Point forecasts based on the limits of the forecast intervals to improve the SPF predictions

Many researchers are interested in making predictions for macroeconomic variables, but few of them studied the accuracy of their forecasts. The problem is essential, especially in crisis periods, because from many forecasts made for the same indicator only one or few are the most accurate. In this research, some alternative forecasts for the annual rate of change for the HICP for EU were developed. Their accuracy was evaluated and compared with the accuracy of SPF predictions. All the proposed predictions for January 2010-May 2012 (those based on a random walk developed for 1997-2009, combined forecasts, the median and the mean of forecasts, predictions based on different econometric models that take into account the previous SPF forecasts) were not more accurate than the naïve forecasts or SPF ones. A considerably improvement of the accuracy was gotten for predictions based on mean error of SPF expectations for 1997-2009 and the previous registered value. This empirical strategy of building more accurate forecasts was better than the classical theoretical approaches from literature, but it is still less accurate than the naïve forecasts that could be made for UE inflation rate. The point forecasts based on the lower limit of intervals built using root mean squared indicator generated an improvement in accuracy, outperforming the SPF predictions and also the naïve forecasts.