More Than Half of Sheltered Homeless Adults Work. The Best Job Program Skips the “Job-Ready” Test.
- When researchers linked a census of the entire 2010 US homeless population to IRS tax records, 53% of sheltered homeless adults and about 40% of unsheltered homeless adults had formal, on-the-books earnings in the same year they were observed homeless.1
- The work doesn’t pay: in every year from 2005 to 2015, fewer than half of the adults who were sheltered-homeless in 2010 earned more than $2,000 a year, and fewer than a quarter earned more than $12,000.1
- The best-evidenced employment help is Individual Placement and Support (IPS): rapid search for regular competitive jobs, an employment specialist, support that continues after hire, and no “readiness” prerequisites. Across randomized trials — mostly with people with severe mental illness — IPS participants were 1.63 times as likely to obtain competitive employment (95% CI 1.46–1.82).2
- The same meta-analysis is precise about the limits: effects on quality of life, global functioning, and mental health all crossed zero. IPS reliably produces jobs. It does not, by itself, transform everything else.2
- Only one RCT has tested IPS with people recently homeless: 90 people just housed by Housing First in Montreal. 34% found competitive work versus 22% (crude p = 0.16; adjusted OR 2.42, 95% CI 1.13–5.16). Every other employment outcome was not significantly different. The authors called IPS “a modestly useful adjunct.”3
- The one trial to trace a housing effect — justice-involved homeless veterans in Dallas — found adding IPS raised employment, and the people who got jobs were then more likely to exit homelessness. The housing effect was real but indirect.4
- The first RCT of IPS delivered through homelessness services is underway in England (~500 people referred, government-funded). Results are expected in late 2027. Until then, “IPS works for homelessness” is an extrapolation.5
The most durable story Americans tell about homelessness is a two-word instruction: get a job. The story assumes the job is the missing piece — that people experiencing homelessness stand outside the labor market, and that the way back in runs through a paycheck.
The data says the arrow points the other way. Most people in homeless shelters have already followed the instruction. It didn’t house them.
The stereotype has it backwards
In 2021, a University of Chicago team led by economist Bruce Meyer did something nobody had managed before: they took a census of the entire US homeless population — everyone counted as sheltered or unsheltered homeless in the 2010 Census — and linked it, person by person, to IRS tax records and administrative benefit data. Not surveys. Not self-reports from one city. The actual earnings records of the actual population.1
Fifty-three percent of sheltered homeless adults had formal labor market earnings in the same year they were observed homeless. Among the unsheltered — the people sleeping in vehicles, tents, and doorways — the figure was still about 40%. These are on-the-books, W-2-and-1099 earnings, the kind that survive an IRS match.1
What the records also show is why the work doesn’t end the homelessness. Tracking the same people across a decade, in every single year from 2005 to 2015, fewer than half of the adults who were sheltered-homeless in 2010 earned more than $2,000 a year — and fewer than a quarter cleared $12,000. That is not a wage that rents an apartment in any American metro, and it wasn’t before or after the homeless spell either. Homelessness, the authors conclude, is a symptom of long-term deprivation — a decade of it, at least — not a year of bad luck that a job offer fixes.1
So the precise question is not why won’t they work. Most do, or try to. The question is what help actually increases work — and what, exactly, more work buys.
What IPS does differently
Employment programs for people with disabilities and disadvantages have traditionally run on a readiness model: assess the person, train the person, run them through job-readiness classes or sheltered work, and place them in a real job at the end, once they’ve proven themselves. Individual Placement and Support — IPS — was developed in mental health care in the 1990s on the opposite bet.2
In IPS, an employment specialist helps the person start searching for a regular, competitive job — the person’s own choice of job — quickly, usually within weeks. Employment support is integrated with whatever treatment or services the person already receives, and it continues after hire rather than ending at placement. And there are no prerequisites. You do not have to be sober first, symptom-free first, or “job-ready” first.5
If that structure sounds familiar, it should: it is the same zero-precondition logic as Housing First, which we’ve written about before. Skip the readiness test, deliver the thing itself, wrap support around it.
What the trials found — and didn’t
IPS has one of the deepest randomized-trial bases in behavioral health. A 2019 meta-analysis of the RCT literature by Frederick and VanderWeele at Harvard found that people randomized to IPS were 1.63 times as likely to obtain any competitive employment as people in usual vocational services (95% CI 1.46–1.82). They also kept jobs longer (d = 0.55) and earned more (d = 0.48).2
The same paper is unusually honest about what IPS does not do, and it deserves quoting in numbers: the pooled effect on quality of life was d = 0.30 with a confidence interval from −0.07 to 0.67; on global functioning, d = 0.09 (−0.09 to 0.27); on mental health, d = 0.03 (−0.15 to 0.21). Every one of those intervals includes zero. IPS is a jobs machine, and the trials show exactly that: jobs. Not a general life transformation.2
Almost all of those trials enrolled people with severe mental illness — not people experiencing homelessness as such. The homeless-specific experimental record is two small trials.
