Housing First Works. Let's Be Precise About What It Works For.

By Common Ladder · May 29, 2026 · 9 min read

In 1992, a psychologist named Sam Tsemberis started an experiment in New York City. The standard approach to chronic homelessness at the time was a "treatment first" staircase: psychiatric stabilization, then sobriety, then skills training, then transitional housing, then — if all went well — a permanent apartment. Most people fell off somewhere on the staircase. Tsemberis reversed the sequence. Give people housing first, unconditionally, then offer wraparound services. By the end of the five-year program, 88% of Housing First participants remained stably housed. In the city's residential treatment system, the comparable figure was 47%.¹

That original study was not a randomized controlled trial, and Tsemberis himself acknowledged its limitations. What happened in the following three decades was unusual for a social intervention: the basic finding replicated. Across sites in the United States, Canada, France, and the Netherlands, under rigorous experimental conditions, the same result kept appearing.⁶ Housing First kept people housed. Treatment-as-usual did not.

That should be the settled part of the conversation. In the current political environment, it is not — and understanding why requires being precise about what the evidence actually shows, where it is genuinely uncertain, and what the critiques get wrong.

The housing stability evidence: remarkably consistent

The largest and most methodologically rigorous Housing First trial in history is the Canadian At Home/Chez Soi study, which enrolled 2,148 individuals who were absolutely homeless and living with serious mental illness across five Canadian cities between 2009 and 2013. It was randomized, the comparison was genuine, and follow-up data extended to six years.

At one year, 73% of Housing First participants resided in stable housing, compared to 31% of treatment-as-usual participants. By the sixth year, high-need participants in Housing First spent 85% of days in stable housing, compared to 60% for the control group — a gap that persisted and widened over time, the opposite of what you would expect if Housing First were merely delaying rather than solving the problem.

These numbers are not outliers. A systematic review and meta-analysis of randomized controlled trials published in the British Medical Journal found consistent evidence of superior housing stability across studies.⁶ Scattered-site Housing First produced long-term stability effects that held across demographic subgroups and severity levels. A smaller Canadian RCT in a mid-sized city (not a major metro, where supportive conditions are often more favorable) replicated the effect at a meaningful scale.

The original Pathways model has now maintained housing retention rates of 85–90% as it has been replicated across dozens of programs over 30 years.¹ For a social intervention, this is an extraordinary evidence base. The housing stability question, taken on its own, is essentially closed.

The cost evidence: favorable but dependent on population

A systematic review commissioned by the National Low Income Housing Coalition found that the economic benefits of Housing First programs in the United States exceed their costs, with societal savings of approximately $1.44 per dollar invested. For the highest-need participants — people with severe mental illness and significant histories of homelessness — the math is particularly favorable: emergency shelter, inpatient psychiatric hospitalization, emergency department use, and criminal justice costs drop sharply when someone is housed. Net annual cost for this group, after service offsets, runs roughly $6,300 per person — down from a gross cost near $20,000.

For moderate-need participants, the cost picture is less favorable. Service utilization drops less sharply, savings are smaller, and the net cost per person rises to around $7,900 annually with a 46% offset. The implication is not that Housing First is cost-ineffective for moderate-need populations — it is still producing stable housing outcomes — but that the cost-savings argument is strongest when the population served is genuinely high-acuity.

Cost-effectiveness analyses in this field carry important caveats: savings from reduced emergency service use accrue to different budget lines (Medicaid, criminal justice, emergency departments) than the costs of housing programs, which typically come from housing and HUD allocations. The social return is real; capturing it within a single agency budget is the structural problem that makes Housing First appear more expensive than it is.

Where the evidence is genuinely uncertain

Two outcomes areas produce mixed results in the Housing First literature: mental health symptoms and substance use.

On mental health: systematic reviews consistently find that Housing First does not reliably improve psychiatric symptoms compared to treatment as usual. At Home/Chez Soi found improvements in community functioning and quality of life for participants, but changes in clinical symptom measures were modest and variable. A 2018 National Academies review found insufficient evidence of a positive relationship between Housing First and long-term improvement in mental health conditions. Researchers Stefan Kertesz and Guy Johnson, reviewing the literature in 2017, characterized the clinical impact as "modest at best."

