Cities, Cities, Health Housing, New York City, Serious Mental Illness
April 21st, 2020 46 Minute Read Report by Stephen Eide

Housing First and Homelessness: The Rhetoric and the Reality

Over the past two decades, a policy known as “Housing First” has come to dominate the government response to homelessness. Housing First has two chief tenets: (1) the most effective solution to homelessness is permanent housing; and (2) all housing for the homeless should be provided immediately, without any preconditions, such as sobriety requirements. The movement to “end homelessness,” in which hundreds of communities have participated, is centered on the implementation of Housing First.

More recently, the Trump administration has begun modifying the federal government’s commitment to Housing First. These changes have been prompted, in part, by the fact that, in California and elsewhere, community efforts to end homelessness have failed even to arrest its increase. Though the changes thus far have been modest, they have been strenuously criticized by advocates who sense a weakening in the Housing First consensus.

This report contributes to the debate over homelessness policy by assessing Housing First’s rhetoric—the claims made by proponents—in light of the available evidence. It argues that proponents overstate the ability of Housing First to end homelessness, the policy’s cost-effectiveness, and its ability to improve the lives of the homeless.


Key Findings

  • Housing First has not been shown to be effective in ending homelessness at the community level, but rather, only for individuals.
  • A Housing First intervention for a small segment of “high utilizer” homeless people may save taxpayers money. But making Housing First the organizing principle of homeless services systems, as urged by many advocates, will not save taxpayers money.
  • Housing is not the same as treatment. Housing First’s record at addressing behavioral health disorders, such as untreated serious mental illness and drug addiction, is far weaker than its record at promoting residential stability.
  • Housing First’s record at promoting employment and addressing social isolation for the homeless is also weaker than its record at promoting residential stability.


  • The U.S. Department of Housing and Urban Development should allow more flexibility from Housing First requirements for communities pursuing homelessness assistance grants through the “Continuum of Care” (CoC) program.
  • State and local Housing First mandates should be reassessed.
  • The homelessness debate should be reintegrated into the safety-net debate.

I. History of Housing First

In response to the emergence of “modern” homelessness in the early 1980s, cities first focused on developing emergency shelter programs. Shelter was emphasized in those years because the rise in homelessness was assumed to be a temporary crisis created by the 1980–82 recession, and, going back to the 19th century, temporary housing had always been part of the response to housing instability challenges.[1] Throughout the 1980s and 1990s, however, the economy improved but homelessness did not decline; in some cities, it increased. Policymakers thus began to reason that a new response was required to meet this new, and apparently structural, socioeconomic challenge.

The first proper homeless services system—as distinct from the preexisting array of safety-net programs and services—is often described as having had a “linear” character.[2] Housing programs for the homeless would be arranged in a continuum of emergency, transitional, and permanent options. Linear-style systems would guide clients out of homelessness gradually, first from the streets to shelter, then to a service-enhanced transitional housing program, and then to permanent housing, either publicly subsidized or private.[3] It was always understood that at least some of the homeless population would need permanent housing benefits—meaning a rental subsidy not subject to any time limits. But the most troubling cases, such as individuals who were mentally ill or had drug addictions, would need services in addition to housing benefits, both for their sake and to ensure the success of the housing intervention.[4]

The linear system was developed during the lead-up to the 1996 welfare reform, the Personal Responsibility and Work Opportunity Reconciliation Act. The same concerns about changing public assistance programs to promote self-sufficiency and minimize dependency also shaped the debate over the early 1990s homeless services system. A 1994 strategic plan by the United States Interagency Council on Homelessness (USICH) to “break the cycle of homelessness” began with an epigraph by President Bill Clinton about how “work organizes life”[5] and, in detailing the purpose of housing programs for the homeless, placed high emphasis on “mak[ing] housing work again.”[6] With so many people cycling between the streets, shelter, and unstable housing arrangements, a welfare reform–style emphasis on work would overcome homelessness recidivism.[7]

Policymakers in the early 1990s were also concerned about the flaws of deinstitutionalization. Transitioning the public mental-health-care system from an inpatient to a mainly outpatient model began in the 1950s, and it proceeded at an especially rapid pace during the 1970s. Deinstitutionalization’s promise of “better care in the community”[8] had been undermined by the spectacle of mentally ill individuals living on the streets who were either former patients in mental hospitals or people who would have been committed to long-term psychiatric care in earlier times. The homeless mentally ill needed not only housing but “structured care and residential support” similar to what had existed in the state hospitals.[9] To correct the mistakes of the past, the homeless mentally ill would need a variety of levels of support, depending on what stage they were at in their psychiatric rehabilitation.

The “linear” character also applied to programs designed to help homeless populations that faced substance abuse, unemployment, and other challenges that had contributed to their homelessness. Heavy focus was placed on the transitional housing model. Transitional housing provides temporary housing, like shelter, but for a longer duration—up to 24 months—and in a more service-enhanced environment.[10] Housing was considered part of an overall effort to repair broken lives and address the problems that caused or strongly contributed to clients’ homelessness.[11]

Press reports and advocates of Housing First often use the phrase “housing readiness” to describe the aim of linear programs. But housing readiness, while certainly used by some participants in the 1990s debate,[12] was not, in every case, how linear-style service providers themselves characterized their ultimate aims. Whereas Housing First providers hold themselves, most of all, to the standard of residential stability—keeping the most clients housed for the longest period—linear-style programs often viewed residential stability as secondary to larger goals of independence or health. Much like how residential treatment programs use temporary housing as a means toward the goal of sobriety, transitional housing providers always aimed at goals beyond mere residential stability.[13] This is why some have described the debate between the two approaches as one of different “paradigms”—the dispute concerns not just the best way to achieve a mutually agreed-upon goal but a dispute over which goals to pursue.[14]

