Community Context, Race, and Foster Care Placement: A Multi-Level Analysis
Method: Using child welfare administrative and U.S. Census tract level data, we analyzed the decision to place children in foster care following a maltreatment investigation (n=22,032) in 34 Chicago community areas over a three-year period. A logistic multi-level regression equation modeled the relationship between individual characteristics (level-one), community characteristics (level-two), and selected cross-level interactions on the likelihood that children would enter foster care. Level-one predictors included type of alleged maltreatment, age, gender, and race of investigated children. Three dichotomous level-two predictors indicated if communities had greater than the city average number of residents who reported that they were African American, Hispanic, or living in poverty. Modeling proceeded in four blocks. Block one, a null model, assessed the degree to which variation in removal decisions correlated with nesting in community areas. The models in blocks two through four, respectively, introduced individual-, community-, and cross-level interaction terms. BIC statistics were used to compare models for relative goodness of fit.
Results: The intraclass correlation found that approximately 11 percent of the variation in removal decisions was related to the community area in which the investigation occurred. After covariate adjustment, results showed that African American children had 1.60 higher odds of entering foster care than their White counterparts. Further, living in communities with a greater than average proportion of African American residents was associated with 1.64 higher odds of removal than communities with lower than average proportion of African American residents. Interestingly, cross-level interactions suggested that African American children living in communities with higher than average number of African American respondents were relatively less likely than African American children living in communities with less than average number of African American residents to enter care. Covariates representing the effects of being Hispanic or living in a community with greater than average proportion of Hispanic residents or residents living in poverty were not significant.
Implications: Results suggest that the decision to place children in foster care may in part be influenced by individual race as well as the community racial composition. In particular, although African American children were more likely than White children to enter care, this effect was vitiated as proportion of African American community residents increased.