Comparing Apples to Oranges? Mixed Support for Ethnic-Racial Equivalence of Ethnic-Racial Identity and Related Measures

• In Event: 3-066 - “What’s Measurement Got to Do with It?” Testing Ethnic-Racial Equivalence of Identity-Related Measures

Integrative Statement

Efforts to draw conclusions regarding similarities and differences in adolescents’ ethnic-racial (E-R) identities and life experiences are constrained by whether available quantitative measures demonstrate psychometric equivalence across groups. Comprehensive tests of invariance for measures of E-R identity and associated constructs are mostly absent from the literature. Drawing from a multi-site study of E-R diverse adolescents in two U.S. regions, we tested measurement equivalence of: (1) the Ethnic Identity Scale-Brief (EIS-B; Umaña-Taylor et al., 2004); (2) the centrality and public regard subscales of the Multidimensional Inventory of Black Identity-Teen (MIBI-t; Scottham et al., 2008); (3) the other-group orientation (OGO) subscale of the Multigroup Ethnic Identity Measure (MEIM; Phinney, 1992); and (4) the Adolescent Discrimination Distress Index (ADDI; Fisher et al., 2000).

Participants were U.S. high school students in the Southwest (N = 1277) and Midwest (N = 471) who completed identical school-based paper-and-pencil surveys. For each site and measure, we fit initial confirmatory factor analysis (CFA) models with a pooled sample of all E-R groups. Next, we fit nested multi-group CFA models with sequentially more restrictive constraints. We tested configural (i.e., factor structure), metric (i.e., factor loadings), and scalar (i.e., intercepts) invariance (Milfont & Fisher, 2010). At each step, we evaluated multiple indices of change in model fit collectively to determine whether constraints resulted in significantly poorer fit (i.e., ≥ .01 change in CFI; rescaled chi-square difference test; Dimitrov, 2010; Satorra & Bentler, 2001). Due to sample size restrictions (Kline, 2010), the Southwest analyses included Native American, Asian American, Latino, and Black adolescents, and the Midwest analyses included Asian American and Black adolescents.

Findings from both regions suggest that comparisons across E-R groups and/or pooling data among all youth of color may not always be possible. For instance, in the Southwest site, the EIS-B and MIBI-t demonstrated equivalence of factor structures and loadings across all four E-R groups and, with slight modifications, both achieved scalar equivalence (Table 1). Yet for the OGO and ADDI, equivalence of factor structures and loadings was only supported across Latino and Black youth. For the Midwest site, the EIS-B, MIBI-t, and OGO all achieved partial scalar invariance across Black and Asian American adolescents (Table 2). However, the ADDI only reached configural invariance (and not metric or scalar), suggesting that combining data for analyses and/or comparing mean-level group differences is not possible across these two groups.

Support for configural invariance across sites and E-R groups suggests that using these four measures is appropriate in E-R homogenous samples (e.g., analyses of Asian American adolescents only). Furthermore, researchers can confidently interpret findings examining EIS-B and MIBI-t subscales in relation to other constructs when studying Native American, Asian American, Latino, and Black youth in the Southwest, and Asian American and Black youth in the Midwest. Yet, comparisons of mean-level differences across groups should be carefully evaluated and may only be possible with certain groups. Findings will be discussed in the context of how they inform practices for studying E-R identity and related constructs.

Authors

• Michael R Sladek, Harvard University
  Presenting Author

• Adriana Umaña-Taylor, Harvard University
  Non-Presenting Author

• Deborah Rivas-Drake, University of Michigan
  Non-Presenting Author

• Elana R. McDermott, Harvard University
  Non-Presenting Author

• Stefanie Martinez-Fuentes, Arizona State University
  Non-Presenting Author