Black–White differences in Child Protective Services involvement: Evidence on the role of differential ‘risk’

Abstract

Black children and families are overrepresented in U.S. Child Protective Services (CPS) systems—the state and county systems responsible for receiving and responding to allegations of child maltreatment—relative to their representation in the U.S. population. They experience higher rates of CPS reports, investigations, substantiations, and child removals than White children (Children's Bureau, 2023, 2024; Edwards et al., 2021) and, conditional on out-of-home placement, spend more time in out-of-home care (Wulczyn, 2020). Moreover, while Black–White differences in CPS involvement have declined substantially over the past 2 decades (Myers et al., 2018; Roehrkasse, 2021; Wulczyn et al., 2023), they remain large: Black children are roughly twice as likely as White children to experience investigations, substantiations, and out-of-home placements over the course of childhood (Kim et al., 2017; Wildeman & Emanuel, 2014; Wildeman et al., 2014; Yi et al., 2023). Native American/American Indian children and families are also overrepresented at all levels of CPS involvement.1 Yet, because true underlying rates of child maltreatment are unknown, research has not established whether these disparities reflect disproportionate rates of maltreatment and, if not, whether they reflect under- or over-inclusion of either group.

It is, perhaps, unsurprising to observe disparities in CPS involvement, especially between Black and White populations. Black–White disparities are well documented for most indicators of health and social and economic wellbeing in the U.S., including income, poverty, wealth, employment, educational achievement and attainment, teen and nonmarital childbirth, family complexity and instability, morbidity and mortality, maternal and infant mortality, neighborhood quality, exposure to violence, and criminal justice involvement (Dagher & Linares, 2022; Darity & Mullen, 2022; Darity et al., 2022; Drake et al., 2023; National Academy of Sciences, Engineering, and Medicine, 2019; Office of the Assistant Secretary for Planning and Evaluation, 2022; Rothstein, 2017). Of particular note, poverty rates for Black children are more than 3 times those for White children (U.S. Census Bureau, 2023).

These disparities stem from historical and contemporary structural and institutional racism, oppression, and discrimination that have pervaded both public policy and social structure in the United States, and have manifested in bias against (differential treatment of or impact on) Black populations, relative to White populations (Darity & Mullen, 2022; Darity et al., 2022; Rothstein, 2017). As a result, compared to their White counterparts, Black populations have a higher likelihood of exposure to inferior educational experiences; segregated and poor-quality housing; poor schools, childcare facilities, and neighborhoods; and environmental toxins, limited and low-quality health and mental health services, violence, police surveillance, and voter suppression policies (Braveman et al., 2022; Yearby et al., 2022). These factors have, in turn, generated widespread concern among scholars, policymakers, advocates, and, in some cases, the public. Adverse trajectories and outcomes in these domains are associated with both child maltreatment and CPS involvement (Font & Maguire-Jack, 2020). Moreover, research has documented a strong inverse relation of income with child maltreatment and CPS involvement (Berger & Waldfogel, 2011; Font & Maguire-Jack, 2020), and most CPS-involved families are low income or poor (Berger & Slack, 2020).

Such Black–White disparities have resulted in Black populations being overrepresented in public systems, including the Supplemental Nutrition Assistance Program; Medicaid; Special Supplemental Nutrition Program for Women, Infants, and Children; Temporary Assistance for Needy Families program; Head Start and Early Head Start programs; Early Intervention Program; Supplemental Security Income program; and Child Support program. Black populations are also highly overrepresented in the criminal justice system. Whereas racial disparities in the factors driving disproportionate involvement in these systems have generated grave concern, disproportionate involvement in them—with the notable exceptions of the criminal justice system (e.g., Blumstein, 2014; Du, 2021; Roberts, 2007), child protective services (e.g., Dettlaff et al., 2020; Roberts, 2022), and, to a lesser extent, the child support system (Edin et al., 2019)—has not, itself, been widely raised as problematic, nor led to widespread calls to dismantle, disrupt, or abolish them (though many engender ongoing debate over their costs and benefits, and calls for reform).

Why, then, might disproportionality in CPS involvement be viewed as prima facie evidence of bias and discrimination toward Black families by CPS systems, while disproportionate involvement in most other social systems is not widely viewed as evidence of bias or discrimination by those systems? We posit that, rather than reflecting concern over disproportionality itself, this reflects concern over the orientation and (perceived or actual) impact of CPS. That is, much like the criminal justice system, CPS is frequently viewed as punitive. Indeed, CPS has been described as a family policing and family surveillance system (Dettlaff et al., 2020; Roberts, 2022). In contrast, most other social welfare systems—despite criticism of aspects thereof—are generally viewed as compensatory (at least in their purpose) attempts to address prior and ongoing sources and impacts of disadvantage and marginalization.

