Homelessness Prevention through the School System: An Impact Evaluation of Hennepin County’s School to Housing Program

The School to Housing (STH) program we study is operated by Hennepin County, Minnesota, which uses the public school system to identify families at risk of homelessness and coordinates services across housing and social support agencies. It is administered by trained case managers and consists of three components: housing search support, short-term rental assistance, and comprehensive case management services. The program is designed to provide consistent and tailored support to families who are precariously housed—meaning doubled up with another household or facing eviction— and at imminent risk of entering the homelessness system.
STH is designed to address both families’ immediate needs for housing and the underlying factors contributing to housing insecurity, such as income instability, lack of access to services, and limited social support. The intervention begins with active support in identifying stable housing options, including help navigating the local housing market. Once housing is secured, participants receive up to nine months of rental assistance. Throughout the program, families receive personalized case management to address barriers to housing stability and promote long-term financial security. Case management includes the development of individualized action plans, regular check-ins, and referrals to additional community-based resources as needed. These services are designed to become progressively more hands-off over time, with the housing support tapered so that families gradually pay more of the rent over time, to help families transition to self-sufficiency. In a second treatment arm, families receive state housing vouchers that are not tapered and continue to be valid even after the case management ceases. Participants assigned to the control group will receive standard referrals to community housing programs.

2025-09-12

2028-10-01

Our analysis will focus on several outcomes related to housing stability and education These outcomes will be tracked at multiple intervals to evaluate both short-term and long-term effects of the intervention. We will also estimate models that exclude the case management period to separately test for any lasting effects of the program.
A. Homeless Services Use
We will use administrative data from Homeless Management Information System (HMIS) for Continuums of Care throughout Minnesota to evaluate the program’s impact on homelessness. HMIS service usage will include the following components:
Any homeless services: We will define an overall homeless services indicator equal to one if a household uses any HMIS-based services and zero otherwise.
Emergency shelter stays: We will analyze both an indicator for whether the household entered an emergency shelter and the total number of emergency shelter days during annual follow-ups after randomization. Days in shelter will be coded as zero for households that do not enter shelter. This outcome measures the program’s effectiveness in preventing shelter entry and reducing time spent in emergency housing.
HMIS helpline calls: We will consider both an indicator for whether the household placed a new call to the homelessness helpline and the number of calls during each follow-up window. This outcome captures continued housing instability and the need for crisis assistance.


B. McKinney-Vento Student Classification
We will examine whether any school-aged child in the household is classified as McKinney-Vento (i.e., homeless or highly mobile) in the school years following randomization. (By definition, all students participating in the intervention should be classified as McKinney-Vento in the first year of the study.) This outcome, which is recorded by schools, captures housing instability outside the formal shelter system.
C. Housing Court
We will study two outcomes using housing court records:
Any housing court record: An indicator equal to one if a household was ever sued in housing court.
Eviction: An indicator equal to one if a household ever faced a formal eviction order.
D. Residential Moves
We will measure housing instability using the following measures:
Any move: An indicator equal to one if a household changed addresses during a given time period.
Number of moves: A count of household address changes in a given time period.
E. School Absences
Using individual student records from the Minnesota Department of Education, we will measure the number of school days missed. This will include both total absences and an indicator for chronic absenteeism. This outcome serves as the primary measure of educational stability and child well-being.
F. School Changes
Using the same student records, we will define an indicator equal to one if a student experiences a non-structural school change during a given period (i.e., unplanned school changes not associated with leveling up from, e.g., middle school to high school).
G. High School Graduation
We will define an indicator equal to one if a student graduates high school and zero otherwise.

Involvement in Other Homeless Services and Programs
Criminal Justice Involvement
Educational Outcomes
Public Benefits Participation
Credit and Financial Well-Being
Child and Family Well-Being

We plan to track involvement in other homelessness-related services and programs using detailed administrative data from the Homeless Management Information System (HMIS), including non-shelter engagements such as coordinated entry enrollments, transitional housing stays, and other service episodes not captured under primary outcomes.
Criminal justice involvement will be assessed through records from the Minnesota Department of Corrections and the Minnesota State Court Administrative Office. Outcomes will include adult incarceration history and juvenile court interactions.
Additional educational outcomes will be evaluated using administrative records from the Minnesota Department of Education, including standardized test scores (Minnesota Comprehensive Assessments), grade completion, and in-school disciplinary infractions such as suspensions.
Public benefits participation will be tracked using data from the Minnesota Department of Human Services, including receipt of SNAP, MFIP (Minnesota’s TANF program), Medicaid, SSI, child care assistance, and disability support programs. These outcomes reflect economic stability and access to the social safety net.
Credit and financial well-being outcomes will be measured using Experian credit reports. Key indicators include credit scores, delinquent balances, debt levels, utility arrears, and mortgage activity.
Child and family well-being will also be assessed using child protection records and children’s mental health service data from the Social Services Information System (SSIS), focusing on involvement in child welfare services and documented behavioral or emotional health concerns.
With the exception of the credit report outcomes, all outcome data will be provided through Minn-LInK, a data center in the Center for Advanced Studies in Child Welfare at the University of Minnesota. Credit report outcomes are from Experian credit reports.

