ادغام در مقابل تقسیم بندی در بازار سهام چین: تجزیه و تحلیل ریسک های بتا متغیر با زمان

This paper assesses whether China's stock market is integrated with the global market during 2000–2010 within the framework of an augmented CAPM. We firstly use Kalman smoothing technique to obtain time-varying global and national systematic risks for the once-restricted A- and unrestricted B-share indices in China's stock exchanges. Then we investigate how these risks are priced while controlling for possible structural changes using the Markov regime-switching technique. We find evidence of partial integration in terms of positively priced global and national systematic risks in most cases. However, the unrestricted Shanghai B-share market is found to be generally segmented from the global market. The degree of integration is therefore not simply a matter of the degree of restriction, confirming that documenting barriers to investment (or the lack of them) is insufficient to prove segmentation (or integration). Given that the domestic systematic risk is still priced, there is scope for international portfolio diversification into China.

It is widely accepted that the issue of whether stock markets are integrated or segmented has implications for financial decisions. Bekaert et al. (2003) summarise that the extent of market integration affects the functioning of equity markets and the diversification ability of local and foreign investors. If a national stock market is segmented from global markets, international investors can gain from risk diversification through portfolios that include stocks traded in this market. On the other hand, from a domestic perspective, such segmentation can require domestic firms to explicitly consider global versus local sources of funds and global versus local hedging opportunities. However, the institutional changes that accompany integration with the global market can lead to positive spill-over into the real economy, by improving corporate governance and reducing the cost of capital, which can in turn increase investment and economic growth. Since the early 1990s, the Chinese authorities have implemented various reforms in their stock exchanges in order to reap the benefits of integration. According to statistics compiled by International Financial Service London (2010), external portfolio investment into China in 2008 was $10bn, in contrast to an overall outflow of $80bn of external portfolio investment from emerging economies in that year. Given that China has become one of the largest recipients of external portfolio investment among the emerging markets, it is of great interest for international investors to assess whether the Chinese stock market has become integrated with the world market in order to gauge potential gains from portfolio diversification into China. From the academic perspective, China provides an interesting case study. The existence of dual classes of stocks, namely the once-restricted A-shares (shares became available to foreign investors in 2002) and unrestricted B-shares (shares available to foreign investors throughout the period under study), will facilitate the evaluation of the effects of capital controls, providing important policy implications for other emerging markets. Many studies have attempted to investigate China's stock market integration through cointegration analysis of stock indices or correlation/dependence analysis of stock returns. Girardin and Liu (2007) suggest a long run international financial integration on the basis of cointegration between the weekly averaged Shanghai A-share and S&P 500 indices during 1992 and 1996 and between the weekly averaged Shanghai A-share and Hang Seng indices since 1997. Johansson (2009) concludes that China experienced an increasing level of integration with several major financial markets in the 2000s on the basis of the market dependences estimated by Copulas. However, Lin et al. (2009), using a dynamic conditional correlation GARCH model, find that the Chinese A-share market has never been correlated with overseas markets during 1992 and 2006 while its B-share market exhibits negligible correlations with the western markets and slight correlations with the Asian markets. Luo et al. (2011) investigate dependences between financial sectors in China and some overseas economies within the framework of a copulas model. They find significant dependences between China and Hong Kong and Singapore, weak dependences between China and Australia, Taiwan and Japan and no dependence between China and Korea or the US in the post-2002 period. Similarly, Li (2012), using a multivariate GARCH, finds that the interdependences between China and the regional markets increase but the conditional correlation between China and the US market remains weak during 1992 and 2010. The above studies are informative indeed with respect to stock market linkages and interdependence between China and the regional and global economies. However, Girardin and Liu (2007) admit that ‘existence of a long run relationship between stock markets may not be a proof of integration but simply of cointegration of fundamentals’ (p. 366). Bekaert and Harvey (1995) suggest that correlation between returns of the local and world markets cannot be used as a measure of stock market integration, because ‘a country could be perfectly integrated into world markets but have a low or negative correlation because its industry mix is much different from the average world mix’ (p. 436). More relevant to our study is the strain of investigations in the context of asset pricing models. In an asset pricing sense, integration is defined as a situation where investors earn the same risk-adjusted expected return on similar financial instruments in local and global markets. Hence the investigation of market integration for a particular economy could, in principle, involve testing whether systematic risk relative to the global market is the only significant factor in a capital asset pricing model. However, the pure international CAPM has typically asserted market integration by assumption, making it an unsuitable framework for testing. Stehle (1977) conducts the first empirical test of segmentation versus integration in an international CAPM that is augmented by including a term to represent national systematic risk. Using the traditional two-pass approach, Stehle (1977) finds that the pricing of US securities is significantly related to a global market portfolio. Within the same framework of this augmented CAPM, Jorion and Schwartz (1986) use a maximum likelihood approach to estimate all the parameters simultaneously and find strong evidence of segmentation in the pricing of the Canadian stocks relative to the North American market during 1963–1982. Mittoo (1992) re-examines the integration between the Canadian and US stock markets during 1977–1986 within the frameworks of the augmented CAPM and a multi-factor pricing model. The evidence is consistent, with both models, that segmentation occurs during 1977–1981 and integration during 1982–1986. Wang and Di Iorio (2007) apply the approach of Jorion and Schwartz (1986) to investigate the segmentation/integration of three China-related stock markets and find that Chinese A-, B- and H-share markets are segmented from the global market during the period from 1995 until 2004. Such studies have commonly assumed that the systematic risks relative to national and global markets are constant over time. However, empirical studies such as Sunder (1980) and Collins et al. (1987) have rejected the assumption of beta stability. The development of multivariate GARCH models in the 1990s enables the estimation of time-varying conditional variances and covariances, offering an alternative to constant national and global systematic risks in the CAPM model. To date, models with time-varying conditional covariance risk have mainly been used to investigate integration between national markets. For example, in the framework of a multivariate GARCH model, De Santis and Imrohoroglu (1997) and Hunter (2006) derive time-varying conditional covariances between some national indices and a global index and assess integration by the testable implication that the prices of covariance risks are positive and equal across the national markets under study. More recent studies such as Hardouvelis et al. (2006) follow Bekaert and Harvey (1995) and include time-varying variance of a national index additionally to represent the domestic factor and exploit the model predictions: if markets are integrated, only the global systematic risk measured by the covariance is priced; whereas under complete segmentation, only the domestic systematic risk measured by the variance will be priced. By applying time-varying weights to the prices of the covariance and variance, they estimate the time-varying probability that national stock markets have changed between segmentation and integration. Assuming that degree of integration depends on the availability of substitute assets, Carrieri et al. (2007) extend the basic idea of Bekaert and Harvey (1995) to partial integration/segmentation cases by calculating an integration index based on the variance ratios of a portfolio of eligible stocks to the country stock market index. However, Ang and Chen (2007) argue that the time-varying conditional covariance and variance derived from the GARCH models are strictly driven by past innovations in returns and do not have independent random components. In this paper, we will use alternative time series techniques to investigate whether China's stock market is integrated with the global market within the CAPM framework. Specifically, we will firstly adopt a Kalman smoothing technique to allow the parameters in the CAPM of Jorion and Schwartz's (1986) style to vary over time. This is a well-established approach to permitting time-varying behaviour of systematic risk in a CAPM; see for example: Ang and Chen (2007) and Mergner and Bulla (2008). Because it is unknown if an international or a domestic CAPM holds in China, we rely on statistics for guidance. We will substitute the estimated time-varying global and national systematic risks into two versions of a two-beta CAPM, namely an international CAPM augmented with a pure national systematic risk and a domestic CAPM augmented with a pure global systematic risk, in order to test for integration versus segmentation. When testing, in the first instance, we will apply OLS and assume a single structure throughout the entire period. The assumption of a single regime will then be relaxed to permit two regimes, “high volatility” and “low volatility”, to account for the possibility that the economic mechanism that generates the stock market returns may have changed from time to time due to China's stock market reforms or global financial crises. The Markov regime-switching technique of Hamilton (1989) will be employed in the tests for integration under the two-state regime. The choice of this two-state regime-switching process is motivated partly by the literature and partly by economic reasoning. Previous studies such as Turner et al. (1989), Schaller and Van Norden (1997), Assoe (1998), Wang and Theobald (2008) and Liu et al. (2012) find strong evidence that both developed and emerging stock markets switch between high-volatility and low-volatility states. Although the two-state regime may seem restrictive, this two-state process is intuitively appealing given its capacity of letting the stock market returns under study depend on the state of market, reflecting policy changes or financial crises. This work is the first to test for China's stock market integration using time-varying systematic risks in a regime-switching setting. Our approach, underpinned by an augmented CAPM, serves well the purpose of testing for integration with the global market for a particular economy. Compared with the conventional tests for stock market integration using cross-sectional time-invariant systematic risks, e.g., Wang and Di Iorio (2007), our application of the Kalman smoothing technique to the derivation of the time-varying systematic risks and our implementation of a Markov regime-switching technique to control for the structural changes in the estimation of the two-beta asset pricing model constitute the methodological innovation of this paper. Our finding that the degree of stock market integration is not simply a matter of the level of capital controls provides evidence to the argument of Jorion and Schwartz (1986) that documenting legal barriers to investment (or the lack of them) is insufficient to prove segmentation (or integration). The remainder of the paper is organised as follows. The theoretical framework is described in Section 2 while the methodology is presented in Section 3. We carry out the preliminary data analysis in Section 4 and implement estimation and tests in Section 5. Section 6 concludes.

