10
Child Maltreatment
Children who experience maltreatment and become involved with the child welfare system are at elevated risk for intergenerational poverty. This chapter provides a demographic overview of maltreatment and child welfare involvement and reviews the correlational evidence linking childhood maltreatment and child welfare involvement to adult poverty-related outcomes. It then reviews what is known about the factors that place children at risk of maltreatment and involvement with the child welfare system. Finally, it evaluates the evidence on effective policies and programs to reduce child maltreatment and child welfare system involvement and on their potential for reducing intergenerational poverty.
WHICH CHILDREN ARE INVOLVED WITH THE CHILD WELFARE SYSTEM?
Involvement with the child welfare system is surprisingly common in the United States. In 2020, the most recent year for which federal data are available (Administration for Children and Families, 2022a), approximately 7.2 million children were reported to child protective services (CPS), and of these about 3.1 million children were investigated or assessed by CPS. These children received foster care services in only about 3% (113,000) of these cases. In thinking about the causes of maltreatment and possible approaches to address it, it is important to bear in mind that most (more than three-quarters) of these investigated cases involved neglect, while the remaining quarter involved some form of alleged abuse—physical, sexual, or psychological.
Expressing the scope of child maltreatment in another way, each year the child welfare system substantiates maltreatment for just under 1% of U.S. children ages 0–17. As shown in Figure 10-1, this overall maltreatment rate has changed relatively little since 2008. But the figure also shows very different rates across racial/ethnic groups, with the highest rates among Black and Native American children and the lowest rates among Asian children. Subgroup rates have changed only in the case of Native American children and, in their cases rates have risen in the last 10 years and now exceed rates for Black children.
NOTES: Children are ages 0–17. The data in this table show rates of maltreatment based on investigations and assessments by CPS that found the child to be a victim of one or more types of maltreatment. The decrease in the rate of victims in 2020 is due in part to the decrease in the number of screened-in referrals during the March through June period. Additional technical notes are available in the annual reports titled Child Maltreatment. These reports are available at https://www.acf.hhs.gov/cb/data-research/child-maltreatment.
SOURCE: Data from Administration for Children and Families, National Child Abuse and Neglect Data System, as reported in Childstats.gov
These kinds of single-year estimates understate the childhood-wide risk of maltreatment. Wildeman et al. (2014) use data from 2011 and estimate that 12.5% of U.S. children will be involved with a confirmed case of maltreatment by 18 years of age. These rates are considerably higher for Black (20.9%), Native American (14.5%),1 and Latino children (13.0%) than for White (10.7%) or Asian/Pacific Islander (3.8%) children. The risk for childhood maltreatment is highest in the first few years of life.2
Disparities by race and ethnicity are strongly correlated with subgroup differences in exposure to poverty and hardship (Eckenrode et al., 2014; Kim & Drake, 2018; Pelton, 2015; Putnam-Hornstein et al., 2021; Thomas & Waldfogel, 2022). Empirical evidence suggests that persistent racial/ethnic bias among actors within the child welfare system may expose Black, Native American, and Latino children to a greater risk of child welfare involvement (Dettlaff et al., 2011; Font et al., 2012).
Most children who are involved with the child welfare system disproportionately grow up in low-income families and are at elevated risk of poverty and other adverse outcomes in adulthood (Bunting et al., 2018; Currie & Widom, 2010; Mersky & Topitzes, 2010). Hence, they are a group at high risk of intergenerational poverty.
CHILD MALTREATMENT AND CHILD WELFARE SYSTEM INVOLVEMENT AS DRIVERS OF INTERGENERATIONAL POVERTY
Adult Correlates of Childhood Maltreatment
A large literature provides descriptive evidence that children who are maltreated (abused, neglected, or in other ways brought to the attention of the child welfare system) fare worse than their non-maltreated or non-child welfare-involved peers. Studies attempt to address selection by controlling for observable characteristics, which leaves open the possibility that outcome differences are partly or entirely caused by unobserved characteristics (Child Welfare Information Gateway, 2019).
___________________
1 The childhood-wide risk estimates for Native American children are based on relatively small sample sizes, which may account for their different relative ranking in the single-year and childhood-wide estimates.
