Files
Abstract
Personality measures are attractive to practitioners in personnel selection context due to their effectiveness in predicting job performance and lack of adverse impact. However, the susceptibility of personality measures to faking has been a concern. A potential problem with the faking research is that measurement invariance of personality inventories across faking and nonfaking groups is rarely examined before personality mean scores are compared. This renders any interpretation of group differences suspect. The current study examined the effect of applicant faking on the measurement properties of the Global Personality Inventory (GPI), a Big Five personality measure. It was found that all the Big Five dimensions had higher latent means for job applicants than incumbents. Emotional Stability had unequal intercepts across applicant and incumbent groups. Conscientiousness and Extraversion had unequal intercepts and uniquenesses across groups. Sources of non-invariance at the scale-level were explored using three different approaches. The three approaches differed in terms of the scales identified for scalar but not uniqueness non-invariance. Further latent variable analyses suggested that the social desirability method bias existed for both the applicant and incumbent samples, but its presence did not affect the Big Five factor loadings as well as the relationships among the Big Five dimensions. Implications of the results were discussed.