Transition equity markets of Central Europe : volatility, predictability, integration
The objective of this thesis is to add evidence from the transition equity markets of Central Europe to the econometric modelling of financial time series by addressing the issues of volatility, predictability and international asset pricing in these markets. In Chapter Two we start from an overview of the transition stock markets by presenting their historical background, basic regulations, statistics, and stock market indices. Chapter Three focuses on the modelling of univariate and multivariate volatility in transition equity markets. Our sample has all the previously documented characteristics of the unconditional distribution of stock returns normally used to justify the use of the GARCH class of the models of conditional volatility. Strong GARCH effects are apparent in all series examined. The estimates of asymmetric models of conditional volatility show rather weak evidence of asymmetries in the markets. The results of the multivariate specifications of volatility have implication for understanding the pattern of information flow between the markets. The constant correlation specification indicates significant conditional correlation between three pairs of countries: Hungary and Poland, Hungary and Czech Republic, and Poland and Czech Republic. The BEKK model of multivariate volatility shows evidence of return volatility spillovers from Hungary to Poland, but no volatility spillover effects are found in the opposite direction. Chapter Four examines the linear and nonlinear predictability of transition equity returns with simple technical trading rules. The application of the moving average trading rules to the data reveals that technical analysis helps to predict stock price changes. Firstly buy signals consistently generate higher returns than sell signals; secondly the returns following buy signals are less volatile than returns following sell signals. The application of the bootstrap methodology to check whether three popular null models of stock returns with linear conditional mean specification replicate the trading rule profits indicates that returns obtained from trading rules signals are not likely to be generated by these models. Comparison of the out-of-sample forecast performance of linear and nonlinear (feedforward networks) conditional mean estimators with past trading signals in the conditional mean equation indicates substantial forecast improvements of the feedforward network regression. Chapter Five addresses the issue of integration of the transition equity markets into the global capital market by testing pricing restrictions of the international CAPM simultaneously for four national equity markets: two developed markets (U.S. and Germany) and two new transition markets (Hungary and Poland). Methodologically, we extend the BEKK multivariate GARCH specification to accommodate GARCH-M effects, and propose an alternative specification of the conditional CAPM, which allows return volatility transmissions between the markets in the system. The results reveal that the world price of covariance risk is positive and equal across the markets. This is consistent with the international CAPM and supports the hypothesis of integration of the transition markets into the global market. However, our further results indicate individual significance of the Hungarian idiosyncratic risk, pointing to some level of segmentation of the Hungarian market. Moreover, the introduction of world-wide information variables into the system reveals that some variation in the excess national returns is still predictable after accounting for the measure of market-wide risk.