The groundwork for Housing First was laid in the late 1970s, when advocates began promoting the term “homelessness,” a term that previously had never been widely in use, to pressure governments to develop more subsidized housing.[15] The belief in housing as a human right—meaning that government is obliged to provide it for anyone who cannot find housing on his own—had many adherents in advocacy circles but was antithetical to the notion of preconditions for housing benefits.[16] Housing First advocates were influenced by the “recovery model,” an approach to mental health that stresses the importance of letting mentally ill people choose their care and treatment regimens.[17] Criticisms that, decades earlier, had been leveled at the traditional asylums by Erving Goffman and others were revived and directed at the linear homeless services system.[18] Housing First advocates believed that linear programs did more to undermine independence than promote it, by placing the homeless in what they viewed as a quasi-institutional living environment. Theories of “community integration” called for decoupling housing benefits and social services for mentally ill clients.[19] Instead of transitional housing, they called for “supported” or “supportive” housing, which generally meant subsidized housing that made services available to tenants but did not require participation or have any other requirements.[20]

These concepts—housing as a human right, the imperative of personal autonomy, even for those with untreated serious mental illness, and community integration— were developed in academic articles in the 1990s and formed the theoretical basis for Housing First.

The empirical basis was developed by Sam Tsemberis, a New York–based clinician who founded Pathways to Housing in 1992. Pathways placed its mentally ill clients, all formerly homeless or at serious risk of homelessness, in scattered-site supported housing units without any preconditions. Tsemberis then did studies, including a rigorous randomized-controlled trial, on their rates of residential stability. He found that, of a pool of individuals suffering from serious mental-health disorders, clients placed in Pathways units stayed stably housed at higher rates than those placed in linear-style programs.[21]

In 2000, the National Alliance to End Homelessness launched the campaign to end the problem in 10 years. “People should be helped to exit homelessness as quickly as possible through a housing first approach,” the organization proclaimed. “For the chronically homeless, this means permanent supportive housing (housing with services)—a solution that will save money as it reduces the use of other public systems. For families and less disabled single adults, it means getting people very quickly into permanent housing and linking them with services. People should not spend years in homeless systems, either in a shelter or in transitional housing.”[22]

This campaign quickly found an ally in the George W. Bush administration, whose secretary of the Department of Housing and Urban Development, Mel Martinez, was the keynote speaker at the 2001 annual meeting of the National Alliance to end homelessness.[23] Under the leadership of USICH executive director Philip Mangano, the Bush administration began the “Chronic Homelessness Initiative,” which encouraged states and localities to create 10-year plans to end chronic homelessness.[24] (Though the formal requirements for “chronic” homeless status have changed over time, the term generally means someone whose experience of homelessness is long-term and who suffers from a disability.) It has been estimated that more than 350 states and localities endorsed, in some fashion, the goal of ending homelessness through a Housing First approach.[25] California, host to the largest homeless population of any state, made Housing First a requirement for state-funded homelessness programs in 2016.[26]

The Obama administration put out a strategic plan to end homelessness in 2010 (updated in 2015).[27] USICH assumed responsibility for defining what it would mean to “end” homelessness and for validating claims made by communities that they had “ended” homelessness for some cohort, such as the chronic or veterans’ population. Targeting resources toward specific homeless cohorts was seen as beneficial in itself and, if successful, a source of proof that ending homelessness, broadly speaking, was achievable.[28]

HUD is the most important agency in federal homelessness policy because of its responsibility to disburse billions in funds for homelessness programs to states and localities. Over time, the federal government has tightened adherence requirements to Housing First for local agencies pursuing homeless assistance funds from HUD. Figures 1 and 2 show how this has led to a dramatic shift in support from transitional housing programs— closely associated with the linear approach—to the permanent supportive housing programs favored by Housing First–oriented systems.

The Trump administration, despite departing from the Obama administration on several safety-net and poverty- policy questions, remained focused on Housing First for its first two and a half years in office. Six months into the new administration, 23 Republican congressmen sent a letter to HUD secretary Ben Carson, asking him to review his agency’s “current procedures” that follow Housing First principles and to “end the recommended scoring guidelines that currently punish programs that prioritize work, education, and sobriety.”[29] Much federal funding for homeless services flows through the Continuum of Care (CoC) grant competition, which is structured around a points system and set of criteria laid out by HUD.[30] In its response letter, HUD asserted that Housing First was an “evidence-based” practice and argued that its current approach was not unduly burdensome on local autonomy.[31] Carson and other prominent administration officials have made many public statements in favor of Housing First.[32] Most critically, HUD’s Notice of Funding Availability (NOFA), the annual document that lays out requirements for access to billions in CoC program funds, kept in the Obama-era language regarding Housing First.

In summer 2019, the Trump administration began to signal a shift. The first notable change came in the 2019 NOFA, which “Provid[ed] Flexibility for Housing First with Service Participation Requirements.”[33] In the section “CoC Coordination and Engagement” (VII.B.1 in the FY18 NOFA, VII.B.6 in the FY19 NOFA), the seven points allocated for embracing “Housing First” were, in FY19, dedicated to “Low Barriers to Entry” (Figure 3). The intention of the change was for localities to discourage service providers from attaching sobriety requirements or other preconditions to clients’ initial entry into a federally funded housing program but allow for their usage in clients’ ongoing participation in programs.

