https://orcid.org/0000-0003-4996-9317
Abstract
This study combines and standardizes multiple sources of administrative data to calculate rates of children in foster care in the United States from 1961 to 2018, more than tripling the length of previously available time series. Results yield novel insights about historical, geographic, and ethnoracial variation in children's experience of living without parents under state supervision. National rates of children in foster care rose from 3 per 1,000 in 1963 to a peak of almost 8 per 1,000 in 1997 before declining to just under 6 per 1,000 in 2018. After stable or increasing racial inequality in the late twentieth century, disparities between Black/African American and White children began to decrease in the twenty-first century in nearly every state, closing entirely in several Southern states but remaining wide outside the South. In many Midwestern and Western states, the extreme overrepresentation of American Indian/Alaska Native children in foster care persisted or intensified.
Introduction
Foster care is the childhood experience of living without parents under the supervision of a child welfare agency, such as with a foster family or in a group home. Placement of children in foster care in the United States is remarkably common. On September 30, 2019, 423,997 children—1 in 172—were in foster care (Children's Bureau 2020). Although this study focuses on point prevalence to analyze historical trends, this measure understates children's cumulative risk of placement. For example, at the end of 2005, 1 in 145 children were currently in foster care, but 1 in 17 had ever been in foster care (Wildeman and Emanuel 2014:3).
Children's experience of foster care is consequential to well-being over the life course. Scholars disagree about whether to attribute observed differences in economic, health, and criminal justice outcomes to foster care itself or to antecedents such as abuse or neglect. For example, research has documented null or positive average treatment effects of foster care compared with parental care (Berger et al. 2009; Font et al. 2019; King 2017) but has found significant negative effects among children at the margin of placement (Doyle 2007, 2008, 2013). Although such findings may appear at odds, they are consistent with the scenario in which foster care provides benefits to children facing severe deprivation under parental care but causes harm when state intervention in families is overly aggressive.
The prevalence of foster care among U.S. children of different races and ethnicities is greatly unequal (Myers et al. 2018). For example, in 2005, 1 in 20 White non-Hispanic children had ever experienced foster care, compared with 1 in 9 Black non-Hispanic children and 1 in 6 American Indian children (Wildeman and Emanuel 2014:3). Group inequality in foster care may result from differences in the underlying prevalence of child maltreatment (Yi et al. 2020), but it also reflects disparities in officials' propensity to surveil and intervene in families of different races and ethnicities (Briggs 2020; Edwards 2019; Fong 2020). Such discrimination, combined with the greater likelihood of negative effects of foster care for children at the margin of placement (Doyle 2007, 2008, 2013), implies that ethnoracial disparities in foster care are likely to drive group inequality in children's life chances.
Despite the prevalence, significance, and inequality of foster care in the United States, data limitations have left critical gaps in scholarly knowledge (Waldfogel 2000; Wildeman and Waldfogel 2014). Before the twenty-first century, statistics on foster care were collected irregularly by a mix of public and private organizations and were published in forms that make comparison across states and years difficult. As a result, landmark studies have relied on time series 7 to 16 years long and have pooled children of different ethnoracial groups (Paxson and Waldfogel 2002, 2003; Swann and Sylvester 2006). This study significantly expands the historical scope and demographic detail of previously available longitudinal data on U.S. children in foster care. It compiles and standardizes multiple administrative sources to report state and national foster care rates from 1961 to 2018, by ethnoracial group from 1982 to 2018. The results place shorter-term studies in broader historical context and highlight how ethnoracial inequality in children's risk of foster care has varied substantially by time and place. The study creates new opportunities for research on the policy causes of child welfare.
