Modeling of Currency Covolatilities

Anotace: The paper deals with dynamic modeling of currency portfolios. In contrast to univariate models of exchange rates and their returns one applies multivariate time series models of the type GARCH that are capable of capturing not only conditional heteroscedasticities (i.e. volatilities) but also conditional correlations for common movements of exchange rates (so called covolatilities). One makes use of recursive estimation algorithms suggested by authors for such models which enable to control, evaluate and manage currency investment portfolios in real time. The main task of the paper is to assess whether the recursive estimation procedures suggested by the authors are applicable for real currency portfolios. It is realized by performing an extensive numerical study for bivariate portfolios of the EU currencies and US dollar concentrating on the role of the Czech crown.