Blog
Homelessness response and the counts, what the city says and what is disputed
The mayor's office reports unsheltered homelessness at its lowest since 2011; here is how the count works, what changed in 2026 and what advocates question.
Fifteen years. That was the headline on May 12, 2026, when the mayor's office announced that unsheltered homelessness in San Francisco had fallen to its lowest level in that long. The claim rested on a single morning's count, taken on January 29, 2026, and it landed in a debate that has run for years: how many people are living outside, how anyone knows, and what the city's programs are really changing. What follows sets out the numbers the city published, how the count works, and the questions critics have raised, without taking a side on a question the data cannot yet settle.
What the city counts, and why
Every two years, San Francisco takes part in the Point-in-Time (PIT) Count, a census of people experiencing homelessness on a single night or morning. The Department of Homelessness and Supportive Housing, HSH, describes it as a biennial count of both unsheltered and sheltered people at a given time. The federal government uses counts of this kind to allocate funding, and cities use them to track change. HSH publishes the results alongside a count of beds, the Housing Inventory Count.
A point-in-time count is a snapshot. It does not capture everyone who is homeless at some point during the year, and it depends on weather, on where counters look and on how the count is organized. For that reason, experts and officials alike compare counts cautiously, and the comparison is hardest when the method changes.
What the mayor's office reported
The mayor's announcement of May 12, 2026 used preliminary data. According to the city:
- Unsheltered homelessness was down 22 percent since 2024, the lowest level recorded since 2011, and about 1,000 fewer people were on the street. One published breakdown puts the unsheltered figure at 3,400, down from 4,354 in 2024.
- Total homelessness was down 4 percent since 2024.
- Sheltered share was 57 percent of the homeless population, described as the highest recorded.
- Tents and structures were down about 85 percent, with fewer than 100 people observed in tents, against nearly 650 in 2024.
- Veterans were down 44 percent, from 585 to 327.
- Youth identified in the count numbered fewer than 600, down 54 percent.
- Families were up 15 percent, with 74 percent of families sheltered. Most unsheltered families lived in vehicles.
The city also pointed to programs it credited for the change: outreach consolidated under the Department of Public Health in April 2026; more than 600 new treatment-focused beds; a crisis stabilization center at 822 Geary Street and a RESET Center for treatment connections; a permitting program for large vehicles that has moved families into housing or shelter; and the Journey Home program, which reconnects people with loved ones outside San Francisco. The mayor's office said it had raised about $50 million in private funding through the Breaking the Cycle Fund.
The count itself is preliminary, and the city said full results were expected that summer.
The questions raised
Advocates and some observers raised several points about the 2026 numbers.
The method changed. The 2026 count was conducted in the morning of January 29, rather than at night, and was reported to use trained city staff rather than volunteers. City officials defended the adjustments as improvements in accuracy. Critics, including the San Francisco Coalition on Homelessness, questioned whether a drop could reflect the change in method as much as a change on the street. Anyone comparing 2026 with 2024 faces that caveat, because the two counts were not taken the same way.
The overall decline is smaller than the unsheltered one. Total homelessness fell 4 percent, a much smaller change than the 22 percent decline in people on the street. The difference tracks how many people moved into shelter and transitional housing, which the city says rose 15 percent. Whether that is progress depends on whether sheltered people move on to stable housing.
Families rose. The city's own figures show family homelessness up 15 percent, and the mayor's release acknowledges that this remains a challenge.
Beds are defined in different ways. The city's claim of a large increase in treatment and stabilization beds has drawn scrutiny over what counts as a shelter bed. Reporting on the issue noted that some facilities that serve people in crisis for short periods are now included. The city's reply, in a spokesperson's words, is that there has been a net gain of more clinically intensive beds. Both sides are describing the same inventory with different definitions.
Delivery gaps. Reports in 2026 noted that a group of tiny cabins the city bought was sitting without residents, and that a campaign pledge of 1,500 shelter beds in the mayor's first six months had been walked back. These are criticisms of execution, not of the count, but they bear on the broader picture.
