A National Study of Targeted Homelessness Prevention

Eligible clients at imminent risk of homelessness are randomly assigned to one of two groups. The treatment group receives Right at Home (RAH) homelessness prevention services, which include rapid, flexible financial assistance (e.g., for rent arrears, forward rent, utilities, or other basic expenses), case management, and legal support. The control group is referred to other resources available in their community. The eligibility criteria at each site are designed to identify people who are unlikely to receive homelessness prevention services outside of RAH. Randomization occurs after a client has been screened for eligibility, walked through the consent process, and completed a baseline survey. Randomization is automated via a computer and case managers have no control over assignment.

2026-07-01

2030-01-01

1. Any homelessness, as indicated by administrative records of program use: an indicator for any new enrollment in an HMIS program where enrollment indicates homelessness (e.g., emergency shelter, street outreach, transitional housing, rapid re-housing, permanent supportive housing) in the period following random assignment.

2. Any homelessness, as indicated by follow-up survey response: an indicator for a housing situation meeting the McKinney-Vento definition of homelessness at the time of follow-up survey.

For the administrative measure, new HMIS program enrollments will be used to construct an indicator of homelessness. Programs included are those where enrollment indicates the client is experiencing homelessness (e.g., emergency shelter, street outreach, diversion, transitional housing, safe haven, rapid re-housing, and permanent supportive housing), as well as any new enrollment in HMIS assessment projects used when these programs are oversubscribed. Prevention program enrollments are not included. For the survey measure, we will use a broader measure of homelessness following the McKinney Vento definition.

This is an individual-level randomized controlled trial conducted at up to ten sites across the United States. At each site, eligible clients who consent to participate are randomly assigned to either receive Right at Home homelessness prevention services (treatment) or to be referred to other community resources (control). Randomization is automated via a computer at the time of enrollment, and case managers have no control over assignment. We will estimate ITT and TOT treatment effects within each site. Treatment effects will be estimated within each site and then pooled across sites using a weighted average (weighting by sample size).

Not available

Randomization is conducted by computer at the time of enrollment. After a client completes the consent process and baseline survey, the case worker initiates randomization through the HMIS enrollment workflow. An automated random number generator assigns the client to either the treatment or control group. Case managers have no control over or knowledge of the assignment prior to randomization. The probability of treatment may vary across sites and, within sites by location (ZIP code/census tract) in order to enable estimation of community-level effects of the program.

Individual (household). The trial is not clustered at the primary level of analysis, though we anticipate building in cross-randomized treatments to elements of implementation such as saturation within a neighborhood, payment method, targeting method, etc. Analysis of any such treatments will be covered in separate analysis plan(s).

No

Not applicable, randomization is at the individual level.

20,000 – 30,000 participants across all sites (individuals/households).

We plan to enroll between 1,000 and 1,500 participants per site in the treatment group, with an equal or greater number in the control group, across up to ten sites. This yields an anticipated total of approximately 10,000 – 15,000 treatment group participants and 10,000 – 15,000 control group participants nationally (20,000 – 30,000 total).

The primary outcome is any homelessness as measured by HMIS administrative records. We base our power calculations on the homelessness prevention study in Santa Clara County (Phillips and Sullivan, 2025). In this study, 7.2% of those assigned to the control group (i.e., not offered targeted homelessness prevention services) experienced homelessness within 12 months of enrollment. Given a total sample of 20,000 (10,000 people in treatment and 10,000 in control) and a baseline homelessness services entry rate of 7.2%, a two-sided test at the 5% level would have 80% power to detect a 1 percentage point (14%) reduction in homelessness 12 months after random assignment. This power calculation is for a simple difference in means of the outcome across groups. If we estimate the treatment effect in a regression framework that controls for baseline characteristics, we reduce residual variance, yielding an even smaller minimum detectable effect. As a benchmark, one-time financial assistance in Chicago reduced shelter entry by 76% within 6 months of calling for assistance (Evans, Sullivan and Wallskog, 2016) and targeted homelessness prevention in Santa Clara County reduced homelessness by 70% within 12 months of randomization (Phillips and Sullivan, 2025). Given these prior estimates, the present national study is well powered to detect even small changes in homelessness. Repeating this analysis for each site, where each site has a total sample of 2,000 (1,000 people in treatment and 1,000 in control), yields a minimum detectable effect of 3.2 percentage point (44%) in the reduction of homelessness 12 months after random assignment. This means that the study is well powered to detect large but reasonable effects on homelessness at the local level as well.