San Francisco in Permits: The City That Slowed Down, 2010–2026
A longitudinal study of 462,783 San Francisco permit records tracing the city from the 2010s construction expansion through the 2020 break and into a lower-volume post-pandemic development regime.
San Francisco research palette uses gold, royal blue and white for a restrained Bay Area editorial system.
Permit records analyzed
462,783
2010 through July 2026
2019 permit records
37,303
Series high, 42.2% above 2010
2020 permit decline
-47.2%
19,690 records vs 37,303 in 2019
2025 net housing addition
2,669
SF Planning Housing Inventory 2025
Key takeaways
- San Francisco’s canonical permit volume rose 42.2% from 26,234 records in 2010 to 37,303 in 2019, then fell 47.2% in 2020.
- The city did not return to its pre-pandemic administrative construction regime: 2025 recorded 23,012 permits, 38.3% below 2019.
- Declared value fell much less than record count in 2020, down 28.0%, showing that the pandemic shock hit permit frequency more severely than the surviving high-value project tail.
- Value concentration remained extreme: the top 1% of valued records represented 61.4% of canonical value in 2019, 72.7% in 2020 and 57.4% in 2025.
- Official housing production also remained constrained in 2025: SF Planning reports a net addition of 2,669 units, 26% below the ten-year average of 3,622.
San Francisco did not simply pause. It changed regimes.
San Francisco entered 2020 after a decade of steadily rising administrative construction activity. PropertyLabs counts 26,234 canonical permit records in 2010, 30,446 by 2013, 34,120 in 2015 and 37,303 in 2019. That 2019 total was 42.2% above the 2010 level. The city’s permit ledger therefore describes a long expansion before the pandemic, not a static market interrupted by one bad year.
The break in 2020 is unusually sharp. Canonical permit records fell from 37,303 in 2019 to 19,690 in 2020, a decline of 47.2%. Declared value fell from approximately $3.67 billion to $2.64 billion, a smaller 28.0% decline. The difference between those two measures is central to this paper. The pandemic disrupted the breadth of permitting far more than it eliminated the value attached to the projects that remained in the administrative pipeline.
The post-pandemic recovery never restored the old volume regime. Records rose to 22,644 in 2021, moved between roughly 21,000 and 23,000 through 2025 and ended 2025 at 23,012, still 38.3% below 2019. The permit ledger therefore supports a structural-regime interpretation: San Francisco moved from a high-frequency 2010s construction system into a persistently lower-frequency post-2020 system.
Population data provide important context but do not mechanically explain the permit series. The U.S. Census counted 805,235 residents in 2010 and 873,965 in 2020, an 8.5% increase. The Census Bureau’s Vintage 2025 estimate is 826,079, approximately 6.0% below the 2020 estimates base. San Francisco therefore experienced the construction break at the same time as a historic population reversal, but permit records alone cannot establish how much of the change came from demand, remote work, financing, construction economics, planning constraints or the composition of the project pipeline.
Official housing evidence is consistent with a city that remained capable of building but at a level below its own recent historical average. SF Planning’s 2025 Housing Inventory reports a net addition of 2,669 housing units in 2025, 26% below the ten-year average of 3,622. The Department of Building Inspection authorized 1,900 net units for construction. The Planning Department fully approved 274 projects proposing 3,723 net units. Those figures show movement through the pipeline, but they also demonstrate the difference between approval, authorization and completed additions to the housing stock.
The 2026 Q1 Housing Development Pipeline makes the same point at larger scale. SF Planning reports 74,888 total new pipeline units, but only 3,301 were under construction. Another 1,219 had site permits issued, 2,422 had building permits approved, 7,856 had building permits filed and 13,036 were applications under review. Fully half of the total pipeline, 37,577 units, sat in major multi-phased projects. San Francisco has a very large theoretical and entitled pipeline relative to the number of units actively under construction.
