A Procedure for Selecting the Best Proxy Variable Used in Predicting the Consumer Prices Index in Romania

The purpose of this research is to propose a procedure for selecting the most suitable proxy variable that explains the best the evolution of another variable and that generates the best predictions. These two criteria suppose the combination of the econometrical approach with the one based on the assessment of forecasts accuracy. In Romania, the BIM unemployment rate proved to be a better proxy than the number of unemployed people for the following reasons: a better lagged model was obtained to explain the evolution of the index of consumer prices (higher R-square and a lower value for Akaike informational criterion) and the predictions based on this model on the horizon January 2012- December 2012 are more accurate (the accuracy measures are lower and the accuracy tests indicates more accurate predictions). The monetary policy interest rate was an exogenous variable for both models. However, naïve forecasts based on random walk are more accurate than the predictions based on the proposed econometric models. So, a double analysis is recommended for the predictions based on econometric model: the selection of the best model and the selection of the most accurate prediction.