In late 2019, prompted by advocates,[34] Congress required HUD to return to the FY18 language for the 2020 NOFA.[35] In the meantime, the Trump administration has been active in questioning Housing First on other fronts. In September 2019, the Council of Economic Advisers (CEA) released a comprehensive report on homelessness policy in America that included a critical discussion of Housing First’s limitations.[36] In December, a new USICH executive director was appointed, Robert Marbut, an adherent of the older, linear approach (“I believe in Housing Fourth”).[37]

The Trump administration has pursued these changes partly because of philosophical objections to the Housing First philosophy but also because so many communities that participated in the campaign to end homelessness, such as Los Angeles and San Francisco, are now dealing with crises of unprecedented magnitudes. The failures of California jurisdictions’ 10-year plans to end homelessness in some form have been covered in a number of press outlets.[38]

California is host to approximately one-fourth of the nation’s total homeless population and half of the nation’s total unsheltered population. Since 2010, California has added more than 25,000 PSH (permanent supportive housing) units, an increase of about two-thirds (Figure 4)—yet the state’s unsh[39] This has inevitably raised questions about the Housing First approach that has been in place through this recent rise in homelessness. Therefore, now is a good time to take stock of Housing First. How effective has Housing First been? Does it deserve the wide acclaim it has received from advocates?

II. “We Know How to End Homelessness”

Housing First has evolved somewhat.[40] Originally, it was associated with providing permanent supportive housing for the chronically homeless. That remains a core priority of Housing First–oriented homeless services systems, but, more recently, USICH and advocates have encouraged governments to view Housing First as a “whole system orientation.”[41] All homeless services, for all homeless populations, temporary and permanent housing alike, are expected to conform with the Housing First philosophy. In addition to expanding permanent supportive housing, the top priority of any Housing First system, emergency shelter should also be provided without any barriers (see, for example, San Francisco’s Navigation Centers, Los Angeles’s Bridge program, and New York City’s Safe Haven shelters).[42] “Rapid Rehousing”—short-term rental assistance to be used for a private apartment—is also seen as part of a Housing First–oriented homeless services system, though it is a temporary benefit.[43] So, too, is providing standard affordable housing— understood as subsidized housing without any time limits—to non-chronic homeless clients, such as families, as long as it is provided without any barriers.[44] Housing First systems work to “align” or “integrate” existing affordable housing programs with homeless services, meaning, for instance, preferential access for the homeless for Section 8 vouchers or newly developed affordable housing units.[45]

Proponents argue for organizing homeless services systems around the principle of Housing First based on scientific evidence, not only, or even mainly, because it is founded on more just or humane principles. In their view, Housing First has been “proven” or “demonstrated” to be superior to alternatives and to be able to end homelessness.[46] In most instances, when a policymaker is making some claim about how “we know how to end homelessness,”[47] they are referring to the social science evidence base behind Housing First.

At their core, these claims are based on studies that have registered high rates of residential stability when homeless individuals, or people at serious risk of homelessness, have been placed in permanent supportive housing units under a Housing First policy. Residential stability may be measured in terms of how many days someone spends in his unit over a particular period, or whether he still occupies his unit at a certain time benchmark.[48]

The “gold standard” in social science research is the randomized-control trial (RCT). In an RCT, researchers examine the effect of some intervention on two different cohorts who are similar in every important respect. Though the literature on Housing First is significant, the number of truly rigorous RCT studies of the approach is relatively small. One 2015 review credits only four, with several more studies having a “quasiexperimental” design.[49] A 2014 survey identified seven RCTs and five “quasi-experimental” studies.[50] A 2017 survey of the literature credits 14 RCTs, based on 12 trials.[51] The best-known RCTs are the Pathways studies discussed earlier and the more recent At Home / Chez Soi, which encompassed five Canadian cities and more than 1,000 participants. One common criticism of the literature on Housing First is that studies often relate few details about the programs under examination (a significant concern for a policy that advocates are trying to scale up and expand nationwide).[52]

Still, despite certain limitations, the Housing First literature has demonstrated that Housing First interventions tend to yield high rates of residential stability.[53] The rates of residential stability are often in the 70%–80% range, for the length of the trial, which typically lasts a couple of years. “Usual care” or “treatment first” comparison groups, by contrast, often register rates below 50%. And, to reemphasize, these studies typically involved “chronic” homeless cases suffering from serious mental illness or some other behavioral health disorder. Whether looking at how many days housed as the measure of residential stability, or how many participants remained in housing at the end of the study, Housing First–style interventions have demonstrated real strength at addressing homelessness.

While it may have been the case 30 years ago that homeless policymakers doubted whether people with untreated serious mental illness and other social challenges could hold on to their housing if those challenges were not addressed first, there is less doubt about that point now. This is the thinking behind claims about how the Housing First literature “proves” how to “end homelessness.”

The ability of Housing First programs to keep the homeless housed at a higher rate than linear-style programs has been acknowledged by, among others, the Trump administration’s CEA.[54] The Trump administration also acknowledges that homelessness is, in large measure, a housing problem.[55] Any community that experiences a shortage of rental units affordable to low-income households will, all other factors being equal, experience higher levels of homelessness than communities with a larger store of such units.[56] Nor is there serious dispute that some of the homeless population, such as those with serious mental illness, will need rental subsidies for the rest of their lives.

But claims that Housing First has been shown to end homelessness elide the distinction between evidence at the individual level and the community level. Housing First advocates’ rhetoric that investing in permanent supportive housing will end homelessness raises hopes of ending homelessness at the community or national level. For example, Los Angeles County’s Measure HHH,[57] which authorized $1.2 billion in bonds to build thousands of permanent supportive housing units, had the working title “Housing and Hope to End Homelessness.” However, as noted above (Figure 4), California’s experience has been increased investment in permanent supportive housing and increased homelessness. Given that, according to advocates, hundreds of localities have adopted Housing First, one might have expected at least a handful of examples of communities where Housing First has eliminated or drastically reduced homelessness in a manner noticeable to the broader public. That has not been the case.

