Abstract

In this paper we evaluate the Central Bank of Chile annual GDP growth forecasts over the period 1991-2009 using a real-time database. We compare the Central Bank of Chile forecasts with those of the Survey of Professional Forecasters (SPF), Consensus Forecasts, and simple time-series models. We compare all forecasts to first and quasi-final GDP growth vintages. We evaluate a number of different forecast properties, including forecast accuracy and efficiency. We report mixed results in terms of root mean squared prediction errors. Depending on the sample period, the forecast horizon and the vintage used in the analysis, forecasts from the Central Bank of Chile may outperform or be outperformed by the benchmarks. Despite these mixed results, differences in root mean squared prediction errors are generally moderate and have no statistical significance. Nevertheless, our efficiency analysis, in addition to the fact that in some periods the forecasts produced by the Central Bank of Chile have been outperformed by alternative forecasts, opens the question about the room for improvement in the accuracy of the Central Bank of Chile forecasts. While the room for improvement may actually exist, our results suggest that this room seems to be small for point forecasts and larger for interval forecasts.