Childhood Poverty and Foster Care Placement: Implications for Practice and Policy

Abstract

Background.

About 6% of US children enter foster care (FC) at some point before age 18. Children living in poverty enter more frequently than non-poor children. Still, it is less clear if specific dimensions of poverty place a child at risk of FC entry.

Objective:

This study aids our understanding of the relationships between poverty and FC entry.

Participants and Setting:

Data were drawn from a large linked administrative data study following low-income and/or children with maltreatment reports at baseline and followed them through 2010 (n=9,382).

Methods.

Separate analyses compared low-income children and children reported for maltreatment. Cox regression analyses were used to account for clustering at the tract level. Poverty was measured at birth, receipt of income maintenance (IM) during the study period, and census tract poverty at baseline.

Results.

The results showed that within a low-income sample, both family poverty and community poverty measures were significant factors in predicting later FC entry. However, when analyses were run comparing children with maltreatment reports with and without baseline AFDC use, the various measures of poverty diminished in impact once the type of maltreatment and report dispositions were controlled. Furthermore, we found that children living in families with more spells on income maintenance were less likely to enter FC.

Conclusions:

Results indicate that specific dimensions of poverty during childhood are associated with later FC entry. The lowered risk associated with a number of spells suggests connections between time limits for income assistance and the risk of entering FC.

Introduction

About 6% of US children enter foster care at some point between ages 0 and 18 (Wildeman & Emanuel, 2014). Despite the systemic and personal costs associated with entering foster care, there have been relatively few studies focused on foster care entry (Drake et al., 2022a; Morrison & Drake, 2023; Perlman & Fantuzzo, 2013; Shaw et al., 2015; Wildeman & Emanuel, 2014). Most, but by no means all, children entering care do so because of child maltreatment (Drake et al., 2020). Placement in foster care is the option of last resort for children reported to CPS (Font & Gershoff, 2020; Jonson-Reid & Drake, 2017). Prevention of placement into foster care is the focus of the Family First Preservation Services Act (H.R. 1892). While this act does not focus on material needs, recent policy discourse related to child welfare reform has brought renewed attention to child and family poverty (Milner & Kelly, 2020). Existing research suggests that family background including socioeconomic status and children’s individual characteristics explain significant variation in placement patterns (Berger, 2006; Berger & Waldfogel, 2004).

Poverty and Entry to Foster Care

Children from low-income families are at higher risk of maltreatment, child protective services (CPS) contact and placement (Andersen & Fallesen, 2010; Berger et al., 2009; Pelton, 2015; Waldfogel, 2004). Barth, Wildfire, & Green (2006) estimated that families of about half of the children in out-of-home care and about a third of children receiving in-home services had difficulties meeting basic needs of their children. The problem of lack of sufficient financial resources and housing were highlighted as risk factors for out of home-placement by other researchers too, though the causal ordering and level of poverty (family vs community) were often not measured (Farrell et al., 2010; Hong & Piescher, 2012; Pelton, 2015;). For instance, 53% of caretakers with children in out-home placements had inadequate housing, and/or lack of sufficient financial resources to meet minimal needs, and/or received some form of means-tested public assistance in 2010 (Pelton, 2015). While even more scant, some research suggests that that cash assistance and relevant concrete supplies provision might reduce the risk of child placement (Ryan & Schuerman, 2004).

The receipt of welfare may be looked at as either a proxy for poverty or as a means of increasing resources to meet basic needs. A recent scoping review of economic supports for working parents found that EITC (Earned Income Tax Credit), TANF (Temporary Assistance for Needed Families), and childcare subsidies may reduce maltreatment, however the evidence on SNAP (Supplemental Nutrition Assistance Program) is limited and mixed (Maguire-Jack et al., 2022). A decrease in welfare generosity was also associated with substantially increased placement, increasing the annual risk of being placed in care by about 1.5 percentage points, corresponding to a roughly 25% increase in their risk of placement (Wildeman & Fallesen, 2017). Specifically, Ginther & Johnson-Motoyama (2017) estimated that TANF restrictions implemented by states increased victims of child maltreatment by 12 % and foster care placements by 13%. On the contrary, expansion of EITC decreased foster care entry rates by 7% per year in states with a state-level EITC, relative to those without (Biehl & Hill, 2018). The models with year- and state-fixed effects, demonstrated that a refundable EITC was associated with an 11% decrease in foster care entries compared to states without a state-level EITC even after controlling for child poverty rate, racial/ethnic composition, education, and unemployment (Rostad et al., 2020).

Racial Disproportionality and Entry to Foster Care

Both child neglect and poverty have adverse consequences of child neglect on child development (Font & Gershoff, 2020). The discussion of the role of poverty is often intertwined with discussions about racial disproportionality in child welfare (Cénat et al., 2021; Drake et al., 2021, 2023; Kim & Drake, 2018; Wulczyn et al., 2013). National data show that Black children are reported to CPS nearly twice as often as White children (Kim et al., 2018; Putnam-Hornstein et al., 2013; U.S. Department of Health & Human Services et al., 2023). Current data show suggest that Black children are reported to CPS at rates consistent with or lower than the high levels of risks they face. Recent national data show that once reported, Black children are substantiated and placed less frequently than White children (Drake et al., 2023). Due to data limitations of our datasets, we will focus on Black and White children while discussing racial disproportionality in the present study.

