COMBINING FORECASTS BASED ON ECONOMETRIC MODELS FOR SHORT RUN MACROECONOMIC PREDICTIONS WITH HIGH DEGREE OF ACCURACY

Abstract

For a certain macroeconomic variable more predictions based on different methods could be made. The essential problem is to establish the most accurate forecast, using different indicators. The econometric modeling is one of the most used forecasting method. A strategy to improve the accuracy of the predictions based on econometric model is to make combined forecasts. In this paper, for inflation rate, unemployment rate and interest rate were made predictions based on ARMA procedures, VAR(2) models and models with lagged variables. For all the analyzed variables in Romania, ARMA models generate more accurate forecasts than VAR(2) models or models with lags.  For inflation and interest rate optimal combination and equal-weights-scheme determined the most accurate predictions, while for unemployment rate ARMA models remain the best forecasting method in terms of accuracy.