The other measures
A count is one measure among several. The Controller's Office published an assessment of the shelter system in December 2024 that counted 33 shelters with 3,228 beds or units, run by 20 nonprofits, as of September 2024. The Controller describes the assessment as a neutral look at emergency shelter that does not audit individual providers. Other numbers come from the city's coordinated entry system, the "front door" that assesses people, prioritizes them and matches them to housing, and from reports to the Board of Supervisors. HSH publishes these on its site, and the department's plan, Home by the Bay, sets goals for 2023 to 2028.
HSH's stated goal is to make homelessness "rare, brief, and one-time." That phrase is the standard the programs are measured against: not whether the number is lower than last time, but whether people who become homeless leave quickly and do not return.
Another measure comes from the 311 system. Complaints about encampments, trash and needles are logged by address and date, and from a 311 call to a cleaned street explains how that data is created. The mayor's office says response times improved by 40 percent after street teams were integrated. Street conditions, like beds and counts, can be measured, and the debate is partly about which measure counts most.
How the system is organized
The city's homelessness response has several layers.
- Department of Homelessness and Supportive Housing. The lead department, at 440 Turk Street, phone 628-652-7700 on weekdays from 8 a.m. to 5 p.m. Its oversight bodies include the Homelessness Oversight Commission, the Local Homeless Coordinating Board and the Our City, Our Home Oversight Committee.
- Department of Public Health. Since the 2026 consolidation, it runs street teams and behavioral health programs. Mental health care and the Street Crisis Response Team and the city's public health clinics and hospitals describe the parts of the system most people meet.
- Shelters and navigation. Emergency shelter, the Interfaith Winter Shelter Program (open November 23, 2026 through March 28, 2027 for adults 18 and over, with dinner and breakfast), support for people living in large vehicles and shelter for women and children fleeing domestic violence.
- Housing programs. Permanent supportive housing, including veterans' housing. On November 10, 2026, the mayor announced $18.6 million in state Proposition 1 funding for a 63-unit project at 629 Post Street, with 44 permanent supportive units for veterans leaving homelessness and 19 Safe Haven transitional units, operated by Swords to Plowshares.
- Oversight of shelters. The Shelter Monitoring Committee tracks conditions and the Shelter Grievance Advisory Committee reviews grievances.
The funding behind much of this comes from several sources. Proposition C, "Our City, Our Home," a tax measure approved by voters, funds mental health and housing work, as Mental health care and the Street Crisis Response Team notes, and the city's general fund and state and federal grants add to it. The budget for the whole city, $16.9 billion for fiscal year 2026-27, is the context, and How to read the San Francisco budget book shows where each program appears.
Why the comparison is hard
A reader who wants to put the 2026 count next to the 2024 count runs into three obstacles at once. The first is timing: the count is a snapshot of one morning, and weather on that morning can move numbers in either direction. The second is method, as described above. The third is what is being counted. The city's headline figure is unsheltered homelessness, but shelter capacity, tents, vehicles and families each move on their own timelines. A drop in tents can mean fewer people on the street, or it can mean that people moved into vehicles or doorways, and the count treats those differently.
None of this means the numbers are wrong. It means the figure that matters, in the end, is the trend across several counts taken the same way. The 2026 count may become the baseline for that comparison, and the 2028 count will be the first test of whether the decline holds.
What residents can do
For a resident who sees someone in need, the guidance is practical. In an emergency, call 911. For an encampment or street condition, call 311. For information on shelter or housing, sf.gov lists services, and HSH's coordinated entry is the entry point for people seeking housing. Volunteers can help through local nonprofits, and A checklist for volunteering with the City of San Francisco lists the ways in.
Reading the next report
The final 2026 results will show whether the preliminary numbers hold. When they appear, a reader can ask five questions of any report on homelessness:
- Who was counted, and when? A morning count and a night count are different instruments.
- Who did the counting? City staff and volunteers see different things.
- What changed in the method since the last count?
- How are beds defined? Short-stay crisis beds and long-term shelter are not the same.
- What happens to people next? A decline on the street matters most if people reach stable housing and stay there.
The mayor's office reads the 2026 count as proof that its policies are working. Its critics read the same numbers as incomplete. Both can be partly right, and the way to judge is to follow the data over the next count and the next, with the same method each time.