Understanding the factors that drive individuals’ residential preferences is a critical issue in the study of racial segregation. An important debate within this field is whether individuals—especially whites—prefer to live in predominantly white neighborhoods because they wish to avoid the social problems that may be more likely to occur in predominantly black neighborhoods (i.e., the racial proxy hypothesis) or because of racial factors that go beyond these social class–related characteristics. Through a multilevel analysis of data from the 2004–2005 Chicago Area Study and several administrative sources, we assess the extent to which the racial proxy hypothesis describes neighborhood satisfaction among whites, African Americans, and Latinos living across a broad range of neighborhood contexts. The racial proxy perspective applies weakly to whites’ satisfaction: whites report less satisfaction in neighborhoods with more minority residents, and only some of their dissatisfaction can be attributed to local social characteristics. The racial proxy hypothesis applies more strongly to blacks’ and Latinos’ satisfaction. In some cases, especially for Latinos, higher levels of satisfaction in integrated neighborhoods can largely be attributed to the fact that these places have better socioeconomic conditions and fewer social problems than predominantly minority communities. At the same time, effects of racial/ethnic composition persist in unique and somewhat divergent ways for blacks and Latinos, supporting the assertion that racial composition matters, above and beyond its relation to social class. Taken together, these findings suggest that individuals balance both socioeconomic and race-related concerns in their residential preferences.
A substantial body of research has demonstrated that racial residential segregation persists across the United States and continues to undermine minorities’ life chances (Charles 2003; Massey and Denton 1993); thus, understanding the forces that cause and perpetuate segregation has been an important goal for social scientists. One line of research on this issue has examined the determinants of individuals’ preferences for neighborhood racial composition, based on the assumption that if stable integration is to exist, individuals must evaluate diversity positively and desire to live in racially mixed places (Lee et al. 1994).
A central debate within this line of research focuses on understanding the factors that drive individuals’ racial residential preferences. Are whites’ preferences for all-white neighborhoods the result of racial prejudice against blacks, or do they arise from class-based concerns about the diminished quality of life, such as lower property values, underperforming local schools, and poor public safety, that they believe occur in neighborhoods where African Americans live? Are blacks’ preferences for racially mixed neighborhoods the result of integrationist social values, or do they stem from the fact that mixed neighborhoods offer class-based improvements in quality of life compared with predominantly black neighborhoods? Previous studies have provided evidence for both the racial proxy hypothesis, which asserts that residential preferences are motivated by quality-of-life indicators associated with race (Ellen 2000; Harris 1999, 2001; Taub et al. 1984), as well as for the hypothesis that residential preferences are driven by explicit beliefs about race (Baybeck 2006; O’Brien and Clough 1982; Stipak and Hensler 1983).
This body of work is limited in two key respects. First, it has not kept pace with increasing racial and ethnic diversity in the United States. Prior work has focused largely on blacks’ and whites’ preferences for living with one another and has neglected other groups’ preferences, as well as how blacks and whites respond to other groups. Some research has indicated the presence of a hierarchy of racial residential preferences (e.g., Charles 2000, 2001; Zubrinsky and Bobo 1996), but these studies did not empirically consider how social class conditions affect individuals’ residential preferences in multiethnic contexts. Second, research evaluating the racial proxy hypothesis has been limited by the use of respondents’ subjective assessments to stand in for objective measurements of neighborhood characteristics (e.g., Harris 2001; O’Brien and Clough 1982). Recent studies have demonstrated that individuals systematically overestimate levels of crime and disorder in neighborhoods with a higher proportion of African American residents (Quillian and Pager 2001; Sampson and Raudenbush 2004); if individuals’ reports of neighborhood characteristics are themselves influenced by racial composition, then analyses that rely solely on subjective assessments of neighborhood quality may yield results that are biased in critical ways.
Our multilevel analysis sheds new light on the extent to which the racial proxy hypothesis describes residential preferences by examining how residents of Chicago and its suburbs rate their neighborhood satisfaction across a broad range of racial and ethnic community contexts. We address the limitations of prior research by investigating (1) the extent to which neighborhood racial/ethnic composition is related to neighborhood satisfaction among whites, African Americans, and Latinos after accounting for administrative measures of neighborhood socioeconomic conditions and respondents’ perceptions of neighborhood quality, and (2) the extent to which the relationship between racial/ethnic composition and neighborhood satisfaction is influenced by independently observed neighborhood characteristics—including socioeconomic status, crime rate, and school performance—or respondent-reported local social conditions.
A long line of research has devoted attention to understanding individuals’ preferences for neighborhood racial and ethnic composition. Much of this work assessed residential preferences by asking respondents to rate hypothetical neighborhoods with different proportions of black and white residents. In general, these studies have found that whites prefer neighborhoods with few or no African American residents, whereas blacks prefer integrated neighborhoods with a sizable black presence (Farley et al. 1994, 1997; Krysan and Farley 2002; Zubrinsky and Bobo 1996). Recent research has also begun to examine how whites and blacks respond to living with Latinos, as well as Latinos’ residential preferences. Whites prefer living with Latinos over living with African Americans, but African Americans prefer integration regardless of whether they live with whites or Latinos. Further, Latinos prefer neighborhoods with both a significant proportion of Latinos and whites and are more willing to live in predominantly white neighborhoods than are blacks, but they prefer predominantly Latino neighborhoods to integration with African Americans (Charles 2000, 2001, 2006; Zubrinsky and Bobo 1996). Many of these studies attempted to control for socioeconomic conditions by asking respondents to imagine they have found an affordable, attractive home in neighborhoods with varying racial compositions, but on the whole, this body of research speaks in limited ways to whether racial residential preferences are propelled by neighborhood social class characteristics.
