How response bias affects the factorial structure of personality self-reports.

Abstract

Various studies have shown that acquiescence can distort the factor structure of personality questionnaires based on the five-factor model.  In the present study, we analysed how acquiescence and social desirability affect the factor structure of a measure based on this personality model and a measure of aggression. We analysed the factor structures of both tests before and after removing both biases in a sample of 532 adolescents aged between 11 and 18 (M= 14.75, SD= 2.1). The factor structure of both tests presented a worse fit to the expected model when response bias was not controlled, and the congruence indexes for the personality and aggression measures showed a moderate (from C= .948 to C= .872) or great (from C= .931 to C= .475) decrease, respectively. Furthermore, acquiescence was largely responsible for these effects, and social desirability effects were only shown on the aggression measure.   Response bias, and especially acquiescence, should be controlled during the development of personality measures to avoid distorting them, especially with samples of people with a high level of acquiescence (for example, those with little education, the young or the elderly). Furthermore, the use of response bias loadings as a criterion for choosing the items minimizes those distortions.