Data and Methods
For all U.S. states and the District of Columbia, four sources of administrative data provide annual point-in-time counts of children in foster care. Data for 1995–2018 come from the Adoption and Foster Care Analysis and Reporting System (AFCARS), operated by the Children's Bureau and accessed through the National Data Archive on Child Abuse and Neglect. I aggregate child-level data to the state-year level. State-level data for 1982–1995 come from the Voluntary Cooperative Information System (VCIS), administered by the American Public Welfare Association. The association did not maintain VCIS data, but I recovered them from authors of prior studies relying on the VCIS (Doyle and Peters 2007; Swann and Sylvester 2006). For 1970 and 1973–1975, state-level data come from reports by the National Center for Social Statistics (NCSS); for 1961–1968, state-level data are from the Children's Bureau Statistical Series. Data from these two sources were published contemporaneously, digitized by Google, and accessed via HathiTrust. Two independent operators manually double-entered information from digital images.
I modify administrative data to standardize the measurement of children in foster care across sources. The VCIS (1982–1995) and AFCARS (1995–2018) both define foster care to include children under the supervision of public welfare agencies and living in relative and nonrelative foster homes, pre-adoptive homes, group homes, childcare and emergency facilities, or supervised independent living arrangements, as well as those who had run away from foster care. I include all children recorded in the VCIS and AFCARS. Figure A1 (online appendix) evaluates the reliability of VCIS and AFCARS child counts in 1995, which I average.
The Children's Bureau Statistical Series (1961–1968) and NCSS (1970, 1973–1975) tabulate children receiving casework or social services from public welfare agencies, by residence. I include all children living in foster family homes, institutions, group homes, independent living arrangements, or “elsewhere,” excluding children living with parents or with relatives not designated as foster family homes. Reliability across sources depends on the assumption that all children in foster care receive casework or social services.
Because annual reporting periods vary across sources, seasonal fluctuation in foster care will affect cross-source reliability. The reliability of VCIS data across states and years is further limited by states' survey noncompliance. For example, some states did not report runaway children. A small number of states' reported compliance changed across years, but many states did not respond to questions about compliance. Unfortunately, extant documentation does not allow for systematic correction of noncompliant reporting (Caliber Associates 1998:iii–iv). These limitations should be considered when interpreting the results.
The VCIS and AFCARS contain information on children's race and ethnicity. For 1982–2018, I measure foster care populations for four non-Hispanic/Latino groups—American Indian/Alaska Native, Asian American/Pacific Islander, Black/African American, and White—and one Hispanic/Latino group. I assign to groups the 2.7% of children in the VCIS with missing ethnoracial information according to the observed distribution of race and ethnicity in the foster care population in each state-year. Before 2000, data collection did not allow for multiracial identification, so multiracial children are grouped according to their primary racial identification. Beginning in 2000, non-Hispanic/Latino groups include only children identifying as single-race; multiracial non-Hispanic/Latino children, who made up 7.6% of all children in foster care between 2000 and 2018, are excluded from group analyses. Interpretation of group-level results, especially for American Indian/Alaska Native children, should consider qualitative variation in child welfare system contact (Lawler et al. 2012).
To calculate rates of children in foster care, I use U.S. Census Bureau state intercensal estimates. I use the population aged 0–17 as a denominator for total foster care rates. Because of data limitations, I use the population aged 0–19 for group rates and rate ratios. I use unbridged data to correspond to the ethnoracial categories used in foster care data over time.
The NCSS, VCIS, and AFCARS contain missing state-year observations. Figure A2 (online appendix) illustrates data missingness. Missing state-level counts will bias estimates of national foster care rates unless states fail to report data completely at random—an unlikely assumption. I use a simple local polynomial smoothing estimator to interpolate missing state-level values, and I combine observed and estimated values to calculate national point and uncertainty estimates of children in foster care. Because estimating national group rates for the twentieth century would require extrapolation, I report national estimates by ethnoracial group for 2000 onward, the years for which ethnoracial data are complete.
Results
Figure 1 plots rates of children in foster care from 1961 to 2018, by state. Although national and regional patterns are evident, state-level trends varied substantially in their timing, magnitude, and even direction. Previous research identified sharp increases in caseloads in the 1980s and 1990s in populous states, such as California, Florida, New York, Pennsylvania, and Illinois (Swann and Sylvester 2006). Figure 1 shows that for their absolute size, these increases were uniquely large over more than half a century. The most recent decade saw important trend divergences in key states. Decreases in the 2000s continued in New York, leveled off in Illinois and California, and reversed in Pennsylvania and Florida. Dramatic increases in states such as Indiana, West Virginia, Montana, and Alaska have brought recent foster care rates to some of their highest observed levels.
