PermitPipeline Preconstruction Activity Index — September 2026

Edition: September 2026 · covering June–August 2026 · data extracted September 28, 2026

The PermitPipeline Preconstruction Activity Index is a year-over-year momentum measure centered on 100: the latest three complete months' permit applications versus the same three months one year earlier. 100 means unchanged; 90 means activity is down 10%. Each city is compared only with itself.

A permit application is an early public indication of intended work, not proof that construction will proceed. PermitPipeline treats it as an early signal.

By Josh Steinman, founder of PermitPipeline. Before building PermitPipeline, he spent nearly 20 years managing construction projects and representing owners.

NYC initial job filings fell and San Francisco building-permit applications rose in June–August

From June through August, initial NYC DOB NOW job filings were down 7.9% year over year and San Francisco building-permit applications were up 13.1%. August alone: NYC initial filings fell 10.1%; San Francisco applications rose 12.8%. In New York the decline in initial filings does not show in total filing counts, because filings attached to existing DOB jobs moved the other way: initial filings were 44.1% of August's 13,252 distinct filings, while total filing volume fell 0.4%. By work type, initial Alteration filings fell 9.5% (17,589 to 15,916, more than the whole citywide net change of 1,575 because other work types moved the other way), while initial New Building filings rose 6.9% (463 to 495).

San Francisco: counting both permit systems

San Francisco has been moving routine permit types out of the Department of Building Inspection's legacy system and into PermitSF, a new online portal whose records are published as a separate open dataset:

Counted from the legacy system alone, as earlier editions did, those moves look like a collapse in activity: June–August applications would read 78.6. From this edition the San Francisco index counts both systems.

Electrical and plumbing permits were never part of the building-permit index; electrical permits are reported separately below, across both systems. Fire-alarm and fire-sprinkler work is left out of both years because PermitSF now issues it as a fire permit rather than a building permit, and on that basis its count jumped from 541 to 777 in June–August (+43.6%), far more than the building work beside it: a wider universe, not more work. Counting it anyway would put the index at 115.7.

Two smaller effects push the reading up, and they are shown here rather than folded in. 55 PermitSF applications were stopped and then refiled on the same parcel within 14 days, so the same job counts twice (stopped records are 5.8% of PermitSF building permits, against 1.6% withdrawn or cancelled in the legacy system a year earlier). And solar and energy-storage applications rose from 22 to 103 as the city gave them their own permit type. Leaving both out, the index would read 110.7 rather than 113.1. How the two systems were checked against each other — double counting, permits split into several records, withdrawn applications and neighborhood mapping — is set out on the methodology page. The August 2026 San Francisco figures below use the same universe.

Finding 1 — NYC filing composition, boroughs and ground-up filings

Initial job filings fell (6,504 to 5,850) while filings attached to existing jobs rose (6,758 to 7,312). The dataset excludes electrical, elevator and limited-alteration jobs, which DOB tracks separately.

Boroughs (initial filings, June–August):

Ground-up (initial New Building filings, June–August):

Finding 2 — the neighborhood map, qualified movers only

Movers shown below passed the stability gate — 60+ (NYC) / 50+ (SF) prior-year filings, an absolute change of at least 15, and the same direction over the six months preceding this window (December 2025–May 2026 versus December 2024–May 2025), periods that do not overlap the June–August window reported here. That is a non-overlapping persistence check, not a test of statistical independence: the two windows measure the same area and are correlated. Areas failing the gate appear in the CSVs and on the city pages but are not shown here. Display rule: up to the six largest qualified moves in each direction, ranked by index; the city pages carry every row that passes the rule. This edition: 45 NYC neighborhoods and 14 San Francisco neighborhoods qualify.

NYC rising:

NYC falling:

SF rising:

SF falling: no SF neighborhood qualified as a faller under the stability gate.

Finding 3 — San Francisco's August universe, itemized

Over-the-counter work, counted across both permit systems, rose 15.5% (1,741 to 2,011), and plan-reviewed additions, alterations and repairs rose 2.4% (84 to 86) year over year in August. Among plan-reviewed projects the median applicant-declared value was essentially unchanged: $149,810 → $150,000 (p25 $31,451 → $40,000, p75 $366,844 → $420,000; n=84 and 86, a positive declared value on 100.0% and 100.0%). The samples include 7 and 9 applications declared at $1, a placeholder often used for deferred submittals or changes to existing permits; leaving them out as a sensitivity check, the medians are $161,254 and $175,000. A populated value is an applicant's declaration, not an audited project budget. Electrical permits, a separate measure of trade-permit activity, were 1,410 in August versus 1,432 (-1.5%), including 8 filed in PermitSF, where standard electrical permits began moving in July.

Finding 4 — permit processing

San Francisco (permits issued in August; elapsed calendar days from filing to issuance, which includes the applicant's own time and leaves out applications still waiting): over-the-counter work across both permit systems, median 0 → 2 days; issued the same day 57.5% → 39.6% (n=1,719 vs 1,806). Plan-reviewed alterations: median 172 → 206 days (n=77 vs 71). In August 2026 the legacy system's over-the-counter permits took a median 2 days (43.6% the same day, n=1,227), and PermitSF building permits took a median 2 days (31.3% the same day, n=579).