Permit value is also highly concentrated. In the PropertyLabs canonical series, the top 1% of valued records represented 61.4% of 2019 declared value, 72.7% in 2020 and 57.4% in 2025. The median valued record was only $14,450 in 2019 and $15,000 in 2025. This is why annual dollar totals cannot stand alone as a measure of construction vitality. A handful of large permits can preserve a strong nominal value year while the everyday volume of administrative construction falls dramatically.
The paper’s central finding is therefore not that San Francisco stopped building. It is that the city’s observable construction system changed scale. The 2010s were characterized by rising permit frequency and repeated multibillion-dollar value years. The 2020 shock broke that pattern. The recovery restored only part of the former volume, while official housing data show approvals and a very large future pipeline coexisting with relatively modest annual completions. San Francisco’s post-2020 problem is best understood as a conversion problem: how much of a large administrative and entitled pipeline becomes actual construction and completed housing, and how quickly.
Key findings
The 2010s permit expansion was persistent
Permit records increased from 26,234 in 2010 to 37,303 in 2019, a 42.2% increase. Volume rose across most of the decade rather than around a single isolated year.
2020 was a genuine structural break
Permit frequency fell 47.2% from 2019 to 2020. Declared value fell 28.0%, indicating that large projects remained more resilient than the broad base of permit activity.
The old volume regime did not return
San Francisco recorded 23,012 permits in 2025, 38.3% below the 2019 high. Every full year from 2021 through 2025 remained far below the 2017–2019 run rate.
The pipeline is much larger than active construction
SF Planning counted 74,888 total new pipeline units in 2026 Q1, but only 3,301 were under construction. Half of the pipeline sat in major multi-phased projects.
Housing delivery remains below the recent norm
SF Planning reports 2,669 net housing units added in 2025, 26% below the ten-year average of 3,622, even as thousands of units remained approved, filed or under review.
What this means
Developers
Entitlement is not the same as executable supply
A large formal pipeline does not guarantee near-term construction. Financing, phasing, infrastructure, market rents, construction cost and project economics determine which approvals become starts.
Lenders
Distinguish broad activity from megaproject value
Permit value is heavily concentrated in the top tail. Market underwriting should pair total declared value with permit breadth, active-construction stage and sponsor-level feasibility.
Policymakers
Measure conversion through the full pipeline
Filed, approved, authorized, under-construction and completed units answer different questions. A city can accumulate a large pipeline while annual housing delivery remains below historical averages.
Data visualizations
San Francisco municipal permit records, 2010–2026
Annual PropertyLabs canonical permit records using issued/activity date. 2026 is through July 30 and is not comparable with full calendar years.
How concentrated is San Francisco permit value?
Share of annual canonical declared value represented by the top 1% of valued records. Derived from permit-level PropertyLabs data.
From construction expansion to pipeline conversion
The first decade of the series is defined by breadth. San Francisco recorded roughly twenty-six thousand canonical permit records in both 2010 and 2011. The annual count then moved higher almost continuously: 27,653 in 2012, 30,446 in 2013, 31,601 in 2014 and 34,120 in 2015. By 2018 the count reached 36,798 and in 2019 it reached 37,303. This is not evidence that every year was an equally strong real-estate market, because permit universes include many types of work. It is evidence that the municipal construction system was processing materially more issued records by the end of the decade than at the beginning.
The declared-value series tells a related but more volatile story. Canonical value rose from approximately $1.18 billion in 2010 to $2.66 billion in 2012 and $3.43 billion in 2013. It exceeded $4 billion in 2015, 2017 and 2018. Yet those totals were never evenly distributed. The 2017 maximum single record exceeded $537 million and the top 1% of valued records represented roughly 77.6% of total canonical value that year. Large-project timing therefore mattered enormously to any annual value comparison.
That concentration is one reason the 2020 break has to be measured in more than one way. Record count fell by nearly half in one year, from 37,303 to 19,690. Total declared value fell by approximately 28%, from $3.67 billion to $2.64 billion. The median valued record, however, increased slightly from $14,450 to $15,000. The broadest interpretation is that the pandemic removed a huge amount of routine administrative volume while leaving a disproportionately large share of high-value project activity in the surviving ledger.