The racial proxy perspective holds that racial composition may indeed matter for individuals’ preferences, but only insofar as it is related to economic and other quality-of-life characteristics of the neighborhood (Harris 1999, 2001; Taub et al. 1984). According to Harris (1999:464), the racial proxy hypothesis maintains that
. . . racial preferences simply represent a desire to live in areas free of crime, deteriorating buildings, ineffective public schools, and other social ills. Because of the concentration of many social problems in neighborhoods with relatively large black populations . . . selecting a “good” environment usually means choosing a predominantly white neighborhood.
Ellen’s (2000) race-based neighborhood stereotyping perspective offers an important variation on the racial proxy hypothesis. Ellen agrees with the main tenet of the racial proxy perspective but argues that a neighborhood’s trajectory of racial change is a more powerful predictor of neighborhood desirability than its current racial composition because individuals—especially whites—use the pace and direction of racial change as a signal that moderately integrated neighborhoods will transition to predominantly minority communities. Subsequently, whites choose to avoid mixed areas because “many white households, rightly or wrongly (and even, perhaps, with some regret), associate predominantly black neighborhoods with diminished neighborhood quality and resilience” (Ellen 2000:47). This perspective is a marked departure from other work on the racial proxy hypothesis (Harris 1999, 2001; Taub et al. 1984) for two reasons. First, Ellen (2000) recognizes that individuals—especially whites—may develop residential preferences based on the assumption that racially mixed neighborhoods inevitably turn into predominantly minority neighborhoods. Second, she acknowledges that there may be slippage between individuals’ perceptions of quality and the actual quality of racially mixed and predominantly minority neighborhoods.
An alternative to the racial proxy perspective is the hypothesis that individuals’ residential preferences are motivated directly by racial factors. This idea has been cast in a variety of ways by extant research, including as “racial prejudice,” “pure race,” and “race per se.” The underlying argument is the same in all cases: that race matters in some way above and beyond its association with socioeconomic characteristics. That is, individuals view their neighbors through the collectively and sociohistorically developed lens of “race” (Bobo and Zubrinsky 1996) and base their neighborhood preferences on these perceptions. Clark (1991, 1992) and others (Patterson 1997; Thernstrom and Thernstrom 1997) have argued that whites and blacks alike prefer the comfort inherent in living among their own groups, and that these benign preferences do not reflect negative attitudes toward other groups. In contrast, some researchers have contended that racial residential preferences are not neutral. Instead, for whites, they are based on stereotypes and prejudice (Bobo and Zubrinsky 1996); for blacks, they are based both on a fear of discrimination from whites (Feagin and Sikes 1994; Krysan and Farley 2002) and, on the other hand, on an endorsement of integration as a social good (Zubrinsky and Bobo 1996). These scholars have divergent views about how racial considerations influence residential preferences; nonetheless, they agree that race is a critical factor for determining how individuals evaluate neighborhoods.
In order to understand the extent to which the racial proxy hypothesis applies to racial residential preferences, we concentrate on the relationship between local racial/ethnic composition and residents’ neighborhood satisfaction. We concentrate on neighborhood satisfaction because it provides information about residents’ evaluations of actual community conditions rather than their opinions about hypothetical neighborhoods. Further, a global measure of neighborhood satisfaction may be preferable to specific questions about how residents evaluate racial/ethnic composition because these focused questions may be subject to social desirability pressures to report openness to living with members of other racial/ethnic groups.1
Empirical studies of neighborhood satisfaction have provided mixed evidence on the racial proxy hypothesis. A number of analyses have found that after accounting for local socioeconomic conditions and perceived social problems, individuals—especially whites—report lower levels of satisfaction in racially mixed neighborhoods than they do in predominantly white neighborhoods (Baybeck 2006; O’Brien and Clough 1982; Stipak and Hensler 1983), which contradicts the predictions of the racial proxy hypothesis. However, Harris (2001) has shown that both whites and African Americans report less satisfaction in places with more black residents, and that these negative evaluations are driven entirely by the zip-code-level poverty rate and individuals’ perceptions of crime, neighborhood deterioration, and bad schools. Further, Ellen (2000) has demonstrated that increases in the proportion of black residents are associated with lower levels of satisfaction among white homeowners but not among white renters or blacks, after controlling for neighborhood poverty, which she interprets as support for the neighborhood race-based stereotyping hypothesis.2
Extending Previous Research
In this article, we build on insights provided by existing research to better understand the relationship between racial/ethnic composition and neighborhood satisfaction. Prior work on residential preferences has focused on blacks and whites and has paid less attention to other racial/ethnic groups, including Latinos; in particular, few studies have explored the degree to which racial/ethnic composition—net of social class characteristics—drives preferences in a multiethnic context. Latinos are a sizable and rapidly growing proportion of the population in metropolitan areas; for example, in the Chicago metropolitan area in 2000, Latinos accounted for 17.4% of the total population, up from 12.1% in 1990 (U.S. Census Bureau 1990, 2000). Thus, contemporary research on residential preferences must take seriously both Latinos’ preferences and how others respond to Latinos. In addition, including Latinos in the analysis can provide leverage for adjudicating between different theories about how race/ethnicity matters. For example, analyzing how African Americans respond to living with whites compared with Latinos can yield insight into whether simple in-group preference drives neighborhood satisfaction—in which case, blacks would respond similarly to neighborhood coresidence with whites or Latinos—or whether fear of racial discrimination is related to satisfaction.
