Findings from Two Longitudinal Surveys on Race, Ethnicity, and Neighborhood Out-Migration

Peter J. Mateyka, Survey Statistician, Housing and Demographic Analysis Division

Housing assistance programs have had success at improving neighborhoods and the lives of individuals receiving assistance, but less success reducing racial and ethnic segregation.

“Validating the White Flight Hypothesis: Neighborhood Racial Composition and Out-Migration in Two Longitudinal Surveys” is a recent article that uses data from two longitudinal surveys, the Panel Study of Income Dynamics (PSID) and the Survey of Income and Program Participation (SIPP), to compare probabilities of neighborhood out-migration for Whites, Blacks, Hispanics, and Asians by neighborhood racial and ethnic composition. “White flight,” or the tendency of White households to move out of neighborhoods as the proportion of racial and ethnic minorities living in the neighborhood increases, is a cornerstone of theories of racial and ethnic residential segregation. Few studies, however, have empirically tested this assumption, and those that have rely almost entirely on PSID. Although PSID is a rich source of longitudinal data on the sociodemographic and economic characteristics of U.S. households, it is based largely on a sample of households originally drawn in the 1960s and their descendants. Research using PSID data has consistently confirmed that White households frequently move out as the number of minorities in a neighborhood increases, but questions persist about how generalizable these findings are to the United States, especially considering the nation’s increasing racial and ethnic diversity since the 1960s.

This study uses data from PSID and internal use SIPP, a nationally representative, recently drawn sample of U.S. households, to model neighborhood out-migration. Using a SIPP internal address file, the researchers attached neighborhood data to SIPP households — the first researchers to use SIPP data in this way. The study has two goals: (1) to replicate past PSID findings on race and neighborhood out-migration using data from both PSID and SIPP to gain insight into racial and ethnic inequality, and (2) to examine contemporary samples of White, Black, Hispanic, and Asian households to expand models of "flight" beyond White populations. The findings have important implications for HUD's commitment to fair housing through the Affirmatively Furthering Fair Housing (AFFH) mandate, which requires HUD to consider residential segregation in the implementation of its programs and take decisive steps to reduce racial and ethnic residential segregation.

Background and Study Design

“White flight,” or Whites’ aversion to living in neighborhoods with sizable numbers of minority residents, has been a central component of theories about racial and ethnic residential segregation for several decades. Some of the most influential work on this topic has focused on White households and how small differences in their preferences for neighborhood diversity could lead to large-scale neighborhood change. In this view, White residents have varying preferences for the number of minority neighbors they will tolerate, although the neighborhood composition eventually reaches a “tipping point” at which all Whites move out of the neighborhood. This process begins when minorities move into a neighborhood and the least-tolerant White residents move out. As more minorities move in to replace these White residents, neighborhood turnover continues until the neighborhood reaches its tipping point.

The researchers sought to empirically verify and extend the White flight hypothesis using longitudinal data from PSID and SIPP, which provide information on household characteristics at various times along with information about the households' subsequent migration behavior. PSID routinely links sampled households to geographic codes for their neighborhoods, allowing researchers to link contextual information on neighborhoods to the household data. This process allows researchers to use PSID to study how neighborhoods shape social outcomes. Although SIPP does not link its household data to geographic codes as part of the normal production process, the researchers were able to obtain an in-house geolinked file covering several years of SIPP data. These data allowed the researchers to replicate and extend previous analyses of White flight that used PSID data.

The authors used the longitudinal nature of the PSID and SIPP surveys to construct 3-month migration intervals out of data from 2009 to 2011. Using geolinked data, the authors identified the racial and ethnic neighborhood shares at the beginning of each migration period and then identified whether households moved over the interval. Through multivariate models, the authors controlled for other factors associated with neighborhood moves, including homeownership, age, education, income, marital status, presence of children, nativity status, and region of residence. To compare estimates from the two surveys, the authors overlap the survey periods and migration intervals, use data on race/ethnic neighborhood composition, and code the independent variables measuring demographic and socioeconomic characteristics similarly. The four main explanatory variables in the analysis indicate the proportion of minority residents of the neighborhood (that is, those who are not non-Hispanic Whites) as well as the proportion of non-Hispanic Black, Hispanic, and non-Hispanic Asian neighborhood residents. The statistical models predicted the probability of neighborhood out-migration for each of the four race and ethnic groups by the main four explanatory variables, controlling for basic characteristics of householders. The researchers used both PSID and SIPP to predict neighborhood out-migration for Whites, Blacks, and Hispanics, whereas only SIPP was used to predict out-migration for Asians because the PSID sample size was deemed too small to generate reliable estimates.

Research Findings

The researchers found that, for White households, SIPP and PSID data show substantively similar migration responses to neighborhood minority racial and ethnic concentrations, with the likelihood of out-migration increasing as neighborhood minority shares grow, although these effects were weaker in SIPP than in PSID. The relationship was nonlinear and most apparent in predominantly White neighborhoods — that is, when the percentage of minorities (non-White residents) in a neighborhood increases from 0 to 20 percent. Because most White households live in neighborhoods with few minorities, this finding suggests that in predominantly White neighborhoods, small increases in the share of minority residents can spur out-migration for some White households. In neighborhoods in which the minority share exceeds 20 percent, the rate of out-migration slowed. When the researchers examined Whites' responses to neighborhood proportions of Blacks, Hispanics, and Asians alone rather than the minority population as a whole, the results were similar: as neighborhood shares of each group increased from 0 to 20 percent, White households were more likely to out-migrate.

The researchers then replicated models of flight for Black, Hispanic, and Asian residents. The results for Black householders were comparable in both PSID and SIPP, indicating that out-migration of Black households increases as the neighborhood share of Hispanic residents rises from 0 to 20 percent. For Hispanic households, however, SIPP and PSID results diverged, potentially reflecting differences in the representativeness of the samples. In the PSID sample, Hispanic households showed an increased probability of leaving neighborhoods as the Hispanic share grew from 0 to 40 percent, whereas the SIPP sample showed a decrease in the probability of Hispanic out-migration for the same neighborhood shares of Hispanics. The PSID and SIPP samples both showed increasing out-migration of Hispanic households as Black neighborhood shares increased, but the relationship was stronger in PSID than in SIPP. In addition, SIPP results reveal that the mobility behaviors of Asian households are largely indifferent to neighborhood racial composition.

Conclusions and Further Research

The similarity of the estimates for White and Black households validates researchers’ longstanding reliance on PSID for understanding the processes and consequences of neighborhood inequality. More broadly, the research highlights the continued salience of race in shaping the migration decision making process and the broader spatial foundations that shape inequality and mobility. Earlier work appearing in PD&R Edge documented the difficulties HUD’s housing assistance programs have encountered in reducing racial and ethnic segregation. Housing assistance programs have successfully improved neighborhoods and the lives of individuals receiving assistance, but these programs have been less successful in reducing racial and ethnic segregation. Housing assistance alone cannot overcome decades of racial and ethnic inequalities in access to quality housing, but this research highlights additional barriers to efforts to reduce residential segregation, including the active resistance of some White households, who may resort to moving to new neighborhoods to avoid living with minorities. “White flight,” combined with supplemental evidence of discrimination in the search for housing, differential access to credit, and restrictive zoning laws points to substantial obstacles for HUD in achieving the stated goals of the AFFH. This research also finds evidence of avoidance of minority neighborhoods by minorities, which suggests a need for more research on the underpinnings of racial and ethnic residential segregation and access to quality housing in an increasingly diverse nation.