Recruitment Early Warning System. Phase II. Volume 1. Research and Development of the Recruitment EWS (Early Warning System).
Abstract
The components most critical to a Recruitment Early Warning System are enlistment supply models which forecast enlistments accurately. In this research, Auto-Repressive Moving Average regression models have been developed for forecasting enlistments of 1-3A Non-Prior Service (NPS) male High School Diploma Graduates (HSDG's) and High School Seniors (HSSR's) and 3B NPS male HSDG's and HSSR'S, estimated with national, monthly data for January 1979 - May 1985. To assess the validity of the models, we conducted out-of-sample forecasting tests for observations in FY 1984-85. The tests were conducted with known values of exogenous variables, except for unemployment which was forecasted. The results indicate that the models adequately forecast enlistments over 12-month intervals. For each Service, forecasting errors are typically only three percent or less for the 1-3A and 1-3 cohorts, over the entire 12-month period; individual monthly forecasts are subject to larger errors, but they cancel over the 12-month period. The forecasting tests covered 12-month periods in which there were no 'regime changes', i.e., changes or introduction of programs or policies which affected the market structure. In periods where regime changes occur, we find that forecasting accuracy deteriorates markedly. Since the Services do change programs or policies from time to time, we devoted considerable exploratory research to the remediation of forecasting error caused by regime change. The results indicate that in three to six months, it is possible to take regime changes into account and restore forecasting accuracy to its high pre-change level. Keywords: Personnel Retention, ARMA Model, Market Assessment.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 30, 1985
- Accession Number
- ADA166567
Entities
People
- Lawrence Goldberg
- Peter Greenston
- Sherry Andrews
- Sigurd Hermansen