We use several models using classical and Bayesian methods to forecastemployment for eight sectors of the US economy. In addition to using standard vectorautoregressiveand Bayesian vector autoregressive models, we also augment thesemodels to include the information content of 143 additional monthly series in somemodels. Several approaches exist for incorporating information from a large number ofseries.We consider two multivariate approaches—extracting common factors (principalcomponents) and Bayesian shrinkage. After extracting the common factors, we useBayesian factor-augmented vector autoregressive and vector error-correction models,as well as Bayesian shrinkage in a large-scale Bayesian vector autoregressive models.For an in-sample period of January 1972 to December 1989 and an out-of-sampleperiod of January 1990 to March 2010, we compare the forecast performance of thealternative models. More specifically, we perform ex-post and ex-ante out-of-sampleforecasts from January 1990 through March 2009 and from April 2009 through March2010, respectively. We find that factor augmented models, especially error-correctionversions, generally prove the best in out-of-sample forecast performance, implying thatin addition to macroeconomic variables, incorporating long-run relationships alongwith short-run dynamics play an important role in forecasting employment. Forecastcombination models, however, based on the simple average forecasts of the variousmodels used, outperform the best performing individual models for six of the eightsectoral employment series.
展开▼
