About 36% of people experiencing homelessness nationally are unsheltered, but the data show wide disparities across states. The figures run from nearly two-thirds (64%) in California to just 4% in New York. These are not two degrees of the same problem — they are two distinct problems that call for different solutions.
HUD's 2025 Annual Homelessness Assessment Report to Congress counted 266,320 people sleeping in unsheltered locations — streets, vehicles, encampments, and other places not meant for human habitation — on a single night in January 2025. That total reflects 35.7% of everyone counted, and a 2.9% decrease from 2024 — the first decline in unsheltered homelessness in nearly a decade, though one driven almost entirely by a single state.
The national figure, though, can be difficult to draw meaning from. What determines whether a person experiencing homelessness in America sleeps indoors has less to do with the scale of homelessness in their community than with whether that community provides a sufficient shelter system — and whether the law obliges it to. Climate and local enforcement policy matter as well, though largely by shaping where people are counted, not whether they have somewhere to go.
A distinction worth holding onto throughout: "sheltered" here refers to emergency shelter, transitional housing, and safe havens — temporary settings that put a roof over someone's head. Permanent housing programs such as permanent supportive housing and rapid rehousing are a separate category entirely, and people living in those housing types are not included in the homeless counts. In the context of this brief, shelter and housing are not used interchangeably.
California reports that 64% of its homeless population is unsheltered; Oregon 61%, Hawaii 59%, Arizona 56%. At the other end, New York reports 4%, Massachusetts 6%, and Vermont 8%. HUD attributes part of this spread to climate — counts occur in the last ten days of January, and warmer states record more people outside — but the report explicitly notes that shelter capacity, access policy, and local housing markets drive the rest.
Share of each state's counted homeless population sleeping in unsheltered locations. Darker shading indicates a higher unsheltered share. Hover for additional data.
Geography is only one axis. Who a person is homeless with matters just as much. HUD reports that 93% of people in homeless families with children were sheltered in 2025 — leaving just 7% unsheltered. Among homeless individuals, the figure is 49%.
| Population Type | Percent Unsheltered |
|---|---|
| People in families with children | 7% |
| All people experiencing homelessness | 36% |
| Individuals (not in a family with children) | 49% |
| People with chronic patterns of homelessness | 59% |
Share of each population counted in unsheltered locations, 2025.
An unaccompanied adult is roughly seven times more likely to sleep outside than someone in a family with children. That gap is largely by design: family shelter has been prioritized in most systems, and where right-to-shelter obligations exist, families typically hold the strongest legal claim to a bed. Unsheltered family homelessness also fell far faster than the overall figure last year — down 13.2%, against 2.9% for unsheltered homelessness as a whole.
One pattern here cuts against the usual convention. While about six in ten people in homeless families with children are in the nation's 50 largest cities, roughly four in ten unsheltered families are in rural CoCs. Unsheltered family homelessness is disproportionately rural — and disproportionately hard to see. HUD cautions that families may take turns sleeping in vehicles or backyards, making them difficult to locate on count night, so the true figure is likely higher than reported.
The pattern is clearest among the nation's 48 major city Continuums of Care (CoCs). Below are the ten largest by total homeless population — the communities carrying the heaviest absolute burden in the country. New York City counted 125,683 people experiencing homelessness, nearly twice Los Angeles and the largest of any CoC in the nation, yet only 3.7% were unsheltered. Los Angeles counted 67,777, of whom 65.5% were unsheltered.
| Major city CoC (10 largest) | Total homeless population | Percent Unsheltered |
|---|---|---|
| San Jose/Santa Clara City & County | 10,711 | 69.8% |
| Oakland, Berkeley/Alameda County | 9,262 | 68.5% |
| Portland, Gresham/Multnomah County | 10,526 | 65.7% |
| Los Angeles City & County | 67,777 | 65.5% |
| Seattle/King County | 16,936 | 57.9% |
| San Diego City and County | 9,905 | 57.7% |
| Phoenix, Mesa/Maricopa County | 9,734 | 53.5% |
| San Francisco | 8,671 | 50.2% |
| Metropolitan Denver | 10,774 | 19.9% |
| New York City | 125,683 | 3.7% |
The ten major city CoCs with the largest total homeless populations in 2025, ordered from highest to lowest unsheltered share.
