La mesure de la personnalité à l’aide du Big Five Inventory français (BFI-Fr) est-elle invariante selon le genre et l’activité des répondants ?

OBJECTIVES The BFI-Fr is an effective tool to assess the five dimensions of the five-factor model of personality. The psychometric qualities are excellent in several validation studies. However, because the inconsistency of certain items may be due to a lack of invariance between groups, we examined this invariance according to the gender and activity of the respondents. METHODS The sample consists of 1904 participants (1232 women and 672 men; 1159 students and 745 professionals). Recruitment was done by \"snowball\" from different studies conducted over four years. Respondents completed the BFI-Fr. The 45 items were corrected for the acquiescence bias. The Item Response Theory methodology was used to test the invariance using the differential item functioning (DIF) test. The T scores were subject to ANCOVA and correlations to compare the differences between subgroups. RESULTS Whatever the subsample, we found the same five-factor structure (similarities were especially high for C). The cumulative explained variances are comparable for all groups. The functioning of eight items varied between men and women; and six between activity groups, however, all but one DIF values were low. No item in domain A varied in its measure. ANCOVA results showed that women have higher A and N scores than men and that professionals have higher C and lower N scores than students. There is no interaction effect between gender and activity. When we removed the non-invariant items, the effect of gender became significant in favor of women on dimension C even if the difference remained minimal. The difference by activity on N became non-significant. There was a very strong link between activity and age and between gender and activity: in the sample, women were more often students and vice versa. CONCLUSIONS Overall and once again, the structure and internal consistency of the Big Five as measured by the BFI-Fr was replicated in new samples. There was a differential functioning of 12 items out of 45 (about a quarter). These differences were small since only one was of moderate magnitude. The deletion of these items did not strongly change the mean comparisons between groups. Therefore, our results do not call into question the usefulness of the BFI-Fr but invite to control the extent of this differential functioning. In all cases (with or without the non-invariant items) the analyses showed significant gender differences for A and N, differences which are typically found in many studies. Students and professionals differ in their levels of C and E. Certain inconsistencies are probably due to the translation from American to French, since the BFI-Fr is based on the lexical approach (in this context, item content was somewhat dependent of languages peculiarities). Thus, the BFI-Fr allows a robust evaluation of the five dimensions but requires a prior control of the invariance of item functioning in order to ensure that the intergroup differences in E-A-C-N-O mean scores do not reflect a psychometric artifact but genuine personality differences.