Scholars who have studied the community-level effects of increased investment in permanent supportive housing have found that: (1) governments may need to create as many as 10 units of permanent supportive housing in order to reduce the local homeless population by one person;[58] and (2) a certain “fade-out” effect is observed whereby the reduction is only temporary. There is no scholarly consensus as to the weakness of Housing First on community-level rates of homelessness. But it does show that scholarship conforms to people’s experiences: more investment in PSH does not necessarily lead to less homelessness.

As noted, many participant communities in the campaign to end homelessness have targeted a specific cohort, such as the chronic homeless or veterans. Utah[59] is perhaps the most touted success story from the campaign to end homelessness. But in a 2015 study, economist Kevin Corinth showed how claims about Utah’s “ending” homelessness can mostly be ascribed to methodological changes and shifting definitions of “chronic” status.[60] In 2009, Utah adjusted its “point-intime” homeless numbers to reflect only the homeless who were counted on a certain day in January, instead of an “annualized” estimate to reflect all homeless throughout the year, and abruptly ceased including transitional housing clients in its count of sheltered “chronic” homeless. Nonetheless, media and public officials continue to tout Utah as a case study in how to end homelessness via Housing First.[61] (USICH does not currently list Utah or any of its localities among the communities that have “ended” chronic homelessness.)[62] Even when the definition of “chronic” homelessness is settled, the number of chronic homeless will always face the challenge of counting the unsheltered population. Counting the unsheltered and documenting their challenges, such as what disabilities they suffer from and how long they have been on the streets, are tasks that continue to be plagued by a range of methodological difficulties that quite possibly will never be resolved.

Problems with data and definitions are one reason for giving pause to claims about the success of the campaign to end homelessness. Another is that, even if homelessness has been “ended” or reduced for one specific cohort, that does not necessarily imply progress toward ending homelessness more generally. Just as many factors cause homelessness, many factors may also be at work in reducing it, such as an improving economy or demographic changes. Many sources have claimed that a recent investment in permanent supportive housing for veterans has reduced veterans’ homelessness, and even ended it in some communities.[63] But a recent study by economist Brendan O’Flaherty demonstrated that the decline in veterans’ homelessness can largely be attributed to the decline in the veteran population of the age at greatest risk of homelessness and the nationwide decline as the nation has emerged from the last recession, not to government policy.[64]

The case of the seriously mentally ill, though less of a priority for USICH (no criteria for “ending” homelessness for this population have been issued),[65] should also be discussed. Housing First supportive housing programs target the seriously mentally ill partly because of a commitment to helping the hardest or chronic cases, but partly because seriously mentally ill individuals qualify for disability benefits. For its influential 2004 study, Pathways to Housing recruited some participants directly from a mental hospital. Indeed, requiring, or strongly urging, supportive housing clients to participate in a money-management program is one of the few infringements on personal liberty that Housing First providers countenance.[66] The number of seriously mentally ill homeless has been virtually flat since 2010, even as the number of permanent supportive housing units nationwide has increased by more than 50% (Figure 5).

USICH defines what it means to “end homelessness” and also evaluates communities’ claims for having done so. The council has published criteria and benchmarks for ending homelessness for four cohorts: veterans, chronic, unaccompanied youth, and families with children and, at present, has recognized about 80 communities for having “ended” homelessness for one of these cohorts.[67]

However, the official language and criteria regarding “ending” homelessness are not uncontroversial. Some have criticized it as “Orwellian.”[68] To give a community credit for having made homeless “rare, brief and one-time,” USICH performs an assessment of that the community’s services system. USICH examines system capacity, relative to need (number of homeless) but also whether that system conforms to Housing First. In other words, if the community’s capacity to house the homeless—as assessed by the government—matches the number of homeless, the government says that the community has ended homelessness. But that does not mean that there are zero homeless people in the community. Ending homelessness in a community does not need to mean zero homeless people.[69]

Figure 6 lists a cohort of communities that USICH currently credits for having “ended” veterans’ homelessness. These communities are, according to the most recent HUD figures, host to more than 2,000 homeless veterans. Communities with modest homelessness challenges more generally are host to as few as one homeless veteran, but others estimate that hundreds of veterans are included in their homeless populations. Most of the communities recognized for having “ended” veterans’ homelessness have at least seen a reduction in veterans’ homelessness since 2011 (the first year that CoC-level veteran data are available), though not all. In 2019, Portland/Gresham/Multnomah County Continuum of Care, the Northwest Minnesota Continuum of Care, and Norman/Cleveland County, OK all reported higher numbers of homeless veterans than in 2011, before they “ended” veterans’ homelessness.

Officials in New York and Los Angeles continue to embrace the goal of ending homelessness, as did some candidates for the 2020 Democratic presidential nomination.[70] But no community has truly ended homelessness using Housing First, and certainly not any community facing crisis-level homelessness. We would not say that a community has ended murder based upon a qualitative analysis of its police department, but rather the absence of murder. If ending homelessness must remain the goal of homelessness policy, governments should define success in a way that can be independently verified by the public. The public can observe homelessness. It cannot easily observe and analyze service systems’ capacity and competence. Thus, ending homelessness should mean the absence of homelessness, as observable to members of the public.