Other Covariates

In addition to poverty, a range of adversities may lead to some groups having a higher likelihood of being placed into foster care (Barth & Xu, 2023).

Child characteristics:

Boys and girls placed at roughly similar rates. In a study using national data, Native American (up to 15.44%) and Black (up to 11.53%) children were at far higher risk of placement (Wildeman & Emanuel, 2014). A few other studies have found small to moderate increased risk specific to Non-White children (Edwards et al., 2021; Fajardo, 2013; Rivaux et al., 2008). However, Putnam-Hornstein and her colleagues (2013) found that Black children were less likely than other socio-demographically similar White children to be referred for maltreatment (RR: 0.95; 95% CI: 0.92 −0.97), substantiated as victims (RR: 0.79; 95% CI: 0.76, 0.83), or placed in foster care (RR: 0.81; 95% CI: 0.76, 0.86) before age 5.

Children in foster care have been found to have higher rates of various mental health, behavioral and health problems (Koponen et al., 2022; Turney & Wildeman, 2016), but studies controlling for risk at entry suggest these issues may predate foster care entry (Berger, 2006). Child age (younger) and emotional, behavior, and medical problems have been noted as risk factors for foster care entry in the literature (Barth et al., 2006; Bhatti-Sinclair & Sutcliffe, 2012; English et al., 2015; Horwitz et al., 2011; Shannon et al., 2023).

Caregiver and family demographics.

Some studies have found that foster care placement was less likely when the parent was a teen or younger (Horwitz et al., 2011; Rivaux et al., 2008). A greater risk was associated with unmarried and single caregivers (Berger, 2006; Rivaux et al., 2008); however, if both parents were identified as perpetrators, there was a higher probability of placement (Fajardo, 2013). The caregiver/family risk factors also included alcohol/drug abuse, caregiver depression, emotional maltreatment, and a number of prior reports (Barth et al., 2006; Carter, 2010; English et al., 2015; Straatmann et al., 2021; Wildeman & Fallesen, 2017).

The present study

Evidence indicates that there is a great variation in how families and children respond to economic hardship (Conger & Conger, 2002). Different assumptions on the interactive processes, perceptions, stresses, and social supports in the family environment indicated that individual characteristics of the child and/or parent/household are not sufficient to explain the occurrence and nature of child maltreatment (Petersen et al., 2014). Two primary mechanisms appear to emerge from the theoretical literature: 1) poverty creates economic and familial stress and increased parental depression and marital conflict and ultimately leads to harsh and ineffective parenting (e.g.: (Conger & Conger, 2002; Conrad-Hiebner & Scanlon, 2015; Newland et al., 2013; Yang, 2015); 2) poverty creates social and/or resource isolation that leads to maltreatment through lack of formal or informal supports that would help alleviate stress or support positive parenting or both (e.g; (Freisthler, 2013; Maguire-Jack & Klein, 2015; Maguire-Jack & Negash, 2016; Morton, 2013).

Compared to maltreatment, we know much less about the role poverty and changes in individual family income may play in out-of-home placement (Wood et al., 2022). It is still unclear how the timing of individual family poverty, community economic context, and welfare benefits may be linked to foster care entry controlling for the demographic and risk factors present in prior studies. As literature emerges regarding the potential for financial intervention to prevent maltreatment and other related risks to child and family well-being (Biehl & Hill, 2018; Rostad et al., 2020; Wildeman & Fallesen, 2017), it is important to understand how varying timing and levels of poverty may influence child welfare trajectories. Understanding this may help us understand the likely benefit of various approaches to poverty alleviation in the prevention of entry into foster care.

The present study seeks to help fill this gap by using longitudinal linked administrative data to advance our understanding of how the early poverty, community level poverty and the receipt of welfare over time may increase or decrease the likelihood of placement controlling for child and family demographics, child maltreatment and known child health or mental health concerns. The present study seeks to answer two questions: (1) Within a low-income sample at baseline, how do community and timing of poverty indicators influence later placement, controlling for whether there was a maltreatment report and other child and caregiver characteristics? This first question allows for the possibility of placement without a prior maltreatment report and the tracking of services provided through means tested programs such as Medicaid. (2) Within a sample of children reported for maltreatment at baseline, how are indicators of family and community poverty associated with foster care placement controlling for the timing, disposition, and child welfare services following the initial report as well as the number of reports over time. This analysis allows for the exploration of foster care entry within a universe of children reported prior to age 12 without constraining the income levels of the sample.

