Inflation Forecasting in Chile

Abstract

This study estimates two models of Chilean inflation with time-varying parameters during the sample period 1990-1999. The first model is based on the Phillips curve and the second represents a small open economy with an inflation-targeting framework. The out-of-sample inflation forecasts of the two models are compared with the out-of-sample inflation forecasts produced by simple time series models à la Box-Jenkins. The most important results are: inflation forecasts produced by models that include the official, pre-announced inflation target are better that those produced by models that do not include it; the Phillips curve model produces better out-of-sample forecasts than the small open economy model; although for the short run the simple Box-Jenkins models produce better out-of-sample forecasts than the Phillips curve model, their forecasting ability deteriorates rapidly in the medium term. In general, including Markov-switching in the two models does not help to explain a significant part of the conditional variance in the forecast error. It is important to stress, however, that as the monetary policy regime enters its steady state in 2001, the relative ranking of the inflation forecast models may change.