stock market integration

2 downloads 0 Views 472KB Size Report
As for the long run causal relations between ASEAN stock markets with the ..... Indonesia, Singapore, Thailand, the Philippines, the US and Japan, respectively.

INTERDEPENDENCE OF ASEAN-5 STOCK MARKETS FROM THE US AND JAPAN By: M. Shabri Abd. Majida, *, Ahamed Kameel Mydin Meerab, and Mohd. Azmi Omarc a

Assistant Professor, Department of Economics, Kulliyyah of Economics & Management Sciences, International Islamic University Malaysia (IIUM). b Associate Professor, Department of Business Administration, Kulliyyah of Economics & Management Sciences, International Islamic University Malaysia (IIUM). c Professor, Department of Business Administration, Kulliyyah of Economics & Management Sciences, International Islamic University Malaysia (IIUM).

Abstract This study empirically examines market integration among five selected ASEAN emerging markets (i.e., Malaysia, Thailand, Indonesia, the Philippines and Singapore) and their interdependencies from the US and Japan based on a two-step estimation, cointegration and Generalized Method of Moments (GMM). Closing daily stock indices starting from January 1, 1988 to December 31, 2006 are used. The study reveals that the ASEAN stock markets are going towards a greater integration either among themselves or with the US and Japan, particularly in the aftermath of the post-1997 financial turmoil. This implies that the long-run diversification benefits that can be gained by investors across the ASEAN markets tend to diminish. As for the long run causal relations between ASEAN stock markets with the US and Japan, the study discovers that Indonesia was relatively independent of both the US and Japan; Malaysia was more dependent on Japan rather than the US; Thailand was relatively independent of the US, but to some extent dependent on Japan; the Philippines is more affected by the US than Japan; and the US and Japan have bidirectional Granger causalities with Singapore.

Keywords: Stock market integration; cointegration; GMM; diversification benefit, ASEAN-5 JEL Classification: C15; F15; G15

__________________ *

Corresponding author: Tel: +603-61964708; Fax: +603-61964850. E-mail: [email protected]

0 Electronic copy available at: http://ssrn.com/abstract=1005287

Please consider my paper for Barclay's Global Investors Australia Prize & Sydney Futures Exchange Prize

1 Electronic copy available at: http://ssrn.com/abstract=1005287

1. Introduction The issue of dynamic interdependence among stock markets has become an important topic in the modern literature of financial economics. The degree of linkages or interdependencies among the stock markets provides important implications on potential benefits of international portfolio diversification and on financial stability of a country (Ibrahim, 2005). Prior to 1970s, empirical studies on market integration document lower correlations among national stock markets (Grubel, 1968; Lessard, 1973; and Solnik, 1974), implying the existence of potential benefits of international portfolio diversification. However, in their survey on available empirical evidences on market integration across national capital markets, Goldstein and Michael (1993) found that the international links have been increasing over the past decade, especially for the stocks traded actively in the major financial centres. The emerging markets are also found to be more closely integrated with markets in the rest of the world, although their integration progress has been far less than the industrial countries. This further implies that the potentialities of portfolio diversification benefits across the world stock markets in the long run tend to diminish. In addition, an increasing integration among the national stock markets further implies that international financial instabilities are easily transmitted to domestic financial markets, a phenomenon called as ‘financial contagion’ (Ibrahim, 2005). Unlike voluminous studies on market integration among developed countries, there have been relatively few studies explore the issue of stock market integration in the ASEAN region. For example, Roll (1995) affirmed that although Indonesia has had an active equity market for a number of years, no empirical studies on this market have appeared in western scholarly journals. In developed economies such as America, Japan and Germany, both market integration and segmentation are documented. For the emerging economies, especially for the ASEAN stock markets, there have been very few empirical analyses done in this area in a few decades ago. However, in recent years, the vast-growing economic activities and the increasing investment opportunities in some emerging markets have attracted investors’ and researchers’ attention. Among the studies on the ASEAN stock market integration have been done by Barus (1997), Palac-McMiken (1997), Wongbangpo (2000), Ibrahim (2000, 2005), Hee (2002), Azman-Saini et al. (2002), Daly (2003), Cheng et al. (2003), Click and Plummer (2005), and Dunis and Shannon (2005). By employing the cointegration techniques, Barus (1997) explored the nature of the interlinkages between Indonesian and other founding members of the ASEAN capital markets during the period 1985-1995. The study found that the Indonesian capital markets were weakly cointegrated with other ASEAN capital markets. This result reveals that a social, political and economic cooperation in ASEAN does not guarantee a cointegrated capital markets among the country members and the benefits of portfolio diversification within the ASEAN capital markets may be understated. By using cointegration approach, PalacMcMiken (1997) found that with the exception of Indonesia all the five founding members of ASEAN markets are linked with each other during the period 1987-1995. Following Palac-McMiken (1997), Wongbangpo (2000) investigated the stock markets in the ASEAN region during the period of 1985-1996 through empirical investigations of the interconnection among stock markets and the interaction between stock and foreign exchange markets. The study documented that the ASEAN stock markets, except for the Philippines, share a long run co-movement, implying that an effective long term diversification of an investor's portfolio among these stock markets cannot be achieved. Ibrahim (2000) explored the degree of financial integration and benefits of portfolio diversification among ASEAN equity markets from the Malaysian perspective from January 1988 to June 1997. From the cointegration analysis and Error Correction Model

