Research

Working Papers

Machine learning tools have the potential to improve the allocation of services to recipients, but there is a limited understanding of how such tools are used by human experts in practice. We use a randomized controlled trial to evaluate the effects of human-algorithm interaction in a high-stakes public services context, Child Protective Services (CPS), where workers have about 10 minutes to decide whether to investigate a family and possibly remove a child from an unsafe home. The trial provides social workers with randomized access to an algorithmic risk score that accurately predicts whether a child will be removed from their home due to maltreatment. We find that giving workers access to the tool reduced child injury hospitalizations by 32 percent and narrowed racial disparities in CPS contact considerably. Surprisingly, despite an improvement in outcomes, workers using the tool were more likely to investigate children predicted as low-risk and less likely to investigate children predicted as high-risk, relative to the control group. Text analysis of social worker discussion notes suggests that algorithmic predictions allow workers to better focus their attention on other salient features of the allegation that may indicate maltreatment. Our results highlight the potential benefits and unexpected impacts of human-algorithm interaction in high-stakes contexts.


AEA RCT Registry: AEARCTR-0006311

NeurIPS 2021 Workshop on Human and Machine Decisions Top-Three Finalist

More than 20 percent of young adult prison inmates in the United States have spent time in foster care, among whom a majority have lived in a congregate care (group-based) setting. Using three decades of administrative data from Wisconsin, we leverage exogenous variation in the relative delinquency status and imprisonment risk of foster care peers to study how peer composition affects youth's future criminal justice system contact, educational attainment, and short-term risky behavior. A one standard deviation increase in peer risk is associated with a modest 2.5 percent increase in a youth's likelihood of dropping out of high school. However, peer risk has no effect on a child's likelihood of entering prison by age 20, nor on a number of other indicators of deviant behavior. Our findings have policy implications for the recent Family First Prevention Services Act, which incentivizes the reallocation of children away from congregate care.

Presented at: AEA Annual Conference, APPAM

Payments to foster parents are among the largest per capita support payments targeted toward disadvantaged children in the United States. These payments vary considerably by state and have been the subject of longstanding policy debates, but the overall effect of payments on children's quality of care is theoretically ambiguous. We study the effect of foster care payments on caregiver labor supply, children’s foster care experiences, and children’s health using two sources of variation: natural variation in increases to state statutory payment rates and age-specific payment discontinuities that vary by state. To measure short-term outcomes, we assemble an extensive 13-year, 39-state panel of payment rates combined with microdata from Medicaid enrollment, claims, the Adoption and Foster Care Analysis and Reporting System (AFCARS), and the American Community Survey. In contrast to the prior literature, we find that increasing foster care payments has only a modest effect on whether a child is placed with a family versus in group-based care: less than a one percent increase in family home placement per $1,000 increase in a state’s annual payment rate. Further, we find little evidence of benefits along multiple dimensions of child well-being. Our findings highlight the limitations of payments to caregivers as a cost-effective strategy for improving children’s quality of care.

Presented at: APPAM, ASHEcon Emerging Scholars, University of Sydney

Work in Progress

Developing Effective Cash Transfers for Youth Aging Out of Foster Care

with Jonathan Tebes

Racial Disparities and Decision Tools in Child Protection 

with Emily Putnam-Hornstein


Additional projects unlisted due to ongoing data agreements!