Race, Class, Gender, and Social Space: Using an Intersectional Approach to Study Immigration Attitudes

Abstract

This study uses an intersectional approach to predict attitudes toward immigrants by examining the intersections of race, class, gender, and social space. With data from the 2004 General Social Survey and the 2000 Census, generalized hierarchal linear modeling generates significant two-way and three-way interactions in predicting attitudes toward immigrants taking jobs, improving the economy, committing crime, and migrating to the United States. Important differences in attitudes between groups and within groups only emerge when particular intersections are considered in the analysis. One implication is that pro-immigrant organizations may gain greater support by devising political strategies from an intersectional perspective.

ACKNOWLEDGMENT

The author would like to thank Thomas Rotolo, Lisa Catanzarite, the editors, and the anonymous reviewers for their valuable suggestions.

NOTES

Notes

¹ Because of data limitations, I was unable to include other racial and ethnic groups.

² I originally left the variables as ordinal scales. Yet after running multinomial models and separate binary logistic regression models for the response categories, the effects of the independent variables were not similar across the five categories of the dependent variables. This means that the pattern of the data for the independent variables on the dependent variables violated the proportional odds assumption of ordered logistic regression. Therefore, any outcomes from ordered logistic regression would be questionable (O'Connell 2006). To relax the proportional odds assumption, I also ran the ordinal variables using generalized ordered logit. Specifically, I used the gologit2 command in Stata 9 with the autofit, lrforce, and cluster options (Williams 2006). This produced more consistency among the response categories regarding the effects of the independent variables. Yet since the significance, strength and direction of the coefficients of these models paralleled the coefficients from statistical techniques for binary dependent variables, I decided not to use generalized ordered logit. Instead, I used generalized hierarchal linear modeling—which is a multilevel modeling technique for binary dependent variables—with a Bernoulli distribution and Laplace estimation (Raudenbush and Bryk 2002). The coefficients generated from this type of modeling are easier to interpret.

³ Eighty-eight respondents (8.76% of the sample) did not report their family income. A covariate analysis revealed no statistical difference between these respondents and other sample respondents in any of the models. I subsequently imputed the missing data based on age, race, sex, educational attainment, region, and the population, the percentage of foreign-born residents, and the percentage of the unemployed in the PSUs. Nonetheless, models that deleted the missing data from the sample rather than imputing them generated nearly identical coefficients in terms of significance, strength and direction to the ones presented below. Tables are available upon request.

⁴ In preliminary analyses, I also constructed a three-way interaction term between race, class, and gender. However, this interaction term was nonsignificant in all of the models, suggesting that nothing changes in the two-way interactions with any change in the third variable (Jaccard and Turrisi 2003). Consequently, I do not include this interaction term in the models presented below.

⁵ Another possible explanation as to why educated blacks are pro-immigrant in these immigrant-filled areas is that these educated blacks are children or grandchildren of immigrants. Further analyses revealed that 9 percent (91 respondents) of the sample claim African ancestry. Two of these respondents self-identify as white, and 89 self-identify as black. Yet only four of them, all who self-identify as black, have at least one parent or one grandparent who is foreign born, and only 10 black respondents total have a parent or grandparent who is foreign born, nine of which have less than a college degree and five of which live in areas with less than 20 percent foreign-born residents. Therefore, overall, these outcomes indicate that educated blacks that live in regions with a comparatively high percentage of immigrants are unlikely to be second or third generation immigrants themselves. This suggests that it is a multiplicative effect between race, education, and space, which may incorporate sympathetic feelings for black immigrants that results in the formation of a pro-immigrant attitude rather than a pro-immigrant attitude being a result of having immigrant parents or grandparents.