Non-Stochastic Argumentation in Predicting Economic Indices

Anotace: This paper studies the use of statistical prognostics in predictions of short-term year-to-year evolution of GDP and other aggregate indices of the national accounts. It shows the utilisation of a non-stochastic prediction range to be used as a prediction tool that, to a certain extent, overcomes the validity of the ceteris paribus principle, on which most of the currently used stochastic approaches are based. The non-stochastic range is a resultant outcome of a wide assortment of time-series models; at the same time, a point forecast for short-term evolution is derived from the said assortment of models. We illustrate our methodology on a year-to-year evolution of GDP indices in France as a time series with a sufficiently large number of observations.