Methods

Data were drawn from a large linked administrative data study following low-income children and/or children with a first screened-in maltreatment report sampled in 1993-1994 followed through various data sources up to 2010 (n=12,409). Families were limited to those with children under age 12 at baseline with no prior history of a maltreatment report known prior to baseline. One child per family was randomly selected from those with first maltreatment reports at baseline and matched to a sample of children families receiving AFDC but not reported by birth year and city/county residence. These data sets shared a unique state identifier allowing for linkage without probabilistic matching. This created three sample groups: (1) baseline reported maltreatment with AFDC, (2) baseline AFDC with no history of maltreatment report, and (3) baseline reported maltreatment with no history of AFDC. Ongoing data linkage included birth and death records, child maltreatment reports and child welfare services, child and caregiver department of mental health Medicaid and non-Medicaid programs, child special education and health records, parental arrest and incarceration, and income maintenance data receipt (AFDC and later TANF). Data included all system contacts by date and included child and parent demographic characteristics as well as diagnoses and/or reasons for system contact. Addresses were available at baseline and geocoded and linked with census information at the tract level. Datasets without state identifiers were linked using a combination of name, birthdate and sex as noted in the data source. All linkages were checked by hand for inappropriate matches (e.g., child death before birth date) as well as expected overlapping system involvement when estimates were available based on prior work. Children who died within seven days of a maltreatment report or entered care at baseline and never exited were excluded from the parent study due to the desire to follow children from a first-system contact who lived in families intact at baseline. The original study received human subjects approval from Washington University in St Louis university as well as approvals from all contributing agencies.

The present study used a de-identified analysis set from the parent study. In order to assure we could track proxy indicators of poverty from birth, the analysis dataset was limited to those with birth record information (n=9,497). Due to the demographics of the region at the time, analyses could only be performed for children identified as Black or White in the study data. Additionally, there were too few child deaths prior to age 18 to be able to conduct meaningful analyses so these cases were excluded. This left a sample of n=9,382.

Many of the poverty and health indicators are limited by data from means-tested programs. Therefore, the data are separated into two sets for the present analyses. The first analyses focuses on the transition to foster care among children in families receiving AFDC at baseline with or without reports of maltreatment (n=7,915). Since children can enter foster care for other reasons than maltreatment, child welfare responses to reports are not included in this analysis. A second analysis focuses on children with reported maltreatment and includes children with and without AFDC history at baseline as well as details in regard to maltreatment reports and child welfare dispositions (n=5,468). All children were followed through systems naturally over time, meaning that children in the baseline poverty-only group might later have reports of maltreatment and that children in the baseline maltreatment-only group might later have families that began receiving income maintenance.

Variables

Dependent variable:

Foster care entry was measured by the date of entry into foster care whether. As noted in other literature (Drake et al., 2022b) children may enter care for other reasons such as parental death or voluntary relinquishment so the first set of analyses includes children without maltreatment reports.

Child and parent demographic variables were taken from the child maltreatment and AFDC records at baseline and included child age at baseline, child sex (only biological sex was available), parent age at birth, child race category (Black or White), and caregiver education (High School (HS) or not). Because of the nature of the sample, over 99% of caregivers in the AFDC groups were female single head of households so caregiver sex was not included in analyses.

Poverty was measured in a variety of ways. All records included proxies of poverty at birth (Medicaid (a joint federal and state program that gives health coverage to some people with limited income and resources), WIC (the Special Supplemental Nutrition Program for Women, Infants and Children) or Food stamps) and linkage to census tract data capturing median family income in a tract at baseline. Children may also have lived in families receiving AFDC/TANF which is measured over time. Because the program changed from AFDC to TANF rules during the study period, we chose to measure participation in a number of spells (start and stop) received rather than total time. In addition, the second analysis limited to children with maltreatment reports at baseline included a variable indicating caregiver employment at the time of the report.

Maltreatment reports included all first-time and subsequent investigated reports of maltreatment (all types of neglect, physical abuse, sexual abuse). For the second set of analyses limited to children with maltreatment reports at baseline, we also included type of maltreatment, case disposition (substantiated or not) and the provision of in-home services by type. Type of maltreatment was coded from exact forms of injury, level of contact related to sexual abuse, and specific forms of neglect into three categories at baseline because of the limited number of emotional abuse reports in the very early years of life and the diversity of what constitutes “Other”. Over time, however, both emotional abuse reports and “other” were included. In-home services include measures of both lower-intensity case management and intensive in-home services designed for cases with more significant safety concerns. Exact dates of reports and services were available for time ordering.

Other child risk factors included variables taken from health diagnoses (Medicaid and/or expanded child health care CHIP, and hospital care), mental health diagnoses (department of MH, health and hospital records), special education by type of disability and birth records indicative of longer-term challenges (abnormal or congenital conditions, very low birthrate). We chose to limit notations of child behavioral or physical health issues noted at or before age 13 as 84% of first entries into foster care happened before age 14. It is not possible to know if a given diagnosis accurately reflects onset of symptoms or difficulties but earlier detection was deemed a more conservative measure of potential challenges to parenting due to a child disability, behavioral health or health issues. For child health, the indicator was limited to ICD 9 codes indicating chronic or serious health conditions (e.g., epilepsy, diabetes, cancer, hospital care for asthma, and congenital conditions). Because child mental health conditions were very limited outside means tested programming, this variable was not included in the second set of analyses.

