1

**THE US AND THE ASEAN-5 STOCK EXCHANGES LINKAGES IN THE PERIODS **
**OF STOCK MARKET TURMOIL **

**Adwin Surja Atmadja**

*Faculty of Economics Petra Christian University *

**Abstract **

This study attempts to answer how the U.S and the ASEAN-5 stock markets’ indices would interrelate during the periods of stock market turmoil. The multivariate time series analyses conducted on the series reveal that there are cointegrating relationships on the series of the two sub-sample periods of the 1997 and the 2002 crisis. However, the study fails to detect any cointegrating vector on the series during the 2007 crisis.

The granger causality tests applied to the series reveal that the number of significant causal linkages between two variables on the series rocketed during the 2007 crisis. In addition, the accounting innovation analysis shows an increase in the explanatory power of an endogenous variable to another in the system during the latest crisis, indicating that the contagious effect of the latest crisis had not only largely influenced, but also dramatically changed the pattern of the short run dynamic interaction of the six capital markets.

Keywords : stock market integration, stock market crisis.

2

**Background **

A number of studies have been conducted on stock market interdependence around the period of financial or stock market crises in Asia-Pacific region. The general consensus is that the degree of integration among countries tends to change with a stronger integration during crisis periods than that before and after the periods (Sheng and Tu, 2000, and Yang, Kolari, and Min, 2003). It is also interesting to note that U.S stock market has played an important role in most national and regional stock markets, including some Asian’s stock markets during the 1997 crisis (Sheng et al, 2000).

3 The 2002 stock market downturn, meanwhile, originally hit the US stock markets. The downturn can be viewed as part of a larger correction in the US stock market triggered by some factors, including the September 11 attacks; an outbreak of accounting scandals; bankruptcy of some dotcom companies. This large stock market downturn, in fact, caused investors’ confidence suffered, and influenced other national stock exchanges.

The (2007) recent crisis also sparked in the U.S. in the second semester of 2007. At the time, the US financial market was deeply suffered from the most significant economic shocks initiated by the sub-prime mortgage crisis leading to the downturn in housing market, and then worsened by the spike in commodity prices (Yellen, 2008). Webb (2009) mentions that this crisis is a representation of hubris or an overconfidence that the previously smooth system will never fail or even collapse. The devastating effects of the US financial market turmoil then widely spread throughout the world.

The focus of this study is particularly on the stock market interdependence among national equity indices in six countries, which are the US; Singapore; Indonesia; Malaysia; Thailand; and the Philippines, during the three crisis periods. The multivariate time series is employed to analyse the degree and the existence of the long-run equilibrium, as well as to explain the short-run dynamic interactions among the indices in three sub sample periods. The study is structured as following: Section One describes the condition of the six stock markets in different periods; Section Two reviews the literature; Section Three discusses the research methodology; Section Four presents the empirical results; Section Five concludes.

4 The basic theoretical concept of financial market integration is laid on the law of one price. In integrated financial markets, the assets with the same risk in different markets will result in the same yield when measured in a common currency (Stulz, 1981). However, if the yields are different across the markets, the arbitrage process will play an important role in eliminating the differences. Operationally capital markets integration refers to the extent that markets’ participants are enabled and obligated to take notice of events occurring in other

markets by using all available information and opportunities, while financial market integration is defined in terms of price interdependence between markets (Kenen 1976).

Roca (2000) states that stock market integration is affected by some factors, such as economic integration (Eun and Shim, 1989), multiple listing of stocks, regulatory and information barriers, institutionalisation and securitisation, and market contagion (King and Wadwhani 1990), which may significantly determine the dynamic relationships among stock markets (Climent and Meneu, 2003), even though in the case of emerging markets, the contagion effect could be smaller than what is widely perceived (Pretorius 2002).

Much research has been carried out in order to find and analyse the existence of integration or long-run equilibrium in stock market across countries. The results are different depending on where, when, and how the research has been conducted. Palac-McMiken (1997) also reveals the existence of cointegration in ASEAN markets1 (Malaysia, Singapore, Thailand, and the Philippines), except Indonesia, during 1987 to 1995. The result is confirmed by Masih and Masih (1999) who report that some of ASEAN countries (Thailand, Malaysia, and Singapore) have a high degree of interdependence with other Asian (Hong Kong and Japan) and developed (the U.S. and the U.K.) stock markets. Furthermore, Masih and Masih (2001) also

1

5 find one cointegration vector among several major Asian stock markets (Hong Kong, Korea, Singapore, and Taiwan) and major developed markets.

