1<!-- Author: Simon-Pierre Boucher — simon-pierre.boucher@uqo.ca -->2# Raw data (not redistributed)34Immutable inputs of this paper. Files are **not committed** (private transaction microdata); checksums are pinned in `SHA256SUMS`.56| File | Origin | Rows | Content |7|---|---|---|---|8| `wp22_clean_transactions.parquet` | WP 22 (`sharp-hpi-quebec/data/processed/clean_transactions.parquet`), derived by WP 22 scripts 01–04 from the SQLite `.backup` of the Vrai Prix production database `M3U96a:~/apps/vrai-prix/data/vraiprix.db` taken 2026-10-21 00:30 America/Toronto (sha256 `2816d8dd…` pinned in WP 22) | 769,995 | registered sales 2021-01-04 → 2026-10-07 joined to the assessment roll (`units`), geocoded, MRNF/MAMH geography, WP 22 cleaning flags (`exclude_reason*`, `in_headline_sample`) |9| `wp22_raw_units_stock.parquet` | same snapshot, table `units` (classification + value columns) | 3,747,008 | housing stock (roll units) used for neighbourhood density features |10| `wp22_raw_market_index.parquet` | same snapshot, table `market_index` | — | Vrai Prix monthly price-per-m2 index by property type (used only for description; **not** given to agents because it is computed with post-cutoff data) |11| `wp22_raw_manifest.json` | WP 22 extraction manifest | — | table/row inventory of the snapshot |1213Vrai Prix (www.vrai-prix.com) is the author's own platform; the registered-sales table is assembled from the public Quebec land register and the municipal assessment rolls (MAMH). Use is authorized for research by the owner; **individual records are not redistributed** (addresses). Agents never receive street addresses: coordinates are rounded to 3 decimals and the municipality/borough name is given instead.1415Immo-Ka listings (M4M64b `~/apps/immo-ka/data/immoka.db`, 479,265 listings first seen 2026-08-12 → 2026-10-27, 195,503 linked to a roll unit) were audited (`docs/data_access_checklist.md`) but are **not used in the main arms**: coverage starts after the test window and ask prices would create an information asymmetry across arms. They are reserved for the prospective-set descriptive analysis (`scripts/60_fetch_prospective.py`).1617To rerun from scratch: copy the WP 22 outputs here under the same names (or regenerate them with the WP 22 pipeline from a Vrai Prix snapshot), then `make data`.18