|The degree to which applicant personality test faking constitutes a real world threat is a topic of considerable debate among industrial and organizational psychologists. Researchers have investigated the faking problem using a variety of methodologies, but have found inconclusive results. One method for studying faking involves the use of impression management scales, which are designed to detect individuals’ use of intentional response distortion. However, most scales designed to detect applicant faking are too lengthy, too general, or otherwise impractical for use in applied settings.
The current applied research involved the development, implementation, and validation of an eight-item impression management scale for use with the Fitability 5a, a Big Five personality test used for screening job applicants. Applicants’ (n = 21,017) scores on the new scale were found to have satisfactory reliability and correlated as one might expect with the five personality scales. Applicants considered to be “fakers” produced meaningful score differences on the agreeableness, conscientiousness, neuroticism, and openness scales, but not the extraversion scale.
Additional tests for measurement equivalence were performed using the item response theory-based differential functioning of items and tests framework developed by Raju, van der Linden, & Fleer (1995). Most personality items (35 of 55) demonstrated differential item functioning (DIF). Only items on the extraversion scale did not exhibit significant DIF. Significant differential test functioning (DTF) was found for each of the scales that contained DIF items. Correction for DTF by eliminating items with significant DIF was impossible, as DIF was uniform across all items in that the high impression management group demonstrated a higher probability of responding positively to the items (or negatively, for neuroticism) than the low impression management group. These findings suggest that applicant faking is a real world threat to the Fitability 5a, because impression management strongly affected the construct validity of personality measure.