گرایش های مشترک و همگرایی؟ بازارهای سهام در جنوب شرق آسیا، 1988-1999

This paper uses two different estimation approaches to demonstrate that equity markets in South East Asia have shown signs of converging during the 1990s. According to the Haldane and Hall (Economic Journal101, 436–443, 1991) method of measuring convergence, a subset of Asian markets had converged by mid-1997. This process appears to have been abruptly halted and somewhat reversed by the recent financial crisis. We find that in general, there are two common trends present in the eight Asian equity market indices modeled here, and also two trends when the US market is additionally included in a Johansen VAR.

The existence of low correlation among returns from different national stock markets has been used frequently to justify the international diversification of portfolios (see for example Levy and Sarnat, 1970, Grubel and Fadner, 1971, Lessard, 1973, Solnik, 1974 and Solnik, 1991). More recent studies have concentrated on volatility spillovers among markets and have examined changes in these spillovers over time. There are clear variations in methodology used to analyze volatility spillovers, see Bollerslev, Chou and Kroner, 1992, Koutmos and Booth, 1995, Booth, Martikainen and Tse, 1997 and Kanas, 1998 for a brief selection. A number of authors have demonstrated that these spillovers vary systematically over time. Indeed, Koch and Koch (1991) and Longin and Solnik (1995) concluded that markets have become more internationally interdependent over time.1 Studies have investigated both ‘mature’ equity markets and so-called ‘emerging’ markets, including several South East Asian markets considered in the present paper2. Many have investigated the effect of the October 1987 stock market crash on equities in ‘emerging’ markets: see for example Lee and Kim, 1994, Choudhry, 1996 and Choudhry, 2000. Unfortunately, there appears to be little common ground between these paper’s findings. Many ‘emerging markets’ have been subject to financial deregulation; nevertheless, Ng, Chang and Chou, 1991 and De Santis and Imrohoroglu, 1997 also fail to reach a consensus on the impact of liberalization on volatility. However, Manning (1999) finds evidence of an increase in equity market volatility and in volatility spillovers among Indonesia, Singapore and Malaysia when comparing the 1997/98 ‘crisis’ period with the pre-crisis 1990s. In view of the preceding discussion and the financial turmoil affecting the region in 1997, further work on South East Asian markets seems justified. The above has almost exclusively been concerned with returns and volatilities in different markets. These are essentially short-run phenomena. However, a related strand of the literature considers long-run relationships among different equity market indices and whether there is any tendency for convergence to occur over time. Most of the recent work in this area has taken into account the unit root property commonly found in equity prices and thus cointegration methods have frequently been applied to multi-country equity market data. For example, Kanas (1998) concludes that the lack of bivariate cointegration between the US market and any European bourse implies gains to international portfolio diversification, 3 which takes us back to the arguments of Levy and Sarnat (1970) and others, some two and a half decades previously. Other papers have applied cointegration techniques to ‘emerging markets’, including those in southeast Asia. Huang (1995) demonstrates that the indices from southeast Asian markets possess unit roots but, in common with Pan et al. (1991), rejects the random walk hypothesis for a number of Asian markets. Chan et al. (1992) also find evidence of unit roots in major Asian markets and find no evidence of cointegration between the Asian markets and the US, or among the various Asian markets themselves. One classic paper considering the long-run relationship among indices is Kasa (1992), who applied the Johansen (1988)Maximum Likelihood approach to stock market data from the US, Japan, Canada, Germany and the UK. The paper is notable for demonstrating that indices from five markets can be approximated by a single common trend. Kasa states “I find it remarkable that a single series, in conjunction with a vector of factor loadings, can do such a good job of tracking these five countries’ stock markets” (p. 115). What the paper demonstrates is that in the long run, these markets do not diverge. As Kasa states in his conclusion, “These results imply that to investors with long holding periods the gains from international diversification have probably been overstated in the literature” (p. 122). One possible reconciliation of Kasa’s conclusion with the previous consensus involves the holding period; it is possible that gains from diversification occur in the short term, but not in the long term. It is also possible that several markets cointegrate, even though bivariate cointegration is absent. A highly relevant study using the Johansen approach is by DeFusco et al. (1996). They find no cointegrating vectors exist among equity indices from Korea, Philippines, Taiwan, Malaysia and Thailand sampled weekly over the