Targeting Placement Rate Disparities with Precision

Background and Purpose: With but a few exceptions, studies of placement rate disparity tend to focus on average Black child/White child placement rate differences. Although a useful first step in the problem-solving process, average disparity rates gloss over the fact that disparity rates vary by place and by sub-population. For example, placement rate disparity shrinks as one moves away from urban centers; placement rate disparity is smaller for younger children and larger for older children. Failure to consider this variation most likely results in one-size-fits-all strategies that miss the nuances that better define the opportunity for locally crafted solutions. Framing solutions around average disparity rates also limits how evaluation research is carried out, thereby increasing the chances any effect of innovative solutions will be missed. Building on the methods discussed earlier in the symposium, this paper describes an approach to identifying place- and population-based disparity rates. The added precision allows local policy-makers to consider customized solutions that are responsive to local challenges.

Methods: Using the multi-level Poisson framework described earlier in the symposium, we analyze a county’s overall placement rate. The overall placement rate is adjusted for poverty rates and an index of socio-ecological diversity. Although we expect a positive correlation between poverty and placement rates, we are interested in counties with similar socio-ecological features but vastly different placement rates. The Poisson model with random effects is used to obtain a county’s Empirical Bayes residual (EBR). The EBR measures the difference between the expected placement rate given the socio-ecological context and the observed placement rate. We repeat the analysis, first for both Black children and White children overall and then for children divided into sub-populations based on race and age.

Results: The results locate each county within a taxonomy that characterizes placement rate deviation. For White and Black children, we organize counties into categories defined as average, above average, or below-average placement rates. The grid we create has nine possibilities. Out of 63 counties, we found 21 counties with an above-average placement rate for White children and a below-average placement rate for Black children. At the other end of the continuum, we found 24 counties with a below-average placement rate for White children and an above-average placement rate for Black children. The remaining counties are dispersed throughout the grid. When the results are stratified by age groups, we find a similar pattern: counties cluster in the grid’s corners where the White placement rate is above average and the Black placement rate is below average and vice versa.

Conclusion and Implications: Measures of the average disparity rate foster one-size-fits-all strategies. We know, however, that local child welfare systems vary substantially in the resource base and other features. Those resource differences give rise to disparities that require a local response cognizant of system strengths and weaknesses. With this study, we show how precision targeting around the intersection of race and age brings greater nuance to the problem-solving local communities must take on if we hope to advance equity.