crisis, however, major markets, such as the US and Japan, reacted to the problems in smaller markets such as Hong Kong and Singapore. In this paper we apply a VAR model to capture the changes in the dynamic relationship among stock markets in Hong Kong, Singapore, Japan and the US. A multivariate stochastic volatility model (SVM) is also utilized to study the correlations in volatility among these markets to determine the impacts of the Crisis. We find strong lead-effect of Hong Kong market returns over the US market returns during the Crisis. Our results also show strong overacting behavior among investors across these markets during the Crisis. We also find that the Crisis period was marked by higher volatilities and stronger spillover effects across these markets.

How Does the Asian Crisis Affect the Interdependencies Between Major Financial

Markets in Asia and the US?

Introduction

Global financial market contagion is well documented (see, for examples, Lin, Engle and Ito, 1994, Karolyi, 1995, Chowdhury, 1994, and Tay and Zhu, 2000, Ko and Lee, 1991). On the Black Monday of October 1987, the New York Stock Exchange’s (NYSE) Dow Jones Industrial Average saw its largest single-day percentage loss in history. The event set off a panic among global financial markets. Markets in Asia, which open about 13 or 14 hours ahead of New York, declined drastically on the following Tuesday. Traders in Tokyo and Hong Kong were selling on the news of the event, which happened in the US, without assessing the local fundamentals. Noise trading and speculative trading has become the driving force for the irregular price movement across borders. The financial market in United States has long been seen as the leader of global financial market. Its rise and fall are seen as the indicator of the global economic health. Thus the contagion effect from the U. S. to other markets should not be a surprise. During the Asian Crisis period, however, this relationship seems to have reversed.

On Monday, October 27, 1997, after losing nearly 20% of its value the previous week over interest rate pressure, the Hong Kong Hang Seng Index (HKHSI) lost 5.8%. On the same day (NYSE opens a few hours after the Hong Kong market closed) the Dow Jones Industrial Average saw its largest single-day loss ever, falling 554.26 points, or 7.18% off of its previous closing value. The Nasdaq and the S&P 500 Index both saw steep declines. The contagion effect seems to have gone the other direction: the declines in Asian stock markets led to the declines in the U.S. market. Is there a change to the

structural relations among major Asian stock markets and the U.S. markets after 1987? If there is, how big is the change? In this paper we attempt to address both issues using data from four major stock markets, Hong Kong, Singapore, Japan and the United States.

The remainder of this paper is organized as following: Section 2 provides an account of the Asian Crisis and in section 3 we provide a review of recent literatures on international stock market interactions. In section 4 we propose a formal model and provide some preliminary data analysis. The empirical results are provided in section 5 and a conclusion in section 6.

2. The Asian Flu

The source of the crisis in Asia can be dated back to as early as July of 1997. The devaluation of the Thai currency against the US dollar on July 2 is the event that set off the crisis. The currency problem in Thailand spread thru East Asia like a flu virus (see Granger, Huang and Yang, 2000). The currencies in the Philippines, Malaysia, Taiwan, Indonesia and Japan all weakened against the US dollar within a month. The Hong Kong dollar is peg to the US dollar, thus free from the effect of the “flu virus”. The peg arrangement between the Hong Kong dollar and US dollar, however, created an opportunity for speculators. As a result, the Monetary Authority in Hong Kong had to carry out several interventions to protect against speculative attacks. Overnight interest rates in Hong Kong rose sharply (150 basis point) in August 15. The Hong Kong stock market sharply lowered the following day. The interventions by the Hong Kong government are seen as the beginning of active involvements in the financial market by the Hong Kong government, previously not seen under the British government’s rule.

From 10/20/97 to 10/23/97, the fear over interest rates and the peg agreement with the US dollar cost the Hong Kong stock market to lose over 23% of its value in four days. The following Monday the US stock markets took a dive with the Asian markets. At the height of the Crisis (November 7, 1997), the Hong Kong stock market lost over 50% of its value compared to its high in early 1997.

