Estimating Personality Constructs from Archival Data

Abstract

Abstract : As part of a leadership research program at the U.S. Military Academy involving cadets in the Class of 1998, this report examined the viability of using archival data on prior cadets to estimate a variety of personality constructs among current cadets. Two sets of archival personality data on prior cadet classes were obtained. The first involved a short form administration of the ABLE inventory to cadets in the Class of 1994. The second involved the administration of the NEO Personality Inventory to the Class of 1996. Scores on the 12 scales contained in these inventories were used as archival criteria. Archival predictors were then sought from other and questionnaire items administered to cadets at the same point in time as the original inventories. For each scale, a different 20-item pool of predictors was developed from the archival items having the strongest zero order correlations with that scale. A series of multiple regression analyses was then used to predict scores on each scale. An average R2 of .39 per scale was obtained after cross validation. Both the original scales and their analogs tended to manifest similar relationships with two external criteria examined, leadership performance and attrition.