A new strategy based on econometric models to improve the forecasts accuracy in Romania

Abstract

An empirical strategy of improving the forecasts accuracy is proposed in this article starting from econometric models with the random component determined using resampling techniques. New predictions were built using those provided by two institutions from Romania specialized in forecasting: the Institute for Economic Forecasting (IEF) and the National Commission of Prognosis (NCP). For the inflation and unemployment rate four different regression models were proposed to build other forecasts. Two of the regression models improved both institutions forecasts for the inflation and unemployment rate on the forecasting horizon 2010-2012, while the other two provided better predictions than the NCP ones. So, the models based on resampled errors is an original way of constructing new forecasts and at the same time a good strategy of improving the predictions accuracy for some macroeconomic forecasts in Romania.