Abstract
This year the PX index, the key price index of the Prague Stock Exchange, celebrated twenty years of its existence. Therefore, one could be indeed interested in analysing historical daily closing quotes of this stock index from an econometric point of view.
In detail, the aim of this contribution is to introduce a particular class of discrete-time state space models and demonstrate that this class is appropriate for such a univariate financial time series. Particularly, it involves regularly applied econometric modelling instruments; it combines a local level model and a linear ARMA process together with conditionally heteroscedastic innovations.
Moreover, the suggested modelling framework is examined in different settings of parameters. The final model is selected with respect to standard information and prediction criteria; it is further investigated and statistically verified by inspecting prediction residuals.
Its empirical performance is compared with other commonly applied methods, i.e. with linear ARMA or benchmark GARCH models.