The second wave of the Crisis comes from South Korea and Japan. For the second part of November 1997, financial markets in Hong Kong, Japan, Korea and the US all experienced high level of volatility never seen before. The period of high volatility continued well into the beginning of 1998, as the financial and political problem in Indonesia began to pick up. In January 1998, financial markets in the region took another hard hit. The effect rippled to the US as expected. As a result, the week of January 4 to 9 1998 saw the worst week in Wall Street history. The ripple effect continued well into late spring of 1998 when signs of recovery started to show in Hong Kong, Singapore, Taiwan and Japan. Currency and financial problems in other Asian countries also stabilized by the early summer of 1998.

3. Literature Reviews

There are a few studies that investigated the interdependencies between the markets considered in this paper before the Asian Crisis. Using intra-day data Lin, Engle and Ito (1994) found bi-directional interdependence in returns and volatilities between the U.S. and Japan stock markets. Chowdhury (1994) used daily data from Hong Kong, Singapore, Taiwan, Korea, Japan and United States to study the interactions among these markets, and found strong interaction between Hong Kong, Singapore, Japan and the

U. S. market. Using weekly data from the same six markets, with time-varying multivariate GARCH model, Tay and Zhu (2000) found similar results.

More recently Phylaktis and Ravazzolo (2002) find that Pacific Basin Countries’ financial markets are integrated with each other in the 90’s However, they also report that only Hong Kong and Singapore market are integrated with the US market. Faff, Hillier and McKenzie (2002) report that emerging markets are exposed to considerably higher volatility risk. As the global financial markets are more interrelated, the volatility risk in emerging markets will no doubt spillover to more developed markets. Therefore, the correlation relations among these markets need to be reexamined. Our paper will also complement the existing works by providing a comparison between the market correlation relations before and after the Asian Crisis.

4. Data Descriptions and Methodology

Data Descriptions

Daily market index data for Hong Kong, Singapore, Japan and the United States are utilized in this study. The market indexes are the Hong Kong Hang Seng Index, Straits Times Index, Nikkei Stock Average and Dow Jones Industrial Average Index for Hong Kong, Singapore, Japan and the U.S. respectively. The data covers the period between December 1994 and November 2000. Excluding holidays in all these markets there are a total of 1439 observations. In light of the discussion in section 3, we break the data into 3 sub-periods. The period before the Crisis goes from December 1994 to June 1997, with a total of 629 observations. The Crisis period covers dates between July 1997, and July 1998, with 274 observations. The remainder of the data, 536 observations, is

considered as the post Crisis period. We break the data into 3 sub-periods based on what we know about the Crisis1.

The Dickey-Fuller test determined that all the index series are integrated of order one. The issue of cointegration should not concern us since we are using a relatively short time frame in each of the 3 periods under consideration2. Thus the returns (log difference) for these markets is used for the analysis in this paper. The graphs for all the series and summary statistics are presented in Figure 1 and Table 1 respectively.

[Insert Figure 1 and Table 1 about here.]

First, from Table 1 we note the sharp decline in mean returns for all four markets during the Crisis period. An average return of .0985 for the Hong Kong market before the Crisis drops to a negative return of -0.2371 during the Crisis period. The average return for Singapore also goes down from -0.0204 to -0.2329 during that same period. The declines for the U.S. and Japan markets are not as drastic, from 0.1121 and 0.0106 to 0.05346 and -0.0838 respectively. This further supports the argument that the Crisis affected export¬oriented markets more severely. There are also uniform increases in return volatility for all four markets. Again, the change is more pronounced for Hong Kong and Singapore.

After the Crisis, Hong Kong and Singapore recover considerably well, while the U.S. market returns actually decline and the Japan market continue to show sign of slowing down. The average returns for the Hong Kong and Singapore markets are in positive territory, and larger than what they were before the Crisis. This should not come

1 However, as indicated in Cheng, Fung, and Chan (2000), the Crisis period is marked by abnormality in pricing behavior and the recovery started in August 1998. Therefore, the division of the data into 3 sub periods is justified.