2

(ECM), he found long run co-movements among the ASEAN and US equity markets. The short run interactions between ASEAN markets were mostly contemporaneous. He also found that the ASEAN markets were highly integrated and the US market exerted significant influence on the ASEAN markets. In a more recent study, Hee (2002) explored the market linkages in South-East Asia over the period 1988-1997 through the use of correlation and cointegration analyses, a timevarying parameter model and the concepts of covered interest parity, respectively. He found no long run relationships among the stock markets of ASEAN; however, correlation analyses indicated that the markets were becoming more integrated. Azman-Saini et al. (2002) empirically investigated the financial integration among the ASEAN-5 (Malaysia, Indonesia, Thailand, Singapore and the Philippines) equity markets through the use of cointegration analysis and Seemingly Unrelated Regression (SUR) based on Toda and Yamamato (1995) Granger noncausality test. The evidence of cointegration among the ASEAN markets was found although they did not find all markets sharing common stochastic trends. An increase in the interdependencies across the ASEAN markets in the aftermath of the 1997 Asian financial crisis is also supported by Daly (2003). Using closing daily indices from 1990-2003, he concluded that there has been a significant increase in the integration between ASEAN stock markets during the post-crisis period. By employing daily closing stock indices over the period of January 1992 to August 2002, Cheng et al. (2003) investigated the linkages across the stock markets of ASEAN-5 (Singapore, Malaysia, Indonesia, Thailand and the Philippines). They found that the ASEAN-5 stock markets are cointegrated before and after the crisis, but not during the crisis. Using cointegration analysis, Click and Plummer (2005) also found that the ASEAN-5 markets were cointegrated. Ibrahim (2005) investigated integration among the ASEAN markets from the perspective of the Indonesian market using cointegration techniques and vector autoregression (VAR) for the periods January 1988 to December 2003. He documented evidence for lack of integration among the ASEAN markets. Finally, Dunis and Shannon (2005) investigated integration among emerging markets from South-East Asia (i.e., Indonesia, Malaysia and the Philippines) and Central Asia (Korea, Taiwan, China and India) with the three ‘established’ markets, the US, the UK and Japan during the post-1997 Asian financial crisis. They found that all the emerging markets have become more closely integrated with the Japanese market. In contrast, the results for the US and UK indicate that several emerging markets have seen their level of integration remain steady or decline over the review period. Unlike most of the above reviewed studies that employed cointegration analysis in exploring market integration, this study adopts two-step estimation i.e., cointegration and Generalized Method of Moments (GMM). The GMM is documented to be a more superior technique of estimation as compared to other estimations. In the theoretical literature, the GMM provides a unified framework for the estimation theory, while in the applied literature; it provides a computationally convenient method of estimation in some models, which are burdensome to estimate with other methods (Hall, 1993). The GMM is potentially more robust than almost all the existing models because it does not suffer from the usual errors-in-variables problem (Zhou, 1999; and Kan and Chu, 1999). In other words, a strong distributional assumption such as the normal distribution of disturbance terms is not necessary in the GMM estimation (Ogaki, 1993). The above-mentioned advantages make the GMM estimation method easy to be implemented in many empirical analyses, because it neither requires strong (often unrealistic) normal distribution assumption of errors nor a specific error distribution

3

assumption; it only needs a suitable moment. Comparing it to Maximum Likelihood estimation, the GMM estimation improves the statistical significance of parameter estimates. This estimation is robust to the error distribution and provides consistent estimates (Lee and Lee, 1997; and Safvenblad, 1997). This motivates the present study to adopt two-step estimation, cointegration and GMM with the intention to produce a more robust finding. This study, therefore, is among the first to use two-step estimation to investigate market integration among ASEAN 1 emerging markets and their interdependencies from the two largest stock markets in the world, the US and Japan. In addition, this study also differs from earlier studies in the sense that it relies on a longer period and more recent data sample and analyses market integration both at bivariate and multivariate 2 levels. The superiority of these adopted estimations could mitigate the shortcomings of the previous studies on this subject, as believed. Does the integration among the ASEAN stock markets move towards a greater integration from the pre- to the post-1997 financial crisis? Who moves the ASEAN markets, the US or Japan? A detailed comparative investigation on the ASEAN stock markets during the preand post-1997 financial crisis is of interest because of the increased economic cooperation in accordance with the ASEAN agreement and the distinguished structure of emerging stock markets. Thus, this study attempts to partially fill this gap in the literature and to provide recent empirical evidence on market integration in the ASEAN region, relying on longer and more recent sample of data and superior model of estimation. The findings of this study may have implications for investors and companies in the international community who internationally diversify their investments and make capital budgeting decisions in this region. The objectives of this study are therefore to: (i) explore empirically the market integration in terms of a long run equilibrium relationship among the five founding members of the ASEAN emerging markets (i.e., Malaysia, Indonesia, Thailand, Singapore and the Philippines and their interdependencies from the US and Japan; and (ii) examine empirically the dynamic causal linkages among the ASEAN emerging stock markets and also their dynamic causal interdependencies from the US and Japan. The rest of this study is organized as follows. In the next section, the empirical framework, cointegration and GMM is outlined. Section 3 describes the data employed in the study. Section 4 provides the empirical results and discussions. Finally, Section 5 concludes the study, provides some implications and proposes some recommendations for further study.