Somewhat contradicts with those of Chung and Liu (1994) and Masih et al. (1999), Chan, Gup and Pan (1992) and DeFusco, Geppert and Tsetsekos (1996) also mention that there is no cointegration between the U.S and several Asian emerging stock markets (Hong Kong, Taiwan, Singapore, Korea, Malaysia, Thailand, and the Philippines) in the 1980s and early 1990s. Interestingly, Pretorius (2002) reports that the stock markets of countries in the same region are more interdependent than those in different regions. Consistent with this finding, Roca (2000) reveals the existence of interdependency among all the ASEAN stock markets in the short run, but not in the long run, during 1988-1995. These findings imply that the interdependence among stock markets is not stable over time as it is also mentioned by Arshanapalli and Doukas (1993).

6 A research conducted by Yang et al (2003) examine the long-run relationship and short-run dynamic causal linkages among the U.S, Japanese, and ten Asian emerging markets using daily data of 1997-1998 periods. They confirm that the stock markets in those countries had been more integrated after the 1997 Asian financial crisis than before the crisis. Both long-run cointegration relationship and short-long-run causal linkages in those markets become stonger during the period of crisis. Meanwhile, Atmadja, Wu and Juli (2009) find cointegration relationship among the eight Asia – Pacific (the U.S, Australia, Japan, Taiwan, Hongkong, China, Korea, and Indonesia) stock indices in the 2007 crisis period, but not in the 1997 crisis. Moreover, they also reveal that the short run dynamic interaction among the indices to be more intensealong the examination periods.

**Research Methodology **

**Data and Samples **

The daily closing stock price indices of the NYSE Composite of New York Stock Exchange-USA (NYSEALL) and the five ASEAN countries, which are Jakarta SE Composite (JAKCOM) of Indonesia; Kuala Lumpur SE Composite (KLSE) of Malaysia; Philippine SE Index (PSEi) of the Philippines; Straits Times Index (STI) of Singapore; and Bangkok - SET Composite of Thailand, would be used as measurement of the countries’ daily stock index movements in the observed periods.

7 1. The 1997 crisis: from July 1997 – March 1999. This period classification is somewhat similar with the one suggested by some researches (Kamin, 1999, Corsetti, Pesenti and Roubini,1999, Sheng et al., 2000, and Yang et al., 2003).

2. The 2002 Stock market downturn period begun in March 2002 and ended in December 2003 when the level of indices sat back at their same level before the crisis.

3. The 2007 crisis: from July 2007 – June 2009, as it is mentioned by Yellen (2008).

**Empirical Framework **

The two most appropriate models of multivariate time series analysis framework that one of which may suitable for this study are VAR and VECM. In the Vector autoregressive model (VAR) all variables are endogenous, and symmetrically treated. A VAR could be very large, however, in standard form, it could be written as :

p

xt = o +

### ∑

i xt-i + ti = 1

VAR requires that all variables be stationary2, and the appropriate lag length is data driven (Brooks, 2002). To define the appropriate lag length, some tests of information criteria will be applied, including the likelihood ratio (LR) test; Akaike Information Criterion (AIC); and Schwarz Bayesian Criterion (SC).

The LR test is based on asymptotic theory and is an F-type approximation. This test actually compares a restricted VAR (less lags) to an unrestricted VAR (more lags), which the null hypothesis is that the restricted model is correct. However, the shortcoming of the test is that

2

8 it may not be useful in small samples. It is worth noting that the LR test is only valid when the restricted model is tested.

Because of the limitation of the likelihood ratio test, multivariate generalization of AIC and SC may be the most suitable alternatives. The minimum values of AIC and/or SC could validly indicate the appropriate lags length, as long as the model’s residual does not suffer from serial correlation problem. Otherwise, the lag length may be too short. Thus, it is necessary to re-estimate the model using lag length that is serially uncorrelated.

In VAR, a block causality test, which is the Pairwise Granger Causality / Block Exogenity Wald Tests based on VAR, would be used to examine whether the lags of one variable enter into the equation for another variable. If y1 granger-causes y2, the parameters of lags of y1

should then not equal zero in the equation of y2. However, it is worth noting that

granger-causality basically means a correlation between the current value of one variable and the past (lags) value of others. Instead, Granger causality simply implies a chronological ordering of movements of the series.

As an alternative of VAR, the vector error correction model (VECM) or cointegration framework analysis could be a correct toolkit to analyze the series, if only the series contain unit root. The VECM basically is a VAR augmented by the error correction term (êt-1), which

takes the form as :

p

xt = o +

### ∑

i xt-i + ′ xt-1 + vt9 Thus, if the parameters of error correction term (ECT), called speed of adjustments () in VECM, are zero, then VECM reverts to a VAR in first differences. Otherwise, the larger the speed of adjustments, the greater the response to previous periods’ deviation from the long run equilibrium. A cointegration relationship is a long term or equilibrium phenomenon, since it is possible that cointegrating variables may deviate from their relationship in the short run, but their association would return in the long run3. Since the VECM result is also sensitive to its lags length, it is essential to use appropriate lag length to get the appropriate outcomes by conducting the lag order selection criteria (LR, AIC, or SC) tests.

