Visualizing Racial Disparity in Child Protective Services Using Racially/Ethnically Disaggregated State-Level Scatterplots

Background and Purpose:

Since the publication of the Third Wave of the National Incidence Studies in 1996 there has been consistent and focused attention on racial disparity in Child Protective Services. Black children have consistently been reported at nearly twice the rate of White children, while Latinx children have generally been reported at rates roughly similar to White children. A number of multivariate studies have focused mainly on Black/White disparity, controlling for factors like child age, gender, family factors and economic variables. The majority of these studies find that once these factors are considered, disparity reduces, disappears, or sometimes even inverts. Many of these studies suffer from drawbacks, however. First, multivariate models can be difficult for many people, including practitioners and decision makers, to understand. Second, many of these models fail to disaggregate samples and variables by race, causing further interpretative difficulties and introducing systematic error. Finally, race is often used as a control variable, which is seen by many as undesirable.

Methods:

We present scatterplots for all 50 states and the District of Columbia, showing the distribution of counties across median same-race county income and single-race child maltreatment report rates. These scatterplots use data from the Child File, a part of the National Child Abuse and Neglect Data System, managed using the Report and Placement Integrated Data System (RAPIDS) and combined with data from the “SDoH4CWS” suite of programs and datasets. RAPIDS and SDoH4CWS are utilities developed by consortiums of universities and available free on LDBase. Consistent with criticisms of prior research, race/ethnicity is not used as a control variable. Instead, scatterplots are racially/ethnically disaggregated, with different datapoints for Black, Hispanic and White residents of each county. For example, there are three datapoints for San Diego County (Black, Hispanic and White), the data point for Black residents of San Diego County using Black report rates and the median household income for Black residents.

Results:

The scatterplots confirm existing findings in the broader literature. Black and Hispanic county datapoints show far lower median income than White datapoints. For most states, Black and White county residents fall on the same trend line. While raw maltreatment rates for Black residents are higher, these higher rates are aligned with their position on a more economically disadvantaged portion of the scatterplot. Third, the so-called “Hispanic Paradox” is strongly evident in virtually all states, with Hispanic residents showing maltreatment rates considerably below the trend line established by Black and White data points.

Conclusion and Implications:

This study shows the utility of the RAPIDS and SDoH4CWS data suites in quickly producing highly interpretable state level findings regarding racial disproportionality. While more interpretable, the findings are also entirely consistent with the majority of pre-existing studies using less accessible multivariate approaches. Practical implications are consistent with those drawn by a number of recent commentators and researchers: Reducing economic stressors in poor families holds great promise in reducing maltreatment in general and racial disproportionality in particular.