Most previous research on this topic analyzes respondents’ evaluations of hypothetical neighborhoods. Whites appear to be more open to integration with Latinos than they are to integration with African Americans (Charles 2000, 2001; Zubrinsky and Bobo 1996). Further, the fact that segregation is lower between whites and Latinos compared with whites and blacks (Iceland et al. 2002; Logan 2001) suggests that whites may be more tolerant of Latinos, although ethnographic accounts indicate that white-Latino neighborhoods are characterized by substantial tension (Flippen 2001; Kefalas 2003). Only one study has investigated whether whites’ preferences for living with Latinos are the result of class-related neighborhood characteristics: Emerson et al. (2001) found, after accounting for the quality of local schools, that trends in property values, or the crime rate, whites prefer neighborhoods with fewer black residents, but their preferences are not sensitive to the proportion of Latinos or Asians.
Evidence on African Americans’ preferences for living with Latinos is also mixed. One study found that African Americans have a slight preference for integration with whites over Latinos (Charles 2001), but other studies have suggested that African Americans prefer integration regardless of the identity of the out-group neighbors (Charles 2000; Zubrinsky and Bobo 1996).
Like whites, Latinos appear to adhere to a hierarchy of residential preferences. They prefer both a substantial coethnic presence and integration with whites, and they are more willing than are blacks to be ethnic “pioneers” in neighborhoods that are predominantly white. Their least-preferred neighbors are African Americans (Charles 2000, 2001; Zubrinsky and Bobo 1996). This distaste for living with blacks may stem from the notion that proximity to African Americans is associated with low social status and limited opportunities for upward mobility and could reflect conflicts that emerge between blacks and Latinos as they compete for scarce economic resources (Bobo and Hutchings 1996; Oliver and Wong 2003).
The studies outlined above provide important information about patterns of residential preferences in multiethnic contexts. However, most of these studies gauged residential preferences by asking respondents if they would be willing, in theory, to live in hypothetical neighborhoods with a variety of racial/ethnic compositions. In relying on assessments of hypothetical neighborhoods, prior research has been limited in two ways. First, respondents’ evaluations of hypothetically mixed neighborhoods may be quite different from their evaluations of actual neighborhoods that are racially mixed. Second, including measures of neighborhood social problems in studies that also ask respondents to evaluate hypothetical neighborhoods places a heavy burden on the respondent; thus, most studies measuring preferences for hypothetical neighborhoods (except Emerson et al. 2001; Krysan et al. 2009) have not addressed the question of whether racial composition matters above and beyond social class characteristics. Our work extends previous research by analyzing respondents’ actual neighborhood satisfaction—as opposed to hypothetical preferences—to understand how whites and blacks evaluate coresidence with Latinos, as well as how Latinos evaluate coresidence with whites and blacks, and by investigating the extent to which these evaluations are based on neighborhood characteristics associated with racial/ethnic composition.
We also argue that earlier studies evaluating the racial proxy hypothesis may have yielded biased results because they relied solely on survey respondents to assess the quality of neighborhood conditions in racially mixed areas. The key problem with this approach is that subjective assessments of neighborhood quality may be closely tied to the racial composition of the neighborhood; for example, previous research has found that individuals report higher levels of crime and disorder in neighborhoods that have a higher proportion of black residents, even after objective levels of crime and disorder are controlled for (Quillian and Pager 2001; Sampson and Raudenbush 2004). If perceptions of neighborhood problems are inflated in neighborhoods with a higher proportion of minorities, then the true effect of racial composition on neighborhood satisfaction may be understated. In light of this concern, we assess how both residents’ perceptions of neighborhood quality and administrative measures of neighborhood quality are related to neighborhood satisfaction.
We explore two overarching research questions designed to assess the racial proxy interpretation of neighborhood satisfaction. First, we investigate the extent to which the relationship between racial/ethnic composition and neighborhood satisfaction can be explained by differences in local socioeconomic conditions and respondent perceptions of neighborhood quality. We estimate the relationship between racial/ethnic composition and neighborhood satisfaction separately for whites, African Americans, and Latinos. For each racial/ethnic group, we begin with a model that includes a measure of racial/ethnic composition and a series of individual-level attributes as controls. Then, we add measures of neighborhood socioeconomic characteristics and subjective neighborhood characteristics. Following previous studies (e.g., Harris 2001; St. John and Bates 1990), to the degree that effects of racial/ethnic composition remain after neighborhood characteristics are accounted for, we will conclude that race is not merely a proxy for social class. Evidence for the racial proxy hypothesis comes from the extent to which neighborhood characteristics explain the relationship between racial/ethnic composition and satisfaction.
Second, we ask whether objective neighborhood quality indicators or subjective perceptions of neighborhood quality are more strongly related to neighborhood satisfaction. We employ data from a subsample of respondents in order to use data on neighborhood crime and school quality that were available only for neighborhoods within Chicago’s city limits. We first estimate a single model of the relationship between racial/ethnic composition and neighborhood satisfaction for the sample of whites, blacks, and Latinos living in city neighborhoods.3 Then we add cross-level interactions between race/ethnicity and racial/ethnic composition to determine the extent to which individuals of diverse racial/ethnic backgrounds respond differently to racial/ethnic composition. In the next model, we add administrative measures of neighborhood quality. In the final model, we add individuals’ perceptions of neighborhood characteristics. If objective neighborhood characteristics predict neighborhood satisfaction and diminish the effect of racial/ethnic composition on satisfaction, we will conclude that the racial proxy hypothesis is supported. But if neighborhood racial/ethnic composition continues to predict neighborhood satisfaction, then we will conclude that explicit racial considerations do matter in shaping residential preferences.