1

Out of ten members of ASEAN (Association of South East Asian Nations), only the five founding members of ASEAN—Malaysia, Indonesia, Thailand, the Philippines and Singapore—are investigated in this paper. The rest of the five ASEAN members, namely: Vietnam, Brunei Darussalam, Myanmar, Cambodia and Laos are excluded from this study in view of the unavailability of data for the study period. Though Singapore is the most developed country in the region, it is still categorized under emerging market (Yang and Siregar, 2001). Masih and Masih (1997) called Singapore as "emerging Newly Industrialized Country (NIC) market". 2 Focusing only on bivariate or multivariate cointegration analysis might miss some important information. The markets are possible to be segmented using bivariate analysis, but they are become integrated using multivariate analysis, or vice versa. This was a shortcoming in the studies of Palac-McMiken (1997) and Jang and Sul (2002), to mention a few.

4

2. Empirical Framework Since the main purpose of the study is to investigate the stock market integration in terms of the long run equilibrium relationship among the stock markets, the adoption of two-step estimation—cointegration analysis and GMM—in this study is suitable for this purpose. Cointegration analysis is able to detect whether the integrated markets exist or not in the sense that there is a tendency among the markets to move together in the long run, while in the short run deviations from the equilibrium is also permissible. On the other hand, the GMM estimation has more flexibility and no stringent assumption compared to other estimations such as the Ordinary Least Squares (OLS) and Maximum Likelihood (ML). It also has a strong distributional assumption such as error terms, ut is not necessarily normally distributed (Ogaki, 1993). Thus, the market integration is tested in the first step with the Johansen’s (1988) and Johansen and Juselius’s (1990), henceforth JJ cointegration methods, whereas the dynamic causal relationships among the ASEAN markets are estimated simultaneously in the second step by the GMM. 2.1. Cointegration Analysis To test for cointegration among the stock markets of ASEAN vis-à-vis the US and Japan, the maximum likelihood approach of the JJ cointegration approach is adopted. Essentially, the JJ test started its cointegration model with a Vector Autoregressive [VAR(k)] model as follows: Yt = A1Yt-1 + A2Yt-2 + ……+ AkYt-k + εt …………….…………….(1) where Yt = (Y1t, Y2t, …….Ynt)'. Subtracting Yt-1 from both sides of the Equation (1) to have: ∆Yt = (A1 – I)Yt-1 + A2Yt-2 + ……AkYt-k + εt then adding and subtracting (A1- I)Yt-2 from both sides to get: ∆Yt = (A1 – I) ∆Yt-1 + (A2 + A1 – I )Yt-2 + ……AkYt-k + εt Again, adding and subtracting (A2 + A1 - I)Yt-3 from both sides to obtain: ∆Yt = (A1–I) ∆Yt-1 + (A2+A1- I) ∆Yt-2 + (A3+A2+A1–I)Yt-3 +……AkYt-k + εt Repeating addition and subtraction in this fashion, the JJ cointegration model can therefore, be formulated as follows: Δ Yt = δ + Γ i Δ Yt-1 + …….+ Γ k Δ Yt-k + ΠYt-k + εt ………..(2)

where Yt is an n x 1 vector of variables and δ is an n x 1 vector of constant, respectively. Γ is an n x n matrix (coefficients of the short run dynamics), Π = αβ′ where α is an n x 1 column vector (the matrix of loadings) represents the speed of short run adjustment to disequilibrium and β′ is an 1 x n cointegrating row vector (the matrix of cointegrating vectors) indicates the matrix of long run coefficients such that Yt converge in their long run equilibrium. Finally, εt is an n x 1 vector of white noise error term and k is the order of autoregression. As our study investigates market integration among the five founding members of ASEAN and their interdependencies from the US and Japan, therefore our n is equal to seven.

5

The above process from a VAR model to equation (2) is called the cointegrating transformation. The long run information matrix Π in this equation is the key to Johansen's cointegration test because its rank r determines the number of cointegrating vectors. If rank (Π) = 0, equation (2) returns to a VAR(k) model in the first differences and the components in Yt are not cointegrated. On the other hand, if Π is a full rank n, all component in Yt are stationary. In a more general case when 1 < rank (Π) < n, the number of cointegrating vectors is equal to r, the rank of matrix Π. Since the rank of a matrix is equal to the number of eigenvalues λi (or characteristic unit roots) that are significantly different from zero, Johansen proposed two statistics to test the rank of the long run information Π, namely: λtrace (r)

= -T

n

∑ ln(1 − λˆ ) …………………………...(3) i

i = r +1

λmax (r, r + 1) = -T ln ( 1 − λˆr + 1) ………………….…….…(4) where λi are estimated eigenvalues (characteristic roots) ranked from largest to smallest. The λtrace in the equation (3) is called the Trace statistics, which is a likelihood ratio test statistics for the hypotheses that are at most r cointegrating vectors. The λmax in the equation (4) is called the Maximal Eigenvalue statistic that tests the hypothesis of r cointegrating vectors against the hypothesis of r – 1 cointegrating vectors. The rank of Π is equal to the number of eigenvalues that are different from zero. If eigenvalues λi's are all zero, then the λtrace and λmax will be zero. To test for the number of cointegrating vectors, this study employs the Johansen and Juselius’s (1990) and Osterwald-Lenum’s (1992) λtrace and λmax statistics that are adjusted for the degree of freedom. Additionally, an important requirement for implementing the JJ cointegration test is that the variables are non-stationary integrated of the same order. Accordingly, prior to the JJ test, the standard Augmented Dickey-Fuller (1979) and Phillips-Perron (1988), henceforth ADF and PP unit root tests are conducted to determine the order of integration for each national stock price. 2.2. Generalized Method of Moments (GMM) In exploring the dynamic interdependencies among the stock markets of ASEAN, the US and Japan, the study estimates the equation (2) by GMM, where the error correction terms are incorporated in the models. Based on Hung and Cheung’s (1995) study, the equation (2) can then be simply reformulated in matrix form as follows: ⎡ΔMay⎤ ⎥ ⎢ ⎢ΔInd ⎥ ⎢ΔSing⎥ ⎥ ⎢ ⎢ΔThai⎥ = ⎥ ⎢ ⎢ΔPhil ⎥ ⎥ ⎢ ⎢ΔUS ⎥ ⎢ΔJP ⎥ ⎦ ⎣