New York (standard plan examination, initial filings; professional self-certified filings — 53% of the initial filings approved in August — are approved in a median of 2 days and are excluded): approvals completed during August took a median 32 days in 2025 and 30 in 2026 (n=2,395 vs 2,279). May cohorts at matched 120-day maturity: 79.5% approved within 120 days in 2025 versus 76.1% in 2026, lower by 3.5 points. The May 2025 cohort is fully matured. The May 2026 cohort reached 120 days on the extraction date, but the city's dataset was last refreshed on September 27, 2026, so 28 standard-plan filings made from May 30, 2026 to the end of May had not yet been observed for a full 120 days; 10 of them had no approval. If every one of them were approved within 120 days, the 2026 share would be at most 76.5% rather than 76.1%, and New Building's would be at most 14.5% rather than 12.8%. Changes are computed from unrounded rates, so displayed percentages may not subtract exactly.

Work-type segments, day-120 approval share. Every work type with at least 100 filings in both cohorts is published. Changes of at least 10 percentage points are highlighted as exploratory findings. These thresholds were fixed in the first edition and are applied unchanged.

One segment meets the highlight threshold in this edition, reported as exploratory: New Building (27.2% to 12.8%). The dataset contains only each filing's current status, so it cannot establish why the remaining filings had not been approved by day 120.

Supporting analysis — contractor concentration

Contractor concentration fell in New York: the 50 busiest general-contractor license numbers held 20.4% of the city's GC permit records, against 21.6% a year earlier (127,256 and 120,492 records; each record is attributed to its applicant's license, and the top 50 are re-ranked in each period). San Francisco's concentration measure is suspended from this edition. The city's contractor records cover only permits in the legacy DBI system, and the work that moved to PermitSF — reroofing and window replacement above all — is among the highest-volume trade work, so the San Francisco share would move for a system reason rather than a market one. Windows: twelve months ending August 2026 against the twelve months ending August 2025.

Revision check — the August 2026 edition, re-read on September 28, 2026

The August 2026 edition was extracted on August 24, 2026 and is immutable. Both cities back-date records, so its windows were re-read on September 28, 2026 to measure how far the published readings have drifted.

Month by month (initial filings / applications, published to re-read): NYC 2025-07: -2; 2026-05: -3; 2026-06: -4; 2026-07: -5; SF 2026-06: +1; 2026-07: -4.

San Francisco restated. The August 2026 edition's San Francisco reading (81.6) was computed from the legacy DBI system alone, which by then no longer received the permit types the city had moved to PermitSF. On this edition's corrected universe (both systems, fire-protection work excluded from both years) the same window reads 109.7: 6,235 applications against 5,683 a year earlier. The table above measures ordinary back-dating on the original basis; the restatement is a correction of the universe.

Revision note. Under the methodology's thresholds this drift requires a revision note on the August 2026 edition: NYC mover: Forest Hills: no longer qualifies; NYC mover: Sheepshead Bay-Manhattan Beach-Gerritsen Beach: no longer qualifies; San Francisco restated: this edition counted only the legacy DBI system, which from February 2026 no longer received the permit types the city moved into PermitSF; counting both systems, May–July reads 109.7 rather than the published 81.6. The note is added to that edition's page with this date; its original text is unchanged.

Data and windows for this edition

Single frozen extraction on September 28, 2026 (edition_2026_09/extraction.json, committed). Current window June–August 2026 against June–August 2025; stability windows December 2025–May 2026 against December 2024–May 2025; latest-month tables compare August 2026 with August 2025; the approval cohort is May 2026 against May 2025. All counts are COUNT(DISTINCT identifier). NYC: DOB NOW Job Application Filings (w9ak-ipjd); initial = a job_filing_number ending -I1; approval timing from approved_date minus filing_date. SF: DataSF's deduplicated-on-primary-address view of the legacy DBI system (f2jc-ivnc), all applications regardless of later status, less over-the-counter fire-alarm and fire-sprinkler work, plus PermitSF building permits (tyz3-vt28, record numbers BLDG-, dated by submission), all statuses; issuance timing from issued_date minus filed_date and from PermitSF submission to issue; freshness and delta-zero reconciliation of the legacy view against raw i98e-djp9 asserted at build; PermitSF parcels mapped to Analysis Neighborhoods through the city's parcel file (acdm-wktn), 2,457 of 2,457 records mapped. Key and mapping coverage (published and gated at ±1 point): contractor-license key present on 100.0% → 100.0% of NYC GC permits; neighborhood mapping present on 99.64% → 99.59% of NYC initial filings (72 and 74 filings unmapped, reported as Unknown and included in every city and borough total) and 100.0% → 100.0% of SF legacy-system applications. The SF declared-value statistics cover exactly permit_type='3' in the deduplicated view, all statuses, one row per permit; plan-reviewed work did not move to PermitSF. Full definitions, thresholds and the revision policy are on the methodology page.