The top-tail statistics reinforce that interpretation. In 2019 the top 1% of valued records represented approximately 61.4% of canonical value. In 2020 the share jumped to 72.7%. A market can therefore look less damaged if one watches only total nominal value. San Francisco’s 2020 permit economy became narrower and more concentrated even though multibillion-dollar value remained attached to issued records.
The rebound in 2021 also demonstrates why a single strong value year does not equal a full recovery. Permit count rose to 22,644, still roughly 39% below 2019, but canonical value rebounded to almost $4.0 billion. Nearly every record carried a declared value in the post-2020 portion of the archive, while earlier coverage was lower. The top 1% represented approximately 74% of 2021 value. The year therefore combined a partial volume recovery, very strong reported nominal value and improved source-field completeness. All three facts matter.
From 2022 through 2025, permit frequency settled into a remarkably narrow band. The annual totals were 21,474, 22,087, 22,777 and 23,012. That stability is important because it argues against treating 2020 as a temporary one-year administrative interruption followed by normalization. Five years after the shock, permit frequency remained far below the 2017–2019 regime. Whatever mix of demand, remote work, project economics, construction costs, regulatory process and source behavior produced the change, the observed system did not revert.
Population moved in the same broad direction after 2020. The decennial Census counted 805,235 residents in 2010 and 873,965 in 2020. The Census Bureau’s Vintage 2025 estimate is 826,079, about 6% below the 2020 estimates base. That reversal matters because development demand is partly a function of households, employment and the use of space. But it should not be used as a one-variable explanation. Housing demand can remain intense even when population declines if household size, vacancy, income distribution, tenure and neighborhood preferences change.
The official housing inventory shows that San Francisco continued adding housing despite the lower permit regime. SF Planning reports a net addition of 2,669 units in 2025, taking the city to approximately 422,958 dwelling units. The annual net addition was 26% below the ten-year average of 3,622. This is a more direct measure of delivered housing than a broad permit count, because the two universes answer different questions. PropertyLabs tracks the administrative construction ledger; SF Planning tracks unit production and loss.
The distinction between stages is even clearer in the authorization data. In 2025 the Department of Building Inspection authorized construction of 1,900 net units. SF Planning also reports 274 fully approved projects proposing 3,723 net units. These figures are not contradictory. They describe projects at different points in the development chain. Approval can precede building authorization, authorization can precede physical construction and construction can precede occupancy. A credible housing-supply analysis must keep those stages separate.
San Francisco’s 2026 Q1 pipeline is enormous on paper. SF Planning counts 74,888 total new pipeline units as of April 2026. Yet only 3,301 units, roughly 4% of that total, were under construction. Site permits had been issued for another 1,219 units, building permits approved for 2,422 and building permits filed for 7,856. Another 13,036 units were in applications under review. The majority of the pipeline was therefore not in active construction.
The largest single category is the major multi-phased project universe, which SF Planning places at 37,577 units, exactly half of the total pipeline. That category includes projects such as Candlestick Point/Hunter’s Point Shipyard, Treasure Island, Parkmerced, Potrero Power Plant, Pier 70, India Basin and Mission Rock. The existence of these projects creates a powerful long-run supply option for the city, but the timing of delivery depends on phasing, infrastructure, capital markets and project economics.
This difference between theoretical pipeline and active production is the central post-pandemic development question. If a city has tens of thousands of entitled or planned units but only a small fraction under construction, the bottleneck is not simply whether land can legally contain more housing. Financing conditions, required returns, construction costs, infrastructure obligations, absorption assumptions, developer balance sheets and execution capacity become part of the explanation.
The permit ledger captures that distinction indirectly. In the 2010s, rising record frequency indicated a broadening volume of administrative construction activity. After 2020, annual records remained much lower even while major projects and housing applications continued to exist. A large future pipeline can therefore coexist with a lower current permit throughput. The two signals describe different time horizons.
