What Housing First Can and Cannot Do

Housing First inverts the older approach. For decades the rule worked like a staircase. A person had to prove sobriety and complete treatment before they earned housing. They moved from the street to a shelter, then to transitional housing, and only then to a place of their own. Housing First treats that order as backwards. It offers a permanent home first, with no precondition, and makes the home the platform from which recovery becomes possible rather than the reward for having already recovered.

In practice the model rests on a few pillars: immediate, low-barrier housing with no sobriety or treatment test; the tenant's own choice over where they live and what help they accept; and a strict separation of the housing from the services, so that support is offered, never required, and missing a counseling appointment cannot cost someone their home.

One distinction does most of the work in the argument. Housing First was designed for people with high and complex needs, the chronically homeless minority entangled with serious mental illness or addiction. The model is delivered as permanent supportive housing with intensive services attached. It was never meant to be the answer to every kind of homelessness. Most people who lose housing are pushed out by cost, not illness, and are better served by shorter-term rapid rehousing. Judging the model by a job it was not built for is the most common mistake in the debate.

On its core test of housing stability, the evidence is about as strong as social policy research gets. The founding New York trial in the 1990s kept roughly four in five participants stably housed, far above the sobriety-contingent group it was tested against. Canada's At Home / Chez Soi study was a randomized trial across five cities with more than 2,000 participants. It found 62 percent of Housing First tenants stably housed against 31 percent in usual care. A 2020 systematic review in The Lancet Public Health, which drew on 72 articles including reports on 15 studies of permanent supportive housing, confirmed the pattern. Six years on, people with high support needs who got permanent supportive housing were 1.42 times as likely as those in usual care to be stably housed.

These are not modest effects, and they have been reproduced across countries, study designs, and decades. This is the most thoroughly replicated result in the entire field. Whatever else is contested, the claim that Housing First durably ends homelessness for the people it enrolls is not.

Here the evidence cuts against the model's loudest boosters. Housing First does not, by itself, make people less addicted or less mentally ill. The founding trial found no difference between groups in alcohol use, drug use, or psychiatric symptoms. A 2019 meta-analysis of randomized trials reached the same conclusion: the effect on mental health was statistically indistinguishable from zero. The researchers' own summary is worth quoting in spirit: use Housing First with confidence to get people housed, but do not count on it to fix addiction or mental illness.

This is a real limit, not a technicality. On this point, critics are right. An apartment removes the chaos that makes treatment nearly impossible, but it is not treatment. The people Housing First serves still need clinical care, and the model is explicit that such care is offered alongside the housing, never as a condition of keeping it. The mistake is to expect the housing to do the clinical work on its own.

Housing First is often sold as a money-saver, and the claim is true in a narrower way than the slogan suggests. The savings are real but concentrated in the highest-cost people, and only when the housing is paired with intensive services. For the most expensive clients, the kind who cycle through emergency rooms, psychiatric beds, and jail, the savings on other services can more than cover the housing. For the trial's moderate-need group, the savings repay only part of what the program spends, so on balance it adds cost.

The Canadian trials make the pattern concrete, and they are more sober than the slogan. For clients with moderate needs, who had case managers rather than an assertive treatment team, the savings on other services, counting costs paid by charities as well as governments, offset only about 46 percent of what the program spent per person. Even for high-need clients with the full team, the program came only close to paying for itself. The savings are real, but for most tenants Housing First is better understood as cost-effective than as cost-saving. It buys housing and stability at a defensible price per day housed, not a check that repays itself.

Many headline savings come from before-and-after studies with no comparison group, and those tend to overstate them. People typically enter the program in a crisis, when they are using a lot of services, and for many of them costs would have fallen anyway. The Canadian trial measured its savings against a randomized usual-care group, so that natural drop is already taken out. A British meta-analysis, pooling two U.S. trials, found Housing First tenants were hospitalized markedly less often. In the Canadian trial, fewer psychiatric hospital stays were the largest single saving for the costliest tenth of participants. The lesson is that the support services, not the apartment alone, are the active ingredient, and underfunding them is what flips the math.

