Community Influences on White Racial Attitudes: What Matters and Why?

Abstract

Tracing the roots of racial attitudes in historical events and individual biographies has been a long-standing goal of race relations scholars. Recent years have seen a new development in racial attitude research: Local community context has entered the spotlight as a potential influence on racial views. The race composition of the locality has been the most common focus; evidence from earlier decades suggests that white Americans are more likely to hold anti-black attitudes if they live in areas where the African-American population is relatively large. However, an influential 2000 article argued that the socioeconomic composition of the white community is a more powerful influence on white attitudes: In low-socioeconomic status (SES) locales, “stress-inducing” deprivations and hardships in whites' own lives purportedly lead them to disparage blacks. The study reported here reassesses this “scapegoating” claim, using data from the 1998 to 2002 General Social Surveys linked to 2000 census information about communities. Across many dimensions of racial attitudes, there is pronounced influence of both local racial proportions and college completion rates among white residents. However, the economic dimension of SES exerts negligible influence on white racial attitudes, suggesting that local processes other than scapegoating must be at work.

ACKNOWLEDGMENTS

Thanks to Tom W. Smith of the University of Chicago National Opinion Research Center for contributing central information.

NOTES

Notes

^a Peter Mateyka performed his portion of the research while at Penn State. He is now a survey statistician with the U.S. Census Bureau. Any views or opinions expressed in the article are the author's own and do not necessarily reflect the views or opinions of the Census Bureau.

¹ The conflict between these patterns and claims that contact can improve intergroup relations is more apparent than real. Locales with large proportions of African-American residents often have high levels of residential segregation, precluding the kind of intergroup contact found to encourage positive attitude change (Allport 1954).

² Objections to using Blumer's theory in survey research (see Esposito and Murphy 1999) were effectively answered by Ulmer (2001).

³ Taking off from Oliver and Mendelberg (2000), subsequent studies by Oliver and his collaborators have incorporated Hispanic and Asian-American presence as potential environmental predictors and have broadened the set of dependent measures examined. Also, level of analysis has moved to center stage: Oliver and Wong (2003), along with Ha (2010), conclude that sizable outgroup representation in the neighborhood encourages positive attitudes among residents, while a large outgroup population in the metropolitan area seem to have a negative influence on attitudes.

⁴ It may be instructive that Moore and Ovadia's discussion shifts between reference to “attitudes” and reference to “norms.” Schuman et al. (1997:2–5) provide a thoughtful response to critics who worry that race attitude measures may largely reflect survey respondents' deference to assumed normative expectations held by researchers, especially in face-to-face interviews and among highly educated respondents. Debates about what attitude measures actually measure are too complex to be extensively considered here. However, the Schuman et al. argument that norms exert powerful influences on social life implies that if indeed racial attitude measures are partially barometers of racial norms, they nonetheless supply us with very important information.

⁵ Another advantage of studying context effects for larger geographical units is that there is less reason to worry about the direction of causation between race composition and attitudes. The chicken and egg problem haunts contextual as well as individual-level assessments of the impact of proximity and contact on intergroup attitudes. Although data analytic strategies have been employed to offer some assurance that selection of diverse environments is not the primary cause of apparent contact effects (see, e.g., Welch et al. 2001; Branton and Jones 2005; Dixon 2006) most analysts acknowledge that reverse causality remains an issue (Oliver and Mendelberg 2000). The role of race composition in white Americans' neighborhood choices is indisputable (see, e.g., Zubrinski Charles 2001, 2006), but it is likely that economic considerations, sentiment, and accident dominate people's selection of metropolitan area or nonmetropolitan county. For most whites, the economic or sentimental cost of avoiding a metropolitan area because of its minority population would be too great. And levels of residential segregation are high enough in virtually all areas (Massey and Denton 1993; Stoll 2005) so that those averse to diversity can satisfy their taste for racially homogenous neighborhoods within the metropolitan area they choose for other reasons.

⁶ A single exception should be noted: One nonmetro Primary Sampling Unit encompasses two counties.

⁷ Anther useful hierarchical linear modeling reference is Snijders and Bosler (2000).

⁸ The logic of such multivariate strategies is powerfully spelled out in the classic work by Rosenberg (1968). Another classic, Cohen and Cohen (1975) outlines the essentials of multiple regression; such recent sources as Agresti and Finlay's text (2008) update this description.

⁹ Southern region and proportion black are confounded (r = .581), and the racial history of the United States leaves some question about how to disentangle the two. Inclusion of each in a multivariate analysis of white racial attitudes generally weakens the observed impact of the other. The region effects seen for Model 1 should be interpreted remembering that they are partial effects, controlling on black population share (as well as population size and metro status).

¹⁰ The corollary of the point made in Note 6 is that assessing the impact of proportion black on white attitudes after controlling for Southern location, as we did here, predictably yields conservative estimates. If the ambiguity about the appropriate treatment of region and race composition in multivariate analyses had been resolved by reporting race composition coefficients without the control for South, we would see substantially larger proportion black effects on most dependent measures; and for one of the three attitude measures that didn't show a significant effect of proportion black in Table 4, social distance, the effect becomes significant at p < .001 (results available from the authors).

¹¹ Standardized slope coefficients are presented in parentheses for all effects reported in Tables 3 and 4, because they allow comparison of the strength of the various partial effects. They can be interpreted as the change in standardized dependent variable scores expected when the independent variable increases by one standard deviation, controlling for other variables in the model.

¹² As noted earlier, Oliver and Mendelberg (2000) suggested that residential segregation may alleviate perceived threat in localities where blacks represent a large share of the population, mitigating the tendency for negative racial attitudes to develop. In supplementary analysis we found only qualified support for this hypothesis: For the stereotyping and emotion measures, the impact of race composition was significantly weaker in metro areas where the (dissimilarity) segregation index was high; but the interaction was not significant for the other six attitude measures, and for five of the six it ran in the opposite direction. Interaction effects between residential segregation and the two socioeconomic composition measures were more likely to run in the predicted direction, socioeconomic composition effects being fainter when segregation was high, but only two of the sixteen interaction effects were significant. (Full results are available from the author.)

¹³ For decades, race scholars have asked whether the tendency for the better educated to give more liberal answers on surveys represented their greater commitment to racial equality or something rather more superficial, even self-serving (see, e.g., Jackman and Muha 1984). The educational composition effect reported here is not open to the same questions: If residents of white communities with a higher proportion of college graduates appear more liberal across the board, the highly educated and the poorly educated alike, something more substantial must be going on than a veneer of liberalism polished by college education.

¹⁴ Other important questions, not our focus here, also deserve continuing research attention. We trust future studies will permit analysts to make confident statements about differences in contextual influences across broader and narrower contextual units, and to extend generalizations about environmental effects to multiracial and multiethnic perceivers and targets.