Abstract
In order to unpack whether and how self-rated health of Hispanics is linked to residential segregation from non-Hispanic whites, this study employs multi-level analysis combining data from the 1997–2002 National Survey of America's Families (n = 16,753 Hispanic respondents across 82 metropolitan areas) with metropolitan area segregation scores derived from restricted-use Census 2000 data. Separate analyses by nativity (U.S.-born vs. foreign-born Hispanics) and ethnic subgroup (comparing Mexicans, Puerto Ricans, and Cubans) are also conducted. Net of individual-level controls, findings reveal a small significant negative effect of segregation on health for all Hispanics and for Mexicans in the sample. Nativity does not have a significant interaction effect. Evidence of a positive segregation/health link is found for Cubans, challenging the assumption that segregation is always bad for minority health. This research highlights the value of multi-level analysis in examining segregation as a social determinant of health, and reveals key Hispanic subgroup differences.
Notes
1 Other research questions of the broader project include: How do the links between health and segregation for U.S. Hispanics compare to those for African Americans? What are the differences across all study groups regarding access to health insurance, or having a usual source of care? Findings are available upon request.
2 The 12 NSAF sample states are: Alabama, California, Colorado, Massachusetts, Michigan, Minnesota, Florida, New Jersey, New York, Texas, Washington, and Wisconsin.
3 Despite their U.S. citizenship but because they are long distance migrants, Puerto Ricans who were born in Puerto Rico are coded as foreign-born Hispanics in this study.
4 To assuage fears of over-specifying the models with too many health-related variables, I conducted each regression with varying combinations of the variables. Including some or all of the health-related variables made little difference to the results; therefore, they are all present in the models reported here.
5 The linearity of dissimilarity was tested in analyses not reported here by coding it as a categorical variable differentiating between low scores (0–0.3), moderate scores (0.3–0.6) and high scores (0.6–1.0). Results confirm that operationalization of dissimilarity as a continuous variable is appropriate.
6 Results with all the stepwise regression results for each subgroup are available upon request.