# USTA Permit Pulse — Methodology

**US Tech Automations Research** · v1 — applies to every Permit Pulse edition

This document is series-level: the sources, gates, and seal discipline below hold for
every edition. Anything that changes month to month — the computation date, the exact
30-day window, per-metro effective windows, and per-metro valuation coverage — is
stated on that edition's own page and in its data file, never here.

## 1. What the Permit Pulse measures

The Permit Pulse is a cross-sectional report of **residential building permits issued** in 8 US jurisdictions over a rolling 30-day window, computed from each city's own public open-data portal. "Residential" means the permit passed our per-city residential gate (see §4) — single-family and small multi-family construction, additions, alterations, and demolitions. Commercial and standalone sub-trade permits are excluded at ingest.

## 2. Data sources

All sources are official city open-data portals, accessed read-only with no authentication:

| Jurisdiction | Portal | Dataset |
|---|---|---|
| Los Angeles, CA | data.lacity.org (Socrata) | `pi9x-tg5x` (Building Permits) |
| San Francisco, CA | data.sfgov.org (Socrata) | `i98e-djp9` (Building Permits) |
| Austin, TX | data.austintexas.gov (Socrata) | `3syk-w9eu` (Issued Construction Permits) |
| Chicago, IL | data.cityofchicago.org (Socrata) | `ydr8-5enu` (Building Permits) |
| New York City, NY | data.cityofnewyork.us (Socrata) | `rbx6-tga4` (DOB NOW: Build – Approved Permits) |
| Seattle, WA | data.seattle.gov (Socrata) | `76t5-zqzr` (Building Permits) |
| Cincinnati, OH | data.cincinnati-oh.gov (Socrata) | `uhjb-xac9` (Building Permits) |
| Scottsdale, AZ | maps.scottsdaleaz.gov (ArcGIS OpenData) | Tabular layer 12 (Building Permits) |

## 3. Collection cadence and sealed provenance

- **Daily ingest:** every day at 08:00 UTC, an automated job pulls a 30-day window of issued permits from each portal (per-source isolation: one city failing does not block the others).
- **Daily seal:** every day at 09:00 UTC — one hour after ingest — each permit observation is sealed into an append-only snapshot store. Each sealed row carries a per-row content hash (SHA-256) and each daily run carries a batch hash, computed **at seal time, before any downstream outcome is known**. Sealed rows are never edited or deleted; re-running a seal is idempotent (duplicate rows are ignored, not rewritten).
- **Why this matters:** the seal makes the dataset tamper-evident. Any figure in a Permit Pulse edition can be traced to specific sealed rows whose hashes prove the data existed in that exact form on that date. We cannot retroactively adjust history to flatter a narrative — the hashes would no longer verify.
- Each edition states its own computation date and the exact issue-date window it covers, at the top of the edition page and in the header of its data file. Editions are never recomputed after publication: a published edition is a fixed observation of what the cities had published on that date, kept at a permanent URL (`/permits/research/permit-pulse/<YYYY-MM>`) so a citation stays valid.

## 4. Residential gates (per-city)

Cities publish permits with different schemas, so the residential filter differs in mechanism and strictness:

- **Structured gates** (a residential flag or class exists in the source data): Seattle (`Single Family/Duplex` class), Austin (residential permit class), San Francisco (use-based gate), Los Angeles ("1 or 2 Family Dwelling" permit types), Scottsdale (SFR permit types), Cincinnati (RCO — Residential Code of Ohio — classification).
- **Keyword gates** (no structured residential column exists): Chicago and New York City permits are included only when the work description self-describes as residential. These are **deliberately conservative**: a residential permit with a vague description is excluded rather than guessed. Chicago and NYC counts are therefore lower bounds on residential activity, by design. NYC additionally excludes renewal permits without changes.

**Consequence: permit counts are not comparable across cities.** They reflect both real activity and per-city permit definitions plus our gate strictness. Compare a city to itself across editions, not to other cities.

## 5. Valuation handling

- Valuations are **applicant-reported** project costs/valuations as published in the public record. They are not appraisals and are not verified by us.
- Reported costs of **$0 or negative are normalized to "missing"** — they are never counted as literal zeros in totals or medians. This matters most for Austin (effectively all residential rows publish $0/blank costs → no valuation aggregates reported) and Cincinnati (frequent $0 costs → reduced coverage).
- Totals are sums over permits with a usable valuation; medians are computed over the same set. Per-metro valuation coverage is disclosed in the data file.

## 6. Known limitations and lags

1. **Limited trend depth.** All-8-metro sealed coverage began in early June 2026. The June 2026 edition is cross-sectional only. Month-over-month comparison begins with the July 2026 edition — the first with a sealed prior edition to compare against — and is stated only against editions we actually sealed. No year-over-year and no seasonality claims are made at any point in 2026: that history does not exist yet, and we do not model it.
2. **Los Angeles reporting lag:** the City of LA publishes permits in batches with lag historically observed up to ~16 days. The most recent days of any LA window may fill in after our snapshot. (This is why our ingest window is 30 days rather than 14.)
3. **Per-metro effective windows:** feeds do not all reach the same last date at snapshot time (portal lag differs per city). Each edition's data file states the effective window per metro, and the edition page flags any metro whose window fell materially short.
4. **Source data quality is the cities'.** Misclassified permit types, typo'd valuations, and late corrections in city portals flow through to us. We seal what the city published on that day.
5. **Residential-only scope.** This is not total construction activity. Large commercial projects are intentionally out of scope.
6. **Seattle demolition values:** Seattle's highest-value line items are demolition permits whose reported value appears to reflect the full redevelopment project value rather than the teardown alone.

## 7. Privacy

All data is public record published by the cities themselves. The Permit Pulse reports aggregates plus address-level facts about properties (address, permit type, valuation, issue date). We do not publish owner or individual names, and our sealed snapshot store excludes them by design.

## 8. CITATION

```
This data is free to use, share, and republish — including commercially —
under the Creative Commons Attribution 4.0 International license (CC BY 4.0),
with attribution to:

  US Tech Automations Research
  https://ustechautomations.com/permits/research

Suggested citation (cite the EDITION permalink, not the current-edition URL,
so the figures you quote stay at the URL you cited):

  US Tech Automations Research, "USTA Permit Pulse — <Month Year>:
  30-Day Residential Construction Permit Activity Across 8 US Metros,"
  <Month Year>.
  https://ustechautomations.com/permits/research/permit-pulse/<YYYY-MM>

Press inquiries: operations@ustechautomations.com
```