We evaluate inflation forecasts from the Survey of Professional Forecasters (SPF) of the Central Bank of Chile. Forecast errors for the period 2000-2008 show an excess of autocorrelation and a statistically significant bias at the end of the sample. We take advantage of the autocorrelation structure of the forecast errors to build new and more accurate inflation forecasts. We evaluate these new forecasts in an out-of-sample exercise. The new forecasts display important reductions in bias and Mean Square Prediction Error. Moreover, these reductions are, in general, statistically significant.
