State Spending on Public Benefit Programs and Child Maltreatment

This was an ecological study of all US states and the District of Columbia during federal fiscal years 2010–2017 (ie, October 1, 2009, through September 30, 2017), excluding Oregon in 2010 and 2011 because of missing maltreatment reporting data. The unit of analysis was state-year observations. The Office of Research Integrity at Children’s Mercy Kansas City deemed our study as involving deidentified data and exempt from institutional board review.

Our primary outcomes were risk-adjusted rates of maltreatment known to the child protection system: reporting (ie, referrals to child protective services screened-in for investigation), substantiations (ie, reports in which children were deemed victims of maltreatment), foster care placements, and maltreatment-related fatalities among children aged <18 years. For maltreatment reports, we examined reporting overall and individually for physical abuse, neglect, and sexual abuse. Data for maltreatment reporting, substantiations, and foster care were sourced from the National Child Abuse and Neglect Data System Child Files.²¹  Counts of fatalities were sourced from the annual Child Maltreatment reports (Department of Health and Human Services, Administration for Children and Families).²²  Population estimates for children were obtained from the American Community Survey (US Census).

We adjusted states’ maltreatment rates for confounding variables selected a priori: childhood poverty, states’ Medicaid expansion status, US Census Division, unemployment rate, teenager birth rate, and the percentages of adults with illicit drug use disorder, adults with alcohol use disorder, adults with a serious mental illness, single mother households, adult educational attainment, and foreign-born population. Historically, some racial and ethnic minorities of children have been disproportionately involved with the child protection system. Therefore, we felt it prudent to also adjust for race and ethnicity demographics. These data were obtained from the American Community Survey (US Census Bureau) and the National Survey on Drug Use and Health (Substance Abuse and Mental Health Services Administration).²³ 

Our primary predictor was states’ total spending on benefit programs per person living below the federal poverty limit (FPL) and was the sum of spending on (1) cash, housing, and in-kind assistance; (2) low-income housing infrastructure development; (3) child care assistance; (4) refundable EITC; and (5) Medical Assistance Programs (MAPs), inclusive of Medicaid and Children’s Health Insurance Program. A full description of these categories of spending and their program sources are detailed in Supplemental Table 4. We aimed to capture state and local government spending on benefit programs that had the capacity to reduce poverty or material hardship, with an emphasis on programs with the largest absolute expenditures and expenditures that varied on a per beneficiary basis from state to state. We also captured total federal spending on the aforementioned benefit programs within each state-year (along with the Supplemental Nutrition Assistance Program)²⁴  to be used as a confounding variable for the association between state spending and maltreatment outcomes.

Expenditures were collected from 6 sources that contributed to the aforementioned 5 types of spending (Supplemental Table 4). The US Census Bureau’s Annual Survey of State and Local Government Finances provided data for state and local spending, outside of federal programs.²⁵  Data for federal programs were obtained directly from program sources and included the Child Care and Development Fund (CCDF),²⁶  TANF,²⁷  SNAP,²⁸  and MAPs.²⁹  Data for refundable EITC were collected/calculated from the Internal Revenue Service and collected directly from some states’ Departments of Revenue (California, Washington, DC, Minnesota, Wisconsin).³⁰ 

Medians and interquartile ranges (IQRs) described the state-year observations. Generalized estimating equations (GEEs) with repeated measures of years risk-adjusted each states’ maltreatment counts using the aforementioned confounding variables. GEE models with repeated measures of years were again used to determine the associations between states’ total spending on benefit programs and risk-adjusted maltreatment rates, with and without adjustment for total federal spending, providing incidence rate ratios (IRRs) with 95% confidence intervals (CIs) for increments of $1000 per person living in poverty. We constructed similar models for each category of spending, adjusting for federal spending within the same category. Negative binomial distributions were used in all GEE models given overdispersion of maltreatment counts.

