Multigroup Comparisons and the Assumption of Equivalent Construct Validity Across Groups: Methodological and Substantive Issues

The purposes of this article are (a) to compare construct validity findings from Campbell-Fiske and LISREL confirmatory factor analyses (CFA) of a multitrait-multimethod matrix for each of two groups, and then to test for construct validity equivalence across groups -- the methodological issue, and (b) to demonstrate how construct validity can differ across groups -- the substantive issue. Multidimensional self-concept (general, academic, English, mathematics) responses to three types of measurement scales (Likert, semantic differential, Guttman) for low- and high4rack grade 11 and 12 students provided the exemplary data base. Methodologically, findings demonstrate the superiority of CFA over the Campbell-Fiske approach by (a) providing more detailed evidence of construct validity within groups, and (b) testing for the equivalency of construct validity across groups. Substantively, the findings illustrate that the assumption of group-invariant construct validity cannot be taken for granted; differences were found in both the measurement and structure of self-concept. Results bear importantly on the validity of findings based on multigroup comparisons.