⎡δ 0⎤ ⎢ ⎥ ⎢δ1 ⎥ ⎢δ 2⎥ ⎢ ⎥ ⎢δ 3⎥ + ∑ ⎢ ⎥ ⎢δ 4⎥ ⎢ ⎥ ⎢δ 5⎥ ⎢δ 6⎥ ⎣ ⎦

k i=1

Γ

i

⎡ΔMay⎤ ⎢ΔInd ⎥ ⎥ ⎢ ⎢ΔSing⎥ ⎥ ⎢ ⎢ΔThai⎥ + Π ⎢ΔPhil ⎥ ⎥ ⎢ ⎢ΔUS ⎥ ⎢ΔJP ⎥ ⎦ t −k ⎣

⎡May⎤ ⎢Ind ⎥ ⎥ ⎢ ⎢Sing⎥ ⎥ ⎢ ⎢Thai⎥ + ⎢Phil ⎥ ⎥ ⎢ ⎢US ⎥ ⎢JP ⎥ ⎦ t −1 ⎣

⎡ν 0 ⎤ ⎢ν1 ⎥ ⎢ν 2⎥ ⎢ ⎥ ⎢ ν 3 ⎥ .......... .......... ( 5 ) ⎢ν 4⎥ ⎢ν 5 ⎥ ⎢ ⎥ ⎣ν 6 ⎦

where May, Ind, Sing, Thai, Phil, US and JP indicate the stock markets of Malaysia, Indonesia, Singapore, Thailand, the Philippines, the US and Japan, respectively. Since our Model (2) or (5) considers the possibility of the past level of parameters to have an effect on current changes in other parameters, the lagged values have to be incorporated in the models. In this study, the Akaike (1974) Information Criterion (AIC) is used to determine the lag length incorporation into the entire tests of this study.

6

It is important to note that for the GMM estimator to be identified there must be at least as many instrumental variables Z as there are parameters θ. Following Lee and Lee (1997), this study used lags of explanatory variables as the instrumental variables. These variables were opted for use because of the difficulty in finding other instrument variables, as our study utilized daily data. These variables are, however, obvious instruments, and in most cases should be included in the instrumental list. Another important aspect of specifying GMM is the choice of the weighting matrix to yield a consistent and robust estimate. To get a robust estimate to heteroskedasticity and autocorrelation of unknown forms, the covariance matrix of the orthogonality conditions is estimated as suggested by Newey and West (1987) using Barlett estimators, while the lag truncation parameter is estimated as suggested by Newey and West (1994) with a fixed bandwidth, 3 following the study of Heinesen (1995). In addition, the prewhitening process is run to soak up the correlation in the moment conditions prior to the GMM estimation.

3. DATA To provide more robust and updated results, this study uses daily 4 closing data of the stock indices, 5 covering at least the last eighteen years, spanning from January 1, 1988 to December 31, 2006. All these indices are denominated in local currency units, extracted from Bloomberg Database. In this study, the stock returns for these markets are calculated from the following indices: (i) the Kuala Lumpur Stock Exchange Composite Index (KLSE-CI) for Malaysia; (ii) the Jakarta Stock Exchange Composite Index (JSX-CI) for Indonesia; (iii) the Bangkok Stock Exchange Trade Index (BSETI) for Thailand; (iv) the Philippines Stock Exchange Index (PSEI) for the Philippines; (v) the Singapore All Equities Index (SAEI) for Singapore; (vi) the Standard and Poor's 500 (S&P500) Index for the US; and (vii) the Tokyo Price Index (TOPIX) for Japan. There are at least two difficulties arise in investigating stock market interdependencies across countries. The first one is the missing observation problem due to different stock market holiday. Since the study extensively incorporates lags in the regressions, missing data is particularly troublesome. 6 Thus, it is desirable to fill in estimate-based information from an adjacent day. Rather than using a sophisticated interpolation, this study follows the studies of Jeon and Von (1990) and Hirayama and Tsutsui (1998) by adopting the method of Occam's razor (just fill in with the previous day's price). Simplistic as it may be, this study justifies this method on the premise that a closed stock exchange does not produce any information on bank holidays. Since no new information is revealed, the previous day's information is carried over to the subsequent day. Another complication in investigating market integration is the problem of noncontemporaneous markets or different trading hours among the markets. Table 1 reports the trading hours of the various stock markets considered. Except the US stock markets, all other markets are exactly or approximately operating in similar time zones with similar opening and closing times. It is therefore important to know the operating hours of one

3

We have also tried the quadratic spectral (QS) kernel estimation and bandwidth selection as suggested by Andrew (1991); the estimation results are very much the same. 4 Instead of weekly or monthly data, which many studies have tended to use, this study uses daily data, as it is more likely to capture potential interactions among stock markets. 5 We opt to use these indices in our study because all the indices are calculated based on capitalizationweighted method. 6 For example, if a regression equation contains six lags; one missing observation would additionally cause six subsequent observations to be dropped.

