B. Cha, S. Oh International Review of Economics and Finance 9 2000 299–322 309
Table 3 Optimal lag length for the autoregressive conditional heteroscedasticity ARCH process for residual
series VAR
Country Lags
x
2
Statistics HONG KONG
US 1
81.40 JAPAN
5 94.26
HONG KONG 1
1.03 KOREA
US 1
80.12 JAPAN
5 84.02
KOREA 8
188.45 SINGAPORE
US 1
74.52 JAPAN
5 93.78
SINGAPORE 1
5.12 TAIWAN
US 1
73.55 JAPAN
5 92.67
TAIWAN 3
227.09 Notes:
1 denotes statistical significance of the ARCH effect at the 5 level. 2 The lag length was determined by the Schwarz Information Criterion SIC.
4. Empirical results
4.1. Variance decomposition With the estimated residuals and the MAR [Eq. 2], we decomposed the forecasting
error variance of the AEM’s return rate in order to compute the relative contribution of shocks in the U.S., Japan, and the AEM itself. First, the innovations were orthogo-
nalized in order to “isolate” a shock in each variable. This was done by expressing the second term of the MAR [Eq. 2] as given by Eq. 6,
o
∞ s
5
A
s
e
t
2
s
5
o
∞ s
5
A
s
GG
2
1
e
t
2
s
5
o
∞ s
5
C
s
u
t
2
s
6 where C
s
5 A
s
G and GG9 is a factorization of the covariance matrix of e. In computing
the orthogonalized innovations u 5 G
2
1
e , the Choleski factorization was used where
the G matrix was chosen to be lower triangular. Then, with weekly return rates, the k-week ahead forecast error of Y at time t becomes as shown by Eq. 7:
C
k
2
1
u
t
1
1
1 . . . 1 C u
t
1
k
5
o
k
2
1 s
5
C
s
u
t
1
k
2
s
. 7
Let C
ij s
be the i, j element of C
s
. Then, the variance of this k-week ahead forecast error is
o
n j
5
1
o
k
2
1 s
5
C
ij s
2
. Therefore, the component of the error variance in the k-week ahead forecast of Y
i
, which is accounted for by innovations in Y
j
becomes
o
k
2
1 s
5
C
ij s
2
o
n j
5
1
o
k
2
1 s
5
C
ij s
2
.
310 B. Cha, S. Oh International Review of Economics and Finance 9 2000 299–322
Table 4 presents the results for variance decomposition, where up to 24-week forecast error variances of the four AEMs’ return rates are decomposed. The explana-
tory power of each country is measured as a percentage, so that the horizontal sum of each row is 100.
While the forecast error variance itself does not depend on the factorization of the covariance matrix of e, the decomposition of this forecast error variance does. With
the Choleski factorization, there is a different factorization for each ordering of the variables. The results in Table 4 are based on the ordering of Y
U ,t
, Y
J ,t
, Y
A ,t
in the orthogonalizing process. In practice, orderings are determined based on a “semi-
structural” interpretation of the model: If movements in A precede movements in B within a single period, A precedes B in the ordering [see Doan 1995]. Therefore,
the chosen ordering of Y
U ,t
, Y
J ,t
, Y
A ,t
features the recursive causal ordering suggested by theory: the U.S. first, then the Japanese market, then the emerging market.
11
Table 4 shows a dramatic increase in the importance of American and Japanese shocks in explaining unexpected movements in stock returns in all four AEMs. This
change came only after the outbreak of the Asian financial crisis. Before the crisis broke out, most of the forecast error variances of the AEMs’ return rates could be
explained by domestic factors. Since the outbreak of the crisis, however, U.S. and Japanese influences explain a significant portion of the forecast error variance of the
AEMs’ return rates. For example, U.S. shocks explained just 6.67 of Hong Kong’s variance in Period I and 7.11 in Period II. In Period III, however, U.S. shocks
accounted for more than 28 of variance. In Singapore, where the U.S. impact was historically largest among the four AEMs 10.78 in Period I and 13.29 in Period
II, U.S. shocks have become even more influential with the explanatory power of 28.07 in Period III.