Brendan O’Flaherty is an economist at Columbia University and has been, for decades, one of the leading scholars of homelessness. He is known for his analysis of how housing-market dynamics account for much of modern homelessness[71] and for refuting the “Dinkins Deluge” thesis that, when New York City provided housing to shelter clients around 1990, it led, through moral hazard, to a significant increase in sheltered homelessness.[72] In a recent review of the literature, including on Housing First, O’Flaherty came to the conclusion that “we don’t know how to end homelessness. Not in the aggregate, anyway.”[73]

III. Cost-Effectiveness

One of the most famous statements in defense of Housing First came in Malcolm Gladwell’s 2006 New Yorker article “Million-Dollar Murray.”[74] This article, which the Bush administration had a hand in setting up,[75] detailed the struggles of a “high utilizer”: a man in Reno, Nevada, whose homelessness and alcoholism placed a costly burden on the local healthcare and criminal-justice systems. The central claim of Gladwell’s article was that homelessness was “easier to solve than to manage” because placing people in permanent housing will lead to less usage of other service systems—most notably, hospitals and jails, thus saving money. Similar cost-savings claims have been central to the rhetoric over ending homelessness.[76]

But in the academic literature, the cost-savings argument for Housing First is treated with more skepticism. Here is an area where RCT-level rigor truly matters. Studies that have a “pre-post” design look at the reduction in costs of hospitals, jails, and so on, that result when a cohort is moved from the streets to stable housing. Homeless people who are put into permanent supportive housing programs often have extraordinarily high health costs immediately before their placement. But someone who costs the healthcare system $100,000 in a given year is not necessarily going to cost the health-care system $100,000 every year of his adult life.[77] The reduction in costs, following a high utilizer’s housing placement, may have as much to do simply with a “regression to the mean” than the virtue of the Housing First /PSH intervention.[78]

Moreover, high utilizers such as Million-Dollar Murray and people with untreated schizophrenia who have lived for years on the street are unrepresentative of the homeless population as a whole. Not only a minority, they are a minority of the chronic homeless.[79] They are certainly unrepresentative of the “working poor” or “down on their luck” homeless often cited in the media. The 2015 Family Options Study, prepared for HUD, examined various housing interventions among a pool of more than 2,000 homeless families with moderate social needs, over a three-year period. The permanent housing intervention was more successful in achieving housing stability than temporary housing interventions, but it was also more expensive.[80]

Governments can’t save costs from people who don’t make much use of expensive service systems, to begin with. Some homeless may have low service costs because they’re “service-resistant,” a particularly significant problem for the mentally ill. Another reason that many of the homeless may be low utilizers is that they live in a jurisdiction with limited mentalhealth and substance-abuse services[81] (states vary dramatically in their investment in behavioral health).[82] “Usual care,” the control with which some studies compare Housing First interventions, can vary widely between jurisdictions. “Usual care,” in the case of New York City, means a $2 billion shelter system. But, in other communities, to build a Housing First–oriented homeless services system might mean building the first homeless services system that they ever had.[83]

This is not to say that homeless services systems shouldn’t focus on “high utilizers,” or that, in some cases, they may yield short-term savings on jails and hospitals for certain individuals. But Housing First’s success with different homeless populations has been cited as evidence of its merit as a systemwide organizing principle, applicable for the entire homeless population.[84] The evidence is weak that a systemwide application of Housing First—for the benefit of the many different types of homeless people—would generate net savings for taxpayers.

Physical Health-Care Systems. Homeless people are generally in bad health, due to rare diseases and illnesses associated with living in conditions not meant for human habitation, high rates of substance abuse, and inadequate treatment for ordinary illnesses.[85] They also make heavy use of emergency rooms and other expensive crisis services. Once they are stably housed, the homeless will be better positioned to avoid the need for costly triage treatment and instead use ordinary outpatient forms of care to prevent their health problems from becoming crises. Housing First programs will thus supposedly achieve better health at lower costs.

Evidence of the health effects of Housing First and permanent supportive housing is far less robust than many suggest. It is fair to argue that no policymaker who wants better health for the homeless can be indifferent as to whether they stay on the streets. But even assuming that Housing First improves people’s physical health, it is not clear that that would mean it saved money. People who live long healthy lives have high health-care costs.[86] Cost-efficiency arguments for smoking-cessation campaigns have been criticized for failing to take into account the fact that nonsmokers live longer than smokers.[87] Perhaps the most reasonable view was expressed in a 2018 survey of the literature by the National Academies of Sciences, Engineering, and Medicine. While still defending the view that “housing in general improves health,” this study came to the overall conclusion that “there is no substantial published evidence as yet to demonstrate that PSH improves health outcomes or reduces health care costs.”[88]

Mental-Health-Care Systems. Arguments that the mental-health-care system, which has always been expensive, holds great potential for cost savings, go back a very long time.[89] Deinstitutionalization promised better care and at a lower cost. On an annual basis, inpatient psychiatric commitment at a state-run facility can run close to $250,000.[90] But civil commitment doesn’t apply to the entire seriously mentally ill homeless population, which is itself a minority of the total homeless population (116,179 out of 567,715).[91] (Million-Dollar Murray was an alcoholic, not a schizophrenic.) Psychiatric hospitals have fixed costs that are difficult to reduce even if a few people avoided being committed as a result of receiving housing benefits.

Criminal-Justice Systems. Jails also have significant fixed costs. Over the last decade, New York City’s jail population has declined by 40% while the Department of Correction budget has increased by one-third.[92] The argument that Housing First saves money on jails dovetails with the critique of the so-called criminalization of homelessness.[93]

There is no question that enforcing quality-of-life ordinances, which are often violated by the homeless,[94] places a fiscal burden on public safety agencies. However, it does not follow that investing massively in permanent supportive housing and drastically scaling back on law enforcement would be fiscally prudent.