The 2025 Housing Inventory also shows a city whose housing stock is already highly multifamily. SF Planning reports approximately 422,958 units at year end, with 39% located in buildings of twenty or more units, 20% in two-to-four-unit buildings and another 18% in five-to-nineteen-unit buildings. Only 22% of the stock was single-family. San Francisco’s housing problem is not that the city has never urbanized. It is that incremental additions are occurring inside an already dense, expensive and physically constrained system.
The annual authorization trend reinforces the slowdown. SF Planning reports 1,900 net units authorized for construction in 2025, below the ten-year average of 2,925. The figure improved materially from 1,031 in 2024, but a one-year rebound does not restore the stronger authorization years earlier in the decade. The official series therefore points in the same direction as the broad permit ledger: some improvement from the trough, but not a return to the old development intensity.
The filed-versus-approved figures add another layer. SF Planning reports 9,614 total units filed in 2025 but only 3,723 total units approved. The agency explicitly warns that applications for the same site can be revised and that revision applications are tracked separately beginning in 2024. Those caveats matter. A surge in filings is not automatically a surge in new independent projects, and a project can appear multiple times across its administrative life. This is exactly why PropertyLabs treats municipal administrative data as evidence requiring denominator discipline rather than as a simple construction counter.
The PropertyLabs archive has its own source limitations. Every San Francisco canonical record has an issued date and activity date, and those fields match in the current extraction. The archive has no populated applied-date field, so this paper does not attempt application-to-issuance timing. It also lacks usable canonical latitude and longitude, community, applicant and contractor fields. Those gaps rule out a defensible parcel-level heat map, neighborhood migration analysis or contractor-network study in the current version.
The source does contain a permit-type code on every record and descriptions on almost every record, but the type field is encoded numerically rather than as a self-describing categorical label. Without a verified municipal codebook mapping each value consistently across the period, this paper does not attach interpretive names to those codes. A chart that confidently labels code 8 as a particular construction class without source validation would be visually attractive and methodologically weak. It is intentionally excluded.
Declared-value coverage changes materially through the series. Only about 75% to 82% of records carried canonical value in many pre-2020 years, compared with more than 99% by 2021 onward. That improvement may reflect source-system changes, normalization improvements or municipal reporting practices rather than a change in economic behavior. Long-run nominal value comparisons should therefore be interpreted primarily as within-source descriptive evidence, with medians and concentration measures used to expose distributional structure.
The post-pandemic decline in permit breadth remains robust despite that value-coverage problem because the record-count series does not depend on declared value. The count fell 47.2% in 2020 and remained 38.3% below 2019 in 2025. That is the cleanest longitudinal finding in the paper. It does not tell us why the city slowed, but it establishes that the administrative construction regime changed.
The official pipeline evidence suggests that lack of conceivable projects is not the only issue. SF Planning can identify 74,888 pipeline units, including nearly nineteen thousand affordable units. The harder question is conversion. How many applications become approvals? How many approvals become building permits? How many issued permits move into active construction? How many starts reach completion? Each transition can fail for different reasons, and policy that improves one transition may leave the next bottleneck untouched.
For lenders and investors, this means the citywide pipeline should not be underwritten as if it were near-term competitive supply. A project with a planning entitlement, a site permit, a full building permit and active vertical construction carries very different timing risk. The same applies to major multi-phased projects whose ultimate unit counts may span many economic cycles. The pipeline is an option set, not a single delivery schedule.
For policymakers, the same distinction changes how success is measured. A city can improve zoning capacity and produce more applications without immediately raising completions. It can accelerate building permits while financing conditions prevent starts. It can produce starts while construction duration expands. The correct dashboard follows units through stages rather than selecting whichever upstream metric looks strongest.