The first came out of Montreal’s arm of the At Home/Chez Soi study. Ninety people who had just been housed through scattered-site Housing First were randomized to IPS or usual services. Over eight months, 34% of the IPS group found competitive employment versus 22% of controls — a difference that was not significant in the crude comparison (p = 0.16) and significant after adjustment (OR 2.42, 95% CI 1.13–5.16). No other employment outcome differed significantly. The authors’ own summary: IPS was “a modestly useful adjunct.” Not our words — theirs.3
The second is more interesting for anyone who cares about housing. At the Dallas VA, researchers randomized justice-involved homeless veterans to a job-search group program alone or the same program plus IPS. The IPS arm was significantly more likely to secure employment — including among veterans who were chronically homeless. And the veterans who secured employment were then significantly more likely to exit homelessness and secure housing. The US Department of Labor’s evidence clearinghouse rates the trial’s causal evidence high. Note the mechanics, though: the housing effect ran through employment. The program bought jobs; the jobs, for some, bought housing.4
Being precise about what this establishes
Here is what the evidence supports saying. For a person who wants to work — and surveys and tax records agree that describes much of the homeless population — IPS is the best-tested way to make employment more likely. The effect shows up in dozens of trials in adjacent populations, in one small trial of recently housed people, and in one trial of homeless veterans where it reached all the way to housing.
Here is what the evidence does not support saying: that employment programs are a homelessness solution. Do the arithmetic the tax records suggest. If a program doubles the odds of obtaining work whose observed median pay for this population runs in the low thousands of dollars a year, it has done something real and valuable for that person — income, structure, a foothold, dignity. What it has not done is close the gap between those earnings and rent, in housing markets that are the primary driver of homelessness rates in the first place.1
This is the same precision we’ve urged about cash transfers and Housing First: interventions work for what they work for. Fund employment programs because jobs are worth having and people experiencing homelessness want them — not because a jobs program will make the shelter census fall. Programs sold on outcomes they can’t deliver get defunded when they don’t deliver them.
The gap that’s about to close
Remarkably, no one has ever run a large randomized trial of IPS delivered through homelessness services themselves — shelters, housing programs, street outreach — rather than through mental-health systems. That trial is finally happening: in England, the Centre for Homelessness Impact and the UK housing ministry have randomized roughly 500 people referred from homelessness services across eight local areas, with competitive employment as the primary outcome and results expected in late 2027.5
Until those results land, every claim that “IPS works for homelessness” is an extrapolation from people with severe mental illness, plus ninety people in Montreal, plus one veterans trial in Dallas. We think the extrapolation is reasonable. It is still an extrapolation, and we’ll report what the trial finds either way.
What this means in Phoenix
If you’re looking for work in Maricopa County right now: our guide to free job help in Phoenix covers the current options — including ARIZONA@WORK job centers and St. Joseph the Worker, which provides job coaching, interview clothes, and transportation help specifically for people experiencing or exiting homelessness. If you’re not sure where to start, call 211. And if housing is the emergency, start with Find resources — a job search goes better from somewhere stable.
Frequently asked questions
Do people experiencing homelessness work?
Yes, at rates that surprise most people. When researchers linked a census of the entire 2010 US homeless population to IRS tax records, 53 percent of sheltered homeless adults and about 40 percent of unsheltered homeless adults had formal, on-the-books earnings in the same year they were observed homeless. The problem is what the work pays: in every year from 2005 to 2015, fewer than half of the adults who were sheltered-homeless in 2010 earned more than $2,000 a year, and fewer than a quarter earned more than $12,000.
What is Individual Placement and Support (IPS)?
IPS is a form of supported employment developed in mental health care. Instead of assessments, training courses, and job-readiness classes before anyone touches a real job, an employment specialist helps the person search quickly for a regular competitive job matching their preferences, and support continues after hire. There are no readiness prerequisites — you do not have to be sober, symptom-free, or housed first. Across randomized trials, mostly with people with severe mental illness, IPS participants were about 1.6 times as likely to obtain competitive employment as people getting usual vocational services.
Does getting a job end homelessness?
Not by itself, in most of the evidence. The IPS meta-analysis found strong effects on jobs, tenure, and income but effects on quality of life, functioning, and mental health that crossed zero. The one homeless-specific trial that examined housing, with justice-involved homeless veterans in Dallas, found that adding supported employment raised employment, and the people who became employed were then more likely to exit homelessness — an indirect path. Meanwhile, most sheltered homeless adults already work in the formal labor market, and homelessness persists because the earnings do not cover housing.
Where can someone get free job help in Phoenix?
Our guide to free job help in Phoenix covers the current options, including ARIZONA@WORK job centers and St. Joseph the Worker, which provides interview clothes, transportation help, and job coaching for people experiencing homelessness. If you are not sure where to start, call 211. Programs change — verify hours and locations before you go.