The same trend of increasing influence also appears in Korea and Taiwan, whose stock markets were historically thought by observers to have moved somewhat inde-
pendently of the developed stock markets. Table 4 shows that before the sudden outbreak of the Asian financial crisis, the U.S. impact was negligible in Korea and
Taiwan. In Korea, the U.S. market had an explanatory power of only 1.36 and 1.98 in Periods I and II, respectively, while in Taiwan, the explanatory power was
1.61 and 1.89 in Periods I and II, respectively. Since the Asian financial crisis, the U.S. explanatory power has jumped to more than 10 and 29 in Korea and
Taiwan, respectively.
The Japanese impact has also increased since the crisis in all AEMs but Taiwan. In Periods I and II, the explanatory power of Japanese fluctuations was only 3.11
and 1.67 in Hong Kong, 1.87 and 4.12 in Korea, and 2.73 and 6.76 in Singapore, respectively. In Period III, the Japanese influence had increased to 4.20
in Hong Kong, 9.08 in Korea, and 8.34 in Singapore. Interestingly, the Asian financial crisis seems not to have increased the Japanese influence on the Taiwanese
stock market.
Prior to the Asian financial crisis, the Singaporean market was the one most affected by U.S. and Japanese shocks among the four AEMs. In Period I, before the October
Crash in 1987, the U.S. and Japanese shocks explained 10.78 and 2.73, respectively,
B. Cha, S. Oh International Review of Economics and Finance 9 2000 299–322 311
Table 4 Percentage of k-week ahead forecast error variance of each Asian market’s return accounted for by the
US, Japanese, and domestic innovations HONG KONG
Period I: 8014,87925 k
Standard Error US
JAPAN HONG KONG
1 4.03813
4.92 2.05
93.02 4
4.12045 6.68
2.45 90.87
8 4.14111
6.67 3.10
90.23 12
4.14115 6.67
3.11 90.23
16 4.14115
6.67 3.11
90.23 20
4.14115 6.67
3.11 90.23
24 4.14115
6.67 3.11
90.23 Period II: 87116,97627
k Standard Error
US JAPAN
HONG KONG 1
3.07704 5.64
1.48 92.88
4 3.12638
7.04 1.58
91.38 8
3.13582 7.11
1.67 91.22
12 3.13590
7.11 1.67
91.22 16
3.13590 7.11
1.67 91.22
20 3.13590
7.11 1.67
91.22 24
3.13590 7.11
1.67 91.22
Period III: 9774,98918 k
Standard Error US
JAPAN HONG KONG
1 5.39934
23.72 1.43
74.85 4
5.68375 28.09
2.97 68.94
8 5.76791
28.62 4.17
67.21 12
5.77079 28.64
4.20 67.16
16 5.77086
28.64 4.20
67.16 20
5.77087 28.64
4.20 67.16
24 5.77087
28.64 4.20
67.16 KOREA
Period I: 8014,87925 k
Standard Error US
JAPAN KOREA
1 2.70818
0.43 1.34
98.23 4
2.73397 1.34
1.71 96.96
8 2.73799
1.35 1.87
96.77 12
2.73802 1.36
1.87 96.77
16 2.73802
1.36 1.87
96.77 20
2.73802 1.36
1.87 96.77
24 2.73802
1.36 1.87
96.77 continued on next page
312 B. Cha, S. Oh International Review of Economics and Finance 9 2000 299–322
Table 4 continued
Period II: 87116,97627 k
Standard Error US
JAPAN KOREA
1 2.91338
1.56 3.31
95.13 4
2.93535 1.86
3.84 94.30
8 2.94337
1.98 4.12
93.90 12
2.94340 1.98
4.12 93.90
16 2.94340
1.98 4.12
93.90 20
2.94340 1.98
4.12 93.90
24 2.94340
1.98 4.12
93.90 Period III: 9774,98918
k Standard Error
US JAPAN
KOREA 1
5.59997 1.37
6.32 92.31
4 5.99069
9.99 9.03
80.98 8
6.08760 10.72
9.12 80.16
12 6.10733
10.89 9.09
80.03 16
6.11202 10.92
9.08 80.00
20 6.11330
10.93 9.08
80.00 24
6.11364 10.93
9.08 80.00
SINGAPORE Period I: 8014,87925
k Standard Error
US JAPAN
SINGAPORE 1
2.68836 6.37
1.79 91.83
4 2.78761
10.84 2.22
86.95 8
2.80737 10.78
2.73 86.49
12 2.80764
10.78 2.73
86.48 16
2.80765 10.78
2.73 86.48
20 2.80765
10.78 2.73
86.48 24
2.80765 10.78
2.73 86.48
Period II: 87116,97627 k
Standard Error US
JAPAN SINGAPORE
1 2.18749
9.16 5.69
85.14 4
2.26603 13.25
6.59 80.16
8 2.26925
13.29 6.76
79.95 12
2.26926 13.29
6.76 79.95