It is important to use caution when interpreting the findings, for three reasons. First, assessing whether residential preferences are based on racial/ethnic composition or associated social problems requires careful specification of local social conditions. It is possible that the analytic models omit some of the wide array of neighborhood characteristics that affect residents’ satisfaction, such as the proximity of parks and other amenities or the age of the housing stock. Second, it may be that individuals who reside in racially/ethnically mixed neighborhoods differ from those who reside in predominantly single-race/-ethnicity neighborhoods on key characteristics. For example, those who reside in integrated neighborhoods may subscribe to ideals of diversity, which might inflate their reports of satisfaction. Third, we analyze data from Chicago and its suburbs, and this focus on a single metropolitan area could limit generalizability to other U.S. cities.
Data and Measures
Data for the analysis come from several sources: the 2004–2005 Chicago Area Study (CAS), the 2000 U.S. Census, the 2006 National Neighborhood Crime Study (NNCS), and the 2002 Chicago Public Schools (CPS) Report Cards published by the CPS Office of Performance (2002). The 2004–2005 Chicago Area Study comprised an area probability sample of individuals aged 21 and older nested within block groups in Cook County, Illinois. Face-to-face interviews were conducted in respondents’ homes, with a response rate of 45%. Our analyses include an adjustment for nonresponse through the use of block group–level nonresponse rate weights.4 Respondents were asked questions regarding their perceptions of their neighborhoods, housing searches, experiences with discrimination, and racial attitudes. We restrict our analysis to those respondents who identified as non-Hispanic white, non-Hispanic black, and Hispanic (who could be of any race), for a sample size of 756 respondents nested within 80 block groups. In addition to weighting for nonresponse, our analyses also include a weight for the probability of selection.
The key outcome of interest in this analysis is drawn from the CAS, as are other measures of individuals’ characteristics and perceptions of neighborhood quality. Neighborhood satisfaction measures how satisfied an individual is with his or her neighborhood as a place to live (1 = not at all satisfied; 6 = extremely satisfied). We also examine respondents’ assessments of both current and future neighborhood social conditions. Measures of current neighborhood conditions include respondents’ reports of how well their neighbors keep up their property (property upkeep problems) and the extent to which there are problems with crime or vandalism in the neighborhood (problems with crime). Both of these variables are measured on Likert-type scales in which 1 corresponds to “never a problem” and 5 corresponds to “always a problem.” Next, respondents report their assessment of the quality (1 = excellent; 5 = poor) of public schools in the neighborhood (school dissatisfaction) and police protection in the neighborhood (police dissatisfaction).
Measures of future neighborhood conditions include respondents’ perceptions of the extent to which the number of blacks will increase in the neighborhood and the extent to which the number of Latinos will increase in the neighborhood (1 = decrease a lot; 5 = increase a lot). Respondents are asked whether, in the next five years, the quality of neighborhood public schools will decline, property values will decline (1 = get much better; 5 = get much worse for both measures), and crime will increase (1 = decrease a lot; 5 = increase a lot). Thus, we measure an extensive set of perceived neighborhood characteristics, with higher values on the scales representing more neighborhood problems (or more minorities entering the neighborhood).5
Individual-level control variables are also drawn from the CAS. We include measures of respondents’ race/ethnicity (non-Hispanic white [omitted], non-Hispanic African American, and Latino), gender (male [omitted] vs. female), age, immigrant status (native-born [omitted] vs. immigrant), marital status (not married [omitted] vs. married), homeownership (renter [omitted] vs. homeowner), number of years living at the same address, the presence of children younger than 18 in the home, income (less than $20,000 per year [omitted], $20,000–$39,999 per year, $40,000–$79,999 per year, and $80,000 or more per year), and education (non-college-educated [omitted] vs. college-educated). We also include a dummy variable that indicates whether a respondent holds city residence or suburban residence.
Measures of neighborhood characteristics are drawn from the 2000 U.S. census. We consider the census block group to be a proxy for the local neighborhood.6 This is a departure from prior work that uses larger units as neighborhoods—for example, the zip-code area (e.g., Harris 2001). We focus on block groups because closer proximity to members of different racial/ethnic groups offers increased possibility of contact and interaction, which may be an important part of residents’ satisfaction. To assess the socioeconomic composition of the block group, we use two sets of measures. We include a measure of the percentage of poor residents and a three-category variable that divides the median household income of the block group into thirds (the lowest third median household income is the omitted category). We also include a standardized scale of residential stability that incorporates the percentage of residents who are homeowners and the percentage of residents who have lived at the same address for five years or longer.
Neighborhood racial/ethnic composition is measured with two different categorical specifications of the percentage non-Hispanic white, percentage non-Hispanic black, and percentage Latino residents in the block group. In models that analyze the relationship between racial/ethnic composition and satisfaction for a respondent’s racial/ethnic in-group—for example, models that analyze how Latinos’ satisfaction varies by the proportion of Latino residents—the measure used is as follows: less than 25% of the in-group (i.e., percentage white, black, or Latino; omitted), between 25% and 75% of the in-group, and greater than 75% of the in-group. In models that analyze the relationship between racial/ethnic composition and satisfaction for a respondent’s racial/ethnic out-group—for example, models that analyze how blacks’ satisfaction varies by the proportion of whites—the measure used is as follows: less than 10% of the out-group (omitted), between 10% and 30% of the out-group, and greater than 30% of the out-group. We use two different specifications of racial/ethnic composition because the unequal distribution of respondents into block groups by race/ethnicity results in categories with sample sizes that are too small to effectively analyze using any single specification. A number of different specifications of racial/ethic composition were used in sensitivity analyses; key results are robust across model specifications.