Sources & footnotes
- Meyer BD, Wyse A, Grunwaldt A, Medalia C, Wu D, "Learning about Homelessness Using Linked Survey and Administrative Data," NBER Working Paper 28861 (2021); summarized in the Becker Friedman Institute research brief (June 23, 2021) and UChicago News, "Employment alone isn’t enough to solve homelessness, study suggests" (June 29, 2021). The study links adults recorded as sheltered and unsheltered homeless in the 2010 Census to longitudinal IRS tax records and administrative benefit data. Figures used here, as stated by the authors’ own summaries: "About 53 percent of the sheltered homeless had formal labor market earnings in the year they were observed as homeless" and "40.4 percent of the unsheltered population had at least some formal employment" that year; "In every year between 2005 and 2015, less than half of adults under 65 experiencing sheltered homelessness in 2010 had more than $2,000 of annual earnings, and less than a quarter had more than $12,000 of annual earnings"; 89% of sheltered and 78% of unsheltered adults under 65 received at least one safety-net benefit in 2010; homelessness "appears to be a symptom of long-term low material well-being — a decade (at least)." Sample restrictions (adults under 65) and the 2010 base year apply to all of these figures; more recent national linked-records estimates do not yet exist.
- Frederick DE, VanderWeele TJ, "Supported employment: Meta-analysis and review of randomized controlled trials of individual placement and support," PLoS ONE 14(2):e0212208 (2019), open access. Pooled estimates from the abstract: obtained any competitive employment RR = 1.63, 95% CI [1.46, 1.82]; job tenure d = 0.55 [0.33, 0.79]; job length d = 0.46 [0.35, 0.57]; income d = 0.48 [0.36, 0.59]. Non-vocational outcomes, verbatim framing "included the null": quality of life d = 0.30 [−0.07, 0.67]; global functioning d = 0.09 [−0.09, 0.27]; mental health d = 0.03 [−0.15, 0.21]. The trial base is overwhelmingly people with severe mental illness receiving IPS through mental-health services; generalization to homelessness populations is an inference, not a finding of this paper.
- Poremski D, Rabouin D, Latimer E, "A Randomised Controlled Trial of Evidence Based Supported Employment for People Who have Recently been Homeless and have a Mental Illness," Administration and Policy in Mental Health and Mental Health Services Research 44(2):217–224 (2017; online December 31, 2015). Ninety people recently housed through the Montreal At Home/Chez Soi scattered-site Housing First program, randomized to IPS or usual services, eight months of follow-up with good fidelity. From the abstract: "fifteen (34%) people in the IPS group versus nine (22%) in the control group found competitive employment (p = 0.16). Adjusted odds of obtaining competitive employment were greater in the IPS group (OR 2.42, 95%CI 1.13–5.16). Other employment outcomes were not significantly different between groups... IPS was a modestly useful adjunct to scattered-site Housing First." Full text is paywalled; figures here are from the published abstract.
- LePage JP, Martin WB, Crawford AM, Rock A, Parish Johnson JA, Washington EL, "Individualized Placement and Support Supported Employment for Justice-involved Homeless and Unemployed Veterans," Medical Care 59(Suppl 2):S195–S198 (2021). VA-funded randomized study at the VA North Texas Health Care System comparing the About Face Vocational Program (AFVP) job-search group alone versus AFVP plus IPS supported employment. From the abstract: participants receiving supported employment "were significantly more likely to secure employment compared with group alone. This significant difference was also found in individuals who identified as chronically homeless. Furthermore, participants who secured employment were significantly more likely to exit homelessness and secure housing, suggesting group membership had an indirect effect on housing status though improved employment outcomes." The US Department of Labor’s CLEAR evidence review rates the causal evidence high, attributing the three-month post-program employment gain to IPS. Effect sizes and per-arm Ns are in the paywalled full text; we report the direction and significance as the abstract states them.
- Centre for Homelessness Impact, "Individual Placement and Support" project page (last updated May 13, 2026): a randomized controlled trial of IPS delivered through homelessness services in eight English areas (Barnsley, Rotherham, Sheffield, Barnet, Fareham, Wakefield, Shropshire, Telford & Wrekin), funded by the Ministry of Housing, Communities and Local Government and evaluated by IFF Research. "Recruitment to the trial has now closed with around 500 people being referred to the service"; primary outcome is movement into competitive employment, measured at 6 and 12 months plus HMRC earnings linkage at 12 and 18 months; "We expect to have the results of the trial in late 2027." The page also carries the plain statement of the gap this brief turns on: "While IPS has shown strong results for people experiencing severe mental health conditions and people using alcohol and other drugs, less is known about its impact in homelessness services." Description of the IPS model (rapid search, employment specialist, ongoing support) is drawn from the same page and from Frederick & VanderWeele (note 2).