Cointegration requires that all variables in a model be integrated with the same order. Ones may use the Engle-Granger (EG) test, which is basically a residuals-based approach, or the Johansen Cointegration Test to test the existence of cointegrated variables. Johansen (1988, 1991) proposed the maximum likelihood based two statistics to test the rank of the long-run information, namely:

max (r, r + 1) = – T ln(1 – r+1)

trace (r) = – T ln(1 – i)

where i are estimated Eigenvalues (characteristic roots) ranked from largest to smallest. The

trace is a likelihood ratio test statistics for the hypotheses that are at most *r* cointegrating
vectors. The max is the maximal Eigenvalue statistic that tests the hypothesis of *r*
cointegrating vectors against the hypothesis of r – 1 cointegrating vectors. If Eigenvalues i’s

are all zero, then the trace and max will be zero. To test for the number of cointegrating

3 _{‘A principal feature of cointegrated variable is that their time paths are influenced by the extent of any }

deviation from long run equilibrium. After all, if the system is to return to long run equilibrium, the

movements of at least some of the variables must respond to the magnitude of the disequilibrium.’ (Enders,

10 vectors, this study employs Johansen and Juselius’s (1990) trace and max statistics that are

adjusted for the degree of freedom. In the case of a cointegration relationship does not exist, a VAR analysis in first difference will then be the correct specification to conduct the estimation.

Following the VECM estimation, the VEC Pairwise Granger Causality / Block Exogenity Wald Tests is applied to reveal block-causality relationship between two variables. If there is a block causality relationship between the both variables, then lags of a variable should be significant in the equation for another.

A direct interpretation of the cointegration relations, may be difficult or misleading (Lutkepohl and Reimers, 1992, Runkle, 1987). As in a traditional VAR analysis, innovation accounting analysis, which consists of Impulse Response and Variance Decomposition Analysis, can provide a solution to the interpretation problem, and might be the most appropriate method to explore the short run dynamic structure of market linkages (Yang et al., 2003). This analysis would answer whether changes in the value of a given variable have positive or negative effect on the other variables in the system, or how long it would affect the variable to work through the system.

11 study employs the generalized4 type of impulse responses analysis since the orthogonalized type is sensitive to the ordering of the variable in the system.

Forecast error variance decomposition analysis, meanwhile, refers to the proportion of the movements in a sequence due to its own shock versus shocks to the other variables. The analysis separates the variation in an endogenous variable into the component shocks to the system. The variance decomposition, therefore, provides information about the relative importance of each random innovation in affecting the variables in the system. It determines how much of the s-step ahead forecast error variance of a given variable is explained by innovations to each explanatory variable. To some extent, impulse responses and variance decompositions offer very similar information.

**Empirical Results **

The ADF tests applied on the series in the three sub-sample periods result in that all series in all sub-sample periods contain unit root, meaning that the series are non stationary. The examination then continues with determining the appropriate lags length of the series by using the information criteria (LR, AIC, and SC) tests. The tests give some conflicting results. However, as the rule of thumb, one should choose the shortest lags length provided by the tests as long as there is no problem of serial correlation. The appropriate lags length are reported in Table 1

4

The Generalized Impulses as described by Pesaran and Shin (1998) constructs an orthogonal set of innovations that does not depend on the VAR ordering. The generalized impulse responses from an innovation to the j-th variable are derived by applying a variable specific Cholesky factor computed with the j-th variable at the top of the Cholesky ordering. Dekker, Sen and Young (2001) found that the generalized approach provided more accurate results than the traditional orthogonalized approach for both impulse response and forecast error variance decomposition analysis.

12

**TABLE 1. Lags Order and Number of Cointegrating Vector Tests**

**Periods ** **Lag Order ** **Number of Cointegrating **
**Vector(s) **

1997 crisis 3 2

2002 crisis 2 1

2007 crisis 4 0

Note:

the tests based on sequential modified LR test statistic (each test at 5% level)

Considering the number of appropriate lags, the number of cointegrating vectors is tested by using Johansen and Juselius’s (1990) trace and max statistics. In case of there are conflicting results between max and trace statistic, Johansen et al (1990) suggest that the trace tends to have more power than the max because trace takes into account all degrees

of freedom (n-r) of the smallest eigenvalues, then the number of cointegrating vectors suggested by the trace statistic would be employed,. With exclusion of linear trend and 95% critical values, Table 1 also presents the test outcomes.