2 Higher still are the estimates of the childhood-wide chance (37.4% overall) of any involvement with the welfare system, which can include cases reported to CPS but not substantiated by an investigation (H. Kim et al., 2017). Racial differences in estimates of lifetime involvement are also substantial. For example, half (53%) of Black children have at least some involvement, whether substantiated or not, with the child welfare system during their childhoods. At 32%, the rate of involvement is also elevated for Latino children. Rates for White (28%), Native American (23%) and Asian/Pacific Islander (10%) children are smaller.
In general, these descriptive studies consistently find that childhood abuse or neglect is predictive of poorer adult outcomes (see Appendix C: Chapter 10 for a more complete review). For example, Bunting et al. (2018) concluded that maltreatment in childhood is related to adverse economic outcomes, such as economic inactivity and income reductions, beyond the contribution of correlated poverty-related risk factors such as maternal education and welfare receipt during childhood. Examining intergenerational income mobility and child maltreatment, Bullinger et al. (2022) found that children residing in counties with lower child maltreatment report rates had a higher chance of intergenerational income mobility than those in counties with higher maltreatment rates. Zielinski (2009) found that adults who had experienced maltreatment as children had lower incomes, elevated utilization of Medicaid, and higher rates of unemployment than those without childhood maltreatment experiences. Notably, adults who had histories of multiple types of maltreatment were three times as likely as those who had not experienced maltreatment to live in poverty.
Consequences of Involvement with Child Protective Services
Only a few studies have used more rigorous methods to compare maltreated children with other children who are most similar to them. For example, Currie and Tekin (2012), using propensity score matching and twin comparisons, found that children who self-reported having been abused or neglected were much more likely to be involved in crime as adults than their non-abused or neglected counterparts. Currie and Widom (2010) used a case-control design and found that children who were substantiated by CPS to have been abused or neglected had lower levels of education, employment, earnings, and assets as adults. Although the methods used in these two studies are stronger than those used in purely correlational studies, at best their results should be considered suggestive of possible causal impacts. Moreover, neither of these studies (particularly Currie & Widom, 2010) was able to separate effects of maltreatment from effects of CPS involvement. For the same reasons that it is challenging to separate maltreatment effects from CPS involvement, it is also difficult to separate the effects of different degrees of CPS involvement. For example, estimating the effect of referral to CPS separate from the impact of substantiation of an allegation is complicated by the fact that substantiation likely involves more severe maltreatment. One exception to this is the decision to remove children from the home, for which there is causal evidence, which we summarize below.
Out-of-Home Care (Foster Care)
Among children with a substantiated case of maltreatment, roughly 3% receive foster care services involving relatives (kinship care), unrelated foster parents, or congregate care. Although they make up only a small share of those who experience maltreatment or child welfare involvement, foster care children are a very important group, as their families tend to be the most disadvantaged and the children themselves tend to face numerous challenges in adulthood.3
There is a large descriptive literature on children who are removed from their homes and placed into some form of out-of-home care. In general, children placed in congregate care fare worse than those placed with families through foster care (Lee et al., 2011). However, as with research on the more general child welfare population, the challenge in this research is to estimate the causal effects of out-of-home placement as distinct from the effects of factors that precipitated the placement. For that reason, a review of descriptive studies is provided in Appendix C: Chapter 10.
The strongest studies of the causal impacts of foster-care placements have used variation in the propensity of CPS investigators to remove children from their parental homes to identify the causal effect of out-of-home placement on later outcomes. In these studies, case assignment to investigators is generally based on their availability rather than characteristics of the case, so variation in investigators’ propensities for out-of-home placements can be leveraged to identify causal impacts of the placements themselves. At the same time, it is important to note that these studies provide evidence about the effect placement has on a child who is at the margin between being removed and not being removed, but they are not informative as to the effect of placement on children whose circumstances at home are so severe that they would prompt placement regardless of the individual investigator or judge.
Taken together, these studies yield a mixed picture on whether out-of-home placements lead to better or worse child outcomes. Doyle (2007, 2008), who originated this method, find detrimental impacts. Specifically, using data from Illinois he found that removal from home between ages 5 and 13 leads to increased delinquency and arrests and reduced labor force participation.