7

market relative to another when performing tests, and taking this into account implies incorporating appropriate lagged values where necessary. Table 1: Trading Hours of Stock Exchanges Stock Exchange

Greenwich Local Time New York Time Mean Time (h) (h) (h) 02:00—03:00 10:00—11:00 21:00—22:00 Kuala Lumpur 03:15—04:30 11:15—12:30 22:15—23:30 06:30—08:00 14:30—16:00 01:30—03:00 03:00—05:00 10:00—12:00 22:00—12:00 Jakarta 06:30—08:00 13:30—15:00 01:30—03:00 Thailand 01:30—10:00 08:30—17:00 20:30—05:00 02:00—04:30 10:00—12:30 21:00—23:30 Singapore 06:30—08:00 14:30—16:00 01:30—03:00 Philippine 01:30—06:30 09:30—14:30 20:30—01:30 0:00—0.2: 00 9:00—11:00 19:00—21:00 Tokyo 04:00—06:00 13:00—15:00 23:00—01:00 New York 14:30—21:00 09:30—16:00 09:30—16:00a Source: Sheng and Tu (2000), Janakiramanan and Lamba (1998) and http://www.pse.org.ph a Time corresponds to previous day.

During our period of study, the stock markets have experienced episodes of stock market crash and financial crisis. To avoid the disturbance of the stock markets "crash" in October 1987 and "financial crisis" in July 1997 on stock market integration tests, the study groups data into two periods, pre- and post-crisis periods. The pre-crisis period covers the period from January 1, 1988 to December 31, 1996. In this period, the data set excludes the period stock market "crash" 1987 and "East Asian financial crisis" 1997 to avoid potentially anomalous observations in the analysis. The post-crisis period runs from January 1, 1998 to December 31, 2006, covering only the period after the 1997 financial crisis. The groupings of data into periods provide answers as to whether the integration among these markets are increased due to the crisis and whether there are any changes in the causal relations between the stock markets during the sample period. Finally, to obtain more information from the tests of market integration among the ASEAN emerging markets, the study follows the suggestion of Allen and MacDonald (1995) to group these markets into 93 portfolio combinations. 7 7

All 93 possible portfolio combinations are tested. They are models: 1. May, Ind, Sing, Thai, Phil, US, JP; 2. May, Ind, Sing, Thai, Phil, US; 3. May, Ind, Sing, Thai, Phil, JP; 4. May, Ind, Sing, Thai, US, JP; 5. May, Ind, Sing, Thai, US; 6. May, Ind, Sing, Thai, JP; 7. May, Ind, Sing, Phil, US, JP; 8. May, Ind, Sing, Phil, US; 9. May, Ind, Sing, Phil, JP; 10. May, Ind, Thai, Phil, US, JP; 11. May, Ind, Thai, Phil, US; 12. May, Ind, Thai, Phil, JP; 13. May, Sing, Thai, Phil, US, JP; 14. May, Sing, Thai, Phil, US; 15. May, Sing, Thai, Phil, JP; 16. Ind, Sing, Thai, Phil, US, JP; 17. Ind, Sing, Thai, Phil, US; 18. Ind, Sing, Thai, Phil, JP; 19. May, Ind, Phil, US, JP; 20. May, Ind, Phil, US; 21. May, Ind, Phil, JP; 22. May, Ind, Thai, US, JP; 23. May, Ind, Thai, US; 24. May, Ind, Thai, JP; 25. May, Ind, Sing, US, JP; 26. May, Ind, Sing, US; 27. May, Ind, Sing, JP; 28. Ind, Sing, Phil, US, JP; 29. Ind, Sing, Phil, US; 30. Ind, Sing, Phil, JP; 31. May, Sing, Phil, US, JP; 32. May, Sing, Phil, US; 33. May, Sing, Phil, JP; 34. Ind, Sing, Thai, US, JP; 35. Ind, Sing, Thai, US; 36. Ind, Sing, Thai, JP; 37. Sing, Thai, Phil, US, JP; 38. Sing, Thai, Phil, US; 39. Sing, Thai, Phil, JP; 40. Sing, Thai, May, US, JP; 41. Sing, Thai, May, US; 42. Sing, Thai, May, JP; 43. Thai, Phil, May, US, JP; 44. Thai, Phil, May, US; 45. Thai, Phil, May, JP; 46. Thai, Phil, Ind, US, JP; 47. Thai, Phil, Ind, US; 48. Thai, Phil, Ind, JP; 49. May, Ind, US, JP; 50. May, Ind, US; 51. May, Ind, JP; 52.May, Sing, US, JP; 53. May, Sing, US; 54. May, Sing, JP; 55. May, Thai, US, JP; 56. May, Thai, US; 57. May, Thai, JP; 58. May, Phil, US, JP; 59. May, Phil, US; 60. May, Phil, JP; 61. Ind, Sing, US, JP; 62. Ind, Sing, US; 63. Ind, Sing, JP; 64. Ind, Thai, US, JP; 65. Ind, Thai, US; 66. Ind, Thai, JP; 67. Ind, Phil, US, JP; 68. Ind, Phil, US; 69. Ind, Phil, JP; 70. Sing, Thai, US, JP; 71. Sing, Thai, US; 72. Sing, Thai, JP; 73. Sing, Phil, US, JP; 74. Sing, Phil, US; 75. Sing, Phil, JP; 76. Thai, Phil, US, JP; 77. Thai, Phil, US; 78. Thai, Phil, JP; 79. May, US, JP; 80. May, US; 81. May, JP; 82. Ind, US, JP; 83. Ind, US; 84. Ind, JP; 85. Sing, US, JP; 86. Sing, US; 87. Sing, JP; 88. Thai, US, JP; 89. Thai, US; 90. Thai, JP; 91. Phil, US, JP; 92. Phil, US; and 93. Phil, JP.