This paper investigates China's asset pricing mechanism with an aim to testing for integration between the Chinese and global stock markets. By estimating the augmented CAPM through the Kalman smoothing technique for the A- and B-share indices of the Shanghai and Shenzhen stock exchanges during 2000–2010, we find that most of the indices’ global and national systematic risks change slowly and the national systematic risk is generally much greater than the global one. These observations suggest that investing in the Shanghai A-share and Shenzhen A- and B-share markets may not significantly increase the global systematic risk of an international portfolio, offering a diversification opportunity. The above implication is supported by the tests for integration versus segmentation within the framework of a two-beta asset pricing model. We found evidence of partial integration at most. Specifically, the national systematic risk is positively priced in the cases of Shanghai A-share and Shenzhen A- and B-share indices under a two-state regime as well as under a single regime. Given that the national systematic risk is still priced in these cases, there is scope for international portfolio investment involving the shares listed on China's stock markets, especially on the Shanghai A-share and Shenzhen A- and B-share markets. The partial integration also implies that the Chinese firms face optimal hedging decisions as well as foreign versus home financing and capital budgeting decisions. We also found that the degree of integration is not simply a matter of the level of legal restrictions imposed on the stock market. It depends on other factors such as nature of shares and the composition of foreign investors on the stock exchanges too. The implications for emerging economies are that the indirect barriers as well as the legal barriers to international investment should be removed in order to reap the benefits of stock market integration. The authorities should make their stock exchanges appealing to international as well as regional investors.