Conditional density and value-at-risk prediction of Asian currency exchange rates

We first demonstrate the simultaneous need for both more general GARCH structures and non-normal innovation distributions for modelling the returns on certain return series such as the highly volatile exchange rates on East Asian currencies against the US dollar. This is accomplished not only via in-sample goodness-of-fit criteria, but also in terms of the precision of Value-at-Risk calculations made on out-of-sample density predictions. Second, a forecasting strategy using weighted maximum likelihood estimation is proposed. We show that it gives rise to considerably improved forecast performance over longer horizons.