The Impact of Cash Transfers for Rapid Rehousing Clients

Beginning in Spring 2024, clients who exit Abode Services’ RRH programs will have the option to participate in a cash transfer program. 2-3 months before RRH exit, case managers will ask their clients if they are interested in participating in a program that provides monthly cash payments of $50 or more for one year.
Once the RRH exit date has been confirmed and the client has chosen to participate in the program, they will be randomized into one of two groups. Those who are randomized into the control group will receive $50 per month for 12 months. The treatment group will receive much larger monthly payments ranging from $800 to $2,000 depending on family type and month in program. These payments are roughly comparable to the average payment made by Abode’s RRH programs ($1,400) and will on average cover one-half to two-thirds of monthly rent ($2,100). Motivated by feedback from prior Abode RRH clients, the payments will be larger in the initial months than in later months to help with bigger expenses such as car repairs.

2024-05-06

2026-05-06

Homelessness

We are primarily interested in whether cash transfers reduce homelessness. The homelessness outcome will be defined as any use of homelessness services within 12 and 24 months after random assignment as recorded in HMIS, measured using an indicator variable. Alternative measures include use of emergency shelters and/or street outreach as identified in HMIS.

Housing Stability
Financial Wellbeing

Our secondary outcome of housing stability will be measured primarily by an indicator for any address move within 12 and 24 months after random assignment based on Infutor address data. We will also measure housing stability via indicators for moving to higher socioeconomic status neighborhoods, former address ending, and new address beginning, as well as neighborhood characteristics of the most recent address. We will also consider an index of these outcomes.

We will measure financial well-being via an indicator for whether a participant’s credit score increased within 12 and 24 months after random assignment as recorded in Experian. We will also use the following alternative measures based on Experian data: credit score, an indicator for a credit score decrease, account balance, delinquent accounts, collections, debt, and credit inquiries. We will also consider an index of these outcomes.

Following the construction of this study as a randomized control trial, those assigned to treatment and control groups should look equivalent to each other on average. Thus, any difference in outcomes between the groups could be attributed to their treatment status. After applicants have been randomized into these groups and the study begins, LEO will monitor outcomes into the future and compare them. For our analysis, we will estimate differences in outcomes between treatment and control group applicants using an intent-to-treat (ITT) framework controlling for baseline covariates. Baseline covariates will come directly from Abode’s administrative data or from data collected during the study intake process.

Specifically, we will estimate the following equation:
yi =β0 + Tiβ1 + xiβ2 + εi
where yi represents key outcome variables, such as returns to homelessness in the time period of interest, and xi represents a vector of observed characteristics for person i. The variables in x will include baseline characteristics such as age, age squared, categorical race variables, and indicators for ethnicity and gender. It will also include randomization strata fixed effects (county of residence and single/adults-only vs. household). The key covariate in the analysis will be the indicator variable Ti which equals 1 if the respondent is assigned to a particular treatment arm and zero if in the respective control group. The term εi is an error term. We will measure the outcomes every year after randomization, determining short- and long-term effects of the intervention. Treatment-on-the-treated (TOT) effects will be estimated via an instrumental variable (IV) model, using assignment to treatment as an instrument for participation in services.

Not available

Computer-based randomization

Individual

No

545 families

545 families

327 control families; 218 treatment families

We will be powered to detect an 8.87 percentage point (31.6 percent) decrease in the probability of becoming homeless 12 months after exit from RRH, assuming an untreated mean of 28 percent. Given the nature of the intervention, we expect take-up to be very high. One group which may not want to participate in the study is those receiving SSI benefits, because the study transfer payments may compromise their SSI eligibility. Since about 4 percent of Abode’s RRH clients receive SSI benefits, we conservatively assume that all SSI recipients in the treatment group will decline to take up the cash transfers, resulting in a 96 percent take-up rate.

Pre-Analysis Plan