Recruitment Early Warning System. Phase II. Twelve Month Forecasts of the United States Unemployment Rate.
Abstract
This report describes the construction and performance of a forecasting model developed to forecast U.S. civilian unemployment. The model is based on a composite leading indicator, tailored to the unemployment rate, which consists of 15 component series on the U.S. economy. The components are weighted by an empirically derived weighting procedure, using bivariate regression analysis. As the objective is a 12 month forecast of unemployment, the weight of each of the component series is derived from its predictive power at a lead of 12 months. Within-sample testing indicates that the USULI12 leads turning points (both peaks and troughs) in U.S. unemployment by two to eleven months. Regression testing confirms the predictive power of the USULI12. To forecast the U.S. unemployment rate, the USULI12 is used as the input variable in a transfer function model with an ARIMA noise structure. In a series of out-of-sample tests designed to examine forecasting accuracy near turning points, the mean absolute error is typically less than a 0.5 unemployment percentage point. Keywords: Unemployment forecasts; Leading economic indicators; Transfer functions; ARIMA models.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1985
- Accession Number
- ADA166565
Entities
People
- R. A. Holmes
- Ross Neil