What we may publish, and what we may not

Property data has a legal surface that market data does not. A dataset can be free, official, machine-readable and still forbidden to this site — and the prohibition is usually one clause deep in a terms page rather than anything announced. This report records those clauses, because a research section that quietly breaks them is worth less than one with gaps.

The rule this section runs on

Every source carries republish: false unless its licence plainly permits redistribution. That means the research cites a source, quotes the individual facts it states, and does not reproduce its rows. Where a claim would require mirroring a dataset to stand up, the claim is not made.

Datasets that forbid what we would want to do

New Zealand is the strictest. LINZ will license bulk ownership, but its Licence for Personal Data requires that the data not be indexed by any public internet search engine and not be passed to unlicensed third parties. A public research page is exactly indexing. So bulk New Zealand ownership cannot appear here at all — while LINZ's district valuation roll, 287,746 properties under CC BY 4.0 with values and floor areas, is perfectly usable. One agency, two licences, opposite answers.

France's DVF carries the same no-indexing condition, and is a rolling five years that is deleted and replaced twice a year. Anything built on it has to be snapshotted at a stated date and cited, never mirrored.

Dubai's "open data" is not an open licence. The Land Department publishes nine datasets under that banner while its site terms authorise use "solely for your personal, noncommercial use", and every query passes a CAPTCHA. Its project-level file — units, percentage complete, escrow account, joinable to the developer register — is the closest thing anywhere to a units-by-developer table, and we may cite single stated facts from it and nothing more.

New South Wales splits the same way as New Zealand. Bulk property sales, every sale since 1990, are CC BY-NC-ND: non-commercial, no derivatives. The bulk land-value file is CC BY 4.0 and is fine. The two sit on the same portal.

Quebec redacts and forbids re-identification. The province-wide assessment roll is the richest open property data in Canada — per-unit land value, building value, year built, floor area, dwelling count, and the date ownership was registered — with owner names removed, and a licence that forbids putting them back. Owner lookups are free and legally public one at a time on the municipal site. Both halves of that have to be respected at once.

British Columbia is the subtlest. BC Assessment's own site forbids commercial use of its values. The City of Vancouver redistributes the same values in its property-tax report under an open licence, 1,553,448 rows. Same numbers, different licensor, different answer — and the difference is easy to miss.

São Paulo's cadastre is share-alike (CC BY-SA) and its vacant-property inspection file is non-commercial, so only the CC0 layers are cleanly usable.

Listing portals: prohibited, and cited anyway

The prices people think of as market data live on listing portals, and their terms forbid taking them. Rightmove states it plainly — "Rightmove prohibits the scraping of its content" — and Realtor.ca, Zillow, Domain and Idealista carry equivalent clauses. None of it is taken.

This costs less than it appears. Portals publish indices and research downloads that are meant to be cited: Rightmove's monthly house-price index, Zillow's research CSVs, the real-estate boards' own market reports. An index is a measurement someone stands behind; a scraped asking price is a number somebody hopes to get. For actual transaction prices the research uses registers that permit it — England and Wales' price-paid file, 5.5 GB of every sale since 1995 under an open licence with commercial reuse allowed; New York's recorded instruments; Cook County's 2.7 million sales; Singapore's daily resale file, licensed "free forever for personal or commercial use".

Hosts that will not open, which is not the same as data that does not exist

Worth recording so nobody re-runs the same attempt: the UK's land-data service redirect-loops without a human sign-in; Ireland's price register sits behind a bot challenge; Spain's cadastre fails TLS chain validation; Mexico City's open-data and GIS hosts refuse connections from outside Mexico; several Gulf hosts time out at the TCP level; and Saudi Arabia's property-enquiry service returns HTTP 200 with a rejection body, which means status codes cannot be trusted there. Ontario's own assessment site sits behind bot protection, and its title portal renders as an empty shell.

A blocked host is a finding about access, not about the underlying record. The city pages distinguish the two: a building whose owner could not be established says whether the register is closed, the search is paid, the licence forbids naming, or the source simply would not open.

Personal data

Only entities are named. Where a register names an individual — British Columbia's transparency registry does exactly that, by design — it is used to establish which company or trust holds the asset, not to publish a person. A family office is named as the office.