8

4. Empirical Findings Table 2 provides summary statistics of the stock returns (i.e., stock prices in first difference) for the ASEAN, US and Japan. It is interesting to note that during the pre-crisis period, all stock markets recorded a positive average daily returns with the exception of the Japanese market in the pre-1997 crisis period. In addition to the stock market of Japan, the Philippines and the US markets also recorded negative returns in the post-crisis period.8 During the eighteen-year period, the Indonesian stock market earned the highest average daily returns of 0.034%, followed by the US (0.022%), Japan (0.013%), Malaysia (0.010%), Thailand (0.005%), the Philippines (0.003%) and Singapore (0.002%). Additionally, the finding that the Indonesian market had the highest returns in the region conforms to the theory of finance, which says that the riskier (more volatile) the market, the higher would be the returns. This evidence is supported by the standard deviation, where the Indonesian stock market recorded the highest, i.e., 0.016. However, the standard deviations of other ASEAN markets range between 0.011 (Singapore) to 0.015 (the Philippines, Thailand, and Malaysia), while the standard deviations from the US and Japan were below 0.010. The findings from the preliminary analysis for the Indonesian stock market are in line with the studies of Palac-McMiken (1997) and Barus (1997). Table 2: Summary Statistics of the Stock Returns Period:

Pre-Crisis Period

Variables Mean Maximum Minimum Std. Dev. Skewness Kurtosis

Thai 0,0003 0,0866 -0,1739 0,0134 -0,8242 19,9584

Ind 0,0006 0,4031 -0,2253 0,0143 8,6103 262,5795

May 0,0005 0,0971 -0,1223 0,0100 -0,7217 21,2547

Phil 0,0004 0,0927 -0,0938 0,0127 -0,0412 9,9735

Mean 0,0001 0,0002 0,0002 -0,0004 Maximum 0,1023 0,1149 0,0585 0,1618 Post-Crisis Minimum -0,0735 -0,0728 -0,0634 -0,0619 Period Std. Dev. 0,0154 0,0139 0,0115 0,0126 Skewness 0,4152 0,6264 -0,1761 2,7364 Kurtosis 8,7356 10,2067 8,9799 42,7267 st st Note: Pre-crisis period starts from 1 January 1988 to 31 December 1996; from 1st January 1998 to 31st December 2006.

Sing 0,0002 0,0564 -0,1037 0,0080 -0,9426 21,0417

US 0,0003 0,0366 -0,0701 0,00652 -0,697 13,779

JP -0,0001 0,0912 -0,0736 0,0094 0,465 14,741

0,0001 -0,0002 -0,0001 0,0466 0,0489 0,0613 -0,0835 -0,0601 -0,0657 0,0111 0,0106 0,0113 -0,3417 0,0226 -0,1437 8,4770 6,060 6,778 while post-crisis period spans

To highlight the short run relations between the movements of the stock markets, the standard correlation coefficients are reported in Table 3. It was used to measure the extent of the association between the stock markets. During pre- and post-crisis periods, all 20 correlations pairs are found to be significantly correlated, at least at the 10% significant levels. Among ASEAN, Malaysia was recorded to have the highest correlation in stock return with Singapore, while Thailand is shown to have the lowest correlated returns with Indonesia during the pre-crisis period. During post-crisis period, Singapore is again found to have the most correlated market returns with Malaysia, whereas Malaysia and the Philippines are found to have the lowest correlated returns in the region. Furthermore, the correlation coefficients of the Indonesian stock returns are relatively lower than in the rest of the ASEAN markets. This indicates that geographically and economically close markets such as Singapore and Malaysia exhibit high correlations of stock returns (0.70), while the correlation coefficients between other stock markets are ranging from 0.05 to 0.50. As for short run relations among the stock markets of ASEAN with the US and Japan, Singapore 8

During a year of the 1997 financial crisis (January to December 1997), except the US stock market, all other stock markets witnessed a negative average daily returns.

9

was recorded to have the highest correlation with the US and Japan in both pre- and postcrisis periods. In addition, comparing to the pre-crisis period, we find a marked increase in short run interactions among pairs of market returns during the post-crisis period. The increase in the market correlation is recorded for 14 pairs from 21 possible pairs of equity returns. This is indicated by the bold figures in the top-diagonal of Table 3. The significance increase in the correlation coefficients in the ASEAN markets indicate that there are short term co-movement among the markets, which suggests that the benefits of any short term diversification, or speculative activities, are limited within the region. Table 3: Correlation of the Stock Returns during Pre- and Post-Crisis Periods Thai

Ind

May

Phil

Sing

US

JP

0.276 0.053 Thai 0.226 0.239 0.494 0.214 Ind 0.048 -0.008 0.192 0.220 0.276 0.138 May 0.409 0.106 0.169 0.498 0.013 0.281 Phil 0.199 0.114 0.214 0.247 0.041 0.202 Sing 0.404 0.084 0.690 0.242 0.130 0.367 US 0.055 0,017 0.100 0.040 0.116 0.108 JP 0.195 0.020 0.279 0.102 0.333 0.119 Note: The bottom diagonal provides correlation coefficients for pre-crisis period, while the top diagonal provides correlation coefficients for post-crisis period. The figures in bold in the topdiagonal indicate the cases of increased correlations.

