vantovault/pittsburgh-str-census-2026

Dataset Card for the Pittsburgh Short-Term Rental Census (Lawrenceville and Bloomfield, August 2026)

Dataset Details

Homepage / methodology: https://vantovault.com/library/pittsburgh-short-term-rental-census/ DOI (concept, resolves to latest version): https://doi.org/10.5281/zenodo.22258181 Also on GitHub: https://github.com/vantovault-stack/open-housing-data Maintained at: Van to Vault (vantovault.com) License: CC BY 4.0 — use with attribution to Van to Vault (vantovault.com) Point of contact: Stephan D., Van to Vault

Dataset Summary

A hand-checked, listing-level count of short-term rentals in four official City of Pittsburgh neighborhoods — Lower Lawrenceville, Central Lawrenceville, Upper Lawrenceville and Bloomfield — built while the City's short-term rental zoning bill (Council Bill 2026-0009) was pending. Every row is a live listing that can be opened and checked. Nothing in the dataset is modeled or estimated.

Two files, two platforms, deliberately kept separate because the counts may overlap and must never be added together:

• airbnb — 275 rows (273 live in scope as of 2026-08-31, plus 2 rows kept after delisting and marked delisted_by). Room type, capacity, host id and the host's in-scope listing count, review count, star rating, coordinates, distance to the nearest neighborhood boundary with an explicit boundary-uncertainty flag, and three observed two-night, two-guest stay totals (mid-September midweek, late-September weekend, December weekend).
• midterm_furnishedfinder — 42 units across 40 properties on Furnished Finder, the 30-day-plus mid-term market. Published asking rent and minimum stay are platform fields, not derived.

Neighborhood assignment uses the City of Pittsburgh's official Neighborhoods layer (WPRDC, CC-BY). Per-neighborhood summary tables, the full method and the running list of limits are on the methodology page above.

Supported Tasks

Reference / lookup dataset for housing policy analysis. room_type and neighborhood support grouping and counting; the three rate_2night_total_* columns support price-window comparisons; host_inscope_listings supports host-concentration analysis. Not a training corpus.

Languages

English.

Dataset Structure

Data Instances

One row per listing (airbnb) or per rentable unit (midterm_furnishedfinder).

Data Fields — airbnb

Field	Description
listing_id, listing_url	Airbnb listing id and its public URL
neighborhood	Official City neighborhood the displayed pin falls in
room_type, property_type	Airbnb's own categories (e.g. Entire home/apt, Private room, Hotel room)
capacity_guests	Stated guest capacity
superhost	true / false
review_count, star_rating	As displayed; star_rating is blank when a listing has no reviews — blank means not observed, never zero
host_id, host_inscope_listings	Host id and how many in-scope listings that host has in this census
lat, lng	Platform-displayed coordinates. Airbnb displaces pins by roughly 150 m; do not treat as an address
dist_to_boundary_m, nearest_other_neighborhood	Distance to the nearest neighborhood boundary and the neighborhood on the other side
boundary_uncertain_any_150m	yes when the pin is within 150 m of a boundary
boundary_uncertain_material	yes when that boundary crossing would move the listing out of scope or across the Lawrenceville/Bloomfield line — 63 of 273 live rows
rate_2night_total_2026-09-15_17_usd, ..._2026-09-25_27_usd, ..._2026-12-11_13_usd	Observed total for a 2-night, 2-guest stay on those dates, including per-stay fees; blank = not bookable on those dates. Not nightly rates; never average across windows
checked_date, checker	When the row was last verified and by whom (claude-draft = AI-assisted first pass; stephan = human spot check)
delisted_by, live_as_of	Delisting date if the listing disappeared before publication; otherwise the date it was last seen live

Data Fields — midterm_furnishedfinder

Field	Description
listing_id, property_id, unit_index	Furnished Finder ids; a property can carry several units (listing_id = property_id_unit)
listing_url	Public listing URL
neighborhood	Official City neighborhood
property_type, unit_class	Platform type and whether the unit is an entire unit or a room
minimum_stay_days, rent_monthly_asking_usd	Published platform fields
sleeps, bedrooms, bathrooms	Published platform fields
lat, lng, dist_to_boundary_m, boundary_uncertain_material, nearest_other_neighborhood	As for the Airbnb file
observed_date	Harvest date

Data Splits

No splits — one flat table per config.

Dataset Creation

Source Data

Airbnb public search (any-dates inventory, then three dated rate harvests) and Furnished Finder public search, both read anonymously. Pagination on both platforms under-delivers against displayed counts, so coverage was completed by quadrant-bounded searches and constructed cursors; counts are therefore floors, not ceilings. A 27-point stratified sample of neighborhood assignments was cross-checked against OpenStreetMap (27/27 agreed). Zoning-district assignment was tested and deliberately not published, because pin displacement makes it unreliable at the parcel scale.

Annotations

Single-maintainer manual review (Van to Vault), August 2026, AI-assisted first pass with human spot checks (checker column).

Considerations for Using the Data

• Counts are floors. Single-search counts undercount; this census closed the known gaps but cannot prove completeness.
• Eight Airbnb rows are hotel-style rooms (room_type = Hotel room), not dwelling units. Counts framed as housing exclude them.
• Rows flagged boundary_uncertain_material = yes may sit in a neighboring neighborhood.
• Rate figures are observed stay totals on specific dates, not nightly rates, and are not comparable across windows.
• Listings churn. Every row carries its check date; re-verify before quoting.
• Not legal, financial or lending advice.

Citation

Van to Vault. Pittsburgh Short-Term Rental Census: Lawrenceville and Bloomfield (August 2026). DOI 10.5281/zenodo.22258181. https://vantovault.com/library/pittsburgh-short-term-rental-census/

Dataset Card Contact

Stephan D. — Van to Vault (vantovault.com)