Application of Exit Disparity Using Targeting

Background and Purpose: When the child welfare system delivers prevention or intervention services for children and families with a goal in mind, where/whom should they prioritize? Considering the limited budget resources and service availability, how do agencies target children/families and target areas? Although the child welfare system is moving towards evidence-based interventions, little attention is devoted to understanding whether service allocations are related to needs or outcomes. Building on the methods discussed in the first two papers, this paper describes a targeting approach based on place- and population-based disparity rates.

Methods: The study used child-level covariates, Census data, placement rates, race-specific child poverty rates, and urbanicity to investigate exit disparities for targeting. The county random effects model was used to calculate county Empirical Bayes (EB) residuals for permanent exits, with stratified models used to calculate sub-group estimates based on race/ethnicity and age groups. To address censoring, we relied on the discrete time hazard model. In addition, the state heterogeneity was measured using state fixed effects.

Results: First, overall county EB estimates were calculated to show the general pattern of county variations. These results show whether exit rates are above or below the state average and whether the county-level deviations are statistically significant. The results locate each county within a taxonomy that characterizes exit rate deviation. Next, county race and age group specific estimates were calculated. The results show the race and county specific deviations (For example, in a given state, the exit rates for Black children deviate from the overall Black child exit rate). From these data, we identify where the overall exit rate is below the average because the Black child exit rate is particularly slow. We also find places where the exit rate would be faster but for the slow rate of exit reported among White children. We find the same results when child’s age is added to the analysis. Finally, we found that state dynamics differ as well.

Conclusion and Implications: With the growing emphasis on evidence-based interventions and the limited resources and opportunities within child welfare, this research provides a way to target services with a clear sense of local circumstances. Our research deals with the question of whether and how much counties deviate from average exit rates after controlling for other factors that also influence permanency exits. Local child welfare systems vary substantially in their resource base and other features. These differences require a local response that is cognizant of system strengths and weaknesses. With this study, precision targeting around the intersection of race and age brings greater nuance to local problem-solving using the customized solutions needed to advance equity.

Finally, across all three papers, we recognize that reducing time in care is an important goal. Ultimately, our main concern sits with the well-being of children. Children need stable caregiving to thrive. We close our symposium with an invitation to discuss how we move the well-being of children to the center of the equity conversation.