2 The cointegration test results, not reported here, confirmed this conjecture. The results are available from the author upon request.

as a surprise because the fundamentals in Hong Kong and Singapore are very strong during the Crisis. Their demise is the result of poor performance of neighboring countries during the Crisis. Once the Crisis is over the market should correct to reflect the actual value of these two markets. The decade-long recession in Japan exposed the many problems with the Japanese banking system (one of the driving forces of the Crisis) thus further weakens the fundamentals. The continuation of decline merely reflects the over¬valuation of the Japanese market over the past decade. The behavior of the U.S. financial market is different than those of the Asian markets during the same period. From Figure 1 we see that the U.S. market is the only market with a clear upward trend throughout much of the Crisis period. The strong economic growth of the 90s is barely affected by the Asian Crisis. The U. S. market is able to come out of the negative impacts of the Asian Crisis relatively quickly partly because the U.S. is the beneficiary of cheaper exports from the Asian countries. The main reason for the growth in stock prices during the Crisis period, however, is the emergency of the Internet Bubble. The burst of the Internet Bubble in early 1999 made the U.S. market performance of 1999 the worst year since the WWII.

[Insert Table 2 about here.]

Table 2 reports the sample correlations among these markets. Note that the correlation between the U.S. markets and the Asian markets should be interpreted as a lead-lag relationship since the Asian markets are one day ahead of the U.S. market. In other words, if the U. S. market return is strongly correlated with the market returns in Hong Kong, it means that the U. S. market return has a strong lead effect on the Hong Kong market return. From Table 2 we can see that the correlation between the market

return in Hong Kong with that of Singapore and Japan is getting stronger. The correlation with Singapore market return went from 0.3675 to 0.5832 during the Crisis. The correlations between these two markets further strengthened after the Crisis to 0.5971. The correlations between Hong Kong and Japan also increased, from 0.2493 to 0.4136 during the Crisis, and further increased to 0.4347 after the Crisis. A similar change occurred between the correlation between Singapore and Japan market returns. The correlations between the U. S. and the Asian markets show a different picture. While the correlations among the U.S., Singapore and Japan increased during the Crisis, the correlation between the U. S. and Hong Kong market returns actually went down during the same period, and the correlation continued to weaken after the Crisis. The increases in correlations for Singapore and Japan with other markets imply that these two markets are more integrated with the other markets.

[Insert table 3 about here.]

Next, we look at how the market variances are correlated with each other in Table 3. Similar to the correlations in returns, the correlation in variance between Hong Kong and the other Asian markets increased during and after the Crisis. The correlation between Hong Kong and Singapore is particularly stronger after the Crisis. This means that investors in both markets are watching closely what is going on in both markets since they are similar in many ways, and what happen to one would likely happen to the other. The correlation of variance between the U. S. and Hong Kong again show a rather different pattern. While there is a large increase during the Crisis, from 0.3504 before the Crisis to 0.6546 during the Crisis, the relationship weakened just as rapidly after the Crisis (down to 0.3399 after the Crisis). The increases in correlation among Asian

markets after the Crisis again show that these markets are more correlated with each other after the Crisis. The correlations among U. S., Japan and Singapore again show large increases after the Crisis, which signals a greater influence of the U.S. market over these two Asian markets.

Methodology

To capture the interactions among the markets under consideration we utilized a vector autoregressive (VAR) model. When the markets in Asia open they will observe the closing value of the U. S. market the day before, therefore a one-period lag return of the U. S. market index will capture the lead effect it has over the Asian markets. On the other hand, if the Asian markets lead the U. S. market, the current period return is needed. Denote HKRt, SPRt, TKRt and USRt as the market returns for Hong Kong, Singapore, Japan (Tokyo) and the U.S. at time t respectively, we consider the following VAR model,

k k k k

HKRt i HKR t i SPRTKR USR

= + – + +++

a 0 a 1 a a a e

2 i t i –

3i t i –

4 1

i t i t

 =  Â

i

1 1 1 =1 i =i =i k k k k

+ – + + + +

b 1 i HKR t i SPRTKR USR

 b  b  b e

2 – – –

i t i 3i t i 4 2

i t i t

 =

i = 1 i 1

=1

i 1

=i

k k

+ – +

 d 1 i HKR t i SPR

 d 2 i t i

– i 1

= i 1

=

k k

+ – +

 l 1 l

i HKR t i SPR

 2 i t i

– i 0

= i 0

=

(1).

