Description

In this article we explore the existence, robustness and size of the contribution that several activity measures make when predicting inflation in Chile. For that purpose, we use backward-looking Phillips curves and make use of a real-time database to get an evaluation of predictive ability consistent with the uncertainty faced by policy-makers when making decisions in real time. Our main results confirm those shown by the recent literature in the USA: the predictive contribution of the GDP measures considered here is episodic, unstable and of moderate size. These results are robust to the estimation being made with either revised or real-time data. This holds true in a context in which final and first vintage GDP measures are shown to differ significantly and to generate quite different inflation forecasts