Investigating the linear predictor-criterion assumption of biodata scaling.

Abstract

Study examines biographical data and Academy training performance for 4,559 Federal Aviation Administration (FAA) Air Traffic Controller Specialists collected from 1986 to 1992. Biodata measure was rationally scaled along personality factors of the Big Five and Occupational Information System (O*NET) general work activities. Study examines a linear predictor-criterion assumption made with rational and empirical scales by using evolutionary theories of personality to model a priori non-linear relationships with job and training performance among FAA candidates. Correlation and stepwise hierarchal quadratic multiple regression analyses revealed most scales had a linear relationship with Academy training performance; however, biodata scales of conscientiousness and mental processes did exhibit slight nonlinearity. High and low level of these dimensions predicted higher training and job performance than mid levels. Some sub-scales of conscientiousness showed the same pattern of results. Additionally, empirical keying of the same rational scales improved criterion-related validity over the purely rational scales and suffered less shrinkage upon cross-validation than a strict empirical key of the full biodata inventory. Implications for construct validity, theory, and practice are discussed.