Do Foster Children Achieve Better Adult Outcomes When Placed in Families Instead of Institutions?

Taylor (2025) examines whether placing foster children in family homes rather than congregate care improves long-term outcomes. He uses national administrative foster care data (2010–2015) linked to survey outcomes at age 21, and applies an instrumental variable based on exits from foster families. He finds that family placement substantially improves outcomes, increasing a combined index of employment, education, and reduced incarceration, homelessness, and substance abuse by about 0.97–0.99 standard deviations. Congregate care nearly quadruples homelessness risk and triples incarceration risk. He also shows that reallocating children to families can achieve a large share of these gains.

Why This Article Was Selected for The Policy Scientist

This article examines a central policy lever in child welfare: how placement decisions shape long-term life outcomes. The question is consequential because foster care systems serve highly vulnerable populations, and placement settings influence trajectories related to incarceration, housing stability, and labor market attachment. This makes the topic directly relevant to broader concerns about inequality, public spending, and social mobility, particularly as jurisdictions continue shifting toward family-based care models under resource constraints. Taylor contributes to this literature by leveraging national administrative data linked to adult outcomes, enhancing scope and external validity. The instrumental variable strategy provides a credible causal design, improving on earlier observational studies.

Full Citation and Link to Article

Taylor, C. (forthcoming). Who gets a family? The consequences of family and congregate care allocation for child outcomes. American Economic Journal: Applied Economics. https://www.aeaweb.org/articles?id=10.1257/app.20210798

Central Research Question

This study asks whether placement in a foster family, as opposed to congregate care, causally improves long-term outcomes for foster youth, and whether reallocating existing family placements across children can generate gains comparable to expanding the supply of foster families. The question is motivated by persistent heterogeneity in outcomes among foster youth and by policy constraints that limit the availability of family placements. Rather than treating foster care as a binary intervention, the paper isolates variation within the system—specifically, the type of placement—to determine how institutional design affects trajectories into early adulthood. It further investigates whether treatment effects vary systematically across children, implying that assignment mechanisms themselves may be a key policy lever.

Previous Literature

The paper builds on several strands of research. First, it connects to the broader literature on the causal effects of childhood interventions on long-run outcomes, including studies by Janet Currie, Raj Chetty, and James Heckman, which emphasize the long-term returns to early-life investments. While much of this work focuses on early childhood, this study extends the analysis to older, highly disadvantaged youth in foster care. Second, it contributes to research on foster care itself, particularly work by Joseph Doyle, which examines the consequences of removal into foster care but does not differentiate placement settings in detail. Third, it relates to studies comparing institutional versus family environments, though prior work often relies on observational designs or non-U.S. contexts. By combining causal inference methods with a focus on placement heterogeneity, the paper advances the literature beyond average treatment effects to consider how allocation rules shape aggregate outcomes.

Data

The analysis uses linked administrative and survey data covering foster youth in the United States. Placement information is drawn from the Adoption and Foster Care Analysis and Reporting System (AFCARS), which provides detailed records on all children in foster care between 2010 and 2015, including placement type, demographics, and reasons for entry. These records are linked to outcomes from the National Youth in Transition Database (NYTD), which surveys youth at ages 17, 19, and 21. The primary sample focuses on children entering foster care between ages 14 and 17, with outcomes measured at age 21 to capture longer-term effects.

The outcome measure is a composite index combining employment or school enrollment, incarceration, homelessness, and substance abuse referrals. This index is standardized to facilitate interpretation of effect sizes. The dataset is notable for its national scope, administrative accuracy in placement records, and linkage to later-life outcomes. However, the survey component introduces potential concerns about non-response and attrition, which the author addresses through weighting and robustness checks. Overall, the data provide a rare opportunity to study both institutional inputs and adult outcomes within a single empirical framework.

Methods

The empirical strategy relies on an instrumental variable design to estimate the causal effect of placement type. The key challenge is that placement into family or congregate care is not random; children assigned to families may differ systematically from those placed in institutions. To address this, the paper uses the exits of other children from foster families within the same county and month as an instrument for whether a given child is placed in a family. The intuition is that when other children exit family placements—due to reunification or aging out—new slots open, increasing the likelihood that incoming children are assigned to families.

This approach identifies a Local Average Treatment Effect for children whose placement is influenced by these quasi-random fluctuations in family availability. The validity of the instrument is supported through multiple tests, including balance on observable characteristics and lack of correlation with prior outcomes. The paper also complements the IV analysis with ordinary least squares estimates using extensive controls, though these are interpreted cautiously due to potential selection bias.

To extend beyond reduced-form estimates, the author develops a structural model based on a generalized Roy framework. This model incorporates heterogeneous treatment effects and allows simulation of counterfactual policies, such as increasing the number of foster families or reallocating existing placements across children. The combination of quasi-experimental and structural approaches strengthens the analysis by linking causal identification with policy-relevant counterfactuals.

Findings/Size Effects

The results indicate that placement in a foster family substantially improves long-term outcomes relative to congregate care. The instrumental variable estimates show that family placement increases the outcome index by approximately 0.97 to 0.99 standard deviations, a large effect size by conventional standards. Disaggregated results reveal significant reductions in homelessness and substance abuse, with suggestive evidence of reductions in incarceration. Employment and enrollment effects are positive but less precisely estimated.

In terms of magnitudes, the estimates imply that congregate care nearly quadruples the probability of homelessness and roughly triples the likelihood of incarceration relative to family placement. These differences underscore the substantive importance of placement setting as a determinant of life outcomes.

The analysis also finds evidence of heterogeneous treatment effects. For example, boys appear to benefit more from family placement than girls, yet are less likely to be placed in families. This mismatch suggests inefficiencies in the allocation process. Structural simulations indicate that reallocating children to families based on observable characteristics could achieve more than two-thirds of the gains associated with a 50 percent increase in the supply of foster families. This finding highlights the potential for policy improvements through better targeting, even in the absence of additional resources.

Conclusion

The paper demonstrates that placement setting within foster care is a critical determinant of long-term outcomes and that both the level and allocation of family placements matter for aggregate welfare. By providing causal estimates using an instrumental variable design, the study addresses a central limitation of prior research, which has often relied on observational comparisons. The integration of a structural model further extends the analysis by showing how policy interventions can leverage heterogeneity in treatment effects.

The findings suggest that improving outcomes for foster youth does not depend solely on expanding system capacity but also on optimizing how existing resources are deployed. The use of comprehensive administrative data linked to adult outcomes enhances the credibility and policy relevance of the results, while the causal inference framework aligns with best practices in empirical economics. At the same time, the absence of experimental evidence leaves open the possibility that future research using randomized or alternative quasi-experimental designs could further validate and refine these conclusions.