As can be seen in Table 1, all series have cointegrating vector, except the one in the 2007 crisis period. The absence of cointegrating vector on the series of the 2007 crisis, in fact, has implication that the indices would not converge to their long run equilibrium. However, it does not necessarily mean that the dynamic short run interrelations are not possible to exist among the indices. The absence of cointegrating relationship has a consequence that the cointegration analysis framework is not appropriate to examine the series. Instead, the VAR in first difference would be the most suitable measurement.

13 vectors. This could happen because the 1997 crisis originally emerged in South East Asia region, greatly affected the ASEAN stock markets, as a result of some indifferences in macro-economic; stock market characteristics; geographical condition of the countries, that most likely had increased the contagious effect of the crisis (Eun et al, 1989; King et al, 1990; Pretorius, 2002).

Considering the outcomes of the Johansen Cointegration test, the cointegration analysis would then validly be used to estimate the series in the 1997 and the 2002 crisis periods only. The NYSE Composite is treated as the world index in these analysis. Based on t-statistic at the 5% level of significance, Table 2 shows that during the 1997 crisis, all indices, except STI and JAKCOM, had significant influence on the first cointegrating relation. Meanwhile, STI; KLSE; and SET significantly affected the second cointegrating vector. In addition, NYSEALL; STI; JAKCOM; and SET significantly contributed to the long run equilibrium of the observed indices in the second crisis period. The significant contributions of KLSE and PSEI to the cointegrating relation in the previous period, however, vanished during this period.

**TABLE 2. Estimates of Cointegrating Vector **

**Cointegrating **
**Equation: **

**PERIODS **

**1997 crisis ** **2002 crisis **

CointEq1 CointEq2 CointEq1

NYSEALL 1.000000 0.000000 1.000000

STI 0.000000 1.000000 -0.503147

(0.18572) [-2.70911]

JAKCOM 0.186180 0.919703 -0.342422

(0.18019) (0.50488) (0.11381)

[ 1.03327] [ 1.82163] [-3.00877]

KLSE 1.308215 2.811024 0.321668

(0.27975) (0.78387) (0.25670)

14

SET -0.543931 -3.410506 0.162479

(0.24413) (0.68406) (0.05878)

[-2.22800] [-4.98568] [ 2.76419]

PSEI -1.394674 -0.951992 -0.188458

(0.22680) (0.63549) (0.16500)

[-6.14936] [-1.49804] [-1.14218]

C -4.332850 -3.224246 -4.570477

Note: cointegration with unrestricted intercepts and no trends in CE and VAR. Standard errors in ( ) & t-statistics in [ ], level of significance 5%

Table 3 presents the speed of adjustment coefficients of the error correction term (i) that

have important implications for the dynamics of the system. A negative value of the significant speed of adjustment indicates a downward long run adjustment, while the positive one implies an upward long run adjustment.

In the 1997 crisis, the speed of adjustment coefficients for the first cointegrating vector for NYSEALL; JAKCOMP; and KLSE are statistically zero, with the critical value of 5%. The meaning is that the first cointegrating vector had no contribution to the convergence of those indices to their long run path, although NYSEALL and KLSE had significant contribution to the first cointegrating vector. In contrast, STI would positively react to a disequilibrium among the other indices. For the second cointegrating vector, the speeds of adjustment coefficients for all ASEAN-5 indices are statistically significant showing that the cointegrating vector had significant contribution to the convergence of the indices to their long run equilibrium.

15 equilibrium as the speed of adjustment coefficient for NYSEALL was insignificant during the period.

**TABLE 3. Speed of Adjustment Parameter of the Error Correction Term **

**Error **

**Correction: ** **NYSEALL ** **STI ** **JAKCOM ** **KLSE ** **LNSET ** **PSEI **

**1997 crisis **

ecm11) -0.001414 0.033795 0.011531 -0.001078 0.055495 0.057323 (0.00742) (0.01286) (0.01721) (0.02040) (0.01650) (0.01329) [-0.19072] [ 2.62792] [ 0.66991] [-0.05283] [ 3.36353] [ 4.31277] ecm2 (2) -7.78E-05 -0.014345 -0.016303 -0.019001 -0.014772 -0.022338 (0.00261) (0.00452) (0.00606) (0.00718) (0.00580) (0.00468) [-0.02981] [-3.17061] [-2.69228] [-2.64725] [-2.54486] [-4.77702]