We then extrapolated our federally adjusted IRRs to the 2017 data to estimate hypothetical reductions in maltreatment that might have been associated with states spending an additional $1000 on benefit programs per person in poverty (maltreatment counts in 2017 × (1 − IRR)). (Maltreatment counts from 2017 for these estimates were 4 258 772 occurrences of children being reported, 719 765 occurrences of children substantiated as victims, 203 298 foster care placements, and 1682 fatalities.) Cost-effectiveness ratios were then calculated as the absolute cost for all states to increase spending by $1000 per poor person divided by the counts of potentially prevented maltreatment outcomes. P values < .05 were considered significant. Descriptive statistics were performed by using SPSS Version 24 (IBM SPSS Statistics, IBM Corporation) and all modeling was performed by using SAS 9.4 (SAS Institute, Inc. Cary, NC).

There were 406 state-year observations across the 8 study years (Table 1). In aggregate, there were 31 198 322 reports for suspected maltreatment, 5 322 401 substantiated reports, 1 526 189 foster care placements, and 12 558 maltreatment-related fatalities. In total, a median of $353.8 billion (IQR: $316.3–$374.2) was spent by states on benefit programs annually: $189.9 billion ($156.5–$203.7) on MAPs; $105.6 billion ($97.4–$110.5) on cash, housing, and in-kind assistance; $51.5 billion ($49.9–$53.6) on housing infrastructure; $4.9 billion ($4.7–$6.9) on child care assistance; and $2.9 billion ($2.7–$3.2) on refundable EITC. In total, a median of $7285 ($6585–$8368) per person living in poverty was spent by the states on benefit programs annually. Individual states’ average annual spending per person living in poverty and spending by category are depicted in Fig 1.

Before adjusting for federal spending, states’ total spending on benefit programs was indirectly associated with maltreatment reporting (neglect only), foster care placements, and fatalities (Table 2). Figure 2 depicts individual states’ total annual spending in relation to their risk-adjusted rates of maltreatment outcomes.

After adjusting for federal spending, states’ total spending on benefit programs was indirectly associated with all maltreatment outcomes (Table 2). Independent of population risk and federal spending, each additional $1000 spent by states on benefit programs per person living in poverty was associated with −4.3% (95% CI: −3.4% to −5.1%) difference in all maltreatment reporting (including significant differences for neglect, physical abuse, and sexual abuse), −4.0% (95% CI: −3.3% to −4.7%) difference in substantiated reports, −2.1% (95% CI: −1.7%, to −2.4%) difference in foster care placements, and −7.7% (95% CI: −6.7% −8.7%) difference in fatalities.

Extrapolating these associations for the 2017 federal fiscal year suggested that an additional $1000 of state spending on the benefit programs included in this study per person living in poverty, a 13.3% increase for that year, would have cost $46.5 billion nationally, and might have been associated with 181 850 fewer reports for suspected maltreatment, 28 575 fewer children substantiated as victims of maltreatment, 4168 fewer children placed into foster care, and 130 fewer maltreatment-related fatalities. Cost-effectiveness ratios were $1.6 million per potentially prevented substantiation and $358 million per potentially prevented fatality.

After adjusting for federal spending within the same category, only spending on cash, housing, and in-kind assistance, child care assistance, and MAPs were associated with less maltreatment reporting and substantiations (Table 3). Housing infrastructure and refundable EITC were not associated with reporting but were directly associated with substantiations. All categories of spending were indirectly associated with fewer foster care placements and fatalities.

In this study of all 50 states and the District of Columbia during federal fiscal years 2010–2017, we found that states’ spending on local, state, and federal public benefit programs, independent of federal spending, was associated with less child maltreatment reporting, substantiations, foster care placements, and fatalities. States’ spending on MAPs, child care assistance, and cash, housing, and in-kind assistance were the categories of spending most consistently associated with differences in Child Protective Service measures of maltreatment. These findings appear to support state investments in benefit programs as a population-level modality for the prevention of child maltreatment. Our measured associations suggest that a 13.3% increase in state spending on benefit programs might hypothetically be related to ∼181 000 fewer reports for suspected maltreatment, 28 500 fewer children being substantiated as victims of maltreatment, 4100 fewer children requiring foster care, and 130 fewer maltreatment-related fatalities.

