Using Large Data Sets to Forecast Sectoral Employment

We implement several Bayesian and classical models to forecast employment for eight sectors of the US economy. In addition to standard vector-autoregressive and Bayesian vector autoregressive models, we also include the information content of 143 additional monthly series in some models. Several approaches exist for incorporating information from a large number of series. We consider two approaches – extracting common factors (principle components) in a factor-augmented vector autoregressive or vector error-correction, Bayesian factor-augmented vector autoregressive or vector error-correction models, or Bayesian shrinkage in a large-scale Bayesian vector autoregressive models. Using the period of January 1972 to December 1999 as the in-sample period and January 2000 to March 2009 as the out-of-sample horizon, we compare the forecast performance of the alternative models. Finally, we forecast out-of sample from April 2009 through March 2018 using the best forecasting model for each employment series. We find that factor augmented models, especially error-correction versions, generally prove the best in out-of-sample forecast performance, implying that in addition to macroeconomic variables, incorporating long-run relationships along with short-run dynamics play an important role in forecasting employment