Pronosticando la inflación mensual en Colombia un paso hacia delante: Una aproximación "de abajo hacia arriba"

Abstract

The hierarchical structure of the Colombian Consumer Price Index (CPI) makes possible to calculate in ation as a linear combination of its subcomponents. We use SARIMA models to forecast each component of CPI and construct an forecast of in ation using a lineal combination of the forecasts of these components, i.e. a "bottom to top" approach. In this paper, we asses the out-of-sample performance of the one-step-ahead forecast of 12 "bottom to top" methodologies. These methods are compared with an aggregate forecast using a SARIMA model. Our results show that a "bottom to top" method to forecast in ation outperforms an aggregate approach for the case of monthly in ation in Colombia.