Equation (1) is then estimated using Seemingly Unrelated Time Series (SUR) method. The estimates for e1t, e 2t, e3t, and e4t are the residuals of their respective resulting

estimated equation. These series are then fitted using a multivariate stochastic volatility model to capture the interactions among the volatilities.

Traditionally a variance decomposition (VDC) and Impulse Response Function (IRF) is utilized to capture the effects of volatilities (innovations). In this paper we take an alternative approach. The residual series of each market is obtained from the VAR model described in equation (1). Then a time-varying multivariate stochastic volatility model (SV) is utilized to capture the effects of innovation across markets. Specifically, we adopted the multivariate stochastic volatility (MSV) model proposed by Harvey, Ruiz and Shephard (HRS, 1994).

Following HRS, with adjustments of some notation already used, let {yt} be a zero-mean time series process such that yt = stet where st is the standard deviation of yt and et is a random variable with zero mean and unit variance. Consequently,

log y2t = Ht + log e2 t, (2)

where Ht = log(s2 t). It is assumed that Ht follows an AR(1) process:

Ht = g + fHt-1 + ht, (3)

where ht is a normal random variable with zero mean and variance s2 h. Also, ht is assumed to be uncorrelated with et. Alternatively, let m = g /(1 – f) and let ht = Ht – m. Then equation (1) and (2) can be rewritten as follows:

log y2t = m + ht + log e2 t = n + ht +xt, (2′)

ht = fht-1 + ht, (3′)

where n = m + E(log e2 t) and xt = log e2 t – E(log e2 t). HRS proposed a quasi-maximum

likelihood (QML) method to estimate equations (2) and (3). They suggested log e2 t be

treated as a normal random variable. The Kalman filter is then applied to generate the quasi-maximum likelihood estimator.

Equation (2) and (3) can be easily generalized to the multivariate case. For example, consider the case of the four time series, the HRS multivariate stochastic volatility model can be written as follows. Let HKV, SPV, TKV and USV denote the market volatility (squared residuals obtained from equation (1)), of the markets under consideration, the multivariate SVM equations can be written as

È HKV ù È n ù È h ù È x ù ,

HK HKt HKt ,

Í ú Í ú Í ú Í ú

SPV n h x ,,

Í ú Í SP ú Í SPt ú Í SPt ú

, = + +

Í TKV ú Í Í

n ú h ú Í ú

x ,,

TK TKt TKt

Í ú Í ú Í ú Í ú

USV USt

,USt ,

Î û Î n h x

US Î

û Î û û

Let x x t = (HK,t , x SP,t , x TK,t , x US,t )¢ and let h h t = ( HK,t , h SP,t , h TK,t , h US,t )¢ , where ¢ denotes the

transpose. The variance-covariance matrices of xt and ht are denoted by Sx and Sh, respectively. Furthermore, it is assumed that xt and ht are uncorrelated with each other.

Note that the off-diagonal parameters of the first matrix on the right hand side of equation (5) will capture the spillover effects of innovations across markets.

5. Estimation Results The Structural Relations

[Insert table 4 about here.]

The results from equation (1) for each of the three periods are reported in Table 4 to 6. The value of optimal lag, k, is determined by Akaike Information Criteria (AIC). Before and during the Crisis the value of k is 2, while the value is 4 after the Crisis. The VAR result before the Crisis is reported in Table 4. We can see that the U. S. market leads all three Asian markets. The influence of the U.S. market over Hong Kong is particularly strong. This is of no surprise in part because the Hong Kong dollar is peg against the U.S. dollar. The U. S. is also the largest trading partner (importer) of Hong Kong. We, therefore, should expect the economic performance in the U.S. to have measurable impacts in Hong Kong’s economic outlook. None of the lagged returns from other markets had any impact on Hong Kong’s market return under conventional statistical criteria.

