Cross-Site Evaluation of a Supportive Housing Demonstration for Child Welfare Involved Families

Background: The Children’s Bureau awarded five-year demonstration grants to five sites across the country to provide permanent, supported housing to high need families involved with the child welfare system. The families, first referred to local programs in late 2013, were homeless or unstably housed and had multiple needs such as substance abuse, mental health needs, domestic violence, criminal histories, and disabilities. Each site established a randomized control trial and has a local evaluator to test both implementation and outcomes of the intervention, which required that the randomly selected families receive housing and services. Sites varied in the type of housing offered, including vouchers, project based, and local affordable housing developments; in the services provided; and in the service delivery mechanisms. 

Across sites, 831 families were referred: 386 were randomly assigned to receive the intervention, and 445 families received services as usual. This paper presents findings from the cross-site evaluation which consider child welfare and housing outcomes of all five sites as one group.

Methods: To consider child welfare outcomes, administrative child welfare data were used to examine the probability of removal (for intact families) and probability and speed of reunification (for families with a child in out-of-home care). For all families, the probability and speed of case closure, and the probability of a new maltreatment report for intact and reunified families is examined. Housing impacts were based on a survey of families from the program and services-as-usual group one year after randomization. Several measures are reported including being housed with a lease (versus doubled-up, in a homeless shelter, living in a place not meant for habitation), number of moves over the past year, evictions, periods of homelessness, and expectations about the permanency of their current housing situation. The survey also measured self-reports of housing quality and neighborhood quality. Impacts were estimated using probit regressions for limited-dependent variables, OLS for continuous variables, and survival analysis for measures of time.

Results: Results based on roughly half of the final sample indicate that, compared with the services-as-usual group, program group families had lower regression-adjusted rates of removal (9% v. 15%), higher rates of reunification (50% v. 30%), and lower rates of new reports of maltreatment (34% v. 42%). Analysis of survey data indicate that, across all measures, the program group had more stable and higher quality housing than the services-as-usual group, although it was not necessarily in better neighborhoods. In particular, 86% of the treatment group had a regression-adjusted rate of being housed with a lease of 86% and only 2% had a homeless spell, compared with 45% and 18%, respectively, for the services-as-usual group.

Implications: Permanent supportive housing clearly accomplishes its goal of providing stable housing to high need families. It also holds promise for reducing families’ involvement with the child welfare system and with time children spend in out-of-home care. Conclusions based on the entire sample will be drawn on the effectiveness of the program overall and how site-level differences may explain its strengths and limitations.