First, as discussed above, academic studies and the experience of jurisdictions in California have demonstrated the weakness of permanent supportive housing programs to reduce homelessness and thus presumably reduce public complaints about disorder. Second, less law enforcement carries costs, including public spaces increasingly occupied by encampments (and their attendant crime and public-health burdens) and attracting more street homeless from neighboring jurisdictions, thus increasing the demand for public services.

In any event, total law-enforcement cost savings would be very difficult to calculate, since jail is a small part of the “use” that homeless make of the criminal-justice system (very few misdemeanor offenses result in incarceration).[95] If 20 men are removed from Los Angeles’s Skid Row by being put in permanent supportive housing, how many cops would the LAPD redeploy? Quite possibly, there would be no savings.

Shelter Systems. San Francisco’s “Navigation Center” costs $100 per bed per night.[96] In New York City, shelter beds for families with children average $201.60 (an 89% increase since FY15) and for single adult shelter beds, the average is $124.38 (a 58% increase since FY15).[97] Shelter costs are high to ensure a certain level of quality, particularly security and on-site social services. For decades, and long before Housing First and its attendant social science literature, advocates claimed that affordable housing is cheaper than shelter.[98] A leading topic of housing policy debate in New York state government concerns “Home Stability Support.” This program would increase the “shelter allowance,” a permanent housing benefit to which public assistance clients are entitled. Proponents of Home Stability Support estimate that a more generous shelter allowance would cost New York City taxpayers about $27,000 less than shelter on an annual basis.[99]

But comparing temporary and permanent housing costs raises “apples to oranges” difficulties. It is complicated to compare a housing benefit that someone may well receive for decades with one that he would receive for only weeks or months. People who receive subsidized housing in tight rental markets are apt to continue using that benefit for a long time.[100] In New York City, the average length of stay for a public housing resident is 23 years.[101] In 2017, the most recent year for which there are data, only about 16% of permanent supportive housing residents moved out, and the share of long-stayers in permanent supportive housing has been steadily increasing over the years.[102] It is extremely expensive to provide a lifetime rental subsidy to someone, which is how permanent housing benefits function in the high-cost jurisdictions that now face the most serious homelessness challenges. It would be extraordinarily expensive to provide such subsidies to everyone, every year, who claims to be homeless in such jurisdictions. It would be much cheaper to provide temporary assistance to the vast majority of the homeless.

Governments that invest heavily in Housing First programs should expect the overall cost of government to rise. For some individuals, or some service systems, there may be cost offsets, but cost offsets are different from savings. A $1 investment in Housing First may be offset by 30 cents in savings on other service systems, but that still means that the government is 70 cents larger. Certainly, cost-effectiveness arguments should not lead anyone to think that Housing First investments will lead to tax reductions or somehow free up money that may be devoted to other purposes. Service systems’ costs are split between various governments and agencies and even nonprofit organizations. (This has been referred to as the “wrong pockets” problem.)[103]

Dennis Culhane, a leading homelessness researcher who was featured in “Million-Dollar Murray,” has subsequently cautioned against the risk of “overstating” the cost-savings argument. In 2008, he criticized the design quality of more than 40 cost studies based upon their small size and selectivity in populations examined, noting that “in general, the larger the sample (and presumably the more representative of adults who are homeless), the lower the average annual costs of services use.” But such studies are beneficial, he says, for showing the efficiency and accountability of homeless services systems and thus “mobiliz[ing] political will.”[104]

It is certainly the case that, in many jurisdictions where homelessness is at crisis levels, the public has shown a marked willingness to raise taxes for homeless services. Some recent, successful ballot initiative campaigns in California, such as Measure HHH (Los Angeles County, 2016), made use of cost-savings rhetoric. Whether those arguments were, ultimately, more important for the voting public than humanitarian considerations is unclear. Some scholars have questioned the benefit of distracting from the humanitarian case for investing in homeless services.[105] Certainly, for those with poor physical or mental health, it is not obvious why reducing health-care expenditures should be a standard of policy effectiveness.

In sum, the truly “evidence-based” view of Housing First, when it comes to cost savings, bears a certain parallel with residential stability. The evidence supports the view that a Housing First intervention may, for certain individuals, reduce costs, at least in the short term. But the evidence does not support any thesis about systemwide cost savings. Housing First has not been demonstrated to be capable of saving costs for entire systems any more than it has been demonstrated to be capable of ending homelessness for entire communities.

IV. The Record on Behavioral Health

HUD estimates that 16% of the homeless population exhibits “Chronic Substance Abuse” and that “Severe Mental Illness” afflicts 20%.[106] Drug addiction and mental illness drive much of the “chronic homelessness” challenge. Permanent housing is seen as a condition of recovery for this cohort.[107] One of the main recommendations that USICH made in its 2017 brief, “Strategies to Address the Intersection of the Opioid Crisis and Homelessness,” was to “Remove Barriers to Housing” by implementing Housing First.[108] But the research is ambiguous as to how much permanent housing, on its own, stimulates recovery.