New York City's position is not an artifact of size. Across all 48 major city CoCs, only Boston reports a lower unsheltered share (2.4%), and Boston's homeless population is roughly one-twentieth the size of New York's. Notably, both cities operate under right-to-shelter obligations. Where a legal duty to provide shelter exists, people are counted indoors; where it does not, they are counted outside. The scale of need does not explain the difference. System design does.
Unsheltered homelessness fell 2.9% in 2025, breaking a nine-year run of increases — but the longer arc still points sharply upward. Unsheltered counts are 36% higher than in 2013 (+70,654 people), even as emergency shelter capacity expanded and transitional housing shrank by nearly two-thirds. The unsheltered share of all homelessness peaked at 40% in 2022 and has since eased slightly to 36%, largely because emergency shelter grew faster than street homelessness — not because street homelessness resolved.
That 2.9% decrease is also far less national than it appears. California's unsheltered count fell by 8,391 people — more than the entire national decline of 7,904. Remove California and unsheltered homelessness across the rest of the country did not fall at all; it rose slightly. Remove California and Florida, the two states with the largest decreases, and it rose 3.4%.
| 2024–2025 unsheltered change | People | Percent |
|---|---|---|
| United States | −7,904 | −2.9% |
| California | −8,391 | −6.8% |
| Florida | −3,706 | −22.0% |
| All states except California | +759 | +0.5% |
| All states except California and Florida | +4,465 | +3.4% |
Change in the number of people counted in unsheltered locations, 2024 to 2025. Territories excluded.
The state-by-state distribution confirms it: 28 states recorded increases in unsheltered homelessness against 21 decreases, and the median state rose 3.8%. The largest increases were Oregon (+2,309, up 16%), Arizona (+1,206, up 17%), and Washington (+811). One reason a single state can move the national figure this much is concentration — California, Oregon, and Washington together account for 56% of all unsheltered homelessness in the country (149,116 of 266,320 people). A national unsheltered statistic is, to a considerable degree, a West Coast statistic.
Longer trajectories diverge just as sharply. Since 2013, Washington's unsheltered count has risen 237%, Arizona's 171%, and Oregon's 123%. Over the same period Florida's fell 53% (28,192 to 13,162) and South Carolina's 55%. Florida is one of the country's most interesting cases: a 22% single-year decline on top of a decade-long downward trend is a sustained trajectory, not a one-year artifact, and it is worth understanding what that state's shelter and rehousing investments did differently.
Geography tells a similar story. HUD reports that unsheltered homelessness declined in every CoC category in 2025. But suburban CoCs show the sharpest changes: sheltered homelessness increased 8.0% while unsheltered fell 7.5% — the signature of new shelter capacity coming online and moving people indoors, rather than of homelessness disappearing.
HUD itself calls the unsheltered count the most volatile figure it publishes, sensitive to weather, local policy, and counting methodology. New York City's own 2026 HOPE count is a striking demonstration: conducted in March under unusually warm conditions rather than on a frigid January night, it produced the city's highest street count in over a decade, an 11% jump. The population likely did not change that much in a year. The method of counting did.
This cuts both ways. A January count in a cold city almost certainly undercounts people outside, who move into cars, public transit, hospitals, or doubled-up arrangements. Communities comparing their own year-over-year numbers should be equally alert to whether their counting methods changed before concluding that local homelessness did.
The policy environment also adds pressure. Two years after City of Grants Pass v. Johnson permitted jurisdictions to penalize sleeping outside even without adequate shelter available, the National Homelessness Law Center counts 350+ cities and 14 states that have enacted measures criminalizing homelessness. Enforcement disperses encampments; it does not reliably reduce the number of people without housing, but it can meaningfully change where — and whether — they are found on count night.
New York City shelters a larger share of its homeless population than any other large community in the country — and has the largest homeless population in the country. Both statements are true, and together they make the central point: shelter determines whether a person is counted indoors or outdoors, while housing determines whether they are counted at all. Sheltered and unsheltered homelessness call for different interventions, different staffing, and different funding. A community at 60% unsheltered faces a shelter and outreach capacity problem. A community at 5% unsheltered faces a shelter exit problem — people are indoors, but they are not moving into housing. Both are real, and a national average of 36% fails to provide a clear characterization. The strategic question for a CoC or provider is not how their rate compares to the national figure, but which of these two problems they are actually solving for — and whether their program model is designed accordingly. Getting that diagnosis wrong is how communities end up funding the wrong solution.