Description

The Central Bank of Chile builds inflation forecasts for several time horizons and using various methodologies. In this paper, we analyze one of these series of short-term inflation forecasts, which we call Auxiliary Inflation Forecasts (AIF), comparing them to forecasts made by private analysts and to forecasts built from simple time-series models. We also evaluate the AIF using encompassing tests and bias and weak efficiency tests. Our aim is to answer two linked questions: first, which is the best forecast series under a specific loss function? and second, are the differences of accuracy between any two series of forecasts totally explained by the differences in the information sets from which they have been built? Our results indicate that the AIF behave extremely well at one- and two-month horizons, but they are less adequate at longer horizons.