Because the vignettes were randomly assigned, confounding would only be an issue if randomization failed. Still, we included several sociodemographic variables as well as implicit racism (1) to increase precision of the estimate of vignette effects and (2) to assess the generality of the vignettes’ effects across demographic groups. The sociodemographic variables (weighted data) were age (mean = 47.6 years; range = 18 to 90 years); gender (50.1 percent female); educational attainment (7.9 percent 12th grade or less with no diploma; 31.3 percent high school diploma or equivalent; 20.4 percent some college; 10.0 percent associate’s degree; 20.4 percent bachelor’s degree; 7.5 percent master’s degree; and 2.5 percent professional or doctoral degree); and race/ethnicity (79.8 percent white non-Hispanic; 11.8 percent black non-Hispanic; 2.1 percent other non-Hispanic; 2.5 percent two or more races, non-Hispanic; and 3.8 percent Hispanic). Because the race of participants identifying as Hispanic cannot be determined from the ANES questions on race and ethnicity, and the “other” and multiracial groups are small and likely heterogeneous, analyses that focus on racial subgroups were restricted to non-Hispanic whites and non-Hispanic blacks. We used analyses available in SPSS 18.0 that produce standard errors that take the complex survey design into account.
