IS THE PRECISION OF THE EX POST FORECAST ERRORS HELPFUL TO CHOOSE A GOOD FORECASTING MODEL?

In this paper we show possible ways of obtaining some information about applicability of forecasting model and its probable accuracy in computing ex ante forecasts using accuracy of ex post forecast errors, which are made at the same time when the extrapolative model is looking for. We will use the Theil coefficient of inequality to show how well the model fits the data during the time period for which the ex post forecasts are computed. The Theil decomposition of mean square of ex post forecast errors will be used to show sources of possible bias in their means or variances. Six months ex post forecasts will be computed for the unemployment rate in Slovakia using the SARIMA model in repeating mode to analyze its stability after adding six new observations and computing six months ex post forecasts.