In the models that use the subsample of CAS respondents who reside within Chicago’s city limits, we use a single measure of racial composition: the linear percentage of white residents in the block group. We employ this strategy because using categorical measures with the smaller sample (n = 392) yields cell sizes that are too small to effectively analyze.7 In addition, the 2006 National Neighborhood Crime Study (Peterson and Krivo 2006) provides a measure of the neighborhood crime rate, which is the sum of the three-year average (between 1999 and 2001) rates of violent crime and property crime. Violent crime includes homicides, rapes, robberies, and aggravated assaults. Property crime includes burglaries, larcenies, and motor vehicle thefts. Because information about the local crime rate is unavailable at the block-group level, we use the crime rate for respondents’ census tract in this analysis. Data drawn from the 2002 Chicago Public Schools Report Cards provides a measure of the percentage of students in a respondent’s city-assigned elementary school who were at or above the national average reading score on the Iowa Test for Basic Skills (Chicago Public Schools Office of Performance 2002). Table 6 in Online Resource 1 presents descriptive statistics for respondents from the full sample of the Chicago Area Study.8
Race/Ethnicity, Neighborhood Racial/Ethnic Composition, and Neighborhood Satisfaction
Table 1 presents models for neighborhood satisfaction among whites. Model 1 shows that whites report lower levels of satisfaction when they live in neighborhoods with fewer whites. These effects are only slightly diminished after neighborhood socioeconomic status, residential stability, and whites’ perceptions of neighborhood characteristics are controlled for in Model 2. Model 3 demonstrates that whites report less satisfaction in neighborhoods with high percentages of African American residents. In this case, whites’ satisfaction follows a nonlinear pattern: in neighborhoods with a moderate proportion of black residents, whites’ satisfaction is no different than it is in neighborhoods with few blacks, but satisfaction drops sharply in neighborhoods where more than 30% of residents are African American. Controlling for neighborhood conditions in Model 4 slightly lowers the negative coefficient for living with a high proportion of blacks. Model 5 shows that whites report lower levels of satisfaction in neighborhoods with a moderate or high proportion of Latinos than in neighborhoods with few Latinos. These negative effects are somewhat reduced by accounting for local social characteristics in Model 6 but remain statistically significant.
In large part, administratively measured neighborhood characteristics do not predict whites’ satisfaction. Median household income and residential stability are unrelated to satisfaction, and only in Model 6 is the poverty rate associated with less satisfaction. However, an array of perceived conditions is linked to whites’ satisfaction. Whites report less satisfaction in neighborhoods where they perceive more problems with crime and where they believe that property values will decline in the future. Concern about school quality appears especially relevant: both current dissatisfaction with local schools and the belief that local school quality will decline in the future are associated with less satisfaction.
Table 2 presents results from models of African Americans’ neighborhood satisfaction. Model 1 shows that African Americans living in neighborhoods that have a moderate proportion of black residents report higher levels of satisfaction than those living in predominantly black neighborhoods. However, blacks living in neighborhoods with few other blacks and those living in predominantly black neighborhoods report similar levels of satisfaction. Accounting for neighborhood conditions in Model 2 reduces the positive coefficient for living in neighborhoods that have a moderate proportion of black residents to nonsignificance, suggesting that blacks’ positive evaluation of racially mixed neighborhoods is driven by the fact that these places have more socioeconomic resources and fewer respondent-reported social problems than predominantly black neighborhoods. Further, the negative coefficient for living in neighborhoods with few blacks becomes significant in Model 2. In other words, blacks’ dissatisfaction in neighborhoods where they are among the few black residents is masked by the fact that these neighborhoods have better social conditions than predominantly black neighborhoods.
Models 3 and 4 suggest that there are no significant effects of the percentage of white residents on blacks’ satisfaction. In Model 5, there is no relationship between the percentage of Latino residents and blacks’ satisfaction, but adding local social characteristics in Model 6 demonstrates that blacks living in neighborhoods with a moderate proportion of Latinos report less satisfaction than they do in neighborhoods with few Latinos. Thus, blacks respond negatively to living with Latinos, but their negative evaluations are suppressed by the fact that these neighborhoods have better social conditions than predominantly black neighborhoods.
Both administratively measured and respondent-reported neighborhood characteristics predict satisfaction for blacks. African Americans report less satisfaction in neighborhoods that have a higher proportion of poor residents and more satisfaction in neighborhoods with high median household income. Additionally, current and future conditions related to crime and policing emerge as central to blacks’ satisfaction: respondents report lower levels of satisfaction in neighborhoods where they perceive problems with crime, dissatisfaction with the police, and future increases in crime. African Americans also report less satisfaction in neighborhoods where they perceive problems with property upkeep.
Table 3 displays results from models for Latinos’ neighborhood satisfaction. Model 1 shows that Latinos report higher levels of satisfaction in neighborhoods with lower proportions of Latinos. After we account for local social conditions in Model 2, the positive coefficients for the proportion of Latino residents each drop by approximately one-third, suggesting that Latinos’ satisfaction is influenced by administratively measured and respondent-reported neighborhood conditions in critical ways. Model 3 shows that Latinos report somewhat higher satisfaction in neighborhoods with a moderate or high proportion of whites compared with neighborhoods with few whites. As was the case in Models 1 and 2, these coefficients are reduced by adding neighborhood characteristics to the model. Accounting for community social conditions in Model 4 reduces the positive coefficient for living in neighborhoods that have a moderate proportion of whites to nonsignificance and reduces the positive coefficient for living in predominantly white neighborhoods by about 40%. However, the positive effect of living in predominantly white neighborhoods persists, indicating that concerns related to both social class and race are at play in Latinos’ evaluations of living with whites. Models 5 and 6 show that Latinos report less satisfaction in neighborhoods with a moderate proportion of black residents than in neighborhoods with few black residents after local social characteristics are accounted for. There are no effects for Latinos living in neighborhoods with a high proportion of blacks: this finding may reflect that only seven Latino respondents in the CAS fall into this category.