On substance use: the results are similarly mixed. Multiple studies, including several within the At Home/Chez Soi data, find no significant difference in substance use outcomes between Housing First and control groups. Participants are housed — they are not more sober as a result of being housed, at least not within the two-to-four-year windows most studies can observe.

These findings are real, and they matter. They also need to be read correctly. Housing First was never designed to be a clinical intervention for mental illness or addiction. Its theory of change is that housing stability is a precondition for health improvement — that you cannot meaningfully address psychiatric or substance use conditions while living in a shelter or on the street. The model's claim is that housing comes first; it does not claim that housing alone resolves behavioral health conditions, or that the services offered alongside housing are sufficient to do so.

The mixed clinical evidence reflects, in part, the uneven quality of supportive services that accompany Housing First programs across the country. A program that places people in apartments and offers minimal follow-on services will produce different clinical trajectories than one that pairs housing with Assertive Community Treatment or Intensive Case Management. Treating all Housing First programs as a single intervention — which much of the critical literature does — obscures this variation.

What the backlash gets wrong

The Cicero Institute's 2024 report "Rejecting Housing First" and related state-level legislative efforts (roughly a dozen states have introduced or passed legislation restricting Housing First requirements for funded programs since 2022) make a consistent argument: Housing First has failed because homelessness has not declined and behavioral health outcomes have not improved. Therefore, programs should require sobriety, psychiatric treatment compliance, or shelter engagement as preconditions for housing.

This argument has three problems.

First, it uses Point-in-Time count trends to evaluate Housing First in exactly the same way it critiques PIT counts as a measurement problem in other contexts. National homelessness counts rising during 2020–2024 reflects housing market inflation, eviction moratorium expirations, and opioid mortality trends — forces entirely outside any program's control. PIT count trends are not a program evaluation. This is true when researchers say it to defend Housing First; it is equally true when critics use PIT counts to attack it.

Second, the treatment-first critique does not engage with the treatment-first evidence. What are the housing stability outcomes for programs that require sobriety or psychiatric compliance before housing? The data consistently show they are worse — not because abstinence-based approaches are inherently wrong, but because the majority of people with serious mental illness and addiction who cycle through shelters will not maintain continuous compliance with conditions. People lose housing under treatment-first models because life — evictions, relapses, hospitalizations, crises — disrupts compliance timelines. The critique of Housing First does not solve this problem; it recreates it with a different label.

Third, conflating Housing First policy (the federal preference for Housing First in CoC program design) with Housing First practice (the quality of services delivered) allows critics to blame an approach for implementation failures it did not cause. A poorly funded program that places people in housing without any accompanying services and calls itself Housing First is not a test of the Housing First model. It is a test of inadequate funding.

The question that isn't being asked

Neither side of the current Housing First debate is asking the most important question: are the services accompanying housing adequate?

The At Home/Chez Soi trial — the gold-standard evidence — paired Housing First with either Assertive Community Treatment (intensive, team-based community psychiatry) or Intensive Case Management, depending on participant need level. These are expensive, high-quality service models. Most Housing First programs in the United States operate with case management ratios and service intensities well below those in the RCT evidence base. The studies show what Housing First can do. They don't show what a Housing First program with three overloaded case managers per 80 clients will do.

This is the honest gap in the Housing First literature, and it is where both advocates and critics tend to go silent. Advocates because acknowledging it requires admitting that current funding levels are insufficient for the model the evidence supports. Critics because engaging it seriously would require proposing service funding rather than simply restricting housing access.

For CoC leaders, the actionable question is: does the caseload and service intensity in our Housing First programs reflect what the evidence actually tested? For funders, the question is whether the programs they support are adequately resourced to deliver services at anything close to RCT quality. For policymakers, the question is whether treatment-first preconditions are actually a proposal to serve people better, or a mechanism for serving fewer people while producing worse housing outcomes.

Housing First is the right approach. The evidence on that is not close. The right approach, underfunded and under-resourced, is not a refutation of the approach. It is a funding decision — one that keeps getting mistaken for a philosophy debate.