Using a randomized controlled trial (RCT) design, LEO and Hennepin County aim to answer the research question: What is the impact of offering housing search support, rental assistance, and personalized case management services on housing stability and educational outcomes for precariously housed families? Study enrollment will begin in September 2025 and continue for three years, with outcomes tracked annually through administrative records during and after the study period.
Eligibility will be determined during the program intake process. Families are eligible for the study if they (1) reside in Hennepin County, (2) have an income at or below 30% of the area median income (AMI), (3) are either doubled up or living in a hotel or motel, and (4) have at least one child under 18 enrolled in a school in one of several eligible districts. Families meeting these criteria will be informed of the opportunity to participate in the study and asked to provide informed consent. Only those who consent will be included in the study population. Consenting to study participation does not affect their odds of being able to enroll in the Hennepin County services.
Once enrolled, participants will be randomly assigned to one of three groups: (1) the School to Housing (STH) group, which receives housing search assistance, nine months of rental subsidies, and case management; (2) the permanent rental subsidy group, which receives lifelong rental support with one year of case management; or (3) the control group, which is referred to other community housing resources. Because demand for STH and permanent rental assistance exceeds Hennepin County resources, random assignment is a fair means of allocating scarce program slots. Randomization proportions may vary over time based on fluctuations in applications and resources.
The research team will track outcomes using administrative data from Minn-LInK and Experian, covering domains such as housing stability, education, credit, public benefits participation, and criminal justice involvement. This study is designed to provide rigorous evidence on the comparative effectiveness of short-term, self-sufficiency-oriented housing interventions and long-term rental support models.
While the core program elements are already being implemented by Hennepin County, this evaluation represents the first effort to rigorously measure their impacts using a randomized design. The results of this study will support Hennepin County in refining service delivery and allocating resources effectively. More broadly, findings will contribute to the growing evidence base on homelessness prevention, particularly for families identified through public school systems.

Not available

Computer-based randomization

Unit of randomization: Household

No

1440 households

1440 households (some outcomes, such as school absences, will have more observations for households with multiple children).

Treatment 1 (STH group): 420 families
Treatment 2 (Permanent voucher group): 155 families
Control: 865 families

The study will conduct enrollment over a period of three years. During this period, we will track interim outcomes while continuing to enroll participants. The primary comparison will be between households who receive the School to Housing program services (Treatment 1), including housing search support, rental assistance, and case management, and those who are assigned to the control group. Additional comparisons will be made between households who receive the permanent rental subsidy vouchers (Treatment 2) and the control group, and between households who receive Treatment 1 and households who receive Treatment 2. For three years of enrollment, we anticipate enrolling approximately 1,440 families, with 420 families in Treatment 1, 155 families in Treatment 2, and 865 families in the control group. The primary housing outcome is HMIS contact. According to estimates from Hennepin County’s 2023-2024 HUD data and the 2023 American Community Survey, we estimate that 9.57% of low-income families with children have contacted the local HMIS hotline. With an expected program take-up of 80%, we are powered to detect a 5.78 percentage point decrease in the likelihood of contact with HMIS for families assigned to Treatment 1, which corresponds to a 60% decline. With an anticipated take-up rate of 80% for Treatment 2, we are powered to detect a 7.54 percentage point decrease, or a 79% decline, compared to the baseline mean. Additionally, we will measure school absences as a primary educational outcome. According to a Minn-LInK brief by Dupuis and Powell (2023) that examined homeless and highly mobile students in Minneapolis between 2014-2015, the mean number of absences in an academic year for this population is 14.49 days. Using standard deviation estimates that are 0.75 - 1.25 times the mean, we are powered to detect a 2.14 to 3.58 day decrease in the number of school absences in an academic year for children in Treatment 1, equivalent to a 15% to 25% decrease. For children in Treatment 2, we are powered to detect a 2.79 to 4.64 day decrease, or a 19% to 32% decrease.