Other studies have found beneficial impacts. Bald et al. (2022) use the investigator method with administrative data from Rhode Island to study outcomes for children who were investigated by CPS before age six. They find that removal leads to significantly higher test scores (an average gain
___________________
3 Another important, and very disadvantaged, group is youth who are involved with both the child welfare and juvenile justice systems, who may be placed out of home by both systems (Herz et al., 2012; Hirsch et al., 2018).
of 1.4 standard deviations in math and reading) and significantly reduced grade repetition (an average reduction of 42.6 percentage points) for girls, with no significant effects for boys.4 The same authors also examine outcomes for children removed at age 6 or later and, in contrast with their results for children removed at younger ages, find no significant effects for girls or boys, with some nonsignificant negative results on educational outcomes for boys.
Baron and Gross (2022) use the investigator method with administrative data from Michigan to study the effect removal has on involvement in crime. They found that foster care placement reduced adult arrests, convictions, and incarceration. Gross and Baron (2022), in a similar study, find that foster care placement led to improved safety and educational outcomes.
Impact studies based on investigator case assignment draw data from only three states and from different time periods, which is an important limitation if the effects of removal vary by state and time. For example, the likelihood that a child will be removed in a given state and period may vary considerably by factors such as whether there has been a recent high-profile child welfare case (e.g., a death), whether the state is under court mandate(s) or supervision, and the capacity, service availability, and quality of preventive and child welfare system services available, potentially including the size and composition of out-of-home placement slots.
More generally, it is probably overly simplistic to estimate a single effect of removal, since the effects on well-being likely vary by the types and quality of services provided, severity of maltreatment experienced, and length, stability, and type of out-of-home placement, as well as child and provider characteristics. Unfortunately, these issues have not been subject to rigorous examination. In addition, it is possible that the impact of out-of-home placement may differ by race and ethnicity if, for example, thresholds for removal (chronicity and/or severity of exposure to maltreatment), services received, types or quality of placements, or the “match” with placement settings and contexts (e.g., neighborhoods, schools) differ by race and ethnicity. Nevertheless, the empirical research to date has not established the extent to which the effects of removal vary by race and ethnicity. Doyle (2007, 2008) provides separate estimates for Black and Latino youth, but these constitute the bulk of his sample and thus, unsurprisingly, the estimates are similar to those for the full sample.
Conclusion 10-1: Children who have been maltreated and (or) involved with child welfare are at elevated risk of intergenerational poverty.
___________________
4 A disadvantage of the investigator-based method is that it can lead to very large confidence intervals for its estimates. This is the case with Bald et al.’s (2022) estimates of impacts on achievement and grade retention.
However, high-quality research provides mixed evidence on the effects of foster care (occurring in only 3% of all child welfare cases) on subsequent outcomes in adolescence and adulthood and almost no evidence regarding the impact of child protective services more generally.
Factors Leading to Child Welfare Involvement
While the literature on the intergenerational impacts of child maltreatment is inconclusive, a parallel literature on the causes of child welfare involvement is more definitive and points to a number of policy approaches that appear successful in reducing involvement. The text portion of this chapter highlights four of the most promising policy-related approaches—income support, Medicaid, nutrition programs, and some community-focused prevention policies. Appendix C: Chapter 10 reviews the literature on four additional possible policy approaches: minimum wages, cash welfare programs, home visitation programs, and early care and education programs.
Parental Income and Employment
An extensive literature has documented that income and poverty are highly correlated with child maltreatment and child welfare system involvement. Indeed, across advanced industrialized countries, low income and poverty are the strongest and most consistent risk factors for child maltreatment and child welfare system involvement, and these relationships are particularly strong with respect to child neglect (Berger & Waldfogel, 2011). A growing body of research, predominantly from the United States, suggests that relatively modest increases in income, particularly among lower-income families, can lead to substantial reductions in child welfare system involvement (Font & Maguire-Jack, 2020b).
Cancian et al. (2013) capitalize on a policy experiment in Wisconsin in 1997–1998, which randomized families receiving Temporary Assistance for Needy Families into two groups, one of which received greater child support benefits and one of which received less. The differences in child support received between the two groups were modest, ranging between $101 and $180 per year. Despite this modest difference in income, the study found that families with more child support income were about 10%, or 2 percentage points, less likely to have a substantiated report of child maltreatment over the course of 2 years. A potential problem with an income-based interpretation of these results is that child support receipt may signal differential willingness or ability of fathers to be involved with their children. A review of the impacts of another major welfare program—Aid to Families with Dependent Children—is provided in Appendix C: Chapter 10.