Many documented Housing First failures are not failures of the model. They are failures of funding and fidelity. The expensive half of the model is the services, and that is the half that gets cut first. When a program keeps the cheap structural pieces, no preconditions, fast move-in, but starves the support layer, it quietly becomes something else: housing only. People are placed and then left without the help the model assumes, and the results suffer accordingly.

The large U.S. veterans' rollout showed exactly this shape, hitting high marks on the quick structural measures while lagging on the services and recovery side. European researchers warn that services carrying the Housing First label without its substance risk bringing discredit on the approach. The practical warning for any community is blunt: never fund housing only for high-need residents. If the services are not funded, it is not the model that failed.

This is the hardest fact for the model's defenders, and the one critics deploy most often. Even as Housing First works for the individuals it houses, national homelessness climbed for years. On a single night in January 2024, a record 771,480 people were homeless in the United States. That was up 18 percent in a year, with chronic homelessness at an all-time high. The January 2025 count brought the first national decline since 2016, down to 745,652, but the drop came almost entirely from families. The part this debate is about barely moved. Chronic homelessness among individuals reached a new record, up 81 percent since 2013. Unsheltered homelessness slipped 3 percent in the year but was still 36 percent above 2013, close to its 2024 record. If the model works, the argument goes, why are those numbers still so high?

Because the model and the count answer different questions. Housing First works on the outflow, the people it moves into homes. It does little about the inflow, the people still falling into homelessness, and that inflow is governed by the housing market, not by any clinical program. When rents outrun wages, more people enter homelessness than any rehousing system can move through it. An individual-level remedy cannot solve a population-level problem set by supply. The sharpest version of the critique goes further than the market. The economist Kevin Corinth found that adding permanent supportive housing barely moves the count at all. Across U.S. communities from 2007 to 2014, his estimates suggest it took ten or more new beds to lower the homeless count by one person. One possible reason he gives is that some of the people it houses would have found their own way out in time. Read carelessly, that sounds like an indictment. Read carefully, it is the same point from the other direction: a program built to house a specific high-need group was never the thing that decides how many people fall into homelessness in the first place. That is not a knock on Housing First. It is a reminder of the job it was never assigned.

Finland bent its national number for more than a decade, and how it did so is the clearest lesson in the debate. Its count of homeless people living alone fell from nearly 8,000 in 2008 to about 3,400 in 2023. Between 2012 and 2016, while that count fell, homelessness was rising almost everywhere else in Europe. Part of the drop came from a change in how Helsinki counted in 2018. Long-term homelessness, the group the national strategy set out to halve, fell by about 70 percent from 2008 to 2024, by municipalities' own estimates, which the national report says may run low. The count has since turned. It rose in 2024, and again in 2025 to about 4,600, the largest one-year rise Finland has recorded. It is still the closest thing to proof that the aggregate count can fall, and a warning that it can climb back.

But the lesson is not that the clinical model is magic. Finland converted shelters into permanent apartments, invested in prevention, and, above all, built and supplied affordable housing at scale through a national strategy. That strategy carried the Housing First name, and the model was genuinely at its center. But the engine underneath it was housing supply. Without the apartments to convert and let, the model would have had nowhere to place anyone. For a place like Asheville, that means a whole community can end homelessness. But doing it depends on the cost and availability of housing, the one thing the clinical model alone cannot change.

For two decades this was a roughly bipartisan answer. It is not treated that way now. In 2025 the federal government began steering its homelessness money away from Housing First and toward treatment-conditioned models, on the argument that the approach failed. Everything above is what that argument runs into: the model being defunded is the better-proven one, and no rigorous trial shows the replacement does better on housing and health together.

That does not make the critics wrong about everything. Housing without the funded services is insufficient, and the street disorder driving the politics is real. They are also right that a rigid federal preference for one model crowded out useful alternatives. For a decade the money rewarded Housing First above transitional beds, interim shelter, and recovery housing, and communities lost room to build the mix their own streets needed. Loosening that is fair. But loosening it is not the same as making a home conditional on treatment, and that is the line the evidence draws. The fix is to pay for the services alongside the housing, which is what the evidence rewards, not to demand treatment as the price of a home. We trace the policy turn itself, and what it puts at stake in Asheville, in Strings Attached.