**2002 crisis **

ecm3 (4) -0.013544 0.041917 0.086862 0.016842 0.047704 0.024071 (0.01743) (0.01587) (0.01770) (0.00993) (0.01623) (0.01425) [-0.77690] [ 2.64118] [ 4.90725] [ 1.69635] [ 2.93950] [ 1.68957]

Note : 5% Level of significance, Standard errors in ( ) & t-statistics in [ ]

As it was discussed in the previous section, a VECM does not seem to be appropriate for estimating the series of the 2007 crisis, since there is no cointegrating vector could be found. To overcome this circumstance, it is commonly suggested that a VAR analysis in first difference would be the correct specification to examine the series. The VAR analysis, however, requires that all variable must be stationary. Therefore, it is necessary to change the non stationary variable into the stationary one by differencing process. Following the alteration, re-identifying the appropriate lag length is a must. Three lags length is then found to be the most suitable lags length to analyze the series using the VAR in first difference. The brief result of the analysis can be seen in APPENDIX 1.

16 The objective of granger-causality test is to examine whether the lags of one variable (y1)

enter into the equation for another variable (y2)

Because the cointegrating relation does not always appear in all sub sample periods, it is necessary to conduct the Pairwise Granger Causality on its both different approaches. For the series containing cointegrating vector, the Pairwase Granger Causality based on Vector Error Correction (VEC) test is employed. Alternatively, the Pairwise Granger Causality based on VAR will test the series without cointegrating vector. The results are presented in Table 4.

**TABLE 4. Pairwise Granger Causality/Block Exogeneity Wald Tests **

**Dependent **

17 Table 4 shows that NYSEALL was the only stock index that significantly granger caused all the ASEAN-5 indices during the three crises. It suggests that changes or movements in the ASEAN-5 indices appeared to lags those of NYSEALL. On the other hand, none of the ASEAN indices, except KLSE and JAKCOM during the 2007 crisis, significantly granger caused NYSEALL. Thus, the past values of those indices were unable to forecast the present value of NYSEALL accurately. In the ASEAN-5’s standpoint, only JAKCOM (in both the 1997 crisis and the 2002 crisis) and PSEI (in the 1997 crisis) significantly granger caused the other ASEAN indices.

Interestingly, in the 2007 crisis, the number of block causality occurred on the series were almost twice as many as those in the previous periods. This outcome provides evidence that the short run interactions among the observed indices seemed to be more intense during the latest period of crisis.

The results of the generalized impulse responses analyses (see APPENDIX 2) show that SET and STI had played significant roles to the movements of the other indices during the 1997 crisis. A shock to SET, where the crisis initially occurred, would result in the second greatest contemporaneous response of STI, JAKCOM, and KLSE after their own shock. Meanwhile, a shock to STI would be reacted by NYSEALL, SET, and PSEI.

18 during the period compared to those before, although NYSEALL showed its dominance among the observed indices in a period later. A shock to NYSEALL had larger impact on all ASEAN indices movements, while NYSEALL gave a little or even no reaction to a shock to the ASEAN-5 indices in the period. Interestingly, STI’s response to a shock to NYSEALL was greater than its own shock. Thus, in general, an immediate response of an index to a shock to another increased during the 2007 crisis, even though the responses would, in fact, fade away quicker than those before.

While impulse response function traces the effect of a shock to one endogenous variable on to the other variables in the system, the variance decomposition separates the variation in an endogenous variable into the component shocks to the system. The forecast error variance decomposition, thus, tells the proportion of the movements in a sequence due to its own shock versus shock to the other variable. This implies that a shock to the i-th variable will not only affect that variable, but can also be transmitted to all other variables in the system.

The results of the forecast error variance decompositions for the six share indices (see

APPENDIX 3) reveal that, in general, the highest percentage of the variance decomposition

for an index is caused by its domestic shocks. Moreover, APPENDIX 3 also shows that there is no specific trend on the value along the sample periods.

19 values of the forecast error variance of STI were somewhat lower during the period compared to those before.

The proportions of the movements of the indices due to a shock to NYSEALL and STI rocketed during the 2007 crisis. NYSEALL greatly influenced STI, SET, and PSEI movements. Meanwhile, STI had more influence than NYSEALL on JAKCOM and KLSE movements. The higher values of this forecast error variance during the latest crisis compared to those before indicate the higher degree of short run interdependence among the variables. In general, it may be concluded that the influence of an index to the movement in another increased during the 2007 crisis, while the percentage value of the error variance attributable to own shocks generally declined.