San Francisco’s 2010–2026 permit history therefore tells a more complicated story than “the city stopped building.” It did not. The city entered the pandemic after nearly a decade of rising permit throughput, suffered an extraordinary break, recovered part of the lost volume and then stabilized at a lower level. Housing approvals and a huge future pipeline persisted, but completed housing remained below the recent historical average in 2025.
The durable research question is now conversion efficiency. The next generation of San Francisco analysis should connect project-level planning applications, building authorization, construction start and completion dates. That would allow researchers to estimate survival rates and time-to-construction across cohorts. The broad permit ledger identifies the regime shift. The next empirical task is to measure where projects are being lost or delayed inside the pipeline.
San Francisco permit activity and declared value, 2010–2025
San Francisco permit activity and declared value, 2010–2025
PropertyLabs canonical records. Declared value is nominal administrative value and coverage varies by year.
| Year | Permit records | Valued records | Declared value | Median value | Top 1% share |
|---|---|---|---|---|---|
| 2010 | 26,234 | 20,646 | $1.178B | $7,900 | 56.8% |
| 2011 | 26,238 | 21,483 | $1.559B | $8,000 | 60.7% |
| 2012 | 27,653 | 22,805 | $2.658B | $8,500 | 72.3% |
| 2013 | 30,446 | 24,984 | $3.430B | $9,500 | 73.9% |
| 2014 | 31,601 | 25,892 | $2.662B | $10,000 | 63.8% |
| 2015 | 34,120 | 27,430 | $4.259B | $10,600 | 71.0% |
| 2016 | 33,733 | 26,588 | $3.534B | $12,000 | 66.5% |
| 2017 | 34,828 | 26,790 | $4.664B | $12,000 | 77.6% |
| 2018 | 36,798 | 27,848 | $4.187B | $12,500 | 68.7% |
| 2019 | 37,303 | 28,137 | $3.673B | $14,450 | 61.4% |
| 2020 | 19,690 | 16,692 | $2.644B | $15,000 | 72.7% |
| 2021 | 22,644 | 22,441 | $3.999B | $15,500 | 74.0% |
| 2022 | 21,474 | 21,340 | $2.383B | $15,000 | 59.7% |
| 2023 | 22,087 | 22,007 | $1.905B | $15,000 | 54.4% |
| 2024 | 22,777 | 22,688 | $2.014B | $15,000 | 54.7% |
| 2025 | 23,012 | 22,858 | $2.470B | $15,000 | 57.4% |
San Francisco housing pipeline, 2026 Q1
San Francisco housing pipeline, 2026 Q1
Official SF Planning pipeline as of April 2026. Categories represent different development stages and are not equivalent to starts or completions.
| Stage | Units | Share |
|---|---|---|
| Under construction | 3,301 | 4% |
| Building permits issued (site) | 1,219 | 2% |
| Building permits approved | 2,422 | 3% |
| Building permits filed | 7,856 | 11% |
| Building permits not yet filed | 9,477 | 13% |
| Applications filed / under review | 13,036 | 17% |
| Major multi-phased projects | 37,577 | 50% |
| Total | 74,888 | 100% |
San Francisco’s post-2020 problem is not the absence of a pipeline. It is the conversion of a very large pipeline into active construction and completed housing.
Methodology
PropertyLabs aggregates `public.permits_canonical` records where `city_slug = san_francisco`. The longitudinal window begins January 4, 2010. Full-year comparisons in this report end in 2025; 2026 is shown only as year-to-date through July 30 and is never compared as a full year.
The current canonical archive contains 462,783 records. All 462,783 have `issued_date`, and `activity_date` matches `issued_date` on every record in the audited extraction. The archive does not contain a populated `applied_date`, so this report does not estimate application-to-issuance duration.
For each full calendar year, PropertyLabs calculates record count, count of records with declared value, nominal declared-value total, median, p90, p95, p99, maximum and the share of annual value represented by the top 1% of records. The top-1% calculation ranks valued records within each year and sums the largest one percent by record count.