Caregiver risk factors were selected based on prior literature and their potential for influencing parenting capacity and availability in the present data. Notation of caregiver development delay, serious and chronic health conditions, or mental health diagnoses were taken from Medicaid health and mental health records based on ICD 9 codes. Caregiver health and disability was almost exclusively derived from means-tested programs and therefore was limited to the first set of analyses focused on lower-income children. When analyses were limited to those with baseline reports of maltreatment, caregiver risk indicators were taken from risk assessment data from baseline reports including notation of alcohol or drug abuse of concern, or notation of intimate partner violence. Unfortunately, mental health indicators on baseline reports were limited to those listed in maltreatment reports under “perpetrator characteristics” so this was not included.

Both analyses included indicators of prior foster care and caregiver criminal involvement. Archived data on foster care was available far enough back to understand whether the caregiver had been in foster care at any point during adolescence as a proxy for intergenerational FC experience. Caregiver criminal justice involvement was measured by any adult arrest prior to entry into foster care or age 14 for those not entering care.

Analyses

Data cleaning and analyses were conducted using SAS 9.4. Survival analyses and Cox Regression were used to control for length of time at risk censoring at time of entry into care, age 18, end of study, or death (Allison, 2010). PROC PHREG was used to adjust confidence intervals for clustering at the census tract level. Bivariate survival curves were examined for possible violations of proportionality and need for time-varying terms to be included. Time-varying terms were retained if they were significant and/or if they impacted model fit. As aforementioned the first set of analyses focused on foster care entry among children living in lower-income families and the second set focused on foster care entry among children with baseline maltreatment reports which accounts for the differing sample sizes.

Results

Some overall descriptive results are presented prior to focusing on the two analytic samples. Combining all three groups (baseline maltreatment with AFDC, maltreatment without AFDC, and AFDC only without maltreatment), about 14% of the sample entered foster care at least once prior to age 18 (n=1,340 out of 9,382). Nearly 6% entered from the poverty-only at baseline group; nearly 10% from the maltreatment-only group and 24% from the group that included both maltreatment and poverty indicators at baseline. Only about 1% of the children who entered care had no record of a maltreatment allegation prior to or on the same date of entry. However, when disaggregated by sample group, 12.9% of those in the baseline poverty-only group entered foster care without a prior report (not shown in Table 1). Only 29% of those who eventually entered care did so after only 1 report; about 33% entered care after four or more reports. Further results are divided into our two analytic samples: (1) children on AFDC at baseline with or without maltreatment reports at baseline (poverty-only group); (2) children with reports of maltreatment at baseline with or without AFDC use at baseline (maltreatment-only group).

Table 1.

Descriptive Statistics: Foster Care or Not in a Low-Income Sample

	Foster Care
	n	Yes n=1,194	No n=6,721
Poor at birth	4166	61.1	51.1	40.79, p<0001
40% child poverty by track	4187	58.8	51.8	19.61, p<.0001
Black	6263	80.9	78.8	NS
Child Female (ref: male)		48.7	48.5	NS
CAN at baseline	4001	80.6	45.2	506.94,p<.0001
Child’s developmental delay	999	9.9	13.1	9.56, p=.002
Child Mental Health	1515	15.7	19.7	10.5, p=.001
Child Chronic or serious health issues	1077	33.2	10.1	68.99, p<.0001
Mom’s foster care experience	283	6.2	3.1	28.04, p<0001
Mom’s substance use (baseline)	300	8.4	3.0	81.06, p<0001
Mom’s mental health (baseline)	101	1.9	0.9	10.44, p=.0012
Mom’ ser/chronic health problems	490	10.1	5.5	37.65, p.<0001
Mom’ cognitive Delay	93	3.0	0.8	41.0, p<.0001
Mom’s education (High School)	5366	56.4	69.7	81.11, p<0001
Mom arrested before	238	6.4	2.4	55.0, p<0001
Ever CAN report	5351	98.8	62.1	625.11, p<000
IA Spell				CMH 286.95,p<0001
1	1161	26.4	12.5
2-3	1902	49.8	42.1
4-5	1548	17.9	25.5
6+	3304	5.8	19.9
Mom mean age at birth	--	24.6 yrs	23.5 yrs	Ttest=−5.10, p<0001*
Child age at baseline	--	3.6 yrs	4.4 yrs	Ttest=7.3, p<0001*

Foster care entry among children in families receiving AFDC.

