The work presented in this dissertation can be grouped around two major themes. The first theme relates to the asset allocation and the second theme relates to risk management. In Chapter Three, we investigate the dynamics of foreign exchange and stock returns based on an extended version of Sentana and Wadhwani (1992) model. This study is mainly driven by the wish to explain two major stylized facts that puzzled the older models. We find evidence to support that only intertemporal variation in the foreign exchange risk premium can be explained by time¡Vvarying covariance priced risk factors. Furthermore, we also find that the first order autocorrelation of both foreign exchange and stock market returns in Taiwan is negatively related to the level of conditional volatility and covariance. This time-varying nature of the serial correlation pattern is consistent with our model where some traders follow feedback strategies. The three nested asset pricing models with four models of conditional second moments are strongly rejected. We conclude that our extended Sentana and Wadhwani model is more adequate in explaining the dynamics of foreign exchange and stock markets. In Chapter Four, we investigate the risk management of futures market and spot market returns. There is widespread evidence that the volatility of stock returns display an asymmetric response to good and bad news. This paper attempted and found the asymmetric behavior co-existence in spot as well as future markets. By using the Asymmetric Dynamic Model (ADC) proposed by Kroner and Ng (1998), we estimated the conditional covariance matrix asymmetric and calculated dynamic optimal hedge ratios. With the help of that asymmetric model, our ¡§out of sample¡¨ dynamic hedging strategy out-performed that of normally dynamic hedging strategies. However, while taking the transaction costs into consideration, the performance was even worse than that of the static strategy.

I find evidence of predictability in out-of-sample data for four risk premia using simple econometric models. Two factor return models are used, an APT model and the Wilshire Atlas. I demonstrate that investors can exploit conditioning information to manage their exposures to risk factors. The results suggest that the investment opportunities set changes in a large and an economically significant way. I show that the growth rate in money supply and trend in stock market valuations are the main drivers respectfully of the risk premia associated with the Book-to-Market and Size factors from the Wilshire model. The predictability results are mixed with respect to Business Cycle Theory. At times investors price business cycle risk while at other times they exhibit herding tendencies.