Testing the Ecological Dissimilarity Hypothesis: The Case of Racial Disparities in Foster Care Admissions

Background and Purpose: The issue of racial disparity in the child welfare system is gaining traction and there is a growing need to understand why children of color are over-represented. Our analysis of the issue is guided by the ecological dissimilarity hypothesis. Used to explain black/white differences in crime data, the ecological dissimilarity hypothesis relies on comparisons of ecologically similar but racially distinct areas to untangle compositional from contextual effects. In the case of placement disparities, we predict that placements rates for white children living in areas characterized by high white poverty rates will be close to placement rates for black children living in areas characterized by high black child poverty rates because the context of poverty levels the risk of placement.

Method: First admissions from 1,034 counties in 14 states from 2000 to 2005 were counted (n=382,772). Data from the U.S. Census were used to derive admission rates/1,000 for black and white children and to describe the county population structure using poverty, family structure, adult educations levels, and residential stability. These attributes of the population were also disaggregated for the black population and the white population.

The county placement rate disparity was calculated as the black child placement rate per 1000 divided by the white child placement rate per 1000. The placement rate disparity was then used as the dependent variable in a series of HLM and GHLM models that assessed the relationship between county population characteristics and the disparity rate. The HLM models provided us with a way to adjust for the nested county samples (e.g., counties were nested within states) and test for slope and intercept differences

Results: There are three primary findings: disparity is lower in counties with higher poverty rates; the slope of the regression line that describes the relationship between white child poverty and white child placement rates shows that white placement rates are higher in areas with higher white poverty (as expected); and, for blacks, there was no significant relationship between black child poverty and the black child placement rate (slope = 0). In addition, results from the HLM model suggest that the findings are consistent across the states.

Conclusions and Implications: No prior research that examines placement rate disparity across multiple jurisdictions has been reported. The findings support the ecological dissimilarity hypothesis in that disparity is lower in ecologically similar but racially distinct areas. However, the finding is limited to poor areas. To the extent that disparities arise because of the differential treatment of black children relative to white children, these findings suggest that differential treatment may be context sensitive such that the treatment of black children in non-poor areas merits particular attention. The latter conclusion is highlighted by the fact that black child poverty and black child placement rates are not related, a finding that overturns the conventional wisdom. With regard to racial disparities, this finding raises a number of critical questions including the possibility that black children are actually underserved in certain contexts.