خطر منابع منطقه ای جهانی در بازارهای سهام: شواهدی از مدل های عامل همراه با چولگی شرطی متغیر با زمان

We examine the influence of global and regional factors on the conditional distribution of stock returns from six Asian markets, using factor models in which unexpected returns comprise global, regional and local shocks. The models allow for conditional heteroskedasticity and time-varying conditional skewness, and are used to measure mean, variance, and skewness spillovers. We find that incorporating time-varying conditional skewness improves the fit of our spillover models, and can alter measurements of variance spillovers. However, time-varying conditional skewness is mostly a local phenomenon; with exceptions, there is little spillover in skewness from global and regional factors.

A thorough understanding of the sources of risk in equity markets is useful for important financial market activities such as risk management, asset allocation, and the development and implementation of regulatory frameworks. We contribute to this understanding by presenting new measurements of the relative importance of global, regional and local components of risk in equity markets. Our measurements are new in two ways: first, we re-estimate volatility spillover using a factor model that, unlike previous models used for this purpose, allows for time-varying conditional skewness. Second, we present additional evidence that distinguishes between downside and upside risks; specifically, we present measurements of spillover in skewness. The evidence we present is from six Asian equity markets, namely Hong Kong, Korea, Malaysia, Singapore, Taiwan and Thailand, using weekly data from the 1990s. Research into interlinkages between stock markets has focused on co-movements in the mean and volatility of returns across stock markets, and has uncovered evidence of spillovers. Eun and Shim (1989), using a VAR model, find interdependence among the daily returns of leading stock markets of the world, with the US stock market being the most influential market. Kasa (1992) finds a common trend driving weekly and monthly returns from the US, Japanese, UK, German and Canadian markets. Hamao et al. (1990) study the interdependence of returns volatility across the US, UK and Japanese stock markets and find that volatility spills over mainly from the US market to the Japanese market, but not the other way around. Lin et al. (1994) find bi-directional dependency between the US and Japanese markets; daytime returns in one market are correlated with overnight returns in the next market to open. Koutmos and Booth (1995) study the US, UK and Japanese markets but differentiate between good and bad news and find, as did Booth et al. (1997) in a study of Scandinavian markets, that volatility spillovers are greater when news is bad, i.e., when the price movement in the latest market to trade prior to opening is a decline. Evidence of co-movements in the mean and volatility of equity returns suggests that factor models, such as those developed in Bekaert and Harvey (1997) and Ng (2000), are useful ways of modeling the behavior of stock returns. Specifying unexpected return to depend on a world factor as well as an idiosyncratic shock, Bekaert and Harvey (1997) find evidence that emerging market volatility is affected by a world factor, and that the influence of the world factor varies considerably over time. Extending this approach to include both a world factor and a regional factor, Ng (2000) finds evidence of spillovers in volatility from the US and Japanese markets to the same six stock markets that we study, with the US market exerting a stronger influence, although the external shocks appear to explain only a small fraction of volatility in these markets. Both Bekaert and Harvey (1997) and Ng (2000) find that liberalization of equity markets changes the proportion of variance caused by external factors. Past studies of mean and/or volatility spillovers have assumed the conditional distribution of stock returns to be symmetric about its conditional mean. Recent work, however, suggests that dynamics in the conditional third moment is an empirically relevant feature of stock returns. Using a model that allows for autoregressive third moments, Harvey and Siddique (1999) present evidence of skewness in the conditional distributions of daily stock index returns in the US, German, Japanese, Chilean, Mexican, Taiwanese and Thai markets, and that this asymmetry in the shape of the distribution depends on the degree of skewness in previous periods. Harvey and Siddique (2000) and Chen et al. (2001) are detailed studies into the determinants and economic significance of skewness in stock returns; stocks that are experiencing relatively high turnover and/or unusually high returns over previous periods tend to be more negatively skewed. Stock capitalization also appears to be important in explaining the degree of skewness in stock returns. Perez-Quiros and Timmermann (2001) relate time-varying skewness to business cycle variation. The skewness in stock returns is economically significant; Chen et al. (2001) demonstrate this by showing that the asymmetry they find in stock returns changes option prices substantially. Harvey and Siddique (2000) incorporate time-varying conditional skewness into an asset pricing model and find that doing so helps to explain pricing errors in portfolio returns using other asset pricing models. Our calculations, reported in Section 2, suggest that ignoring conditional distributional asymmetries can lead to substantial mis-measurement of the probability of large negative returns. The presence of time-varying conditional skewness in equity returns raises a few questions concerning the measurement of the influence of global, regional and local factors on individual stock markets. For instance, will incorporating time-varying skewness into an analysis of spillovers provide substantially different measurements of the relative importance of world and regional factors on the volatility of domestic equity returns? Furthermore, can we improve our understanding of spillovers by also measuring spillovers in skewness? This would give us some insight into downside-risk and upside-“risk”, where downside-risk is measured by the probability of large unexpected negative returns relative to the probability of similarly sized unexpected positive returns. In this paper, we investigate spillover effects within the context of a factor model with time-varying conditional skewness. First, we assume that the spillover effects are constant over time. Next, in the light of previously reported evidence that liberalization and other changes in the environment in which stock markets operate influence the extent of spillovers, we consider a model where spillover effects vary with important developments in the six markets. We begin with some preliminary data analysis in Section 2, where we document evidence of time-varying asymmetry in the markets that we study. The evidence we present here justifies our use of a time-varying skewness framework for studying spillover effects. The evidence also highlights the importance of studying the extent of spillovers in skewness. The models that we employ for studying spillovers are described in detail in Section 3. These models are similar to those employed by Bekaert and Harvey (1997) and Ng (2000) in that unexpected returns comprise world, regional and local shocks, with the difference that these shocks are now characterized not just by time-varying conditional volatility, but also by time-varying conditional skewness. Empirical results are presented and discussed in Section 4, and Section 5 concludes.

We present new measurements of the relative importance of global, regional and local components of risk in equity markets, an issue with implications for important financial market activities, using factor models that allows for time-varying conditional skewness. The evidence is from six Asian markets, namely Hong Kong, Korea, Malaysia, Singapore, Taiwan and Thailand, using weekly data from the 1990s, and using world and regional indexes as proxies for world and regional factors. We explore spillovers in terms of mean, volatility and skewness. We estimate a constant spillover model, and a model that permits the degree of spillover to change in the post-financial crisis period to control for possible structural change as a result of regulatory and other changes that took place during this period. We compare our results with several alternative specifications. In particular, we compare our results to models without time-varying conditional skewness, and to models that use alternative regional indexes. We find that incorporating time-varying conditional skewness makes statistically significant improvements to the fit of our spillover models, so that questions such as “are measurements of variance spillovers affected?” and “are there skewness spillovers?” become relevant. Our results suggest that incorporating time-varying conditional skewness can alter measurements of variance spillovers. However, time-varying conditional skewness seems to be mostly a local phenomenon; with some important exceptions, there is little spillover in skewness from global and regional factors. One interesting avenue for future research is to explore spillover effects with time-varying conditional skewness at the daily (or higher) frequencies. Also, more research into the economic reasons behind asymmetry in the conditional distribution of stock returns is needed. Despite our lack of knowledge of the causes of time-varying conditional skewness, the results in this paper show that studies of spillovers and linkages between equity markets will benefit from incorporating predictability in conditional skewness.