The Society for Social Work and Research

2014 Annual Conference

January 15-19, 2014 I Grand Hyatt San Antonio I San Antonio, TX

Non-Relative Foster Care Or Kinship Care? Estimating the Effects of Placement Type On Child Achievement and Behavior

Saturday, January 18, 2014: 5:00 PM
HBG Convention Center, Room 001B River Level (San Antonio, TX)
* noted as presenting author
Sarah A. Font, MSW, PhD Student, University of Wisconsin-Madison, Madison, WI

            Child welfare policies at the state and federal levels have evolved to incentivize kinship care placements as a less expensive, more family-like alternative to non-relative foster care. However, the evidence suggesting that children in kinship care fare better than children in non-relative care is limited and generally fails to account for selection bias. This study uses an instrumental variables approach, in combination with propensity score weighting, to produce an unbiased estimate of the effect of kinship placement on children’s achievement and behavior.

Data and Methods

            This study uses the first cohort of the National Survey of Child and Adolescent Well-Being, which includes a national sample of child welfare investigations and an additional sample of children who have been in foster care for 1 year. The analysis sample for this study restricted to children who spent some time in kin or non-relative foster care, and have complete data on the outcome measures at waves 1 and 4 (N=950).  

  The outcomes examined include three achievement measures and two behavioral measures. The achievement measures are the math and reading test scores from the W-J Mini-Battery of Achievement, and the composite score from the Kaufman Brief Intelligence Test. The behavioral measures are the internalizing and externalizing behavior subscales from the Youth Behavior Checklist. The primary independent variable is a dichotomous indicator of whether the amount of time a child spent in kinship care met or exceeded 50% of total time spent in out-of-home care. Control variables include child demographic characteristics, age at entry to care, regional indicators, and risk factors.

            Residualized change models and simple change models are used to estimate the effect of kin placement on W4 outcomes, controlling for that outcome at baseline (W1).  To reduce the likelihood that the estimates reflect selection bias, I then employ two additional techniques. First, I estimate models using inverse propensity for treatment weights, to better correct for differences in child characteristics. Second, I use agency level variation in preferences for kinship care and need for placements to instrument kin placement.


            Findings suggest that increased time spent in kinship care (as opposed to non-relative placement) has a significant negative effect on reading scores. Estimates suggest no statistically significant effects of kinship placement on the other achievement measures or on behavior.


            The results of this study suggest that, after accounting for selection bias, no positive effects of kinship placement are found on child achievement or behavior. Findings suggest that efforts to increase reliance on kinship placements may not provide the expected benefits to children in out-of-home care. At the same time, however, with the exception of reading scores kinship placements are not associated with worse outcomes than non-kin placements.