Latinos’ perceptions of social conditions appear to be more significant drivers of satisfaction than are administratively measured neighborhood characteristics. As was the case for whites, in only one case do local socioeconomic characteristics predict satisfaction for Latinos (in Model 6). Across models, Latinos report lower levels of satisfaction in neighborhoods where they believe school quality and property values will decline in the future. This finding suggests that perceptions of neighborhood change are more relevant for Latinos’ satisfaction than are perceptions of current neighborhood conditions.
Objective and Subjective Neighborhood Characteristics and Neighborhood Satisfaction
Table 4 presents models from the subsample of CAS respondents that analyze how administratively measured neighborhood characteristics compare to respondent-perceived neighborhood characteristics in predicting satisfaction. Model 1 shows that the relationship between the percentage of white residents and satisfaction in the city subsample is similar to the relationship found in the overall sample for whites and Latinos: in neighborhoods with a higher percentage of white residents, individuals report higher levels of neighborhood satisfaction. Model 2 adds cross-level interactions between individual race/ethnicity and the percentage of white residents. The cross-level interactions test whether the relationship between percentage white and neighborhood satisfaction is significantly different for African Americans and Latinos as compared with whites. The small, nonsignificant coefficient for Latinos indicates that the relationship between percentage white and neighborhood satisfaction is similar for Latinos and whites; both groups report higher levels of satisfaction in neighborhoods with a higher percentage of white residents. The negative coefficient for African Americans demonstrates that the positive relationship between the proportion of white residents and satisfaction is weaker for African Americans as compared with whites.
In Model 3, objective neighborhood characteristics are added. Neighborhood poverty is associated with less satisfaction, and higher median household income is associated with more satisfaction. There is no significant relationship between local schools’ reading scores or the local crime rate and neighborhood satisfaction. However, including these characteristics reduces the size of the coefficient for percentage white by almost half, indicating that objective neighborhood social class–related characteristics are an important part of the relationship between racial composition and satisfaction. The coefficients from the cross-level interaction indicate that the positive relationship between racial composition and satisfaction continues to be weaker for African Americans as compared with whites and Latinos.
Model 4 includes respondents’ perceptions of current and future neighborhood conditions. Respondents’ perceived property upkeep problems, problems with crime, police dissatisfaction, and perceptions of future increases in crime are all associated with less satisfaction. Including these respondent-reported measures does not reduce the size of the coefficient for the percentage of white residents. The coefficient for the cross-level interaction between a black resident and the percentage of white residents is reduced to marginal significance, indicating that small differences persist in the ways that blacks and whites respond to the percentage of white residents after accounting for both administratively measured and respondent-perceived neighborhood social conditions. However, the coefficient for the effect of percentage of white residents on black’s satisfaction is 0.002, or nearly zero (i.e., the main effect for percentage white is 0.015 and the interaction term is −0.013), suggesting that, on the whole, blacks’ satisfaction is not affected by the proportion of whites in city neighborhoods.
Discussion and Conclusion
Our multilevel analysis brings together several sources of data to investigate the extent to which individuals’ racial residential preferences are driven by the prevalence of social class–related problems in minority communities. The results offer some support for the racial proxy perspective, or the idea that residents’ dissatisfaction in neighborhoods with a higher proportion of minority residents is driven by poor social conditions in those places. Whites report lower levels of satisfaction in neighborhoods with a higher proportion of minorities; in the city subsample, socioeconomic characteristics account for almost half of this racial composition effect. Similarly, Latinos report lower levels of satisfaction in neighborhoods with more minorities, and much of this relationship is driven by the greater socioeconomic resources and fewer respondent-reported social problems in these neighborhoods. African Americans report higher levels of satisfaction in moderately integrated neighborhoods, and these reports can be entirely attributed to the fact that these neighborhoods have better living conditions than predominantly black neighborhoods.
However, after accounting for community social characteristics, distinct effects of racial/ethnic composition persist, supporting the idea that there is something about race, above and beyond social class, that propels neighborhood satisfaction. In the full sample of respondents, whites report less satisfaction in neighborhoods with higher proportions of minority residents, and their dissatisfaction is diminished by perceived and observed social conditions in only trivial ways. Even after local social characteristics are accounted for, Latinos report lower levels of satisfaction in neighborhoods with higher proportions of Latinos or African Americans. Further, African Americans report lower levels of satisfaction when they live with Latinos, and they report the least satisfaction in neighborhoods where they live with the fewest blacks, despite the better socioeconomic conditions and fewer social problems present these neighborhoods.
Taken together, the findings suggest that the racial proxy hypothesis weakly describes whites’ neighborhood satisfaction, but that it is more descriptive of minorities’ neighborhood satisfaction. For African Americans and Latinos, socioeconomic conditions are fundamental to understanding the relationship between racial/ethnic composition and satisfaction. Although in some cases, unique effects of racial/ethnic composition persist after considering neighborhood characteristics, in other instances, accounting for local social conditions diminishes the effects of racial/ethnic composition to nonsignificance. While neighborhood characteristics are also an important part of whites’ evaluations of racial composition, especially within the city limits, it is noteworthy that in no case does accounting for local social conditions eliminate the effects of racial/ethnic composition. Instead, for whites, other neighborhood characteristics appear secondary to racial considerations in determining satisfaction.