4.1. Tests of the Unit Roots Hypothesis In order to obtain credible and robust results for any conventional regression analysis, the data to be analyzed should be stationary (Gujarati, 1995). Hence, to test for stationarity, the ADF and PP tests are performed based on model with constant and trend. Table 4 reports the ADF and PP tests statistics that examine the presence of unit roots (non-stationary) for all stock indices. Table 4: Unit Root Tests Variable

Pre-Crisis Period Level First-Difference

Post-Crisis Period Level First-Difference

ADF PP ADF PP ADF PP ADF PP -2,873 -2,837 -53.63*** -53.67*** -1,987 -1,997 -11.68*** -34.04*** May -2,740 -2,543 -9.98*** -50.53*** -2,175 -2,003 -8.38*** -33.97*** Ind -1,579 -1,706 -15.58*** -54.32*** -1,531 -1,756 -12.22*** -34.61*** Thai -2,742 -2,547 -12.50*** -54.56*** -2,495 -2,617 -14.33*** -34.94*** Sing -2,436 -2,282 -20.80*** -49.22*** -2,309 -2,224 -19.17*** -33.88*** Phil -2,171 -2,391 -21.55*** -56.77*** -2,559 -2,325 -35.20*** -35.68*** US -2,307 -2,320 -14.82*** -53.60*** -2,862 -2,581 -14.35*** -34.84*** JP Note: *** denotes significance at the 1% level. The lag lengths included in the models are based on the Akaike Information Criteria (AIC). The above tests of ADF (Augmented Dickey-Fuller) and PP (PhillipsPerron) are based on model with constant and trend.

The study finds that all stock indices contain a unit root, implying that the null- hypothesis of the presence of a unit root at level cannot be rejected even at the 1% significance level. Since the indices are found to be non-stationary at levels, the first differences for whole models are taken. The same tests are applied to the first differences of the indices and the results show that all the indices become stationary after differencing once. This result indicates that all index levels are integrated of order one, I(1) and, therefore, we can proceed to the cointegration analysis with these indices because they are all integrated in the same order as required for cointegration. For our present analysis, this, therefore, serves as a prerequisite for our empirical models.

10

Having identified that all stock indices are stationary at first difference, we now proceed to test for cointegration, aiming at investigating whether there exist long run relationships among the stock markets. In order to get more insights into the long run interdependencies of the ASEAN markets from the US and Japan, as explained earlier, 93 portfolio combinations of stock markets are grouped to test for cointegration. In addition, these portfolio combination groupings allow us to test cointegration both in bivariate and multivariate analyses. 4.2. Cointegration Analysis Out of 93 portfolio combinations, Table 5 reports the cointegration results only for three portfolio combinations among the stock markets of ASEAN, the US and Japan. 9 Of the 93 portfolio combinations during pre-crisis, only three cointegrating portfolio models are found. They are Model 12 (May, Ind, Thai, Phil, JP), Model 19 (May, Ind, Phil, US, JP), and Model 21 (May, Ind, Phil, JP). This indicates that during this sample period, the stock markets of ASEAN, the US and Japan shared a long run equilibrium. However, unlike the stock markets of Malaysia, Indonesia, the Philippines and Japan which are found to be cointegrated in all models, these three ASEAN markets is only found to be cointegrated with the US market in one model, Model 19. Again, this indicates that only the stock markets of Indonesia, Malaysia, the Philippines and Japan documented a strong significant cointegration, implying their stronger linked to Japan rather than to the US. Table 5: Cointegration Tests on Each Portfolio Combinations Portfolios Combination

Null Hypothesis r≤0

May, Ind, Sing, Thai, Phil, US, JP

r≤1 r≤2 r≤3 r≤4 r≤5 r≤6

Pre-Crisis Period TS 134,61 90,53 61,92 37,14 22,23 9,13 2,17

MES 44,09 28,61 24,78 14,91 13,10 6,96 2,17

Post-Crisis Period TS 195.88** 113,47 71,86 51,89 34,89 18,70 6,53

[8]

May, Ind, Thai, Phil, JP

r≤0 r≤1 r≤2 r≤3 r≤4

90.16* 47,87 29.07* 11,73 2,91

[2] 42.29* 18,80 17,34 8,82 2,91

83,77 50,39 28,61 16,43 6,64

[4]

May, Ind, Phil, US, JP

r≤0 r≤1 r≤2 r≤3 r≤4

87.94* 47,65 23,35 11,46 4,21

MES 82.40** 41,62 19,97 17,00 16,19 12,17 6,53 33,38 21,78 12,18 9,79 6,64 [2]

40.28* 24,30 11,89 7,25 4,21

98.24* 57,77 36,00 18,53 6,64

40.47* 21,77 17,47 11,89 6,64

[4] [4] Note: * and ** denote significance at the 5% and 1% levels, respectively. r denotes the number of cointegrating vectors. The numbers in the [.] show the optimal lag length based on the Akaike Information Criteria (AIC). TS and MES refer to Trace Statistic and Max-Eigen Statistic tests, respectively. 9

Detailed results are not reproduced here to conserve space. They are available from the authors upon request.

