Combinación de brechas del producto colombiano

This paper combines eight estimations of the Colombian output gap for the time period between the first quarter of 1994 and the third quarter of 2012. The combined forecast densities of the gap are obtained by using bivariate VAR models that include the different gaps and inflation, and using three weighting schemes: logarithmic scores, continuous probability range scores (CRPS) and mean square error criteria (MSE), these output gap densities are useful because they provide an insight of the gap's central tendency, while characterizing the uncertainty. The results suggest that the combined densities under these three schemes for one, two, three and four quarters ahead are not miss-specified. Additionally, the combined density based on logarithmic weights has the best performance for all horizons. For two and three quarters ahead there are significant higher scores for the logarithmic weights methodology that those calculated by CRPS and MSE weights.