Using Propensity Score Matching for Estimating the Casual Effect of Parental Substance Abuse On Child Welfare Outcomes

PURPOSE: Parental substance abuse is detrimental for children’s development (McNichol & Tash, 2001) and it increases the chance of child maltreatment (Famularo et al., 1992; Jaudes et al., 1995). Additionally, maltreated children with substance abusing parents stay in the child welfare system longer and experience poorer well-being and permanency outcomes (Kelleher et al., 1994). Through a quasi-experimental design, this study takes a beginning step to answer the following research questions: (1) whether parental substance abuse as a removal reason has a causal effect on foster children’s permanency outcome? (2) whether parental substance abuse as a removal reason has causal effects on certain types of permanency outcomes, such as reunification, living with other relatives, adoption, and guardianship? 

METHOD: This quasi-experimental study applies the method of propensity score matching which addresses the issue of selection bias and balances mean differences between children with and without substance abusing parents. The population of the study is 276,266 children who exited the child welfare system during fiscal year 2009 from the Adoption and Foster Care Analysis Reporting System (AFCARS) dataset. Children in the treatment condition (those who are removed from their caregivers due to parent drug/alcohol use) and children in the control condition (those who are removed due to other reasons) are matched on study covariates, such as age, gender, race, lengths of stay, number of placements, and number of removals. The permanency outcome in the matched sample of 99,389 children is estimated with an OLS regression analysis. The sub-analyses on the four types of permanency outcomes are carried out: reunification, living with other relatives, adoption, and guardianship.

RESULTS: Findings from the major analysis indicate that there is no significant causal effect of parental substance abuse on children’s permanency outcome while other covariates are controlled. According to the sub-analyses on different types of permanency, we find statistically significant causal effect of parental substance abuse on reunification, living with other relatives, and guardianship, but no significant effect on adoption of children’s permanency outcome. Children with substance abusing parents are less likely to be reunified with their parents (p < .001), but are more likely to live with other relatives (p < .001) and achieve guardianship (p < .001). However, the effect sizes of the outcomes are small.

IMPLICATIONS: Based on the statistically significant findings, this study has implications for social work practice and policy. Integrated services with substance abuse treatment and child welfare services should be implemented to assist developing permanency planning for family reunification (Marsh et al., 2006). Since children with substance abusing parents are more likely to achieve outcomes of living with other relatives and guardianship, family support and family preservation services can target kinship care families which can serve as more legally permanent placements. The implication for research also promotes the increasing use of propensity score matching to eliminate biases when randomization is not ethically available in the field of child welfare. Due to the small effect size of the outcomes, future research should address other influential factors affecting children’s permanency outcomes.