Comparing the findings for three different racial/ethnic groups offers some leverage for understanding why race matters for neighborhood satisfaction. One possibility is that, all things being equal, people have benign preferences for living with others of the same race/ethnicity (Clark 1991, 1992). If neighborhood satisfaction were driven by in-group preferences, all racial/ethnic groups would have negative responses to living in neighborhoods with few coethnics. In addition, individuals would respond similarly to residence with out-group members of various backgrounds. However, Latinos report marginally more satisfaction in neighborhoods with fewer coethnics, and blacks and Latinos respond differently to coresidence with one another compared with coresidence with whites. These two sets of findings are inconsistent with the idea that in-group preferences are the key to understanding neighborhood satisfaction.
A different line of research has posited that the reason race matters for residential preferences has more to do with the ways in which individuals experience their group position within the U.S. racial/ethnic hierarchy (e.g., Almaguer 1994; Bonilla-Silva 2004; Feagin 2000). Whites, based on their position at the top of the hierarchy, may feel threatened by coresidence with minorities because geographic proximity implies that their superior social position may be diminished (Bobo and Hutchings 1996; Bobo and Zubrinsky 1996; Park 1924) and may subsequently endorse negative stereotypes of minorities or adhere to racialized views of neighborhood social conditions (Krysan 2002b). The results of our analysis—that whites report less satisfaction in neighborhoods with more minority residents, and that these reports are not linked to socioeconomic conditions in the full sample and cannot be fully explained by neighborhood conditions in the city subsample—are in line with this interpretation of whites’ preferences.9
Compared with whites, minorities occupy a lower group position on the racial/ethnic hierarchy; thus, race matters differently for their residential preferences than it does for whites. Previous research suggests that African Americans fear hostility and other forms of discriminatory treatment from whites when they live in a neighborhood with few other blacks (Feagin and Sikes 1994; Krysan and Farley 2002; Timberlake 2000). The finding that blacks report less satisfaction in neighborhoods that have few black residents than they do in predominantly black neighborhoods supports this idea. However, this theory is not supported by the findings for Latinos. Instead, Latinos report more satisfaction in neighborhoods with higher proportions of whites than they do in predominantly Latino neighborhoods, implying that Latinos are more comfortable than African Americans being ethnic pioneers in all-white neighborhoods (Charles 2000, 2001, 2006). This finding may result from the fact that the majority of the CAS Latino sample is made up of immigrants, which could minimize concerns about discrimination.10 Glenn (1986) has argued that immigrants pay less attention to labeling “discrimination” and instead focus on the skills, behaviors, and conditions necessary for upward socioeconomic mobility.
Indeed, for Latinos—who occupy a position in the middle of the racial/ethnic hierarchy compared with blacks, who are at the bottom of the hierarchy—upward social mobility may be more attainable and therefore a more relevant concern. Latinos’ satisfaction is more responsive to local social characteristics than either whites’ or blacks’ satisfaction, which supports the idea that Latinos are centrally concerned with upward social and economic mobility. Further, if residential proximity to whites is one indicator of upward mobility, Latinos may evaluate coresidence with whites as an opportunity to make gains in social status. Another explanation, however, highlights concerns about omitted variable bias. It may be that the persistent positive effect of living with a higher proportion of whites or a lower proportion of Latinos is related to unmeasured aspects of the neighborhood environment that are related to racial composition, such as better amenities, improved local services, or increased political power. African American and Latino residents’ negative responses to living with one another may also be linked to their positions in the racial/ethnic hierarchy. Whereas Latinos may evaluate coresidence with whites as an opportunity to improve their social standing, they may view coresidence with blacks as diminishing their social status. Further, African Americans and Latinos may compete for scarce economic and social resources, which could create tensions in shared neighborhood space.
The analysis also demonstrates that social problems are an important part of negative evaluations of neighborhoods with more minorities, as predicted by the racial proxy hypothesis. For each racial/ethnic group, different types of perceived neighborhood problems come to the forefront as important predictors of neighborhood satisfaction. Latinos are most concerned with school quality; African Americans are most concerned with crime and policing; and whites have the longest list of concerns, which includes almost all types of perceived problems that were included in our survey. It may be that what constitutes “neighborhood problems” is race/ethnicity specific, and that different groups use different markers to make decisions about what constitutes a “good” or “bad” neighborhood. In addition, Ellen’s (2000) argument about the centrality of future neighborhood quality to satisfaction is strongly supported: for all groups, perceptions of future increases in social problems are linked to lower levels of satisfaction. However, for no group is there evidence to support her assertion that residents’ perceptions of minority group increase are associated with lower levels of satisfaction.
Our analysis also demarcates the effects of residents’ perceptions of social problems and objectively observed social problems on neighborhood satisfaction. With the exception of neighborhood socioeconomic status indicators, neither of the other two objective social class–related neighborhood characteristics (neighborhood crime and school quality) predicts individuals’ satisfaction. However, individuals’ perceptions of neighborhood characteristics—especially those related to crime—are related to lower levels of satisfaction. It is instructive that these likely racialized perceptions of crime and social problems (Quillian and Pager 2001; Sampson and Raudenbush 2004) are more strongly related to neighborhood dissatisfaction than is an objective measure of crime. Given that perceptions of social problems operate differently from objective social problems in this analysis, future research should take great care in relying on residents’ perceptions of social problems as indicators of true neighborhood conditions.