** **

**Conclusion **

** **

The study finds two cointegrating vectors on the series of the 1997 crisis, and one cointegrating vector on the series of the 2002 crisis. This implies that the stock markets were interdependent and had long run equilibrium in the periods. The VECM estimation results show that most indices had significant contribution to the cointegrating relationship during the both crises. However, the study also detects no indication of cointegrating relationship on the indices during the 2007 crisis. These findings prove that the US and the ASEAN-5 stock markets integration had been removed by the 2007 crisis, and confirm that stock market interdependence is unstable and tend to change overtime.

20 innovation analysis results give evidences that STI apparently had more explanatory power to the other ASEAN indices’ movements during the 1997 crisis, while NYSEALL had played dominant role in the 2007 crisis. These outcomes also clarify that the short run dynamic interactions among the indices seem to be more intense during the latest period.

21

**BIBLIOGRAPHY **

Azman-Saini, W.N.M., Azali, M., Habibullah, M.S. and Matthews, K.G. 2002. Financial
Integration and the ASEAN-5 Equity Markets. *Applied Economics. Volume 34. *
2283-8.

Arshanapalli, B. and Doukas, J. 1993 . International Stock Market Linkages. Evidence from
the pre- and post-October 1987 Period. *Journal of Banking and Finance. Volume *
17, No. 1. 193-208.

Arshanapalli, B., Doukas, J. and Lang, L. 1995. Pre and post-October1987 Stock Market Linkages between U.S and Asian Markets. Pacific-Bacin Finance Journal. Volume 3, No. 1. 57-73.

Atmadja, Adwin S., Wu, Yanhui. dan Juli, Wan. 2009. Market Integration and Financial
crisis**. **New Evidence from Asian Pacific Markets. Indonesia Capital Market Review
*4th Edition. Forthcoming. *

Brooks, C. 2002. *Introductory Econometrics for Finance. Cambrige: Cambrige University *
Press.

Chan, K.C., Gup, B.E. and Pan, M. 1992. An Empirical Analysis of Stock Prices in Major
Asian Markets and the United States. *Financial Review. Volume 27, No. 2. *
289-307.

Cheng, W.K., Leng, G.K. and Lian, K.K. 2003. Financial Crisis and Intertemporal Linkages Across the ASEAN-5 Stock Markets. FEA Working Paper. No. 2003-4.

Chung, P. and Liu, D. 1994. Common Stochastic Trend in Pacific Rim Stock Markets.
*Quarterly Review of Economics and Finance. Volume 34. 241-259. *

Climent, F. and Meneu, V. 2003. Has 1997 Asian crisis Increased Information Flows between International Markets. International Review of Economics and Finance. Volume 12, No. 1. 111-143.

Corsetti, G., Pesenti, P. and Roubini, N. 1999. What Caused the Asian Currency and Financial Crisis? Japan and the World Economy. Volume 11. 305-73.

Daly, K.J. 2003. Southeast Asian Stock Market Linkages: Evidence from Pre- and Post-October1997. ASEAN Economic Bulletin. Volume 20, No. 1. 75-85.

DeFusco, R.A., Geppert, J.M. and Tsetsekos, G.P. 1996. Long-run Diversification Potential in Emerging Stock Markets. Financial Review. Volume 31, No. 2. 343-363.

Dekker, A., Sen, K. and Young, M. 2001. Equity Market Linkage in the Asia Pacific Region.
A Comparison of the Orthogonalized and Generalized VAR Approaches. *Global *
*Finance Journal. Volume 12. 1-33. *

22 Enders, W. 2004. Applied Econometric Time Series. Second Edition. New York: John Wiley

& Sons.

Eun, C.S. and Shim, S. 1989. International Transmission of Stock Market Movements.
*Journal of Financial and Quantitative Analysis. Volume 24 , No. 2. 241-56. *

Hee, N.T. 2000. Economic Integration in South East Asia. PhD Dissertation. The University of Wisconsin - Madison WI. MI: UMI Dissertation Services, A Bell & Howell Company, Ann Arbor.

Hee, N.T. 2002. Stock Market Linkages in South East Asia. *Asian Economic Journal. *
Volume 16, No. 4. 353-77.

Ibrahim, M.H. 2000. Financial Integration and Diversification among ASEAN Equity Markets: a Malaysia Perspective. Capital Market Review. Volume 8, No. 1. 25-40. Ibrahim, M.H. 2005. International Linkage of Stock Prices: the Case of Indonesia.

Management Research News. Volume 28, No. 4. 93-115.

Johansen, S. 1988. Statistical Analysis of Cointegration Vector. *Journal of Economic *
*Dynamics and Control. Volume 12, No. 2-3. 231-54. *

Johansen, S. and Juselius, K. 1990. The Full Information Maximum Likelihood Procedures
for Inference on Cointegration with Applications to the Demand for Money. Oxford
*Bulletin of Economics and Statistics. Volume 52, No. 2. 169-210. *

Johansen, S. 1991. Estimation and Hypothesis Testing of Cointegration Vectors in Gaussian Vector Autoregressive Models. Econometrica. Volume 59, No. 6. 1551-80.

