Measuring individual-level racism/discrimination: The state of the art

Background: Documenting the impact of individual-level racism on health becomes increasingly important as societies become increasingly multiethnic. Self-reported discrimination is the generally-accepted best approach for measuring individual-level racism, but few measures of discrimination have been validated across multiple racial/ethnic and language groups. Further, it is not known whether discrimination is experienced similarly across groups. To accelerate progress in eliminating racial/ethnic health disparities, a valid measure is needed. The aim of this multi-year, multi-method project is to develop an effective measure of racial/ethnic discrimination and to evaluate its impact on health in a large, randomly-sampled, multiethnic population. Our study goal is to create a self-reported racial/ethnic discrimination measure that is appropriate for use with racially- and ethnically-diverse groups.

Methods: We designed a split-sample field test within the 2007 California Health Interview Survey to evaluate the two leading approaches to measuring discrimination among 6 major racial/ethnic groups in the United States. The first approach directly asks respondents if they encountered discrimination due to their race/ethnicity (1-stage approach). The second approach initially asks respondents about their experiences of discrimination, and then if they were due to their race/ethnicity or something else (2-stage approach). Our analysis of these data used descriptive statistics (e.g., Chi-Square tests) to assess differences in reporting experiences of discrimination by approach and by racial/ethnic group. We used Item Response Theory to test for Differential Item Functioning (DIF) between racial/ethnic groups.

Results: The field test had a sample of 7,401 respondents (≥18 years). We identified variation in reporting of discrimination experiences by approach and by racial/ethnic group. The 1-stage approach yielded a broader range of rates of discrimination (31-76%); the 2-stage approach yielded higher rates (68-79%). DIF analyses revealed variability in the nature and structure of discrimination. For example, discrimination based on speaking with an accent can dominate the experience of discrimination against groups for whom English is a second language for a large percentage of the subgroup (e.g., Latinos, Asians). In short, we found that the nature and context of discrimination varies across racial/ethnic groups.

Discussion: Our study is the first to directly test the leading approaches to measuring self-reported discrimination, and the first to psychometrically evaluate them in a multi-ethnic sample. Our interpretation of DIF suggests that differences in the discrimination experiences by different ethnicities is not a problem with items that needs to be corrected, but reflects the variation of types of discrimination experienced by different race/ethnic groups.