FEASIBILITY OF DISTINGUISHING SUPERVISORS' POLICIES IN EVALUATION OF SUBORDINATES BY USING RATINGS OF SIMULATED JOB INCUMBENTS,

Supervisors in 4 specialties differed in the importance they attached to selected qualifications in subordinates, when these subordinates were known to them only by patterns of trait scores. The technical problems considered in the statistical analyses involve some recent developments in the application of regression techniques. Groups of 50 supervisors in AFSCs 43171, 43190, 64670, and 64770 rated a set of 250 profiles of trait scores representing subordinates in AFSCs 43151, 43171, 64650, and 64750. Each supervisor assigned 250 worthto-the-Air Force ratings, using a 9-point scale and fixed allotments of each value. These ratings served as criteria, with the trait scores as predictors, providing a regression equation for each rater his policy. Using the Judgment Analysis JAN technique, the policies were grouped, and the loss in prediction computed when moving from 50 different equations policies to a single equation for a given specialty. Average predictions were high, and the grouping loss small. Comparing JAN to factor analysis techniques, it was concluded that the JAN method was efficient, but that a factor strategy method aided in qualitative interpretation of rater policies. Demographic and personal data collected on the raters showed no meaningful relationship to their policies.