The issue of measurement invariance revisited

The purposes of this study were twofold: (a) based on the analysis of covariance structures, to test for the equivalence of a well-known measuring instrument across two culturally diverse groups and (b) provided with findings of nonequivalent items, to identify possible determinants of their noninvariance; post hoc investigations included tests for evidence of item bias using an ANOVA-based approach and examination of graphical displays of item response and distributional patterns. Example data comprised item responses to the four nonacademic subscales of the Self Description Questionnaire I (SDQ-I) for Australian (N = 497) and Nigerian (N = 439) adolescents. Despite similarly specified and well-fitting factor structures for both cultural groups, findings revealed evidence of both measurement and structural noninvariance. Results underscore previous caveats regarding interpretation of instrument equivalence; they add also to the growing body of skepticism that queries whether measuring instruments can ever be totally equivalent when used in cross-cultural comparisons.