This study is limited in several ways. Most importantly, if individuals choose to move out of racially and ethnically mixed neighborhoods or avoid moving into these neighborhoods in the first place, some of the association between racial/ethnic composition and satisfaction could remain unobserved. In this case, our analysis could underestimate the true extent to which neighborhood satisfaction is diminished by coresidence with minorities. Given this potential bias, it is especially striking that whites report lower levels of satisfaction in neighborhoods with more minorities. Further, if the proportion of minority residents is correlated with other measures of quality of life that are not included in the analysis—perhaps neighborhoods with more white residents have better local services (e.g., trash pickup), have better amenities (e.g., parks and grocery stores), or are more centrally located—then the demonstrated effects of racial/ethnic composition may reflect these unmeasured aspects of quality of life rather than a pure “race” effect. Nonetheless, our analysis includes many of the local characteristics that have been shown by prior research to be important to local quality of life, most notably concerns about crime, school quality, property values, and neighborhood deterioration.
Despite the limitations of the analysis, our work provides new evidence that both social characteristics and race/ethnicity play important roles in shaping neighborhood satisfaction across racial/ethnic groups. The fact that socioeconomic conditions and local social problems are part of respondents’ negative evaluations of racially/ethnically mixed neighborhoods indicates that improving the quality of life in these neighborhoods may contribute to stable integration. However, the results also demonstrate that racial and ethnic composition—apart from other neighborhood characteristics—continues to be a critical factor in how individuals evaluate their neighborhoods. Thus, efforts to build and sustain stable integrated neighborhoods cannot rely solely on improving neighborhood conditions to be effective, but must also address how individuals, especially whites, view different-race/-ethnicity neighbors and how they perceive racially and ethnically mixed neighborhoods.
The authors gratefully acknowledge Robert Adelman, Nancy Denton, and the anonymous reviewers for their comments on this paper. In addition, the second author’s writing group colleagues—Tyrone Forman, Amanda Lewis, Omar McRoberts, and Beth Richie—provided invaluable assistance, feedback, and suggestions on earlier versions of this paper. Support for data collection came from the National Science Foundation (SES-0317740), the University of Michigan, the Ford Foundation, and the University of Illinois at Chicago. Data analysis was supported by a grant from the National Institute of Child Health and Human Development (1R03HD051677-01A1), as well as the Population Research Center at the University of Chicago. The article’s contents are solely the responsibility of the authors and do not necessarily represent the official views of the National Institute of Child Health and Human Development.
Despite concerns about social desirability, whites’ reported preferences for living in neighborhoods that are predominantly white are quite consistent (e.g., Charles 2000; Farley et al. 1994; Krysan and Farley 2002; Zubrinsky and Bobo 1996).
A separate set of studies has found that both hypothetical mobility preferences (e.g., Krysan 2002a) and actual mobility behavior (e.g., Crowder 2000) are influenced by local racial composition, net of social class–related characteristics. Most recently, Krysan and Bader (2007) analyzed blacks’ and whites’ preferences for living in actual communities in the Detroit area and demonstrated that whites are very unlikely to consider moving to communities that are not predominantly white, but that African Americans are unaffected by the percentage white in a community, net of social class characteristics. Finally, Krysan et al. (2009) used an innovative video experiment to show that net of local social class–related characteristics, whites view all-white neighborhoods as most desirable, and that the effect of racial composition is smaller among blacks, who identify racially mixed neighborhoods as most desirable.
For the analysis of the subsample of city residents, we use a single model that pools together whites, blacks, and Latinos because small sample sizes prevent us from reliably estimating the relationship between racial/ethnic composition and satisfaction separately for each group.
Specifically, respondents are weighted based on the overall nonresponse within their segment (block group), so that individuals residing in neighborhoods with particularly low segment-level response rates are weighted more heavily than those in neighborhoods with high segment-level response rates. Because our dependent variable of interest is neighborhood satisfaction, it is a reasonable assumption that individuals living in the same neighborhood (block group) will be highly correlated on our key dependent variable; thus, our weight accounts for differential rates of nonresponse across block groups and helps minimize, though of course does not eliminate, the effects of the low response rate in some segments.
Table 5 in Online Resource 1 shows the mean level of perceived social problems by the proportion of white, African American, and Latino residents in the neighborhoods. Indeed, residents of neighborhoods with higher proportions of minorities tend to report more social problems.
In supplemental analyses, we reestimated all models using tracts in place of block groups and found that the results are robust across both specifications.
We replicated these models with a linear measure of the percentage of Latino residents and found the same substantive results reported in the text. However, we were unable to estimate these models with a linear measure of the percentage of black residents because of the extremely small numbers of white respondents in Chicago neighborhoods who live with moderate or high proportions of black residents.
Table 6 shows descriptive statistics for our analytic sample, which comprises 622 respondents. About 17.7% of the sample was dropped because of missing information on perceived neighborhood social problems; most of this missing information came from the school dissatisfaction variable. The analyses were re-run without the school dissatisfaction measure, and the coefficients did not differ in sign or significance from those reported in the text.
It is notable that whites report less satisfaction in neighborhoods with a high proportion of African American residents but similar levels of satisfaction in neighborhoods with a low or moderate proportion of black residents. This finding may reflect the idea that whites are tolerant of coresidence with blacks as long as blacks do not exceed a certain proportion of neighborhood residents, as suggested by “tipping point” theories of whites’ residential choices.
Charles (2006) has emphasized the importance of Latinos’ nativity status in shaping residential preferences. She found that recent immigrants to the United States express the greatest preferences for living with coethnic neighbors, but that native-born Latinos and immigrants who have lived for a longer time in the United States report increased preferences for living with white neighbors. In the CAS sample, about two-thirds of the Latinos are immigrants. Because of the relatively small sample size (i.e., 204 Latinos), we are unable to systematically assess whether the relationship between racial/ethnic composition and satisfaction differs for immigrant versus native-born Latinos.