Kamin, S.B. 1999. The Current International Financial Crisis. How Much is New?. *Journal *
*of International Money and Finance. Volume 18. 501-14. *

Kenen, P.B. 1976. Capital Mobility and Integration. A Survey. *Princeton Studies in *
*International Finance No. 39. USA : Princeton University. *

King, M and Wadwhani, S. 1990. Transmission of Volatility between Stock Markets. Review
*of Financial Studies. Volume 3 , No. 1. 5-33. *

Liu, Y.A., Pan, M.S. and Shieh, J.C.P. 1998. International Transmission of Stock Price
Movements. Evidence from the U.S. and Five Asian-Pacific Markets. *Journal of *
*Economics and Finance. Volume 22 , No. 1. 59-69. *

Lutkepohl, H. and Reimers, H.E. 1992. Impulse Response Analysis of Cointegrated System.
*Journal of Economic Dynamics and Control. Volume 16. 53-78. *

23 Masih, R. and Masih, A.M.M. 2001. Long and Short-term Dynamic Causal Transmission amongst International Stock Markets. Journal of International Money and Finance. Volume 20. 563-87.

Palac-McMiken, E.D. 1997. An Examination of ASEAN Stock Markets. a Cointegration Approach. ASEAN Economic Bulletin. Volume 13 , No. 3. 299-311.

Pretorius, E. 2002. Economic Determinants of Emerging Market Stock Market Interdependence. Emerging Markets Review. Volume 3. 84-105.

Roca, E.D. 2000. *Price Interdependence among Equity Markets in the Asia-Pacific Region. *
*Focus on Australia and ASEAN. Ashgate. Aldershot. *

Runkle, D.E. 1987. Vector Autoregressions and Reality. *Journal of Business and Economic *
*Statistics. Volume 5, No. 4. 437-442. *

Sheng, H. and Tu, A. 2000. A Study of Cointegration and Variance Decomposition among
National Equity Indices before and during the Period of the Asian Financial crisis.
*Journal of Multinational Financial Management. Volume 10 , No. 3. 345-65. *

Stulz, R. 1981. On the Effects of Barriers to International Investment. Journal of Finance. 36. 923-934.

Wongbangpo, P. 2000. *Dynamic Analysis on ASEAN Stock Markets. PhD Dissertation. *
Department of Economics Southern Illinois University at Carbondale, Carbondale –
IL. MI: UMI Dissertation Services, A Bell & Howell Company, Ann Arbor.

Yang, J., Kolari, J.W. and Min, I. 2003. Stock Market Integration and Financial Crises. the Case of Asia. Applied Financial Economics. Volume 13. 477-486.

24

**APPENDIX 1 **

**Vector Autoregression Estimates **

DLNYSEALL DLSTI DLJAKCOM DLKLSE DLSET DLPSEI

DLNYSEALL(-1) -0.195603 0.444076 0.362900 0.216314 0.330846 0.470399

DLSTI(-2) 0.027650 -0.091301 -0.036468 -0.014135 -0.043481 0.022143 (0.08158) (0.06553) (0.07092) (0.03892) (0.06055) (0.05304)

DLJAKCOM(-3) -0.016671 -0.051907 -0.086522 -0.038121 -0.057379 -0.071334 (0.07144) (0.05739) (0.06211) (0.03409) (0.05303) (0.04645) [-0.23335] [-0.90450] [-1.39297] [-1.11840] [-1.08196] [-1.53583]

DLKLSE(-1) -0.293605 -0.134793 -0.124978 -0.135580 -0.133783 0.059341 (0.11638) (0.09348) (0.10118) (0.05552) (0.08639) (0.07566) [-2.52291] [-1.44194] [-1.23522] [-2.44185] [-1.54867] [ 0.78433]

DLKLSE(-2) -0.405527 -0.178074 -0.166502 -0.144808 -0.314605 -0.061157 (0.11652) (0.09360) (0.10130) (0.05559) (0.08649) (0.07575) [-3.48033] [-1.90259] [-1.64359] [-2.60483] [-3.63735] [-0.80733]

DLKLSE(-3) -0.061016 0.228061 0.119153 0.176802 0.087526 0.127988 (0.11607) (0.09323) (0.10091) (0.05538) (0.08616) (0.07546) [-0.52569] [ 2.44612] [ 1.18076] [ 3.19268] [ 1.01588] [ 1.69612]

