Naturally Occurring Selection: Using Applicant Pool Data to Estimate Job Relevant Range Restriction

Due to factors that include self-selection and recruitment, job applicant pools can be more homogeneous than the general working population. Using GMA data based on the Wonderlic Personnel Test, we examined applicant pools, finding them on average substantively more homogeneous than Sackett and Ostgaard’s (1994) early investigation. This indicates a major impact on range restriction corrections as using broader population norms instead of job applicant pool estimates typically overestimates variance accounted for by 26%. Consequently, the common practice of using population norms during range restriction helps explain why many of our field’s core or fundamental validity coefficients are implausibly high, at times indicating perfect prediction. Using a synthetic validity driven simulation of the U.S. labor market, this applicant pool homogeneity can be retraced to emergent or naturally occurring selection (i.e., where job candidates are appropriately matched to occupations prior to more formal selection procedures) that accounts for 2% to 3.5% of the performance variance. The simulation, showing a high degree of construct validity, supports several core theories of occupational choice, with educational requirements as the primary mechanism of applicant pool homogeneity for GMA.