As such, we argue that whether one views disproportionality in CPS involvement—itself—as driven by problematic CPS-specific, or society-wide, biases depends largely upon whether one views CPS as punitive or compensatory; in other words, whether one views it as helping or harming (in the sense of assisting or punishing) children and families. Racially disproportionate system involvement is appropriate and productive when it compensates for prior and ongoing marginalization, oppression, disadvantage, and sources thereof; it is inappropriate and unproductive when it increases marginalization, oppression, and disadvantage, or otherwise harms involved populations. Thus, knowledge of both whether disparities in CPS involvement are consistent with disparities in actual child maltreatment, and whether CPS is punitive or compensatory, is essential to understanding and addressing them. Scholars, policymakers, advocates, CPS involved children and families, and citizens vary in their assessments of both factors.

We acknowledge that the child welfare literature entails arguments on both sides: that CPS is helpful or harmful to children and families. However, the most rigorous estimates of causal effects of CPS involvement (predominantly out-of-home placement) have produced inconsistent estimates in terms of direction, magnitude, and statistical significance (Bald et al., 2022; Berger et al., 2017; Doyle, 2007, 2008, 2013; Font et al., 2018, 2019, 2021; Grimon, 2023; Gross & Baron, 2022). As such, we view the evidence on whether CPS helps or harms children, on the whole, as inconclusive. Moreover, we suspect that CPS involvement has heterogenous effects on short- and long-term child safety and wellbeing that vary considerably by child and family circumstances, behaviors, and functioning, as well as by the levels of CPS involvement (investigation, substantiation, case opening, service receipt, child removal) children and families experience, their length of involvement, the types and quality of services they receive, how well those services meet their needs, their engagement in them, and the features of the local child welfare system itself. Notably, however, quantitative research has focused almost exclusively on the impacts of CPS involvement on children. We are aware of only one rigorous study to estimate plausibly causal effects on parental wellbeing. Grimon (2023) found CPS involvement to increase maternal participation in mental health and substance abuse treatment and decrease short-term CPS re-referrals, but also that out-of-home placement increases maternal incarceration in the short term and CPS re-referrals in the long term. She found little evidence that CPS involvement affects fathers in these domains, with the exception of decreasing re-referrals.

Whereas rigorous quantitative evidence has not conclusively determined whether CPS helps or harms children and families, a growing body of rigorous qualitative evidence indicates that parents perceive CPS as causing harm to them and their children, families, and communities through adversarial, stigmatizing, and traumatizing interactions, and by engaging in racist, discriminatory, and biased practices (Fong, 2020, 2023; Hanna & Rogers, 2022; Merritt, 2020, 2021; Miller et al., 2012; Roberts, 2022). The rich data collected in these studies provide unique insights into families’ lived experiences and highlight crucial failings in how CPS approaches, engages, and serves them, underscoring the need for system reorientation. They also raise salient concerns about potential bias within CPS. Yet, despite the contributions of qualitative studies to our understanding of the experiences of child welfare involved families, they do not, themselves, provide insight into whether CPS involvement causally impacts child safety or child and family wellbeing, nor whether bias within CPS causes racial disparities in CPS involvement. This simply reflects that identifying causal effects requires counterfactual conditions that facilitate rigorous comparison of outcomes between CPS-involved and non-involved families, or between CPS-involved Black and White families, and analytic strategies that allow the effects of CPS involvement or racial bias in CPS actions to be isolated from all other influencing factors. Such approaches are not relevant to qualitative inquiry.

In the remainder of this essay, we review the dominant hypotheses regarding the causes of Black–White disparities in CPS involvement and the quantitative evidence with respect to each. Throughout, we draw heavily on our recent work (Jones Harden & Berger, 2025), which provides a more extensive review of these issues. We argue that the evidence that differential risk is a key contributor to Black–White differences in CPS involvement is substantial, but that this does not discount the roles of racism and bias, either within or outside of CPS. That is, we find no compelling reason not to expect (or suspect) that both risk and bias contribute to racial disparities in CPS involvement at each level thereof. Moreover, we view differential risk as a direct result of differential historical and contemporary constraints, opportunities, experiences, and outcomes for Black and White populations. Given that CPS systems operate within the larger U.S. social structure, we assume that, at the very least, the same patterns of racism, discrimination, and bias operate within CPS as in other aspects of U.S. society.