In a 2019 law review article, Sara Rankin, of Seattle University School of Law, argued in favor of Housing First based on “the reality that people need basic necessities like food, sleep, and a stable place to live before attending to any secondary issues, such as getting a job, budgeting properly, or attending to substance use issues.” She wrote that the “Housing First approach views housing as the foundation for life improvement and enables access to permanent housing without prerequisites or conditions beyond those of a typical renter.”[109]

However, a 2017 survey of the literature by researchers Stefan G. Kertesz and Guy Johnson judged Housing First to have demonstrated, at best, modestly beneficial clinical impacts.[110] The Trump administration’s CEA acknowledged the research on Housing First residential stability but argued: “For outcomes such as impacts on substance abuse and mental illness, Housing First in general performs no better than other approaches.”[111] The 2018 study published by the National Academies of Sciences, Engineering, and Medicine found no strong evidence of Housing First and improvement of mental disorders, as have other surveys.[112]

Stated otherwise, the evidence for Housing First and behavioral health is far weaker than for residential stability. Some Housing First proponents, committed to the harm-reduction philosophy of recovery as a choice, are forthright about Housing First’s modest ability to address behavioral health disorders.[113] Harm-reduction policy calls for prioritizing the remediation of symptoms and the harmful effects of disorders such as opioid addiction over trying to root out or overcome the underlying disorder. More commonly, though, advocates display a rhetorical suggestiveness about the link between permanent housing and behavioral health that seems intended to convince the public of evidence that does not exist.[114]

V. Self-Sufficiency and Social Isolation

Originally, Housing First was mainly associated with the chronic homeless population who had disabilities— most notably, serious mental illness. Hence, employment outcomes were not of leading interest. But as the theory of Housing First has evolved to take on a “systemwide orientation,” applicable to the entire homeless population, it has come to be applied for cohorts that might be considered potential members of the working class. Permanent housing benefits are often likened to a “platform” from which, after having secured stable housing, people can go to pursue various other goals, such as health and employment.[115] “Optimize self-sufficiency” is an official goal of HUD’s NOFA.[116]

As noted, the large-scale Family Options Study (2015) showed robust rates of residential stability for the families receiving a permanent housing intervention. Accordingly, the study has been seen as supportive of Housing First, particularly as regards the “whole systems” orientation. But it also found evidence that housing subsidies, instead of granting recipients the freedom to focus more on employment and less on their housing instability challenges (à la the “platform” theory), actually led to diminished work effort.[117] In sum, housing subsidies increased rates of housing stability (and, as noted, at a greater cost than other interventions) but not self-sufficiency.[118] This was a troubling finding, since lack of work was one of the major social challenges faced by homeless families that participated in the study.[119] A 2012 article about Housing First cautioned that “subsidized housing may create disincentives for employment … and for independent housing … much in the way that disability benefits and public income support have been found to be associated with less employment.”[120]

Another outcome worth evaluating is social isolation, a significant cause of homelessness. HUD has noted that while, nationwide, about 13% of the U.S. population is a member of a single-person household, 65% of the sheltered homeless population is.[121] “Community integration” was one of the original goals of Housing First, which criticized the quasi-institutional character of the linear homeless services system.[122]

ProPublica’s “Right to Fail” report in late 2018, and the accompanying documentary released by Frontline in February 2019,[123] suggested that Housing First may serve more to increase social isolation than address it.[124] The report profiled a few seriously mentally ill clients of a supported housing program in New York, and how an excess of independence led to decompensation and even death. These individuals were, in some cases, stably housed, but living in apartments strewn with waste, swarming with bugs, and living with untreated infections and other health problems, and extremely isolated. “Right to Fail” did not specifically target Housing First—these were former residents of adult homes who had been placed in independent living under court order. Still, the report demonstrates that many mentally ill adults are, on the one hand, not eligible for institutionalization but, on the other, plainly not prepared for independent living.

The ProPublica study cannot be dismissed as simply anecdotal.[125] Several peer-reviewed articles and studies have questioned whether Housing First has lived up to its initial promise of “community integration.”[126] Others, to be sure, have defended it.[127] But the least that can be said is that whatever some Housing First program may have managed to achieve with respect to community integration, the evidence is far weaker with respect to that outcome than has been measured with respect to residential stability.

VI. Conclusion

The claim that Housing First is “proven” is an attempt to take homelessness policy out of the realm of ordinary policy debate. “Evidence-based” rhetoric means to suggest that homelessness policy is simply different: alternatives to Housing First are illegitimate because they are not grounded in science in the way that Housing First is. This is not accurate. Homelessness policy questions should not be considered more settled than questions of mental health, public safety, or any other element of poverty or social policy.

It is crucial to parse claims about what is evidencebased about Housing First and what is founded on humanitarian concerns, intuition, ideology, or some other factor. There is no evidence-based proof of Housing First’s ability to treat serious mental illness effectively, or drug or alcohol addiction. Housing First is not a reliable solution to social isolation, a very significant cause and effect of homelessness.

Claims made on behalf of the campaign to end homelessness—that Housing First has ended veterans’ homelessness, chronic homelessness, or homelessness at the community level—are not based in “evidence,” as that term is normally understood, and they rely on a highly technical (and dubious) definition of “ending” homelessness.

A common refrain among advocates is that “‘Housing First’ does not mean ‘Housing Only.’”[128] This is not an evidence-based claim. The claim could be verified only through a broad and thorough analysis of Housing First’s implementation across scores of programs across the nation. Surely, some programs are far more inventive in getting service-resistant clients to accept treatment and services than are others. A supportive housing program that systematically fails to engage any of its clients is, practically speaking, a “Housing Only” program. The literature about how Housing First programs function is far too sparse to validate that “‘Housing First’ does not mean ‘Housing Only.’”

There is, however, reasonable evidence to suggest that Housing First–style interventions will promote residential stability, and quite likely at a higher rate than programs that provide housing on a time-limited basis and/or rely on “barriers,” at least over a oneto two-year horizon. But an intervention is different from a policy or service system. An intervention could be one program among many. The evidence does not support the idea that Housing First should be made an organizing principle of homeless services systems. Arguments for Housing First on a systemwide basis may be defended based on intuition or humanitarian concerns, but they are not evidence-based.

