Impact Evaluation of Housing Problem Solving

Housing Problem Solving is a new service being introduced in Santa Clara County which is designed to either prevent individuals from becoming homeless, or to respond quickly when individuals become homeless. When an individual is connected with Housing Problem Solving services, a trained Housing Problem Solving staff member will engage in a Housing Problem Solving conversation with the eligible callers who are randomly assigned to the treatment group. Housing Problem Solving conversations are unique to the caller’s situation and the specialist will explore best approaches and resources with the callers to solve their current housing crisis. Specialists may connect callers to services, such as for childcare, employment, mental health/counseling, food bank or clothing, and other resources. They may also provide mediation services with employers or landlords, family or friends, partners, roommates, or others to help callers retain or secure housing, whether temporary or permanent. If needed, financial assistance is provided to cover expenses such as childcare, moving costs, rental assistance, security deposit, utility assistance, transportation, and other expenses.

2022-01-04

2022-09-01

The primary outcomes we hope to measure in this project are the impacts of Housing Problem Solving on preventing individuals from entering shelters, housing stability, and credit repayment.

Homelessness:
-Primary measure: Housing Services Receipt
-Constructed using the local area’s HMIS (Homelessness Management Information System) data. Following Evans et al. (2016), we will create indicators for any services by month t following randomization, where t ranges from one to 12.
Housing Stability:
-Primary measure: Any address change since random assignment
-Constructed using address histories in the Infutor data. When analyzing the effect of HPS on Infutor outcomes, we will restrict the sample to the set of individuals who match to Infutor prior to random assignment. We expect roughly 50 percent of individuals to meet this criteria.
Credit repayment:
-Primary measure: Total balance in collections
-Constructed with Experian data listing this value over time. When analyzing the effect of HPS on Experian outcomes, we will restrict the sample to the set of individuals who match have a credit record prior to random assignment. We expect roughly 50 percent of individuals to meet this criteria.

Secondary outcomes in which we are interested include whether individuals achieve permanent or temporary housing, criminal justice outcomes, government benefit utilization, and ER visits.

We will measure our secondary outcomes using administrative data. For example, we will link our records with data on individuals’ use of government benefit programs to determine whether Housing Problem Solving has an effect on government benefit participation.

This study will be conducted using a randomized controlled trial to evaluate whether receiving Housing Problem Solving services has a positive impact on individuals’ housing stability and shelter entry outcomes. When individuals are put into contact with Housing Problem Solving services using the hotline, they will be randomly assigned to receive Housing Problem Solving or not. This randomization will take place directly in the Qualtrics form used for study intake. Those who are assigned to receive HPS, the treatment group, will be immediately provided with services. Those who are not assigned to receive HPS, the control group, will receive standard services from Santa Clara County.

Individuals in the treatment and control group will be tracked in administrative data for two years after they receive Housing Problem Solving services. We plan to track outcomes such as housing services utilized, address history, and creditworthiness and use of credit. We also hope to track secondary outcomes including criminal justice outcomes and ER visits.

Not available

Randomization will take place directly in Qualtrics. All individuals will be assigned a random number between 0 and 1, and Qualtrics will assign individuals to receive Housing Problem Solving or not based on the value of the random number they are assigned. This method will also allow us to alter the proportion of clients who receive services. For example, we will begin the study by setting all clients with a random number below 0.25 to receive services, while clients with a random number above 0.25 will not receive services. If in the future, we would need to change these proportions, we can change the threshold random number to receive services. Both individuals who consent to the study and those who do not will be randomized in Qualtrics and will have the same chances of receiving Housing Problem Solving.

Individual

No

3000 individuals

3000 individuals

-1500 individuals in treatment
-1500 individuals in control

Housing services: Assuming that 35 percent of people who utilize HMIS services return to receive additional services within 3-12 months and a target sample size of 3,000 individuals, we would be able to observe a change of 4.95 percentage points in the likelihood of returning to receive additional HMIS services. Housing Stability: To conduct these power calculations, we rely on Infutor data collected for a different project conducted by LEO, the Youth and Family Homelessness Prevention Intervention. In that study, roughly 10 percent of control group individuals experienced an address change following random assignment. Given that proportion and the expectation that half of the sample will link to Infutor data, we would be able to detect a 4.76 percentage point change in the likelihood of experiencing an address change. Balance in Collections: To conduct these power calculations, we rely on Experian data collected for a different project conducted by LEO, the Padua Pilot. That sample includes nearly 400 low-income individuals from the large metropolitan area of Fort Worth, Texas who sought services from the local Catholic Charities, similar to the sample for our current project. In this Padua sample, the average balance in collections is $4,480 and the standard deviation on the residual balance on collections (controlling for three pre-quarters of balance on collections) is $3,160. Given these parameters and our anticipated Experian-matched sample size (N=1500), we would be able to detect a difference of $457.