The core drivers of racial disparities in CPS involvement in the U.S. are hypothesized to be differences in surveillance and racialized surveillance bias by potential reporters; racism and racial bias within CPS; racial disparities in risk for child maltreatment; and structural racism, discrimination, and bias at the societal level. We briefly summarize the evidence vis-à-vis each below. We emphasize that they need not be seen as competing or alternative hypotheses. Rather, they provide complementary insights into how various mechanisms may influence racial disparities at each stage of CPS involvement.

There are strong theoretical reasons to expect that greater surveillance of Black and, more generally, low-income communities, is a core driver of racial disparities in CPS involvement (Boyd, 2014, 2022; Dettlaff & Boyd, 2022; Fong, 2019), particularly at the report stage. Research has shown that these communities experience disproportionate surveillance by law enforcement (Boyles, 2015; Braga et al., 2019; Brunson & Gau, 2014), which accounts for about 20% of CPS reports (Children's Bureau, 2024). Black and low-income young children are also more likely to be referred to CPS by medical personnel (Edwards et al., 2023), which account for approximately 11% of CPS referrals (Children's Bureau, 2024). Disproportionate participation by Black families in social welfare benefit programs and social services may also be salient. Yet, because contemporary social welfare benefit programs rarely require in-person enrollment, assessment, or home visits, it is unlikely that benefit receipt substantially increases surveillance. Social service participation, including home visiting and early childhood education and care programs and, perhaps, public healthcare coverage and utilization, may increase exposure to mandated reporters. Social service and mental health providers jointly account for about 16% of reports and medical providers for about 11% (Children's Bureau, 2024). However, considerable evidence indicates that disparate surveillance (exposure to mandated reporters) by race explains only a small portion of racial disparities in CPS involvement (Chaffin & Bard, 2006; Drake et al., 2009, 2017; Jonson-Reid et al., 2009; Kim et al., 2018). Of additional concern, potential reporters may, all else equal, be more likely to report Black families than White families. Research has been unable to rigorously test this possibility (Harris, 2021). On the whole, then, whereas greater surveillance of Black and low-income families may play some role in racial disparities in CPS involvement, it is unlikely a driving factor. At the same time, there is little evidence on differences in decisions to report Black and White families, conditional on exposure to reporters and identical observed behaviors.

A second hypothesis posits that racism, bias, and discrimination within CPS primarily drives racial disparities therein. Rigorous quantitative research, here, is also limited. However, racial disparities are largest at the report stage, which occurs outside of CPS, and are mitigated rather than exacerbated at deeper levels of involvement, which occur within CPS (Baron et al., 2023; Drake et al., 2023, 2024; Myers et al., 2018). These findings suggest that CPS actions are not driving and may help allay Black–White disparities.2 Several studies have shed additional light on the potential role of racism and bias within CPS in screen-in (investigation), substantiation, and child removal decisions. For example, research comparing whether CPS decisions to investigate reports in the context of artificial intelligence-generated risk scores, which are intended to inform screen-in decisions, has indicated that caseworker decisions to open an investigation result in lesser racial disparities than differential risk scores would suggest (Cheng et al., 2022; Stapleton et al., 2022), implying that CPS actions may reduce racial disparities in investigations relative to reports.

Turning to substantiations, Font et al. (2012), using national data, found no differences in substantiation decisions (or caseworker ratings of risk or harm to the alleged victim) for Black and White families, net of case characteristics and maltreatment-related risk. Further, they found that Black caseworkers are more likely than White caseworkers to substantiate Black families, relative to White families, as well as to rate Black alleged victims as experiencing greater risk of harm than White alleged victims, suggesting that racial disparities in CPS actions do not reflect biased actions by White caseworkers toward Black families. In contrast, Dettlaff et al. (2011), using CPS administrative data from Texas from 2003 to 2005, found no difference in substantiation rates between Black and White families when adjusting only for sociodemographic and case characteristics, but found that Black families are more likely to be substantiated once caseworker assessments of family risk are also considered (the study did not consider caseworker race).

Finally, two particularly rigorous studies examined the potential role of racial bias in child removals. Baron and colleagues (2023) leveraged administrative data from Michigan and employed a quasi-experimental strategy to examine the impact of race-blind child removal decision-making, such that the team responsible for making removal decisions is unaware of the family's race, and found no impact on racial disparities in removals. Baron and colleagues (2024), also using a quasi-experimental approach and administrative data from Michigan, found higher removal rates for Black than White children, particularly among those at greatest risk of future maltreatment, and that this pattern reflects racial bias resulting in under-placement of White children at high risk of future maltreatment rather than over-placement of similarly high-risk Black children.