16 2.26926
13.29 6.76
79.95 20
2.26926 13.29
6.76 79.95
24 2.26926
13.29 6.76
79.95 continued on next page
B. Cha, S. Oh International Review of Economics and Finance 9 2000 299–322 313
Table 4 continued
Period III: 9774,98918 Standard Error
US JAPAN
SINGAPORE 1
5.01981 27.16
0.04 72.80
4 5.47872
27.41 7.37
65.22 8
5.61658 27.95
8.22 63.84
12 5.62759
28.06 8.32
63.61 16
5.62881 28.07
8.34 63.59
20 5.62893
28.07 8.34
63.59 24
5.62895 28.07
8.34 63.58
TAIWAN Period I: 8014,87925
k Standard Error
US JAPAN
TAIWAN 1
2.55266 0.95
0.33 98.72
4 2.65281
1.73 0.41
97.86 8
2.75171 1.62
1.16 97.22
12 2.76080
1.61 1.22
97.17 16
2.76183 1.61
1.22 97.17
20 2.76196
1.61 1.22
97.17 24
2.76198 1.61
1.22 97.17
Period II: 87116,97627 k
Standard Error US
JAPAN TAIWAN
1 5.25462
0.03 4.56
95.41 4
5.36258 1.73
4.89 93.38
8 5.38120
1.89 5.12
92.99 12
5.38129 1.89
5.12 92.99
16 5.38129
1.89 5.12
92.99 20
5.38129 1.89
5.12 92.99
24 5.38129
1.89 5.12
92.99 Period III: 9774,98918
k Standard Error
US JAPAN
TAIWAN 1
3.14478 28.41
0.02 71.58
4 3.30596
31.75 2.59
65.66 8
3.51622 29.96
2.40 67.63
12 3.54105
29.75 2.45
67.80 16
3.54589 29.70
2.45 67.85
20 3.54683
29.69 2.45
67.86 24
3.54701 29.69
2.45 67.86
of the unexpected movement in the Singaporean stock returns. In Period II, the U.S. impact increased to more than 13. After the Asian financial crisis erupted, Singapore
is still the most affected by U.S. and Japanese shocks among the four AEMs. Together, U.S. and Japanese influences explain more than 36 of the variation in the Singa-
314 B. Cha, S. Oh International Review of Economics and Finance 9 2000 299–322
porean stock returns. Another interesting observation is that in Hong Kong and Singapore, the U.S. impact dominates the Japanese impact in all three periods. On
the other hand, the Japanese impact exceeds the U.S. impact in Period III in Korea and in Periods I and III in Taiwan.
These results suggest that an emerging market’s sensitivity to shocks from developed markets is related to its degree of openness. As is well known, there are no restrictions
on equity investment for either foreigners or domestic residents in Hong Kong and Singapore. These two AEMs also permit foreign currency to be either imported or
exported. Through these venues, shocks from the U.S. can be more directly transmitted to these two Asian financial centers. The higher level of restrictions and regulations
on international capital flows in Korea and Taiwan appear to have made their stock markets less sensitive to foreign shocks.
Given their initial insulation from foreign shocks, one would expect the opening of markets after the crisis to have a greater impact on Korea and Taiwan than on
Hong Kong and Singapore, which were already quite open before the crisis. This prediction is borne out by the data in Table 4. While U.S. influence after the crisis
increased by 4.0 and 2.1 times in Hong Kong and Singapore, respectively, its influence in Korea and Taiwan increased by 5.5 and 15.7 times, respectively. It is particularly
interesting to note that the U.S. impact has increased more in Taiwan than in Korea, even though Taiwan was able to ward off the crisis and Korea wasn’t. Such a result
appears to be related to the fact that the Japanese impact increased by more than five times in Korea after the crisis, while it has been reduced in Taiwan. Since it was
Japanese banks’ refusal to roll over the Korean short-term liabilities that triggered the Korean financial crisis, the Japanese influence would naturally be much bigger
in Korea than in Taiwan. Conversely, the relative influence of the U.S. market should be smaller in Korea than in Taiwan.
