Forecasting Inflation in Chile With an Accurate Benchmark

In this article we analyze the accuracy and stability of a number of inflation forecasts for Chile. We place special attention on forecasts coming from Extended Seasonal ARIMA (ESARIMA) models. Our analysis considers the sample period from January 2000 to November 2008. We compare ESARIMA forecasts to survey-based forecasts and to those coming from traditional time series benchmarks available in the literature. We consider one to six months ahead forecasts for comparisons with the traditional time series benchmarks and one, three and twelve month ahead forecasts for comparisons against the Survey of Professional Forecasters (SPF). Our results show that ESARIMA based forecasts display lower out-of-sample Mean Square Prediction Error than forecasts coming from traditional benchmarks, almost without exception. We obtain mixed results when comparing ESARIMA forecasts to survey-based forecasts in terms of forecast accuracy: the SPF provides more accurate forecasts for one month ahead forecasts, provides slightly less precise forecasts at the three month horizon and is less precise than half of our ESARIMA models at the 12 month horizon. Finally, we also note that our ESARIMA forecasts are more stable than traditional time series methods and also than survey based forecasts when prediction is made three and twelve months ahead.