Table 1 below illustrates bivariate descriptive statistics for the sample (n=7,915). Among child demographics, only age was significant-with younger children at baseline more likely to enter care. Children with known cognitive, health, or behavioral health diagnoses were more likely to be among those who entered foster care. A similar pattern was noted for mothers with diagnoses in these areas. In addition, children who entered foster care were more like to have a mother that had at least one arrest at baseline and/or before foster care or had a prior episode in foster care during adolescence. Children who entered care were more likely to have an indicator of poverty at birth and live in poorer census tracts. Children in families with 4+ welfare episodes were less likely to enter care than those with fewer episodes.

A Cox regression model adjusting for clustering at the census tract was used to examine foster care entry controlling for demographics and significant factors in the bivariate analyses. First a model was run using only basic sample descriptors (Wald Chisq=1548.01, df= 13, p<0001) (Table 2). Next a model was run including all child, caregiver and poverty dimensions (Wald Chisq=2441.54, df=24, p<0001) (Table 2: Full Model).

Table 2.

Cox Regression Model of Foster Care Entry in a Low-Income Sample

	Model with Demographics only				Full Model
	Estimate	P>chisq	Hazard Ratio	95% CI	Estimate	P>chisq	Hazard Ratio	95% CI
Black	−0.031	0.898	0.969	.60-1.57	−0.034	0.891	0.97	0.59-1.57
Child’s age	−1.27	<.0001	0.28	.25-.32	−1.12	<.0001	0.33	0.29-0.37
Mom’s age	0.024	<.0001	1.02	1.02-1.03	0.017	0.0006	1.02	1.01-1.03
Mom’s education (high school)	−0.471	<.0001	0.62	0.55-0.71	−0.332	<.0001	0.72	0.63-0.82
Sample group	1.26	<.0001	3.53	2.84-4.37	1.04	<.0001	2.82	2.26-3.53
Poor at birth	0.249	0.02	1.28	1.04-1.58	0.391	0.002	1.48	1.15-1.90
IM spells	−0.31	<.0001	0.73	0.70-0.77	−0.33	<.0001	0.72	0.68-0.75
40% Child Poverty tract	0.906	<.0001	2.47	1.81-3.38	0.74	<.0001	2.1	1.56-2.81
Black*sample	0.216	0.014	1.24	1.04-1.48	0.194	0.03	1.21	1.02-1.45
Black*40% child poverty by track	−0.4	0.009	0.67	0.49-0.91	−0.36	0.02	0.7	0.52-0.94
Child’s developmental delay					−0.62	<.0001	0.53	0.44-0.65
Child’s health					0.977	<.0001	2.66	2.31-3.07
Child’s mental health					−0.04	0.728	0.96	0.77-1.20
Mom’s developmental delay					0.649	0.004	1.91	1.23-2.97
Mom’s health					0.139	0.271	1.15	0.90-1.47
Mom’s mental health					0.343	0.06	1.4	0.99-2.00
Mom’s substance use					0.511	0.0005	1.67	1.25-2.22
Mom’s foster care experience					0.338	0.002	1.4	1.13-1.74
Mom arrested before					0.566	<.0001	1.76	1.34-2.31
Number of CAN reports					0.125	<.0001	1.13	1.11-1.16
Time*Age	0.01	<.0001	1.01	1.01-1.01	0.01	<.0001	1.01	1.01-1.011
Time*Sample	−0.005	<.0001	0.995	0.993-0.996	−0.005	<.0001	0.995	0.994-0.996
Time*40% child poverty by track	−0.004	<.0001	0.996	0.994-0.998	−0.003	0.002	0.997	0.995-0.999

The addition of child and caregiver risk factors had a relatively small impact on the associations of baseline poverty on foster care entry. In both models, older children were less likely to enter care and racial designation was not significant as a main effect. In both models, living in an area at baseline with 40% or more child poverty and indicators of poverty at birth increased likelihood of entering foster care-though the effect for poverty at birth was stronger in the second model. Interactions indicated that the effect of baseline community poverty diminished somewhat over time and was less for Black children. In both models, children with caregivers with an HS diploma or more were less likely to enter foster care. Children living in families with more spells on income maintenance were less likely to be in foster care.

In the full model, children with known serious or chronic health concerns were over two times more likely to be among those who entered foster care. Indicators of caregiver developmental delay, substance abuse at baseline, prior foster care history and history of arrest at baseline all increased the likelihood of foster care entry. A greater number of reports of child abuse or neglect was associated with higher likelihood of entering foster care (about 11% increase per report).

Foster Care Following a Maltreatment Report

Next analyses were re-run comparing children who had at least one report of abuse or neglect alone at baseline to be able to examine the potential impact of maltreatment type, case disposition and services use on foster care entry (n=5,468; 1,185 entered FC care). Some of the control variables differ for this analysis since the group is not limited to those receiving means-tested services at baseline. Caregiver substance use and family IPV concerns were available as notations in the maltreatment report investigation, and are included here along with other variables not restricted to means tested programs.