The result for the Singapore equation displays a different pattern. Both of the first and second lag market returns from Hong Kong and the U. S. markets affected the market returns in Singapore. The size of the effect is dependent on the market value. The market value of the U. S. is far greater than that of Hong Kong and Singapore, while Hong Kong also has a sizable advantage over Singapore in that regard. Therefore, while the Hong Kong market returns do have lead effect on Singapore market returns, the lead effect from Hong Kong is far smaller than that from the U.S.

By inspecting the third equation, we observe that the Japanese market is not affected by Hong Kong or Singapore market and is led by the U.S. market. As for the result for the U.S. market, there is little surprise either. Only the first lag of the Japan market and the U.S. market had any explanatory power over the U.S. market returns at

the 90% confidence level. And the combined effect is only a mere 10.7%. The above result is similar to the finding of Lin, Engle and Ito (1994) using intra-day data. [Insert Table 5 about here.]

The results during the Crisis are shown in Table 5. The result for the Hong Kong equation changed considerably. First, the one period lag of the Hong Kong market return is now significant near the 95% level and has a negative sign. This would suggest that investors in Hong Kong market are over reacting in any given trading day. Then they adjust to their previous mistakes the next day. While the impact from Singapore is still statistically insignificant, it had gained ground, from -0.0083 to 0.07585, and significance level (from t-statistics of -0.17115 to 1.0209). The lead of the U.S. market over Hong Kong also increased during the Crisis. Again, the two period lag impact of the U. S. market return shows that investors in Hong Kong were overreacting strongly (an adjustment of over -0.304).

The Singapore market also saw some dramatic changes during the Crisis. The impacts of the one period lag in the Hong Kong market return over Singapore has strengthened considerably, while the results from its own lagged values also show signs of overreacting from investors in Singapore. The led of the U.S. market over the Singapore market also increased. Furthermore, it appears that investors are not overreacting to the U.S. market as they did before the Crisis. The Japan market continued to have little impact over the Singapore market.

The impacts of the Asian Crisis can be observed in the changes that occur in the Japan and the U.S. equations. From Table 5 we can see that the t-statistics for the coefficients of the one period lag for Hong Kong and Singapore are both significant at the

90% level. Although the effects from Hong Kong and Singapore to Japan are small (only -0.07646 and 0.07821 respectively), it does tell us that the Japan market is now affected by other Asian financial markets. Similar to the changes occurred in Hong Kong and Singapore, the influence of the U.S. market over the Japan market increased and there is overreaction by investors in the Japanese market.

The impacts of the Crisis were most dramatic on the U.S. market. First, both the current and two periods lag value of the Hong Kong market returns have effects over the U. S. returns. This means that the Hong Kong market now leads the U. S. market. The effect is still significant after two periods. Furthermore, the lead-coefficients for both Singapore and Japan market increases, along with the significance level. These suggest that there are more interactions between the U. S. markets and the Asian markets during the Crisis and investors in the U. S. are taking a closer look at what is going on in Asian financial markets.

[Insert table 6 about here.]

Table 6 reports the results for the period after the Crisis. The results for the period after the Crisis are very different from those before and during the Crisis. For one, the optimal lag length determined by AIC is four, instead of two for the periods before and during the Crisis. Two, the correlations between markets also changed drastically.

The first noticeable change is the dependency of the Hong Kong market on Singapore market becomes significant at the 90% level (3 lags, at -0.16537). The reason for the strengthening of the ties between Hong Kong and Singapore market will be explored further later in this section. Also note that a similar change occurred between Hong Kong and Japan market. Both the one and four lag of Japan market returns has

strong explanatory power on Hong Kong return. While the lead effect of the U. S. market over Hong Kong market diminished, the combined effect of the U. S. market is actually stronger than before. This result is not captured by the sample correlation analysis discussed in the previous section.

From the Singapore equation it is clear that the Hong Kong market has much lower influence over Singapore market after the Crisis. Though the Hong Kong market still leads the Singapore market in a positive way. The lead decreased from over 23% to only about 10% (counting only those with over 90% significance level). Again, we saw a fundamental change in the correlation between Hong Kong and Singapore. As for the relationship between Singapore and Japan, the result is similar to the previous two periods. There is little evidence of dependency between Singapore and Japan after the Crisis. On the other hand, the influence of the U. S. market increased after the Crisis. This suggested that investors in Singapore are paying more attention to what is happening in the U.S. The reason for this change, and the changes occurring in Hong Kong, will become clear when we look at the changes that occurred in the U.S. market.

