In the Search of a Univariate Benchmark Model to Forecast the Unemployment Rate

In this work the precision and stability of the forecasts of Chile's unemployment rates are analyzed. Said models were obtained by a family of SARIMA (Seasonal Autoregressive Integrated Moving Average) models, between February 1986 and February 2010. The SARIMA projections are compared with the ones originating from univariate models, including the benchmark predictive ones. Simultaneously and ARFIMA (Autoregressive Fractionally Integrated Moving Average) model was adjusted, owing to the signs of persistence that the unemployment indicator shows in its behavior; nevertheless, starting from the estimation methods developed by Reisen (1994), Geweke et al. (1983), and Whittle (1962) integration parameters greater than 0.5 were obtained, which empirically upholds the proposal of addressing the unemployment rate as a non stationary series. The evaluation of the predictive capacity of the models is centered in the forecasts out of the sample of 1, 6, and 12 months from then on. The results indicate that the RECM out of the sample of the SARIMA projections is less than the one from the considered univariate methods