Feasibility of Estimating Personnel Turnover from Survey Data - A Longitudinal Study

This report investigates the validity of career intent and job attitude statements for predicting reenlistments among 54,803 airmen in 101 enlisted specialties. Statements were made while completing Air Force occupational surveys during the period 1966 to 1971. Frequency and percentage distributions characterizing item responses and career decisions of the sample are given. Relationships between career intent/job attitude responses and actual reenlistment decisions were studied using multiple linear regression techniques. Job attitudes were found to be substantially related to career decisions but did not provide unique contributions to predictions of career decisions when used in conjunction with career intent statements. A model-seeking exercise identified a second-degree polynomial model with career intent and time-in-service interaction vectors as the most appropriate prediction model. Separate prediction equations were obtained for a number of selected subsamples representing two, three, and five-digit Air Force specialty code and aptitude requirement groupings. Homogeneous grouping exercises were performed to illustrate a procedure for reducing the number of subsample prediction equations to a minimum subset. Practical examples of forecasting reenlistment rates are illustrated. The regression model selected for the demonstration used weights derived from the total sample to predict the percentage of actual reenlistments for thirty specialties. Overall, approximately 17 percent of the sample remained in service.