Figure 2 combines state-level observations and estimates to calculate national foster care rates. Table A1 (online appendix) reports plotted values. National rates of children in foster care rose from 3.0 per 1,000 in 1963 to a peak of 7.9 ± 0.1 in 1997. Rates then fell consistently to 5.3 per 1,000 in 2011, before rising again to 5.9 in 2018.
Figure 3 plots ethnoracial ratios of state foster care rates, comparing group rates to those for the total child population. After high or increasing relative rates in the late twentieth century, the overrepresentation of Black/African American children in foster care declined in the twenty-first century in almost every state, disappearing entirely in several Southern states. The persistent underrepresentation of White children also began to erode broadly. In recent years, White children in Florida, Georgia, Louisiana, and Mississippi were slightly overrepresented in the foster care population for the first time on record.
Inequality among American Indian/Alaska Native, Asian American/Pacific Islander, and Hispanic/Latino children depended greatly on history and geography. In upper-Midwestern and Northwestern states, the extreme overrepresentation of AmericanIndian/Alaska Native children in foster care continued from the late twentieth century into the twenty-first. In certain states, though, it compounded, such that in Minnesota in 2018, American Indian/Alaska Native children were overrepresented in the foster care population by a factor of 15.8.
Figure 4 shows national group rates and rate ratios. Tables A2 and A3 (online appendix) report plotted values. Rates of foster care among Black/African American children decreased from 17.3 per 1,000 in 2000 to 9.3 in 2018. Despite this decline, in 2018, Black/African American children were 1.66 times more likely to be in foster care than the average child. American Indian/Alaska Native children's over-representation in the foster care population grew from a factor of 2.00 in 2000 to 2.85 in 2018. White children remained underrepresented in foster care, but relative to the average child, their risk grew from 0.59 in 2000 to 0.88 in 2018. Hispanic/Latino children's contribution to the foster care population remained stable over two decades, and Asian American/Pacific Islander children's low risk of foster care decreased. Independently of change in the distribution of race and ethnicity in the U.S. child population, the ethnoracial composition of children in foster care has shifted significantly in the twenty-first century.
Conclusion
National rates of children in foster care in the United States grew unevenly between 1961 and 2018. Marked decreases in the early twenty-first century reversed in recent years, with unprecedented increases in some states. Longstanding Black–White disparities in foster care narrowed in nearly every state and closed entirely in some. But children's experience of living without parents under state supervision remains deeply divided by race, ethnicity, and geography. Because foster care has negative long-term causal effects at the margin of placement (Doyle 2007, 2008, 2013), and because state actors intervene disproportionately in marginalized families (Briggs 2020; Edwards 2019; Fong 2020), inequality in this childhood experience is likely to contribute to stratification throughout the life course.
Complementing other recent efforts outside the United States to construct and publish historical data on child welfare (Degli Esposti et al. 2019), the National Data Archive on Child Abuse and Neglect has published portions of the data used in this study (Roehrkasse et al. 2021), with plans to publish additional data. Longer time series and group data offer important new opportunities for demographers, sociologists, economists, and social work and human development scholars seeking to understand the causes of inequality and change in child welfare. Researchers should extend studies of U.S. trends over the late twentieth century (Paxson and Waldfogel 2002, 2003; Swann and Sylvester 2006), examining whether factors such as female imprisonment, violent crime, and welfare generosity were also key drivers of foster care placement in the more distant and more recent past. Research should also inquire into whether the causes of children's placement in foster care vary by ethnoracial group.
Acknowledgments
I thank Christopher Wildeman and an anonymous reviewer for helpful comments; Michael Dineen, Joseph Doyle, Christopher Swann, and John Vogel for sharing data; and Chiara Fontaine, LeAnn McDowall, Sarah Sernaker, and Matthew Sheen for research assistance. The Children's Bureau provided funding. Any errors are my own.