period 1989–93. Pan et al. (1999) also conclude that their sample of six equity indices4 do not cointegrate, but they do find evidence in favor of common volatility in the long run. However, Masih and Masih (2000) find a single cointegrating vector existing among indices from a number of South East Asian markets5 using a sample of daily data from 1992 to 1997 and a lag length of 7. These results are notably at variance with both Kasa and the present study (see below); it would appear that these conclusions might follow from an inappropriately small lag length in the VAR. A number of recent papers have considered the linkages between macroeconomic variables and equity markets, see for example, Kwon and Shin, 1999 and Bracker, Docking and Koch, 1999. The latter paper computes annual series of Geweke (1982) measures of feedback for different pairs of equity indices using daily returns data, and then considers possible macroeconomic determinants of these feedback measures by using pooled regressions. They study nine markets including Japan, Hong Kong, Singapore, Australia and the US: these particular markets are grouped to form a ‘Pacific Rim’ block. However, in the pooled regressions estimated on all nine markets, they find a significant positive coefficient on a Pacific Rim dummy only when modeling the bilateral measures of same-day correlation; there is no significant Pacific Basin effect when considering measures of (dated) feedback. In the context of the present paper, these results hardly suggest that the Pacific rim markets have converged. In contrast to the above, the Kasa paper demonstrates that the cointegration rank in the Johansen VAR modeling his sample of stock indices is relatively high so the long-run co-movement of international markets can be characterized by relatively few series or common trends. However, as pointed out by Serletis and King (1997), cointegration analysis does not consider whether convergence is occurring or when it occurred; the conclusion is that in one sense, markets have converged if they are described by a single common trend. Haldane and Hall (1991) propose an entirely different approach to convergence relying on time-varying parameters estimated by the Kalman Filter. Parameter restrictions implied by convergence can be considered by inspection of the time-varying parameter estimates. In essence, this technique identifies convergence of markets within Asia with common variation with respect to an external market (such as the US). For example, Serletis and King (1997) consider whether convergence had occurred within European Union stock markets, the external market being the US. In the present paper, we follow Serletis and King (1997) and apply both the Johansen Maximum Likelihood approach and the Haldane and Hall Kalman Filter technique to consider the co-movement of equity markets in South East Asia, also taking the United States to be the external market6. The 10 countries sampled include Japan, Hong Kong and Indonesia beside the US: casual inspection of Fig. 1(a) and (b) identifies differences among the fortunes of these markets over the 11 years from January 1988 to February 1999 whether returns are calculated in domestic currency or in US dollars. The question posed in the title is therefore very real: have these markets converged, and if not, have they shown any tendency for convergence over this 11-year period? Is there any evidence for convergence during any sub period of the data set? What is the impact of the severe shocks accompanying the Asian financial crisis commencing in 1997? What is the role of the US market? Are the conclusions sensitive to the definition of returns in local currency or in US dollars?

This paper has analyzed data from nine Asian equity markets sampled at weekly and monthly frequencies over an 11-year period. Two very different methodologies have been applied and both techniques take into account the unit root properties of the data but address different, albeit related, economic issues. The Johansen maximum likelihood approach determines the number of cointegrating vectors, and therefore the number of common trends, which exist among these equity indices. If a single common trend is found, then arguably, convergence has already occurred. However, we find a minimum of two common trends in these data indicating partial convergence of the indices. In these circumstances, the Johansen approach says little about the dynamics of convergence although it clearly signals the interdependence of these markets. In fact, the Haldane and Hall Kalman Filter technique identifies two periods of convergence of the Asian markets, namely 1988–1990 and 1992–mid-1997; divergence occurring both in 1990–1992 and during the recent Asian crisis. It would also appear that a sub-sample of the Asian markets, which excludes Japan, Korea and Taiwan, had achieved a common relationship with the (external) US market by mid-1997. According to the Haldane and Hall measure of convergence, these markets had converged over the period mid-1992 to mid-1997. 15 However, this process was abruptly reversed by the onset of crisis, even though many Asian financial markets were rocked by the crisis almost contemporaneously. 16