Upload README.md with huggingface_hub

README.md

@@ -0,0 +1,92 @@
+---
+license: cc-by-4.0
+pretty_name: Garage-Conversion and ADU Building Permit Timelines for Four U.S. Cities
+language:
+  - en
+tags:
+  - housing
+  - building-permits
+  - accessory-dwelling-unit
+  - open-government-data
+  - urban-planning
+  - los-angeles
+  - san-diego
+  - san-francisco
+  - seattle
+size_categories:
+  - 10K<n<100K
+configs:
+  - config_name: records
+    data_files: permit-records.csv.gz
+    default: true
+  - config_name: cohorts
+    data_files: permit-cohorts.csv
+  - config_name: timelines
+    data_files: permit-timelines.csv
+---
+
+# Garage-Conversion and ADU Building Permit Timelines for Four U.S. Cities
+
+64,427 residential garage-to-ADU and garage-to-habitable-room conversion
+permit applications for **Los Angeles, San Diego, San Francisco, and
+Seattle**, with measured application-to-issuance waits, application-year
+cohort completion rates, and small-cell-suppressed timeline aggregates.
+
+Every number is reproducible from the committed, dependency-free build
+script in the [source repository](https://github.com/asafichaki/permittimelines),
+which fetches directly from each city's official open-data portal. No number
+is typed by hand. Version 2026.1, retrieved 2026-08-24.
+
+Record coverage follows each city's published history (San Francisco from
+1980, Seattle from 2000, San Diego from 2001); Los Angeles begins in 2020 by
+design, because its application dataset starts there.
+
+Also available as an R package:
+`pak::pak("asafichaki/permittimelines")`.
+
+## Files
+
+| File | Rows | What |
+|---|---|---|
+| `permit-records.csv.gz` | 64,427 | One row per unique application: city, public permit id, applied/issued dates, days to issue, project (`adu` / `habitable-room`), review track, permit type group |
+| `permit-cohorts.csv` | 43 | Per city × application year: submitted, issued, resolved share, publishable flag |
+| `permit-timelines.csv` | 72 | Per city × dimension × group: n, p25/median/p75/p90 days, within-30/60/90/180/365 shares, suppression flag |
+
+## Method
+
+1. Cohorts are formed on the **application year**, never the issue year.
+2. Records are **deduplicated on the permit number** first.
+3. Review tracks are **never pooled** into one median.
+4. Any aggregate cell under 30 records is **suppressed**.
+5. Only cohorts with roughly 75% of applications resolved enter the
+   aggregates; younger cohorts are censored toward fast permits.
+
+## Interpretation caveats
+
+- The measured wait is the homeowner's calendar wait, **not** a statutory
+  compliance measure: California Government Code 66317 runs its 60-day ADU
+  clock from a *completed* application and tolls on applicant delay; these
+  records carry only the submission date.
+- An application with an empty `issued_date` had no issued permit when the
+  records were retrieved. That may be a stalled review or an abandoned
+  project; the records do not distinguish the two, and none are marked
+  denied.
+- Cross-city level comparison is not apples-to-apples; the trend within a
+  city is sound.
+
+## Privacy
+
+No names, addresses, coordinates, or free-text work descriptions. Only
+public permit identifiers, dates, and categorical fields.
+
+## Sources and citation
+
+Original sources, all public record: [LADBS](https://data.lacity.org/d/pi9x-tg5x)
+([applications](https://data.lacity.org/d/gwh9-jnip)),
+[San Diego DSD](https://data.sandiego.gov/datasets/development-permits-set1/),
+[San Francisco DBI](https://data.sfgov.org/d/i98e-djp9),
+[Seattle SDCI](https://data.seattle.gov/d/76t5-zqzr).
+
+Prepared version: [Renology permit timeline explorer](https://www.therenology.com/tools/permit-timeline).
+No city endorses this preparation. Cite the city agencies as the original
+sources and Renology as the prepared version.