In sum, we find little compelling evidence that CPS investigation decisions exacerbate racial disparities, and some evidence that they may reduce such disparities. The limited quantitative evidence on substantiation decisions has produced mixed findings but suggests, at the very least, that Black–White substantiation disparities are not driven by disproportionate decisions to substantiate Black families by White caseworkers, nor by differences between Black and White caseworkers in decisions to substantiate White families. The evidence on child removals is also mixed, indicating both that race-blind removal decisions have no impact on racial disparities in removals and that White children in the most unsafe households are less likely to be removed than Black children in the most unsafe households. Together, these findings suggest that bias within CPS is unlikely a primary driver of racial disparities investigations and substantiations. It is less clear that this is the case for child removals but, if so—and, if out-of-home placement serves to increase safety for children at greatest risk of abuse or neglect—then racial disparities in CPS child removal decisions may be better protecting Black children than White children.

A third hypothesis speculates that disparities in CPS involvement are driven by disparities in risk for abuse and neglect. This is, perhaps, the most straightforward hypothesis to test, given that risk factors are more readily observed than bias and discrimination, which are frequently proxied by the difference by race in an outcome that is unexplained by observed factors (model covariates). Empirically, it is straightforward to compare the magnitudes of unadjusted Black–White differences to their magnitudes once adjusted for observed risk factors, and to assess whether they are partially or fully attenuated (with the remaining differential reflecting unobserved factors, including bias and discrimination).3

As noted above, research has documented large racial disparities—favoring White populations—in a range of risk factors for child maltreatment, including low-income and poverty status. Moreover, a growing body of research, leveraging experimental (Cancian et al., 2013) and quasi-experimental designs (Berger et al., 2017; Bullinger et al., 2023; Rittenhouse, 2023; Wildeman & Fallesen, 2017), has demonstrated a causal effect of income on both CPS involvement and child maltreatment indicators, including parental abusive and neglectful behaviors and child death. Economic scarcity is closely linked to a host of individual, family, community, and structural factors that have resulted from historical and contemporary racism and oppression and are associated with maltreatment risk (Skinner et al., 2021, 2023). That is, factors such as parental stress, health and mental health challenges, substance abuse challenges, family and neighborhood violence, criminal justice involvement, and compromised parenting practices and behaviors are disproportionately common among low-income and poor families (Karriker-Jaffee, 2013; Magnuson & Duncan, 2019). They have also consistently been found to largely explain Black–White differences in CPS involvement and actions (Barth et al., 2022; Coulton et al., 2007; Drake et al., 2011; Jonson-Reid et al., 2009; Jones-Harden & Slopen, 2022; Maguire-Jack et al., 2022; Molina et al., 2012; Wadsworth et al., 2016). In other words, accounting for such factors substantially reduces the magnitude of Black–White disparities in CPS involvement and actions and, in some cases, fully eliminates or reverses them (Drake et al., 2011, 2023, 2024; Jonson-Reid et al., 2009; Putnam-Hornstein et al., 2013, 2022).

A concern here is that low-income and poverty status, and associated maltreatment-related risk factors, serve to explain Black–White disparities in CPS involvement or actions because reporters or caseworkers confound economic scarcity with child maltreatment, particularly child neglect. We are unaware of research to explicitly examine this possibility for mandated or voluntary reporters. However, CPS screens out a large fraction of reports—more than 50% in 2022 (Children's Bureau, 2024)—because the information obtained fails to reach legal thresholds for investigation. Although there is no available evidence on the proportion of cases that are screened out due to their allegations reflecting poverty alone, it is possible that this pattern may reflect overreporting of families at low risk of maltreatment and, perhaps, primarily struggling with economic scarcity. If so, it appears that CPS systems decline to investigate many such cases. Studies have further indicated that mistaking economic scarcity alone for potential abuse or neglect is unlikely a widespread pattern within CPS (Palmer et al., 2024; Children's Bureau, 2024). Nonetheless, the data are clear that families from low-income backgrounds constitute the majority of those involved with CPS (Berger & Slack, 2020; Pelton, 2015).