4.2. Impulse response functions Next, we compute the dynamic impulse responses of the AEMs’ return rates to
random shocks in the U.S. and Japanese return rates. In Eq. 6, computing the dynamic responses is equivalent to tracing out its coefficients C
s
. Specifically, the
response of Y at t 5 k to an initial shock of size l in the u vector is C
l
k
. Similarly, the response of Y at t 5 k to a shock of size one to u
j
is the j-th column of C
k
. Thus, C
ij k
is the dynamic impulse response of Y
i
in k quarters to a positive shock of one standard deviation in Y
j
, which is regarded as a measure of dynamic causality from Y
j
to Y
i
. Fig. 3 presents impulse responses for the AEMs’ return rates to a 1 positive
shock in the U.S., Japanese, and domestic return rates. The responses are computed for up to 24 weeks and drawn using the same scale. Figures on the vertical axis are
in percentage terms. Table 5 also presents the minimum and the maximum values of these responses.
Fig. 3 shows a striking contrast between the pattern of the responses before and after the crisis. In all four AEMs, the responses to the U.S. and the Japanese shocks
are bigger and more persistent after the outbreak of the Asian financial crisis. In
B. Cha, S. Oh International Review of Economics and Finance 9 2000 299–322 315
Fig. 3.
Responses of
stock returns.
316 B. Cha, S. Oh International Review of Economics and Finance 9 2000 299–322
Fig. 3.
continued
B. Cha, S. Oh International Review of Economics and Finance 9 2000 299–322 317
Fig. 3.
continued
318 B. Cha, S. Oh International Review of Economics and Finance 9 2000 299–322
Fig. 3.
continued
B. Cha, S. Oh International Review of Economics and Finance 9 2000 299–322 319
Table 5 Minimum and maximum values of impulse responses
Responses to a 1 US Shock Period I
Period II Period III
MIN MAX
MIN MAX
MIN MAX
HONG KONG 20.156
0.241 20.013
0.244 20.524
0.274 KOREA
20.072 0.077
20.110 0.100
20.698 0.234
SINGAPORE 20.067
0.323 20.023
0.306 20.381
0.382 TAIWAN
20.002 0.080
20.022 0.347
20.232 0.316
Responses to a 1 Japanese Shock Period I
Period II Period III
MIN MAX
MIN MAX
MIN MAX
HONG KONG 20.173
0.247 20.046
0.021 20.185
0.249 KOREA
20.005 0.108
20.054 0.063
20.183 0.539
SINGAPORE 20.147
0.115 20.036
0.068 20.216
0.672 TAIWAN
20.183 0.038
20.078 0.078
20.041 0.185
Responses to a 1 Domestic Shock Period I
Period II Period III
MIN MAX
MIN MAX
MIN MAX
HONG KONG 20.053
1.020 20.065
1.015 20.073
1.158 KOREA
20.071 1.020
20.035 0.055
20.110 1.183
SINGAPORE 0.000
1.021 20.055
1.016 20.174
1.171 TAIWAN
0.001 1.021
20.052 1.016
20.093 1.171
Periods I and II, the responses are small and tend to start to dampen after two weeks, before dying out in six or seven weeks. In Period III, however, the responses of the
AEMs’ return rates to the U.S. and the Japanese shocks are larger and last much longer. The responses to the U.S. shocks persist longer than those to the Japanese
shocks during Period III.
The reactions to shocks in the developed countries’ stock markets did not always have predictable effects on the AEMs. For example, the 1 unexpected increase in
U.S. return rates caused Korean rates to oscillate from 10.7 to 21.27 in Period III. The Korean market also shows a bigger response to Japanese shocks than the
Taiwanese market in Period III, which is consistent with the results of the variance decomposition analysis.
5. Conclusion