11

Furthermore, throughout post-crisis period, 54 cointegrating portfolio models 10 are found among the ASEAN, US and Japanese stock markets, representing 58.06% out of the total portfolio combinations. Unlike in the pre-crisis period, during post-crisis the stock market of the US generally played a more dominant role in the region as compared to Japan. However, Thailand and Indonesia 11 are found to have a strong cointegration either with Japan or the US (see Model 64: Ind, Thai, US, JP; Model 65: Ind, Thai, US; and Model 66: Ind, Thai, JP). It is interesting to note that the Philippines is the only market in the region that is concurrently cointegrated with the largest markets in the world, the US and Japan, as shown in Model 91 (Phil, US, JP). Taking into account both pre- and post-crisis periods into the consideration, the US is still found to be a more dominant market than that of Japan in the ASEAN region. Malaysia and Singapore is found to have more cointegrating portfolio combinations compared to other ASEAN stock markets, implying their leading role in the region. Malaysia and Singapore have a long run equilibrium with the US. 12 At the same time, Malaysia and Thailand also shared a long run equilibrium with Japan. It is interesting to note here that the stock markets of: (i) Malaysia 13 and Singapore are more linked with the US; (ii) Thailand is more linked with Japan; and (iii) Indonesia and the Philippines are cointegrated with both the US and Japan. One general conclusion that can be drawn from this finding is that the ASEAN stock markets are moving towards a greater integration either among themselves or with the US and Japan. This finding is in line with many previous findings that documented the world capital markets have been increasingly integrated and that co-movements among them have been rising (Goldstein and Michael, 1993; Blackman et al., 1994; Ghosh et al., 1999; and Jang and Sul, 2002). The ASEAN stock markets have become more cointegrated after the 1997 financial crisis, findings similar to those of Sheng and Tu (2000). In the aftermath of the crisis, Sheng and Tu (2000) found that the cointegrating relationship has improved in the Asian countries as a whole. One possible reason is because of financial deregulation adopted by these countries to recover from the crisis. As discovered by Chowdhury (1994), the Asian markets have become more integrated in recent years due to their financial deregulation. The financial liberalization did increase integration both locally and internationally, but did not drive up local market volatility (Bekaert and Harvey, 1997). The cointegration test results show that the number of significant cointegrating vectors increases after the 1997 Asian financial crisis, a result that is consistent with the contagion effect. Our findings further imply that though there is a limited room to gain benefits from international investment diversification in the ASEAN stock markets, these diversification benefits would diminish. However, the absence of cointegration would simply rule out the existence of a long run equilibrium tending relationship, but this did not invalidate any 10

They are models: 1, 2, 3, 4, 5, 6, 7, 8, 10, 11, 13, 14, 15, 16, 17, 18, 19, 20, 22, 23, 24, 25, 26, 28, 29, 31, 32, 33, 34, 35, 36, 37, 40, 41, 43, 44, 46, 47, 48, 49, 50, 55, 56, 58, 61, 62, 64, 65, 66, 67, 70, 73, 76, and 91 (see Footnote #7). However, detailed results are not reproduced here to conserve space. They are available from the authors upon request. 11 Out of nine cointegrating portfolio combinations that contain the ASEAN and Japanese markets, the Thailand stock market is found to have more cointegrating portfolios (8 models), followed by Indonesia with 7 cointegrating portfolio models. 12 This may be due to the fact that Malaysia and Singapore have higher levels of financial liberalization such as on capital control, exchange rate, and interest rate liberalization. This finding is line with the findings of Daly (2003), who finds a high integration among the Malaysian and Singaporean stock markets. 13 This finding is, to some extent, similar to the finding by Ghosh et al. (1999). They found that Malaysia shared a long run equilibrium relationship with the US stock market, while Indonesia and the Philippines were linked to the Japanese market.

12

short run relationships that may arise due to profit-seeking opportunities (Dwyer and Wallace, 1992; and Yang and Siregar, 2001). The presence of cointegration among the ASEAN stock markets and the markets of ASEAN vis-à-vis the US and Japan rejects the non-causality among them and therefore implies that at least one of the stock markets reacts to deviations from the long run relationships. 4.3. Bivariate Causalities Table 6 summarizes the results of the bivariate Granger causality tests based on the standard F-test framework. The study finds that during the pre-crisis period, Singapore Granger-caused all other ASEAN markets, while Indonesia is Granger-caused by other ASEAN markets with the exception of Thailand, implying the leading and lagging roles of Singapore 14 and Indonesia in the region. The Thai market is found relatively to have no causalities with the rest of the ASEAN markets and Japan. In short, in this period we find that there are no causalities running between the Philippines and Thailand, Indonesia and Thailand, and Malaysia and Thailand. Table 6: Summary of Bivariate Granger Causality Tests Pre-Crisis Period Post-Crisis Period Ind ======= Thai Ind Thai May ======= Thai May Thai Phil ======= Thai Phil Thai Sing ======> Thai May ======> Ind May Ind Phil ======> Ind Sing Ind Sing ======> Ind Sing ======> Phil Sing Phil Phil May Sing May JP

Suggest Documents