A Cox Regression model adjusting for clustering at the census tract was constructed to look at foster care entry within this sample (Wald Chisq 1723.38, df=26, <0.001). Generally, children reported before age 2 were more likely to enter care than those of ages 2-3, ages 4-5, or 6 and older. An interaction term indicated that age variation in placement was lessoned for Black children. Child risk factors (developmental delay and serious and chronic health conditions were similar to the model among those noted as poor at baseline. Caregiver risks captured in both models were similar with the exception of caregiver arrest. Because noted IPV was only available from maltreatment report investigations this variable is unique to this model. Generally, the poverty indicators operated in the same direction but were not significant with the exception that those that did not receive income maintenance throughout the risk period were less likely to enter care. An interaction term with poverty at the tract level at baseline suggests that the risk of placement decreased slightly per month of time a child does not enter care.

Children whose first report was substantiated were more likely to enter care in the future while those initially reported for sexual abuse were less likely to enter care compared to neglect-only reports. Also, children in families receiving in-home case management services following the first report were less likely to enter foster care, but that was not true for those who were provided intensive in-home services.

Discussions

While children most commonly enter foster care following a report of abuse and neglect, that is not always true. Our analyses suggest that different aspects related to family and community poverty may be at play within a low-income sample as compared to within a separate sample of children reported for maltreatment with and without histories of poverty. Within a low-income sample, both family poverty and community poverty measures were significant factors in predicting later foster care entry. However, when analyses were run comparing children with maltreatment reports with and without baseline AFDC use, the various measures of poverty diminished in impact once the type of maltreatment and report dispositions were controlled. The only poverty-related variable that remained significant was living in a family that never received AFDC/TANF which is a proxy for higher income.

A greater number of reports of child abuse or neglect was associated with a higher likelihood of being in foster care. Given the desire to consider foster care placement as a last resort option (Font & Gershoff, 2020), it may be that multiple contacts are required in many cases to move forward to placement. Or repeated reports may be reflective of a worsening trajectory present that eventually leads to a decision to place the child.

Although somewhat different caregiver variables were available depending on the sample group, in both cases having a caregiver who completed high school was protective while notation of caregiver substance use and history of foster care as an adolescent were risk factors. Although it was not possible to assess the type of substance used, it is interesting to note that in the last 10 years foster care entry rates have closely tracked the opioid crisis numbers (Meinhofer & Angleró-Díaz, 2019). This suggests that substance use continues to play a significant role in placement. In the present study, a caregiver’s history of foster care as an adolescent is used as a proxy for prior caregiver history of FC. There are a number of potential reasons that prior childhood maltreatment or system involvement may be associated with difficulties in parenting (Madigan et al., 2019). It is not clear how intervening with youth in care related to positive relationship-building or trauma-informed treatment might offset later risk related to parenting. As educational outcomes for youth in care tend to be poor (Moyer & Goldberg, 2020), a prior history of foster care may also be a signal of economic difficulties as youth transition to parenthood.

Indication on the maltreatment report assessment that there were concerns about IPV was associated with placement. In some states, IPV is an actual reason for maltreatment, but IPV was not a reason for maltreatment report in the state where this study was conducted. The overlap between IPV and maltreatment is well-documented (Guedes et al., 2016; Guedes & Mikton, 2013; Zolotor et al., 2007), but much less is known about its association with placement decisions. Future research should examine whether it relates to family safety concerns or perhaps difficulties protecting a child from an abusive partner. Unfortunately, this variable was only available in maltreatment report information. Future research should seek to understand how this may relate to foster care entry in low-income populations that may not involve a CPS report.

Generally, children reported before age 2 were more likely to enter care than those of ages 2-3, ages 4-5 or 6 and older. This is consistent with greater vulnerability related to age as this is also the highest risk category for maltreatment and death (U.S. Department of Health & Human Services et al., 2023).

The main effect for race category in the present study was not significant in either the poverty model or the model of entry following a maltreatment report. However, there were some significant interaction terms. In the full model that included children without reports of maltreatment, living in a high poverty neighborhood was associated with higher risk, but this was reduced for Black children. One of the difficulties in interpreting the community poverty effect is the relative likelihood of living in concentrated poverty by race in this region. Black children are several times more likely to live in concentrated poverty (Drake & Rank, 2009; Maguire-Jack et al., 2020; Putnam-Hornstein et al., 2013; Shapiro et al., 2015), so its unclear if the fact that the impact seems lower for Black children might be associated with specific conditions that may lead White children to live in those contexts. In other words, Black family location can be traced to decades of racist policies related to real estate and income (Oliver & Shapiro, 2013) so White families in those contexts may have more endogenous risks. A second interaction term indicated that the higher risk among poor children associated with at least one report of maltreatment was slightly higher for Black children. However, in the model of placement among children with maltreatment reports, race was not significant. An interaction term indicated that the lessoned likelihood of later placement for older children was somewhat moderated for Black children. At least one prior study indicated that among poor children, Black children were less likely to be reported prior to age five (Putnam-Hornstein et al., 2013). Future research should examine if reports of alleged maltreatment involving older Black children are indicative of more risk than other older children. The relationship between child risk factors (developmental delay and serious and chronic health conditions were similar in both sets of analyses. While any increased burden on caregivers would be hypothesized to create more difficulties in parenting, only chronic health issues were significant in the multivariate model. Studies have found that children in foster care generally have poorer health compared to other children (e.g., (Turney & Wildeman, 2016) Turney & Wildeman, 2016). It is less clear how pre-existing health conditions may actually increase the risk of placement based on prior work. Our study sample predates the Affordable Care Act (United States, 2010), so it is unclear if there may have been benefits in regard to the prevention of foster care in improving health care for lower income children. More research is needed that can address whether access to health and respite care for serious or chronic illnesses may have a preventive function.