Other research has capitalized on natural experiments in which policy benefits differ over time and place, providing exogenous sources of variation in families’ income. Berger et al. (2017) draw on state-level differences in the generosity of Earned Income Tax Credit (EITC) benefits and find that $1,000 in additional income was associated with an 8% to 10% (0.58 to 0.70 percentage point) lower rate of child welfare contact and a 3% to 4% (1.0 to 1.2 percentage point) lower rate of parental behaviors proxying child neglect.
Klevens et al. (2017) use state-level variation in EITC payments between 1995 and 2013 to examine impacts on rates of hospital admissions for pediatric abusive head trauma. They find that states with a refundable EITC had 3.1 fewer hospital admissions for pediatric abusive head trauma per 100,000 children than states without a refundable EITC, but they found no difference in such hospital admissions between states with only a nonrefundable EITC and states without an EITC. Kovski et al. (2022) find that $1,000 in additional income available to families from EITC and Child Tax Credit benefits led to a 5% lower state-level rate of child maltreatment investigations in the short term.
Research on the effects of economic support policies (and income) on child maltreatment has not explicitly examined heterogeneity by race/ethnicity. However, given that populations of color, and Black and Native American populations in particular, are disproportionately likely to have both low family incomes and be involved in the child welfare system-involved, economic support policies have the potential to disproportionately benefit these groups and, thereby, to reduce racial/ethnic disparities in child maltreatment and child welfare system involvement. Pac et al. (2023), using estimates of the causal effect of income on child maltreatment, estimate that reducing child poverty by 40% to 46% by applying the policy proposals provided in National Academies (2019a) would result in 11% to 20% fewer child maltreatment investigations. Because these policies have greater poverty-reducing impacts on Black and Latino children, Pac et al. estimates suggest that the effects on children’s involvement in the child welfare system would be two to three times larger among Black and Latino children than White children (Native American children were not separately analyzed).
Medicaid
Whether expanded access to health care through public programs for low-income families is likely to increase, decrease, or have no effect on reported child maltreatment or child welfare system involvement is theoretically ambiguous. Because health care providers are required to report cases of possible child maltreatment, greater family interactions with them may result in increased reports to CPS.
At the same time, greater access to health care for children and (potentially) parents may result in decreased child maltreatment and/or CPS involvement for several reasons. First, such access may reduce parental stress about their own or their children’s health care needs and the cost thereof, including by reducing out-of-pocket medical expenditures and medical debt. Second, to the extent that parents access treatment for their own health and mental health needs, they may be better equipped to provide safe and consistent care for their children. Third, to the extent that children’s health and mental health needs are treated (early), they may be easier to parent and thereby be less likely to experience maltreatment. Fourth, health care providers may serve as a referral source to other programs spanning parenting behaviors, child development, and food and nutrition, which may help reduce a family’s likelihood of maltreatment or child welfare system contact. As a result, access to low-cost or free health care may help families meet their children’s health and mental health needs and, potentially, developmental and material needs as well, which may result in better care for children and a decreased probability of being reported for child abuse or neglect.
Three methodologically strong studies have taken advantage of geographic variation in the Medicaid expansions enabled under the Affordable Care Act. In a county-level analysis of California, Pac (2019) finds that expanded access to Medicaid was associated with an 11% reduction in physical abuse investigations but no differences in overall investigations or investigations for other types of maltreatment. Brown et al. (2019) use data from 2010 to 2016 on the state-by-state expansions of the Affordable Care Act and find an 11% decrease in neglect investigations in Medicaid-expansion states relative to non-expansion states in the post-expansion period, but no effect on abuse investigations. Finally, McGinty et al. (2022) find that state Medicaid expansions were associated with reductions in neglect investigations of 13% for 0–5-year-olds, 15% for 6–12-year-olds, and 16% for 13–17-year-olds, although they find no effects for physical abuse. Together, these studies suggest a potential causal relationship between Medicaid access and reductions in child welfare investigations, for neglect, but not necessarily abuse.