Declared-value coverage is materially lower before 2021 than after 2021. The report therefore does not interpret long-run changes in total value as pure changes in physical construction. Value results are presented with coverage, median and concentration statistics so the reader can distinguish source completeness and upper-tail effects.
External validation uses official sources only for the claims presented here: U.S. Census Bureau QuickFacts for population, the San Francisco Planning Department 2025 Housing Inventory for housing stock, production and authorization, and the 2026 Q1 Housing Development Pipeline for stage-of-development counts.
The current canonical San Francisco extraction has no usable latitude/longitude, community, applicant or contractor fields. No parcel heat map, neighborhood trend map or firm-concentration graphic is produced. The permit-type field is present but encoded numerically, and no semantic labels are assigned without a verified municipal codebook.
Sources and limitations
Primary permit evidence: PropertyLabs canonical San Francisco municipal permit archive. External context: U.S. Census Bureau and San Francisco Planning Department official publications.
- Date field:
- Canonical issued/activity date
- Note:
- Normalization, source hierarchy, deduplication and administrative-value limitations.
- Date field:
- Census / Vintage 2025 estimate
- Note:
- 2010 Census 805,235; 2020 Census 873,965; July 1, 2025 estimate 826,079.
- Date field:
- Calendar year 2025
- Note:
- Housing stock, completions, authorizations and Planning approvals. Report documents revision and duplicate-application caveats.
- Date field:
- Pipeline status as of April 2026
- Note:
- 74,888 total new pipeline units with stage breakdown. SF Planning cautions that source accuracy and status definitions vary.
- A permit record is an administrative event, not a completed building, housing unit or realized construction expenditure.
- Declared value is nominal, is not adjusted for construction-cost inflation and may not equal final project cost.
- Value-field coverage changes across the series, especially around 2020–2021, so total-value comparisons require caution.
- The canonical archive has no populated application date, so this paper cannot estimate application-to-issuance timing.
- The current extraction lacks usable geocoordinates and neighborhood fields, so spatial claims are intentionally omitted rather than inferred.
- The official SF Planning pipeline includes projects at very different stages and contains large multi-phased developments whose units may be delivered over long periods.
- Population, permit volume and housing production are correlated contextual series; this paper does not claim a causal relationship among them.
- 2026 PropertyLabs data are incomplete through July 30 and are excluded from full-year trend conclusions.
Frequently asked questions
How many San Francisco permit records did PropertyLabs analyze?
462,783 canonical permit records from January 2010 through July 2026. Full-year trend conclusions use 2010 through 2025.
How much did San Francisco permit activity fall in 2020?
Canonical permit records fell from 37,303 in 2019 to 19,690 in 2020, a decline of 47.2%.
Did San Francisco permit activity fully recover after the pandemic?
No. Permit records rose to 23,012 in 2025 but remained 38.3% below the 2019 series high.
How large is San Francisco’s housing pipeline?
SF Planning reported 74,888 total new pipeline units in 2026 Q1, with 3,301 under construction and 37,577 in major multi-phased projects.
How many housing units did San Francisco add in 2025?
SF Planning reported a net addition of 2,669 units in 2025, 26% below the ten-year average of 3,622.
Does a permit count equal new housing units?
No. A permit record is an administrative record and the dataset includes many types of construction work. Housing-unit production is measured separately by SF Planning.
Citation & provenance
- Publisher
- PropertyLabs Research
- Last updated
- August 2026
- Methodology
- Read the methodology
Suggested citation: PropertyLabs Research. "San Francisco in Permits: The City That Slowed Down, 2010–2026." Updated August 2026. https://propertylabs.org/resources/san-francisco-city-in-permits-2010-2026
Explore the underlying permit data
Every city this report analyses, with its full municipal permit record: filings, work class, declared value coverage, and the neighbourhood and street hubs beneath each one.
Get city-level permit data access
Every number in this report is designed to trace back to municipal-source permit data and PropertyLabs normalization logic inside the product.
Share insights from this research
Follow PropertyLabs Research