Previous investigations have consistently supported an association between more generous benefit programs and less maltreatment, particularly for neglect.^{11–13 ,15,17,19,20,31}  Our results broadly align with these previous findings. We found that an additional investment of $1000 per person living in poverty was associated with 4.3% less reporting, 4.0% fewer substantiated reports, 2.1% fewer foster care placements, and 7.7% fewer maltreatment-related fatalities. Using quasi-experimental designs among TANF recipients, Cancian et al found that an additional $100 of income per year was associated with a 2-percentage point reduction in investigated reports of maltreatment and Beimers et al found that for each $100 of income per month there was an associated 2-percentage point reduction in substantiated maltreatment.^10,14  Previous estimates are not available for maltreatment-related fatalities, but McLaughlin et al found that an additional $1000 per capita in federal transfers to states to fund benefit programs was associated with an ∼7.7% reduction in overall infant mortality.³²  Further study is needed to quantitate benefit program effects at the individual level and to examine how benefit program eligibility and receipt may moderate the effectiveness of other individual-level prevention modalities, such as home visiting.

Our cross-sectional study suggests that $1.6 million and $358 million in spending on benefit programs were associated with 1 fewer substantiated maltreatment victim and 1 fewer maltreatment-related fatality, respectfully. These cost-benefit ratios do not appear advantageous until one considers the substantial lifetime economic burdens associated with child maltreatment, estimated at $830 928 per nonfatal victim and $16.6 million per fatal victim (inclusive of short- and long-term health care costs, criminal justice costs, child welfare costs, special education costs, monetized quality-adjusted life years, and value per statistical life).³  Using these previous estimates, we suggest that an additional annual investment of $46.5 billion from states in benefit programs might be associated with $25.8 billion fewer costs related to substantiated maltreatment and fatalities or up to $153.2 billion fewer maltreatment-related costs if also considering all children investigated for maltreatment. This latter estimate is based on the application of costs for nonfatal victims to children investigated but not substantiated as victims, an assumption previously made by others,^2,3  given that unsubstantiated and substantiated children have similar long-term outcomes.^33–35  It should be stressed, however, that the majority of potentially avoided economic burdens related to reductions in maltreatment would not be realized within the same fiscal year as spending. Conservatively assuming only $51 000 of near-term health, child welfare, and criminal justice costs associated with nonfatal maltreatment,³  states’ additional $46.5 billion investment in benefit programs might only return $1.5 billion to $9.3 billion in near-term reductions in maltreatment-related economic burdens (substantiations and reports, respectively). Although benefit program costs were high, potential reductions in economic burdens related to prevented maltreatment may partially offset some program costs. It should be stressed however, that reductions in maltreatment need not fully offset programs costs and that prevention of maltreatment would be just one of many secondary positive outcomes of public benefit programs.

It is important to highlight how our estimates should be interpreted and the potential for ecological fallacy. Although we standardized states’ spending relative to the size of their impoverished populations, precise beneficiaries of spending or changes in spending varied by program, year, and state. For instance, TANF largely benefits families in deep poverty, whereas MAP eligibility criteria are often inclusive of individuals above the FPL. Likewise, some benefits are contingent on earned income and/or employment (eg EITC, CCDF), criteria selecting for populations with already higher incomes. Children and families at the greatest risk for maltreatment, therefore, may not always be the largest beneficiaries of states’ investments in benefit programs. Also, maltreatment can and does occur outside of the context of poverty, and supplementary benefits and income may prevent maltreatment with varying effectiveness across the spectrum of income. For these reasons, our estimates apply broadly to state budgets and maltreatment at the population-level and cannot be translated to the individual level.

