Bay EvidenceInvestigative research · Bay Area public systems-22%
San Francisco unsheltered homelessness
Reported change from 2024 to 2026
Bay Evidence evidence brief
Beyond Visibility
Reconciling San Francisco's 2026 Point-in-Time Reductions with Longitudinal Housing Retention
The 2026 results do not describe one uniform improvement. They show different movements across cross-sectional measurement categories: fewer people observed in tents and on sidewalks, more people counted in sheltered settings, and more families without stable housing. The records do not establish that the same people moved between those categories.
Evidence status · updated September 13, 2026: This brief does not contain person-level longitudinal housing-retention data. It cannot establish that visible street reductions produced durable housing exits or isolate the effect of shelter expansion, enforcement, mobility, or count-method changes. The downloadable July PDF retains its original subtitle while this interpretive framing is under revision.
San Francisco appears to have reduced visible street homelessness faster than it reduced homelessness itself.
One plausible interpretation combines expanded shelter and interim capacity, encampment and vehicle interventions, continued barriers to permanent housing, severe economic pressure on families, and count-method changes. The available records do not isolate the contribution of each factor or show that visible street reductions produced durable housing retention.
Reported changes
Four numbers, two different stories
-85%
People identified in tents or outdoor structures
Reported decline from the 2024 count
+15%
Families experiencing homelessness
465 families counted in 2026, up from 405
-4%
Total homelessness in San Francisco
A much smaller decline than the unsheltered change
01
The street decline is larger than the overall decline.
The 22 percent reduction in unsheltered homelessness occurred while total homelessness fell only 4 percent. That aggregate divergence is compatible with different sheltered, unsheltered, vehicle, and other category patterns, but the PIT tables do not track individual movement or explain why the categories changed.
02
Tent disappearance is not the same measure as housing resolution.
The roughly 85 percent decline in people identified in tents occurred during a period of added shelter capacity and increased encampment enforcement. Available public data do not identify the contribution of either factor or establish that a removed tent produced a durable housing exit.
03
Family homelessness follows a different economic pathway.
Families can appear in vehicle, doubled-up, or family-shelter settings that differ from visible sidewalk encampments. Broader research identifies rent burden, eviction risk, child-care costs, immigration-related instability, and scarce family-sized subsidized units as plausible pressures; the PIT tables do not establish which pressure drove the 2026 increase.
04
Local systems did not move in lockstep.
San Francisco, Alameda County, Contra Costa County, and neighboring jurisdictions reported different directional changes. Shelter supply, housing production, enforcement, and count methods may contribute, but the published counts do not isolate their effects.
05
Methodology limits the strength of causal claims.
San Francisco changed the timing and field procedures for its 2026 count. Until the city publishes a bridge analysis showing how the new method compares with 2024, the reported decline should be treated as significant but not perfectly comparable.
Causal assessment
Concurrent trends can be produced by different parts of the homelessness system.
Shelter expansion
City records document added shelter and interim capacity during the comparison period. The PIT tables do not establish how much that change contributed to the sheltered or unsheltered totals.
Encampment enforcement
Encampment enforcement increased after Grants Pass during the comparison period. That sequence is documented, but available records do not establish that enforcement caused the tent decline or quantify regional displacement.
Permanent-housing bottlenecks
Available placement totals do not show an increase commensurate with the visible street reduction. Comparable person-level retention data are needed before characterizing the difference as a durable-exit gap.
Family housing economics
High rents, low vacancy, income volatility, and limited family-sized affordable units are plausible upstream pressures identified in broader evidence. The 2026 PIT tables do not attribute new family homelessness to any one factor.
Vehicle homelessness
Families living in cars and RVs are counted differently from tent residents and can remain hidden from shelter waitlists and conventional outreach.
Count design
A single-night estimate is sensitive to timing, volunteer deployment, occupancy assumptions, weather, enforcement activity, and who can be observed or interviewed.
A PIT decline is not automatically a durable housing outcome.
The test of progress is not only whether fewer people are counted outdoors on one January morning. It is whether entries into homelessness fall, permanent housing exits rise, returns to homelessness decline, families remain connected to schools and services, and people are not simply displaced across neighborhoods or county lines.
The available evidence supports a finding of real operational change, but not a claim that San Francisco has reduced every form of homelessness by 22 percent.
Selected sources
Source record
Evidence Brief No. 2
Read the complete analysis, references, methodology, and policy implications.