| Modeling the dependence across financial markets and estimating the market risk of a portfolio are important topics in financial economics. This dissertation includes three essays, which discuss these two topics from different aspects.;Cointegration and causality analysis is frequently used in investigating the dependence or comovement of international stock markets. However, the existing literature on cointegration analysis seldom takes into consideration the possible presence of structural breaks and time-varying volatility effects. In causality analysis, the causality in mean tests are frequently carried out but not the causality in variance tests. In the first essay, I examine the cointegration and causal relations among five major stock markets. The recently developed techniques for investigating unit roots, cointegration, structural break, time-varying volatility, and causality in variance are applied.;In the existing literature on risk management, many papers compared the performance in estimating Value-at-Risk of various approaches, such as GARCH models, historical simulation and extreme value theory models. However, stochastic volatility models were seldom included in the list of models to be compared. The second essay compares tithe-varying volatility models, which includes GARCH and stochastic volatility models, with the Extreme Value Theory (EVT) model by applying them to a Value-at-Risk (VaR) framework. For the GARCH model, the parameter estimates are obtained by maximum likelihood. For the stochastic volatility model, I resort to the recently developed highly efficient MCMC method, and for the extreme value theory model, the conditional EVT modeling method is employed. By the backtest, the performance of these three models is evaluated in the VaR framework.;Various types of approaches are used in existing literature in investigating dependence across stock markets. However, most of them only examine the degree of dependence, and the multivariate normal distribution is frequently made. The degree of dependence alone is not adequate for a better portfolio choice, and the multivariate normal distribution assumption is unrealistic for financial series. The third essay investigates the structure of dependence across international stock markets by mixed copula approaches, which nest both symmetric and asymmetric dependence patterns. The findings in this essay have important implications to both risk management and asset allocation. |