DLSET(-1) -0.129921 -0.223772 0.006069 -0.082932 -0.170231 -0.102307 (0.07948) (0.06384) (0.06910) (0.03792) (0.05900) (0.05167) [-1.63466] [-3.50508] [ 0.08783] [-2.18705] [-2.88541] [-1.97998]

25

DLSET(-3) -0.082612 -0.112591 -0.104078 -0.015746 -0.026827 -0.010035 (0.07910) (0.06353) (0.06877) (0.03774) (0.05871) (0.05142) [-1.04446] [-1.77213] [-1.51349] [-0.41726] [-0.45691] [-0.19516]

DLPSEI(-1) -0.025680 -0.038321 0.027957 0.057543 0.025716 0.047445 (0.07492) (0.06018) (0.06514) (0.03575) (0.05562) (0.04871) [-0.34275] [-0.63673] [ 0.42918] [ 1.60976] [ 0.46239] [ 0.97403]

DLPSEI(-2) -0.034663 0.052607 0.038345 0.046933 0.071260 0.059935 (0.07508) (0.06031) (0.06528) (0.03582) (0.05573) (0.04881) [-0.46168] [ 0.87230] [ 0.58743] [ 1.31018] [ 1.27861] [ 1.22788]

DLPSEI(-3) -0.011395 -0.001747 0.007863 -0.014698 0.036463 -0.052252 (0.06542) (0.05255) (0.05688) (0.03121) (0.04856) (0.04253) [-0.17419] [-0.03325] [ 0.13825] [-0.47091] [ 0.75089] [-1.22858]

C -0.001534 -0.000289 0.000484 -0.000191 -0.000122 -0.000194

(0.00098) (0.00079) (0.00085) (0.00047) (0.00073) (0.00064) [-1.56402] [-0.36628] [ 0.56758] [-0.40856] [-0.16806] [-0.30427]

R-squared 0.100398 0.230530 0.194794 0.242115 0.216041 0.425265

Adj. R-squared 0.068013 0.202829 0.165807 0.214831 0.187818 0.404574 Sum sq. resids 0.244826 0.157969 0.185061 0.055730 0.134902 0.103478 S.E. equation 0.022128 0.017775 0.019239 0.010557 0.016426 0.014386 F-statistic 3.100085 8.322097 6.719972 8.873936 7.654892 20.55367 Log likelihood 1251.111 1364.811 1323.735 1635.179 1405.771 1474.588 Akaike AIC -4.748018 -5.186168 -5.027881 -6.228051 -5.344012 -5.609203 Schwarz SC -4.592360 -5.030511 -4.872223 -6.072394 -5.188355 -5.453546 Mean dependent -0.001003 -0.000794 -0.000142 -0.000470 -0.000581 -0.000802 S.D. dependent 0.022921 0.019908 0.021064 0.011915 0.018226 0.018643

28

2 0.004777 0.000365 0.001490 -0.001146 -0.000904 0.000757 3 0.000860 0.000794 0.000915 -0.001529 0.000588 0.000490 5 0.000219 0.000217 -0.000135 -0.000473 -0.000521 -2.09E-05

** Response of SET: **

Period NYSEALL STI JAKCOM KLSE SET PSEI

1 0.006666 0.007750 0.003085 0.001150 0.012429 0.000000 2 0.005809 -0.002181 0.000943 -0.001291 -0.002059 0.000338 3 0.001850 0.001117 0.002206 -0.003042 0.000344 0.000771 5 0.000599 -0.000452 -0.000214 -0.001004 -0.001076 -0.000264

** Response of PSEI: **

Period NYSEALL STI JAKCOM KLSE SET PSEI

31

Period S.E. NYSEALL STI JAKCOM KLSE SET PSEI

1 0.016426 16.46728 22.25885 3.527946 0.489780 57.25614 0.000000 2 0.017754 24.80051 20.56146 3.301644 0.947907 50.35219 0.036286 3 0.018295 24.37809 19.73609 4.562671 3.658049 47.45349 0.211600 5 0.018515 24.54167 20.01009 4.552055 3.874974 46.79101 0.230203

** Variance Decomposition of PSEI: **

Period S.E. NYSEALL STI JAKCOM KLSE SET PSEI

1 0.014386 2.411756 5.405031 4.543029 1.731093 2.338218 83.57087 2 0.018574 38.00412 4.868794 3.979295 1.105116 1.797539 50.24514 3 0.018835 36.96001 5.365022 4.388070 2.087949 2.205803 48.99315 5 0.018948 36.58706 5.336441 4.671211 2.192331 2.513771 48.69918 Note : Cholesky Ordering: LNSTI LNJAKCOM LNKLSE LNSET LNPSEI