The result of governments adopting Housing First as a “whole-system orientation” has been to discredit, or at least drastically de-emphasize, approaches to homelessness other than permanent housing. Less than one-fifth of the homeless population is “chronic”[129]—the population for whom Housing First was initially developed. The more that the homeless problem is described as people “down on their luck,” the less logical is the claim that permanent housing is the solution. Housing First is an entirely inappropriate intervention for the working poor, examples of which include participants in “Safe Parking” programs[130] (which is to say that, in addition to reckoning with the limitations of Housing First for the chronically homeless, permanent housing is not always an appropriate solution to street homelessness).

What kind of homeless services system do we want? That is ultimately what the Housing First debate is about. As noted, the reduction in transitional housing units is a striking example of the influence of Housing First. But it is impractical to try to design a homeless services system without programs that have features similar to transitional housing. The homeless population has many problems other than housing instability. As such, there is a certain logic to trying to address these problems along with housing instability and give them equal emphasis while doing so. That logic, though, runs contrary to the logic of Housing First, which, particularly in its original articulation, insisted on the separation of housing and social services.

In the criminal-justice world, “problem-solving courts” such as drug and mental-health courts are not simply concerned with adjudicating charges. They also deal with the addiction and untreated serious mental illness of people involved in the criminal-justice system.[131] Similarly, the linear approach to homelessness had much more of a problem-solving orientation than the current Housing First system—focused, as it is, on keeping the most people housed for the longest period of time.

But if homeless services systems don’t work on problems other than housing instability, other systems will. Indeed, the line between emergency shelter and transitional housing can get blurry. New York City’s family shelter system, for instance, in many ways resembles transitional housing more than traditional notions of emergency shelter.

Before Housing First, the homeless population was offered a robust variety of housing and service options that reflected their diverse needs. This so-called linear system viewed permanent supportive housing and other low-barrier housing programs for the homeless as valuable to a continuum of service options.[132] But when too much emphasis is placed on low-barrier options, governments must ask whether they are designing a truly inclusive homeless services system.

Clearly, some clients will be best served by providers that emphasize sobriety and work. In the world of addiction services, many providers use social pressure to encourage sobriety. Is it illegitimate or not “evidence-based” for residential treatment programs to offer temporary housing coupled with sobriety requirements?[133] What’s more important— achieving a year of sobriety or a year of housing stability? A program that sets no goals other than “residential stability,” and that specifically does not require sobriety, will not be able to use social pressure to encourage sobriety. The same issue arises for programs that try to turn their clients into responsible fathers and economically independent members of their communities. As an example: Joe Biden’s presidential campaign has called for reinvesting in transitional housing programs to facilitate prisoner reentry.[134]

Housing First is the dominant policy framework for homeless services. Yet, after years of implementation, communities are not close to ending homelessness. If homeless services systems can’t focus as much on substance abuse, unemployment, and other social ills as they do on residential stability, those challenges will simply be left to other social-services systems. In light of these facts, a certain reorientation is justified.


1. HUD should allow more flexibility from Housing First requirements for communities pursuing homelessness assistance grants through the “Continuum of Care” program.

There are about 400 CoC agencies across the nation. HUD directs billions in Homelessness Assistance Grants through these agencies to on-the-ground service providers. Federal homeless services funding was structured in this manner in deference to localism.[135]

When the CoC program was set up in the 1990s, it was “designed to meet the multi-faceted needs of homeless persons in the nation’s communities.”[136] In many communities, the local “CoC” is the lead policymaking organization on homelessness. As Housing First requirements have tightened, however, the CoC program has been criticized for departing from its original spirit and adopting a “one-size-fits-all” approach to homeless services.[137] Many criticisms of HUD’s application of Housing First principles have come from religious organizations, which have, for more a century, played a significant role in addressing homelessness.[138] The federally directed restructuring of homeless services has had a significant impact at the community level. Examples of highly regarded service providers that have experienced cuts include Community Housing Innovations, the largest provider of homeless services on Long Island,[139] and the New York City–based Doe Fund.[140] Other providers have ceased pursuing HUD funding or been pressured—by the federal government, ultimately— to make programmatic changes contrary to their priorities.

2. State and local Housing First mandates should be reassessed.

Homelessness is highly concentrated in certain urban areas, as are major homeless services systems. California and New York are hosts to about one-third of the total permanent and temporary year-round beds for the homeless.[141] Thus, state and local policies may, in some cases, matter even more than federal funding requirements. State Housing First mandates, such as California’s SB 1380,[142] should be reassessed in light of the need to develop homeless services systems reflective of the needs of the entire homeless population.

3. The homelessness debate should be reintegrated into the safety-net debate.

Housing First has separated the debates over homelessness and the safety net more broadly. In its approach to poverty, the Trump administration has tried to promote the expanded use of work requirements for safety-net programs.[143] While there is a serious debate over the appropriateness and effectiveness of work requirements for noncash programs such as Medicaid and the Supplemental Nutrition Assistance Program, there is a broad acceptance of their legitimacy in the case of public assistance. In homeless policy circles, by contrast, there is broad opposition to the use of work requirements, as well as drug testing, program-participation requirements, and adherence to treatment regimens.

As a result of Housing First’s influence, the question of upward mobility for the homeless is discussed far less often than it is for the poor. Policymakers speak with modesty about such grandiose goals as ending poverty. But with respect to ending homelessness, they are expected to accept not only the nobility of that goal but its practicality. As a result, Housing First has come to function as a harm-reduction approach not only for behavioral health but also for poverty. Someone placed in permanent supportive housing may have ended his homelessness, but he is only managing his poverty.


See endnotes in PDF


Are you interested in supporting the Manhattan Institute’s public-interest research and journalism? As a 501(c)(3) nonprofit, donations in support of MI and its scholars’ work are fully tax-deductible as provided by law (EIN #13-2912529).