Once analyses were limited to children with at least one report of maltreatment, a number of child welfare-specific variables were significant. Children whose first report was substantiated were more likely to enter care. Sexual abuse-only reports were less likely to result in foster care placement. Past research suggests that sexual abuse cases are more likely to trigger mental health care due its impact (Collin-Vézina et al., 2013). It is also possible that sexual abuse cases are more likely to involve partners or family members that can be more easily excluded from contact that reduces the need for care.

Children in families receiving in-home case management services following the first report were less likely to enter foster care, but that was not true for those who were provided intensive in-home services. Intensive services are not offered until a child is at risk of placement (Bezeczky et al., 2020). A recent systematic review and meta-analysis also showed that placement outcomes reported at family level did not demonstrate a significant reduction in out-of-home placements (Bezeczky et al., 2020). It may be that less crisis-oriented cases that are provided protective referrals may have greater chances of remaining at home. Unfortunately, national reporting data combine intensive and less intensive in-home services making it impossible to examine how outcomes may vary based on these differences. More work needs to be done in regard to the type and duration of services aimed at keeping families intact, particularly with the renewed focus of the Families First Prevention Act.

Implications

Implications for poverty policy vary by metric. For example, the importance of poverty at birth may lend support for early intervention (such as home visiting) paired with material resources. Not receiving any income assistance (proxy measure for higher income) was associated with less likehood for entering FC among those children with maltreatment reports. Interestingly, if the family had ever received the AFDC, lowered risk was associated with an increased number of spells. This finding may be a proxy for a family’s ability to access needed resources over time and is consistent with more recent data suggesting connections between time limits for income assistance and risk of maltreatment (Beimers & Coulton, 2011; Ginther & Johnson-Motoyama, 2017; Maguire-Jack et al., 2022; Spencer et al., 2021). Our findings may be indicative of positive effects of expansion of various income assistance programs that have been found to be associated with reductions in child maltreatment and entering FC (Berger et al., 2017; Cancian et al., 2013; Conrad et al., 2020; Kovski et al., 2022; Rostad et al., 2020).

Poverty in the community may impact families in a number of ways. Community poverty is associated with an increased risk of poor health due to environmental exposures (Braveman et al., 2022). This in turn may place additional stress on parenting. Community poverty may also be a proxy for lack of services to support at risk families (Freisthler, 2013; Klein, 2011; Morton, 2013). High poverty areas may also show disproportionate impact of trends in substance use like the opioid crisis (Chapman, 2022). As poverty interventions are tested, it will be important to note interactions with the level of economic distress in the community and availability of services as well as whether or not there are longer-term reductions in foster care. Furthermore, the results showed that children living in families with more spells on income maintenance were less likely to be in foster care. Thus, it may be that continued economic support may be more effective than brief material needs or other singular interventions. More research is required to understand the relative benefit.

Limitations

Despite several strengths of this research paper including longitudinal data analysis, multiple measures of poverty over time, and samples that allow for exploring first-time entries into care, there are significant limitations as well. First, our sample was restricted to one geographic region (Midwest, urban geographical area). Given the variation in child welfare policy across states it is difficult to know how well these findings may generalize. The demographics of the study region also limited the ability to assess variation by a greater array of racial/ethnic designations. At the time of sample selection, there were very few children of racial/ethnic groups other than Black or White in the region generally and engaged in social services systems specifically. While the coding of race by agencies may have diminished the ability to detect more variability in the sample it is also consistent with the regional demographics of the time. Due to data limitations, we also were not able to measure whether a family exited welfare receipt due to sanctions or lack of need. We also lacked information related to family assets or income outside participation in welfare. Child and family health, behavioral health, and IPV characteristics are limited by the data sources. For example, health records only include diagnosed conditions for which a service was paid. Caregiver substance misuse or IPV risk is limited by whether or not such factors are detected and are therefore not reflective of unmet needs or conditions that did not appear relevant to the service system in question. Nor is it possible to understand informal support or support received through community-based organizations that are not included in administrative data. Nonetheless, our study adds to the scant literature on foster care entry and is consistent with other trends and work related to current policy. This is also one of the few studies that was able to include a poverty-only comparison group which is important given that not all children enter foster care following a report of maltreatment.