Nationally, 15% of all substantiated families exhibit alcohol abuse, 24% drug abuse, and 27% domestic violence; by comparison, only 8% exhibit substandard, overcrowded, or unsafe housing, or homelessness (Children's Bureau, 2024). In addition, a study of neglect investigations in California indicated that the vast majority of investigated families—75% investigated for any type of neglect and 99% investigated for physical neglect—exhibit substance use challenges (41%), mental health challenges (18%), domestic violence (21%), and/or concurrent allegations of physical abuse, sexual abuse, parental absence, or abandonment (29%; Palmer et al., 2024). On the whole, the available evidence indicates that Black families are more likely than White families to be low-income and poor; that mental health, parental substance misuse, and domestic violence challenges are disproportionately common among low-income and poor populations (Karriker-Jaffee, 2013; Magnuson & Duncan, 2019); that these factors are closely linked to child abuse and neglect (Skinner et al., 2021, 2023); and that they are present among the vast majority of CPS involved families (Children's Bureau, 2024; Palmer et al., 2024). Furthermore, disparities by income and race in self-reports of maltreatment-related behaviors are similar in magnitude to disparities by income and race in CPS involvement (Baldwin et al., 2019; Slopen et al., 2016; Steele et al., 2016; Thomas & Waldfogel, 2022; Thomas et al., 2023).

With respect to the fourth hypothesis, no study of which we are aware has established whether structural racism, discrimination, and bias at the societal level directly explain racial disparities in CPS involvement and actions. This reflects that current research designs and methodologies are insufficient to isolate potential effects thereof (Boyd, 2022). Nonetheless, we see no grounds for dismissing the role of these factors. Rather, we posit that historical and contemporary structural racism, bias, and discrimination in U.S. society, including its policies and institutions, have resulted in racial disparities in income, poverty, and associated risk factors for child maltreatment. These disparities, in turn, largely account for Black–White differentials at all levels of CPS involvement. We interpret this as indicating that racial disparities in CPS involvement and actions are driven by society-wide, rather than CPS-specific, patterns of structural racism, bias, and discrimination, which have resulted in substantial Black–White differences in maltreatment-related risk that constitute the predominant mechanisms through which Black–White disproportionality in CPS involvement manifests.

Our assessment of the research literature leads to several implications for public policy. Because Black–White disparities in CPS involvement reflect underlying differences in maltreatment risk that are driven by factors outside of CPS, it is unlikely that CPS reform alone—barring purposeful differential actions by race, which may disproportionately comprise Black children's safety—will substantially reduce Black–White disparities in CPS involvement and actions. Rather, the paramount challenge for CPS is to ensure—in both reality and the perceptions of CPS-involved children and families and the public—that it helps rather than harms children and families. This will require wholesale system reorientation to demonstrate that CPS is a supportive and trusted partner with which to ensure children's safety and promote their wellbeing. To this end, CPS must provide interventions that engage and retain families, and that explicitly value and benefit them. This includes delivering concrete resources and supports, as well as effective parenting interventions to facilitate raising children in safe and stable homes. It will also require substantial public investment to expand availability of and access to substance abuse, mental health, domestic violence, and other services to promote the well-being of parents and children, as well as increased flexibility in use of current CPS funding streams. Given historical perceptions of CPS, and system capacity and budgetary constraints, such change is likely to be difficult.

Moreso than CPS reform, substantially reducing racial disparities in CPS involvement will require substantial investment outside of CPS systems. While not a panacea, more generous income supports, such as a fully refundable universal monthly Child Tax Credit, have the potential to substantially reduce child maltreatment, CPS involvement, and racial disparities therein (Pac et al., 2023). Income transfers are straightforward to deliver, efficient, and have relatively immediate effects and, while expensive, their long-term societal benefits far outweigh their costs (Garfinkel et al., 2022). Although economic supports will likely disproportionately reduce maltreatment among the lowest-risk families rather than those at most risk of substantiation and child removal, reducing the proportion of such families who are CPS involved can both spare low-risk families from an intrusive investigation process and free up CPS resources to be concentrated on the highest risk families. Beyond income supports, reducing maltreatment, CPS involvement, and disparities therein will require substantial public investment to ensure that high-quality substance abuse, mental health, domestic violence, and parenting programs and services are widely available and accessible before families become CPS involved. Notably, such programs are relatively expensive and characterized by limited take-up, engagement, retention, and immediate success. However, their long-term benefits are likely to outweigh their costs. In short, there is no cheap or easy solution for reducing child maltreatment, CPS involvement, and racial disparities therein. Doing so will require extensive public and political will and investment, both within CPS systems and, especially, beyond them.