Assessing the Impact of Faking on Binary Personality Measures: An IRT-Based Multiple-Group Factor Analytic Procedure

Abstract

This article proposes a model-based multiple-group procedure for assessing the impact of faking on personality measures and the scores derived from these measures. The assessment is at the item level and the base model, which is intended for binary items, can be parameterized both as an Item Response Theory (IRT) model and as an Item Factor-Analytic (FA) model. The specific solution proposed is bidimensional, and the propensity to fake is specifically modeled as a second factor. The approach we propose allows (a) previous separate results obtained either from IRT-based or from FA-based studies to be related and (b) some hypotheses on measurement invariance and structural changes to be assessed in more detail. The procedure was used in a study based on 2 personality scales that were administered in 1 group under standard instructions and in another group under faking-good instructions. For both scales, results suggested that strong measurement invariance was attained. Also, for both scales substantial mean changes were obtained in the propensity-to-fake factor but not in the content factors. The implications of the results and future research directions are discussed.