The temporary dependency of the Japan market on Hong Kong and Singapore market during the Crisis is short lived. The lead of the U.S. market over the Japan market also declined compared to that of the Crisis period. It is, however, still much larger than what it was before the Crisis. It suggests that the Japan market is also more integrated with the other financial markets after the Crisis.

From the U.S. market equation we see a result that is different than before or during the Crisis. The first noticeable change is the lead effect of the Hong Kong market is nonexistence after the Crisis. The relationship is also weakened when compared to that

before the Crisis. The opposite happens to the relationship between Singapore and the U.S. The lead effect of the Singapore market over the U.S. market is now significant. In some sense, Singapore is playing an ever more important role after the Crisis.

Next, we turn to the impacts of innovations across markets during these three different periods. The residual series were obtained from each of the three periods analyzed. The squared values are then fitted with equation (4) and (5). The results are as follow.

For the period before the Crisis, the equations we obtained are,

È 0.0864 ù

È HKV ù È h ù È x ù

(0.4433)

,,

t Í ú HKt HKt

Í ú Í ú Í ú –

SPVÍ 0. 5814ú h x

, ,

t

Í ú (0. 2117 ) Í SPt ú Í SPt ú

= Í ú + +

Í TKVú 0.3089 Í h ú Í

,

t TKt

Í (0.0661) ú

Í ú Í ú

USV Í 1.6353ú h –

t USt

,

Î û Î û

Î (0.1505 ) û

È h ù È 0.997

,

HKt

Í ú Í

h 0.448 ,

Í SPt ú Í =

Í h ú Í 0.041

TKt ,

Í ú Í

USt ,

Î h û Î 0.251

0.002 0.001 0.003

– –

0.130 0.054 0.001 –

0.039 0.080 0.005

0.026 0.020 0.047 –

ù È h ù

HKt

,1 HKt

,

ú Í ú

h h

SP

t , 1 ,

ú Í SP t ú

.

+

h

^ O h ^

TKt TKt

– ,

,1 ú Í ú

h h

USt

,1 USt

,

– û Î û

The value in parenthesis under the numbers in the first matrix in equation (6) is the standard error. The covariance matrix of disturbances is,

È 2.044 ù

Í 0.142 0.023 ú

S = Í ú ,

x Í 0.150 0.022 0.022 ú

Í Î 0.162 0.011 0.012 0.013 ú

û

È 0.002 ù

Í – 0.024 1.327 ú

S = Í ú

h –

Í 0 .006 0 . 152 2.091 ú

Í Î 0.046 – 0.016 – 0.117 2.292û ú

16

From equation (6) we can see that the Japan market had the highest volatility level while the U. S. market had the lowest. The diagonal values in equation (7) capture the persistence of their respective equation. The Hong Kong market clearly has the highest persistence among them, while the Japan market has the lowest. Also, note that there are clear asymmetric spillover effects. An increase in volatility in Singapore lowered the volatility in Hong Kong by a small amount (-0.002) only. At the same time an increase in volatility in Hong Kong would cause the volatility in Singapore to increase by 0.448. A similar relationship exists between Hong Kong and the U. S. market. The covariance matrix also revealed that the volatility in Hong Kong contains mostly noise and little information, while the reverse is true in the other three markets.

During the Crisis period, the equations are,

È 3.466 ù

È HKV

t ù (0.1220) HKt

È h ù È HKt

x ù

Í ú Í 4.077 ,,

ú Í ú Í ú

SPVÍ

t

Í ú = (0. 1047) ú h ,SPt

SPt

Í ú Í

+ + x ,

ú (8)

Í TKV

t ú Í

Í 1.304

(0.1944 ) ú Í h ú Í

TKt

,,

ú ,

x ú

TKt

Í Í USV

Î ú

û ú Í

Î – 0.451Î

(0.0865 ) Í ú Í

ú USt

,,

û h û Î ú

US t

x û

È h ù – –

È 0.817 0.087 0.035 0.018 ù È h ù È h ù

HKt

, –

,1HKt

HKt ,

Í ú

h SPt Í – –

0.104 0.337 0.096 0.388 ^ O ^ O

h SPt – h ú

SPt

Í ú = ,,

Í ,1 ú Í

ú Í + ú . (9)