Food and Nutrition Programs
The committee identified three rigorous studies examining the impact of food and nutrition programs on child maltreatment. Two focus on the Supplemental Nutrition Assistance Program (SNAP); the third considers both SNAP and the Special Supplemental Nutrition Program for Women, Infants and Children (WIC). Johnson-Motoyama et al. (2022) use variation in the number of state SNAP income generosity policies in
effect (broad-based community eligibility, excluding child support received from income calculation, providing transitional SNAP benefits for families leaving cash welfare, simplified reporting of changes in household circumstances) to examine the effects of SNAP policy generosity and caseload size (participation) on child maltreatment rates across states and over the 2004–2016 time period. Results indicate that the effect on maltreatment of each additional income generosity policy and of a 5% increase in SNAP caseload size are similar in magnitude, with each resulting in an 8% to 9% reduction in child maltreatment investigations, 9% to 10% reduction in substantiation cases of maltreatment, and 9% to 15% reduction in foster care placements.
Bullinger et al. (2021) leverage within-monthly group variation by Census block in proximity to a SNAP-authorized retailer to estimate monthly within-Census block group changes in child maltreatment rates in Connecticut between 2011–2015. They find that, in large rural locales, each additional SNAP-authorized retailer in a Census block group is associated with a 4% decrease in maltreatment reports and an 11% decrease in maltreatment substantiations.
In their investigation of the effects of both SNAP and WIC participation on child maltreatment, Lee and Mackey-Bilaver (2007) compare siblings using individual-level administrative data on Medicaid-enrolled children in Illinois. They find that participation in SNAP only, WIC only, and both programs is associated with reductions in substantiated child maltreatment of 7%, 11%, and 9%, respectively.
Community-focused Prevention Programs
Several large-scale community-focused interventions have been launched in recent decades aimed at improving parenting and family functioning and reducing child maltreatment. Such programs typically include both universal (community-level) and targeted components, such that particularly at-risk families have opportunities to engage in more intensive services than their less at-risk counterparts. That is, they tend to offer a continuum of interventions intended to address the needs of particular subgroups of families through direct service provision and/or by assisting families to access existing programs and services (Berger & Font, 2015). Such programs are difficult to evaluate rigorously, and the evidence on their efficacy is quite limited. Moreover, they tend to be difficult to implement and sustain, as well as relatively expensive.
A notable exception, however, is the Positive Parenting Program (Triple P), which has been implemented in a growing number of communities and has been subject to the most rigorous evaluation of such programs to date. A summary of impact evaluations is available from the California
Evidence-Based Clearinghouse for Child Welfare.5 An example is Prinz et al. (2009), an evaluation of Triple P in 18 counties that were randomly assigned to either Triple P implementation or services as usual. The evaluation found that, 2 years after implementation, Triple P counties experienced substantial reductions in substantiated child maltreatment, out-of-home placements, and hospital and emergency room admissions for child maltreatment-related reasons. In a more recent quasi-experimental study, Schilling et al. (2020) compare county-level child welfare and hospital discharge data in North Carolina for 34 counties that implemented Triple P and 66 counties that did not, for the period 2008–2015. Results indicate that implementation was associated with a 4% reduction in county-level child welfare investigations and a 7% decrease in foster care placements. The study found no effects on hospital admissions for child maltreatment-related reasons.
Conclusion 10-2: Causal evidence on factors leading to maltreatment and child welfare involvement is limited, although most evidence points to household economic hardship as elevating the risk of child welfare involvement and to income support and income-support policies reducing risk for child welfare involvement. Evidence on the likely favorable impacts of Medicaid and food and nutrition program eligibility is also relatively strong.
INTERVENTIONS REDUCING CHILD MALTREATMENT
Evidence on prevention policies and programs offers some promising avenues for reducing the number of child maltreatment cases reported to or substantiated by child welfare services. On the other hand, the literature on the longer-run causal impacts of the various elements of the child welfare system on poverty in adulthood is not strong enough to identify the changes to the system that would reduce intergenerational poverty. This leads us to consider the following promising prevention approaches as indirect approaches to reducing intergenerational poverty:
- The most consistent evidence of causal effects on reduced child maltreatment is for direct income transfers to low-income families.
- Consistent evidence of reductions in child maltreatment is also found in strong studies of the impacts of the recent Medicaid expansions occasioned by the Affordable Care Act.
___________________
5 For more information see https://www.cebc4cw.org/
- Expansions of eligibility and benefit levels in food and nutrition programs such as SNAP and WIC have also been linked with reductions in child maltreatment.
- Some community-level interventions such as the Triple P appear to be promising approaches for reducing child maltreatment.
