A STRATEGY TO IMPROVE THE GDP INDEX FORCASTS IN ROMANIA USING MOVING AVERAGE MODELS OF HISTORICAL ERRORS OF THE DOBRESCU MACROMODEL

In this paper article, two strategies based on the econometric approach are proposed in order to improve the forecast accuracy of GDP index in Romania. First, the index is predicted starting from an econometric model that reflects the relationship between the GDP index and the GDP deflator. Then, the errors of these forecasts are computed. On the other hand, the errors result directly from an econometric model that shows the relationship between the GDP index forecast errors and the GDP deflator prediction errors. The data series are historical errors of forecasts based on the Dobrescu macromodel. The forecasts errors of the GDP index based on the Dobrescu macromodel historical errors for 2009-2011 are lower than the errors taken directly from the proposed econometric model. However, the Dobrescu macromodel provided a better accuracy for the GDP index. If the historical errors are predicted using updated MA(1 )models, the one-step-ahead forecasts are the most accurate, this being a suitable strategy to improve the prediction accuracy.