Analysis of financial time series: estimation and forecasting by non- and semi-parametric methods

This thesis proposes a semiparametic copula methodology for modelling, estimating and forecasting tail risks of stock-bond portfolios of Australia and G7 countries. These tail risks have vastly increased as a consequence of the recent global financial crisis. Additionally, this thesis introduces a nonparametric methodology for estimating and forecasting S&P 500 returns volatilities. An attractive feature of this method is that it does not make restrictive assumptions and data speaks for itself. In comparison to existing methods, our proposed method captures the underlying movements in the market volatilities and produces superior out-of-sample forecasts of global risk, particularly during the crises.