Class Isolation and Affluent Americans’ Perception of Social Conditions

In addition to Oliver (1999, 2001), others have examined the effects of suburban living on political behavior (e.g., Gainsborough 2001). Yet it is important not to directly equate suburbanization with class isolation. It is now the case that more of America’s poor live in the suburbs than in inner cities, which was not true as recently as 2000 (Kneebone and Berube 2013). As a consequence, affluent Americans may live in suburbs and still live in close proximity to those of lower socioeconomic status.

Though Cruces et al. (2013) attribute people’s tendency to extrapolate from neighborhood conditions to the representativeness heuristic, other heuristics may contribute to this tendency as well. For example, the availability heuristic, which involves canvassing memory for relevant examples, may lead people to extrapolate from neighborhood conditions to form perceptions of social conditions if neighborhood conditions are foremost in their memory (Tversky and Kahneman 1973).

Due to an error in survey administration, an incomplete version of the survey that does not contain the necessary variables was administered to an additional 4880 respondents in 2010. Though these respondents are present in the publicly available data, I exclude them here.

No other variable is missing for more than 4 % of respondents.

While this subgroup analysis reflects the limits of a small sample size, it also provides suggestive evidence of substantively important effects. Specifically, there is preliminary evidence to suggest that class isolation may negatively affect affluent blacks’ participation in community-based efforts to address harmful social conditions. More research with a larger sample of affluent blacks is needed to test this potential finding, but if it were to hold, it would have important implications for black politics. Cooperation between affluent and non-affluent blacks to address harmful social conditions is part of the foundation of black political life (Dawson 1994), yet it may be imperiled should class isolation continue to rise. See Supplementary material Appendix p. 12 for more details.

The number of affluent respondents from other minority groups in the data, such as affluent Asians and affluent Latinos, is even fewer than the number of affluent blacks, preventing subgroup analyses of these other groups.

Throughout the text I use the term “metropolitan community” to reference the communities defined by the Knight Foundation and Gallup for the survey (see Supplementary material Appendix Table A1).

Full details on question wordings, coding, and distributions are available in the Supplementary material Appendix for all variables used in the analysis (p. 3).

The crime item question is as follows: “On a five-point rating scale, where 5 means extremely low and 1 means extremely high, how would you rate the level of crime in your community?”

1 on the original scale is converted to 0 on the new scale, 2 is converted to 25, 3 is converted to 50, 4 is converted to 75, and 5 is converted to 100.

This information was learned through extensive conversations with the director of the survey.

Notably, this measure compares the affluent’s perceptions to the perceptions of the non-affluent, rather than to objective reality itself. This aligns with H2, which concerns the positivity of the affluent’s perceptions relative to everyone else in society, rather than the accuracy of the affluent’s perceptions per se. While it is not the focus of this paper, objective reality—the aggregate experience of all members of a metropolitan community—is likely to lie in between the perceptions of the affluent and non-affluent. As the non-affluent are typically more vulnerable to each of the four social conditions under study than the affluent (Adler and Newman 2002; Benach et al. 2014; Levitt 1999; Reardon 2013), their average rating is likely to overestimate the true severity of social conditions by not accounting for the experience of the affluent. Conversely, the average affluent respondent would be expected to underestimate the true severity of social conditions by not accounting for the experience of the non-affluent, leaving the objective reality in between the perceptions of the two groups. The affluent’s tendency to underestimate the objective severity of social conditions may increase as their isolation from the non-affluent increases, a hypothesis that warrants investigation in future research.

I also check to see that results are consistent when the full range of the variable is used.

The diversity of Americans’ voluntary activity is also reflected in data collected by the Corporation for National and Community Service. Among those who volunteered with a group or organization, most volunteered with religious organizations (36 %), followed by educational and youth service organizations (27 %), with social or community service organizations—the category most directly related to addressing harmful social conditions—coming in third at 14 % (CNCS 2010).

Results for voter registration and voter turnout should be interpreted in light of the finding that over-reporting often biases self-reported measures of electoral participation (Bernstein et al. 2001). Ninety-six percent of affluent respondents report being registered to vote, while 88 % of affluent respondents in the SOTC survey report voting in a local election in the last 12 months. This issue is not unique to the SOTC survey: In the 1990 American Citizen Participation Study, one of the few other surveys to ask about voting in local elections, 80 % of affluent respondents (with incomes adjusted for year) report having voted in a “local community election.” By comparison, the best available data (see Oliver et al. 2012, p. 65) suggest that actual turnout in local elections for all social class groups ranges from below 35 % when there is no concurrent federal election to upwards of 75 % when there is a concurrent presidential election (there is no available data with which to measure actual turnout in local elections among the affluent in particular).

Measures of individual-level partisanship and ideology are not available in the Soul of the Community survey. I note that past studies of the effects of economic segregation have not included either as controls (Oliver 1999). I also test for bias from the omission of these controls by rerunning the main analyses controlling on whether a respondent lives in a “red state” or a “blue state” as a proxy for individual-level partisanship and ideology (see Supplementary material Appendix p. 16).

The lowest level of geo-coding available in the survey is the county.

Two SOTC communities, Miami, FL, and Palm Beach, FL, share the same MSA.

In measuring the isolation index, affluence is defined as having an income larger than four times the poverty threshold for a family of four ($20,650), making the limit $82,600.

Class isolation and income inequality are only moderately correlated among the MSAs in the dataset, preventing issues with multicollinearity (Supplementary material Appendix p. 6).

Results are identical in direction and significance with and without standardized non-binary independent variables.

The same also holds true in a multilevel ordered logit model.

An alternative interpretation of these results is that respondents are ignoring the survey taker’s instructions (p. 10) by providing ratings of the crime level that are only meant to apply to their neighborhood rather than their larger metropolitan community, leading to a strong relationship between Neighborhood Experience and perceptions of crime. If this were the case, respondents could not be said to be extrapolating from neighborhood conditions to form perceptions of social conditions. Yet this alternative interpretation is contradicted by results showing that MSA-level factors, most notably Percent Black, also have significant effects on crime perceptions. Respondents appear to be taking MSA-level factors into account in rating the crime level, suggesting that they are complying with the instructions to provide ratings that apply to their larger metropolitan community rather than their neighborhood alone.

See p. 20 in the Supplementary material Appendix for regression results from the fixed effects model.

The regression results from which this figure is derived are available in the Supplementary material Appendix (p. 21).