Conclusion

The present study explored the role of poverty at the family and community levels may impact entry into foster care, controlling for child protection history and other child and family characteristics. Our results showed that different aspects of family and community poverty may play a role at differing time within a low-income sample and that the impacts are somewhat less among children who already have a report of maltreatment. Furthermore, we found that children living in families with more spells on income maintenance over time were less likely to be in FC and the in-home services they receive matters. Our findings also highlight the importance of child health prior to FC entry. It is hoped that these findings will contribute to discussions of poverty reduction and healthcare access among vulnerable populations. It is also hoped this work will encourage replication in other regions with greater ethnic/racial diversity as well as across states with differing approaches to prevention of foster care as outlined in their Families First Prevention plans and an better data on in-home services (https://www.childwelfare.gov/topics/systemwide/laws-policies/federal/family-first/).

Table 3.

Descriptive Statistics for Foster Care Entry Among All Children with CPS Reports at Baseline

Variable	N	In Foster Care	Not in Foster Care	Chi-sq/Ttest	P<value
Black	3518	76.3	61.3	86.13	<.0001
Female	2610	48.0	47.7	0.4	0.83
Child age
Infant	1377	38.0	21.9	106.60*	<.0001
2-4	1127	19.7	20.8
5-7	447	6.8	8.5
8+	2517	35.6	48.7
Child’s developmental delay	709	9.7	13.8	12.76	.0004
Child’s health	466	27.6	13.7	649.90	<.0001
Mom’s age	--	24.8	24.3	−2.21 (ttest)	p=.03
CAN+AFDC (CAN: ref)	4001	24.0	9.95	131.92	<.0001
Poor at birth	2395	26.7	15.3	108.62	<.0001
40% child poverty by tract	2364	53.6	40.6	60.97	<.0001
Mom’s education (High School)	3685	57.1	70.0	66.59	<.0001
Mom arrested before	194	5.8	3.0	20.16	<.0001
Mom’s foster care experience	214	6.8	3.2	30.12	<.0001
Family IPV	75	2.8	1.0	20.89	<.0001
Mom substance use	322	11.8	4.4	88.29	<.0001
Neglect only	3420	21.6	18.1	9.78	.002
Physical abuse only	1441	18.4	20.9	4.25	.04
Sexual abuse only	422	10.4	21.1	27.39	<.0001
Any mixed type CAN	185	32.9	19.8	19.14	<.0001
Substantiated	1267	32.1	16.7	143.56	<.0001
In-home family case management services	460	5.0	9.3	20.34	<.0001
In-home family preservation services	90	2.4	1.4	5.37	.02
IM onset				7.51*	.006
Never	1165	7.4	24.8
0-1 yr	2409	59.4	40.2
2-3 yr	629	11.0	11.6
4-5 yr	516	9.4	9.4
6-9 yr	629	11.4	11.5
10+ yr	120	1.4	2.4
# CAN Reports				142.07*	<.0001
1	2203	28.9	43.2
2	1192	19.7	22.3
3	692	13.6	12.4
4	455	10.3	7.8
5+	926	27.4	14.3

Table 4.

Cox Regression Model of Foster Care Entry Among those with CPS Reports: Full Model

Variable	Estimate	P value	HR	CI
Black (White: ref)	−.158	.320	0.85	0.62-1.17
Female (male: ref)	.048	.415	1.05	0.93-1.18
Age at CAN report	−1.212	<.0001	0.30	0.26-0.34
Child development delay	−.535	<.0001	0.59	0.47-0.73
Child Health	1.54	<.0001	4.67	4.01-5.45
Mom’s education (High School)	−.244	.0006	0.78	0.68-0.90
Mom’s age	.018	<.0001	1.02	1.01-1.03
Mom’s arrest record	.097	.561	1.10	0.79-1.53
Mom’s foster care experience	.331	.006	1.39	1.10-1.77
Mom’s substance use	.539	<.0001	1.71	1.40-2.10
Family IPV	.934	<.0001	2.55	1.80-3.59
Poor at Birth	.153	.220	1.17	0.91-1.49
40% child poverty by tract	.194	.106	1.21	0.96-1.53
Never IM	−.761	<.0001	0.47	0.36-0.61
IM under age 2	.044	.733	1.04	0.81-1.34
Number of CAN reports	.048	.047	1.05	1.01-1.10
Substantiation	.916	<.0001	2.50	2.13-2.93
Neglect (ref.)
Physical abuse only	.069	.374	1.07	0.92-1.25
Sexual abuse only	−.536	.0005	0.58	0.43-0.79
Mix CAN	.234	.106	1.26	0.95-1.68
In-home case management services	−.655	<.0001	0.52	0.38-0.71
In-home family preservation services	.429	.052	1.54	0.997-2.37
Black*CAN*child’s age	.145	.006	1.16	1.04-1.28
Child development delay*Substantiation	−.518	0.012	0.59	0.40-0.89
Time*Tract poverty	−.002	0.001	0.997	0.995-0.999
Time*child’s age	.010	.0006	1.01	1.01-1.012