Housing First Policy: What 3 Decades Show (2026)

The objection to housing first policy is intuitive and almost always the same: give an apartment to someone using drugs and you have funded the drug use. Fix the person, then house them. It sounds like discipline.

It has been tested against the alternative in randomized trials, and the intuition loses. Not narrowly. The counterintuitive part is that the sequence itself is the active ingredient.

What is Housing First, exactly?

Sam Tsemberis founded Pathways to Housing in New York in 1992 on a simple inversion. The prevailing model was a staircase: shelter, then transitional housing, then permanent housing, with sobriety and treatment compliance required at each landing. Fail a step and you go back down.

Housing First removes the staircase. Permanent housing comes first, with no sobriety requirement, no treatment mandate, no minimum income. Services are offered continuously and voluntarily. Tenancy is not conditional on accepting them.

Three decades of implementation across the U.S., Canada, and Europe have followed. Finland went furthest, converting its shelter system to permanent units and cutting homelessness for more than a decade, the longest sustained decline in the EU. That trend reversed in 2024, and Finnish homelessness rose about 20% in 2025 (ARA).

Why does the sequence matter so much?

Because everything treatment requires, homelessness destroys.

Storing medication requires a place to store it. Keeping an appointment requires knowing where you will sleep the night before. Managing a psychiatric condition requires sleep, and shelters are loud, crowded, and time-limited. Recovery from addiction requires stability, and the street supplies none.

Asking someone to complete treatment before receiving housing asks them to solve the harder problem first, without the conditions that make it solvable. The staircase model was not cruel by design. It was sequenced backwards, and it filtered out the people with the highest needs at every landing.

What did the randomized trial find?

At Home/Chez Soi is the largest randomized controlled trial of Housing First ever run. Between October 2009 and June 2011, researchers recruited more than 2,000 homeless adults with mental illness across Vancouver, Winnipeg, Toronto, Montreal, and Moncton, randomly assigned them to Housing First or to usual services, and followed them for up to 24 months. The cost figures below come from the 950-person high-need arm.

Random assignment is what makes this hard to argue with. The two groups were not selected for motivation or readiness. A coin decided. The gap that opened between them is attributable to the model, not to who signed up for it.

Does it cost more or less?

Less than opponents assume and more than advocates sometimes imply. The honest answer is that most of it pays for itself.

The cost analysis of the high-need arm, published in Psychiatric Services in 2020, found about 69% of the intervention's cost offset by reductions elsewhere: emergency shelter nights, hospital admissions, and other crisis services that homelessness generates. Across the trial's sites, every $10 invested returned roughly $9.60 in avoided costs for high-need participants and $3.42 for moderate-need participants.

Read that carefully, because the framing matters for anyone deciding where money should go. Homelessness is already expensive. Emergency rooms, jail beds, ambulances, and shelters are the current budget line, and they buy no exit. Housing First redirects most of that spending toward something that produces a housed person at the end. The remaining 31% is the real price of the policy, and it is small next to what the status quo costs to accomplish nothing.

Where does Housing First fail?

Two criticisms survive scrutiny, and pretending otherwise weakens the case.

The first is arithmetic. Housing First cannot place people into housing that does not exist. The National Low Income Housing Coalition counts a shortage of 7.2 million affordable and available rentals for extremely low-income households, or 35 units for every 100 who need one. The Center on Budget and Policy Priorities finds only about one in four eligible households receives any federal rental assistance, with waits running years in most large housing agencies. A program designed to hand someone keys, operating in a market with no keys to hand out, will underperform its trial results. Finland's success came with a building program attached. Most American implementations did not get one.

The second is clinical. Critics argue the model under-invests in treatment for addiction and serious mental illness, and that voluntary services leave some participants housed but unwell. Federal funding criteria began shifting toward treatment-participation requirements in 2025, moving money in the Continuum of Care program away from permanent housing and toward mandated services and short-term interventions. States and provider groups sued, and a federal court injunction has blocked parts of the plan. The dispute is live and unresolved.

The evidence on that second point is more mixed than either camp says. Housing First produces large, replicated gains in housing stability. Its effects on substance use and psychiatric symptoms are smaller and less consistent across studies. That is a real limitation. It is also not an argument for the staircase, which performed worse on housing stability and did not deliver better clinical outcomes either.

What would make housing first policy work at scale?

Units. That is the whole answer, and it is unglamorous.

HUD's 2025 point-in-time count found 745,652 people homeless on a single night in January, including 266,320 unsheltered. The count fell 3.4% from 2024, the first decrease since 2016, and still sits about 27% above 2013. Housing that many people means producing hundreds of thousands of units, in a country that removed millions of low-cost rentals from its stock and made building replacements difficult through zoning and through the reasons it stopped building at all.

The supporting machinery already has names: housing vouchers, funded for a quarter of who qualifies, and public housing, which the country spent forty years demolishing. The pressure feeding the whole system is the cost-burden line that 22.7 million renter households now sit above, and the direct link between housing costs and homelessness.

Housing First is not charity and it is not naivety. It is the finding that stability precedes recovery rather than following it, and thirty years of data keep confirming it. What defeats the model is never the model. It is a country that decided to fund a quarter of the assistance its own rules say people qualify for, zoned out the cheap housing the policy depends on, and left the wage floor at $7.25 since 2009 while rents climbed past every safety margin a working household had. Then it looked at the shortfall and called the policy a failure. The evidence says the policy works. The units are what never showed up, and that is one of the clearest chapters in the broken American Dream.