Í h ú

TKt ,-

Í 0.049 0.016 0.996 0.035 ú Í h ú Í

TKt

,1 h ú

TKt

,

Í ú Í ú Í ú Í ú

h USt

Î û ,-

– –

Î 0.094 0.105 0.105 0.718 h US t

,1

^ Œ ^ Œ h US t

, û

The covariance matrix of disturbances is,

È 1.272 ù

Í 0.431 0.146 ú

S = Í ú ,

x Í 0.013 0.004 1.920 ú

Í – – –

Î 0.173 0.058 0.032 1.705û ú

17

È 0.265 ù

Í ú

0.213 1.421 –

.

S = Í ú

h Í 0.029 0.006 0.004 ú

– –

Í ú

– –

Î 0.055 0.032 0.006 0.013û

Comparing equation (6) and (8) we see a dramatic increase in the volatility level in all four markets. This shows that the Crisis is marked by period of high volatility. The changes for Hong Kong and Singapore are much bigger than that of Japan and the U. S. The anti-log trend value went from 1.09 and 0.58 to 32.01 and 58.94 for Hong Kong and Singapore respectively, while the numbers are 1.36 and 0.19 to 3.68 to 0.64 for Japan and the U.S. respectively. The currency crisis in Asia is expected to have larger impacts on these two smaller economies. Equation (9) also shows that, with the exception of Hong Kong (still has the highest persistent parameter, yet lower than before the Crisis), the market volatilities are more persistent during the Crisis period. Asymmetric spillover effects still exist, but their relationship has changed. First, spillover effects generally increased during the Crisis period. Secondly, the strong spillover effects from Hong Kong to Singapore and the U. S. are not there any more. Replacing the influence of Hong Kong over Singapore is the U.S. market. Finally, we note that the volatilities of the other markets, besides Hong Kong, become noisier and less informative.

We now turn to the last period, the period after the Crisis: the equations are,

È h ù È 0.943 0.042 0.004 0.014 ,

HKt

Í ú Í

h – 0.009 0.974 0.0002 0.013 ,

Í SPt ú Í =

Í h ú Í 2.089 2.332 0.002 0.011

TKt – – – ,

Í ú Í

Î h USt – –

Î 2.319 1.792 0.023 0.071

The covariance matrix of disturbances is,

È 2.269

Í 0.299 2.110

S = Í

x Í 0.047 0.051 0.002

Í 0.049 0.186 0.005

– – –

Î

È 0.002

Í 0 .001 0.004

S = Í

h Í – 0.041 0.021 2.032 Í Î 0.035 0.078 0.049

ù È h ù

HKt

,1 HKt

,

ú Í ú

h h

SP

t

, 1 ,

ú Í SP t ú

.

+

h

^ O h ^

TKt TKt

– ,

,1 ú Í ú

h h

US t

,1 US t

,

– û Î û

First, we observed that even though the volatility trends for all three of the four markets had gone down compared to that during the Crisis, the overall volatility trend is still much higher than that before the Crisis (anti-log values of 7.84, 4.94 and 3.09 respectively). Furthermore, the volatility level for the U.S market is actually higher (anti¬log trend value of 1.58 after the Crisis verse 0.64 during the Crisis). From equation (11) we saw a much different spillover effects and persistence across these markets. First, note that there is little spillover of volatility from any market to Hong Kong and Singapore. These two markets, however, have very high persistence compared to Japan and the U.S. On the other hand, spillovers from Hong Kong and Singapore to Japan and the U.S. markets are high. Note, however, that they almost cancel out each other, leaving both Japan and the U. S. market with lower persistence. The covariance matrix of disturbances confirms that the volatilities of Hong Kong and Singapore contain little information and the reverse is true for Japan and the U. S. In terms of stability of the volatility, Hong Kong, Japan and the U.S. have similar values before and after the Crisis. On the other

hand, Singapore market volatility becomes much more unstable after the Crisis. In fact, the stability of Singapore market volatility is almost the same as the Hong Kong market volatility after the Crisis.