MAPs under the Affordable Care Act have been estimated to reduce poverty to a similar degree as all other mean-tested benefits programs and tax credit programs combined.³⁶  Medicaid expansion has also been reported to reduce economic burdens and psychological distress among low-income parents, specifically.³⁷  It follows then, that states’ spending on MAPs would be associated with reductions in maltreatment, as we found in our analysis. Our results are qualitatively similar to Brown et al who found that neglect reporting was ∼10% lower in states that chose to expand Medicaid, the only other study to test this association.¹⁸  Children investigated for maltreatment impose substantial economic burdens on the Medicaid system directly, accounting for ∼9% of all expenditures among children.³⁸  Therefore, in addition to the other benefits that MAPs provide to recipients, health care systems, and states’ economies,³⁹  further expansion of states’ Medicaid and Children’s Health Insurance Programs might also have the potential associated benefits of reducing maltreatment and lowering future program costs.

Our results suggest that states’ decisions to financially invest in quality child care for low-income families may be related to lower population rates of maltreatment. The potential for child care subsidies to reduce child maltreatment has previously been demonstrated, specifically for measures of neighborhood-level availability and paid family leave policies.^40–42  Child care assistance, in addition to reducing low-income families’ out-of-pocket expenses, may also increase maternal employment and income,^43,44  potentiating direct program effects. Recently, however, only 1 in 7 eligible children received CCDF benefits, the largest federal-state program subsidizing child care among low-income families.⁴⁵  Despite these unmet needs, levels of funding are currently low, only accounting for ∼1% of states’ annual spending on benefit programs in this study. The potential for child care assistance for low-income families to contribute to the prevention of child maltreatment appears to be an opportune area for policy discussion and future study.

EITC benefits have previously been associated with reductions in self-reported child welfare involvement,¹⁷  reports for neglect (but not abuse),⁴⁶  and foster care entry.^15,16  Klevens et al³¹  reported a marginally significant effect for EITC reducing abusive head trauma. Our results align with these previous findings, suggesting that states’ spending on their own EITC programs may be related to lower rates of foster care placement and maltreatment-related fatalities, but not reporting or substantiations.

This study had limitations. States’ generosity in spending on benefit programs may have correlated systematically with their inclusivity in identifying maltreatment, a relationship that would have favored this study not finding significant associations. In this study, we exclusively examine child maltreatment captured by the child protection system. We may have under- or overestimated the association between spending and maltreatment given that not all of child maltreatment is reported to Child Protective Services.^47,48  Although the Annual Survey of State and Local Government Finances database captured all state and local government spending, it was not feasible to include all applicable federal benefit programs, with the most notable omission being Supplemental Security Income. State refundable Child Tax Credit and Child and Dependent Care Tax Credit were given consideration but ultimately not included given that the expenditures were nominal compared with other categories of spending, and data could not be located for some state-year observations.

Our study appears to support an association between state fiscal investments in public benefit programs and population-level reductions in child maltreatment. Future cost-benefit analyses of benefit programs need to account for potential reductions in maltreatment as their associated economic burdens might offset some program costs, minimally in the near-term, but substantially longer-term. Future investigations should also explore if benefit programs, as they are currently structured, either increase or decrease racial and ethnic disparities in maltreatment, and health equity more broadly.

We would like to thank Leesa Brown for her assistance in data collection and her unwavering kindness. The analyses presented in this publication were based on data from National Child Abuse and Neglect Data System Child Files, 2010–2017. These data were provided by the National Data Archive on Child Abuse and Neglect at Cornell University and have been used with permission. The data were originally collected under the auspices of the Children’s Bureau. Funding was provided by the Children’s Bureau, Administration on Children, Youth and Families, Administration for Children and Families, and US Department of Health and Human Services. The collector of the original data, the funder, NDACAN, Cornell University, and the agents or employees of these institutions bear no responsibility for the analyses or interpretations presented here. The information and opinions expressed reflect solely the opinions of the authors.

CCDF

Child Care and Development Fund

CI

confidence interval

EITC

Earned Income Tax Credit

FPL

federal poverty limit

GEE

Generalized estimating equation

IQR

interquartile range

IRR

incidence rate ratio

MAP

Medical Assistance Program

TANF

Temporary Assistance for Needy Families