6. Conclusion

The U.S. financial market has long been the leader of the global financial market. A change in the market value in the U.S. market would lead to a change in market values for most of the financial markets around the world, while the reverse was not true. As the global financial markets become more integrated, the leadership role of the U. S. market also have changed. The change might have been accelerated by the recent financial Crisis in Asia. Most noticeable is the steep decline in the U. S. market value following a similar decline in the Hong Kong market value occurred in late October of 1997. In this paper, we utilize a two-step procedure to capture the effect of the impacts of the Asian Crisis on the interdependencies between 3 major financial markets in Asia and the financial market in the U. S. The VAR model captures the interactions before, and changes that occurred, during and after the Crisis. We found strong lead-effect of Hong Kong market returns over the U.S. market returns during the Crisis. But this is only a short-lived change. Our results also show strong overacting behavior among investors across these markets during the Crisis. While the Singapore market does show small lead effect over the U.S. market after the Crisis, the U.S. market resumed its dominating role over the Singapore market. That being said, our result does show that the Asian and the U.S. financial markets are more integrated after the Crisis. The increasing ties in trade relations among these

markets seems to be the main driving force behind the improved integration, a point echoed by Granger et al (2000).

For the second step of the analysis we employed a time-varying multivariate stochastic volatility model to capture the transmission of innovations across these markets. We found the Crisis period was marked by higher than usual volatilities and stronger than usual spillover effects across these markets. Though the volatilities for these markets seemed to have subsided after the Crisis, they’re still higher than what they were before the Crisis. Furthermore, we found that an interesting change occurred in the spillover effect between smaller markets (Hong Kong and Singapore) and larger markets (Japan and the U.S.). The Singapore market is behaving more and more like the Hong Kong market in terms of volatility trends and noise contained in their respective markets. The spillover effect of innovations from the Hong Kong market to Japan and the U. S. markets are similar to that of the Singapore market, except they are in opposite direction. In all, our results suggest U.S. firms that are facing international financial market risks must pay more attention to smaller open markets such as Hong Kong and Singapore.

Reference

Cheng, Louis T.W., Joseph K.W. Fung, and Kam C. Chan, 2000, Pricing dynamics of index options and index futures in Hong Kong before and during the Asian financial crisis, Journal of Futures Markets, 20, 145-166.

Chowdhury, A.R., 1994, Stock market interdependencies: evidence from the Asian NIEs, Journal of Macroeconomics, 16, 629-651.

Faff, R. W., D. Hillier, and M. D. McKenzie, 2002, Modeling return and volatility exposures in global stock markets, working paper, RMIT University, Australia.

Granger, C. W.J., B.N. Huang, and C.W. Yang, 2000, A bivariate causality between stock prices and exchange rates: evidence from recent Asian Flu, The Quarterly Review of Economics and Finance, 40, 337-354.

Harvey, A., E. Ruiz, and N. Shephard, 1994, Multivariate stochastic variance models, Review of Economics Studies, 61, 247-264.

Karolyi, G. Andrew, 1995, A multivariate GARCH model of international transmission of stock return and volatility: the case of United States and Canada, Journal of Business and Economics Statistics, 13, 11-25.

Ko, Kwang-Soo and S. B. Lee, 1991, A comparative analysis of the daily behavior of stock returns: Japan, the U. S. and the Asian NICs, Journal of Business Finance and Accounting, 18, 219-314.

Lin, Wen-Ling, R. F. Engle, and T. Ito, 1994, Do bulls and bears move across borders? International transmission of stock returns and volatility, The Review of Financial Studies, 7, 507-538.

Phylaktis, K. and F. Ravazzolo, 2002, Measuring financial and economic integration with equity prices in emerging markets, working paper, City University Business School, U.K.

Tay, Nicholas S.P. and Zhen Zhu, 2000, Correlations in returns and volatilities in Pacific-Rim stock markets, Open Economics Review, 11, 27-47.