Land Registry Price Paid Data, explained
Every registered sale in England and Wales since 1995 is public. What the file contains, what it leaves out, why the trailing months are always wrong.
HM Land Registry publishes the price of every residential sale registered in England and Wales since January 1995. It is the reason a UK house price can be checked rather than believed, and it is the ground under most of what MarketCode does with markets. It is also misread constantly. This is what is in it, what is not, and how to use it.
What a row contains
One row per registered sale: the price paid, the date of transfer, the address as registered, the postcode, the property type (detached, semi-detached, terraced, flat, other), whether it was a new build, the tenure (freehold or leasehold), and two flags that matter more than they look.
The first is the category. Category A is a standard market sale at full value. Category B covers everything else: transfers under a power of sale (repossessions), right-to-buy sales, and transfers where the price is not a market price. A median that includes category B is not a market median. The monthly volume series counts both but computes prices on A only, and says so.
The second is new build. The first sale of a new flat carries a premium the second sale does not. Treat a new-build first sale as a comparable for the resale market and you will overvalue the resale.
What it leaves out
Scotland and Northern Ireland, which have their own registers. Sales that were never registered, which is rare now but common before compulsory registration was complete. Commercial property, which appears only where it is registered as residential. And, for a property that has not changed hands since 1995, everything: no row does not mean no value, it means no sale in the window. The sold-price history of a property says "no registered sales since 1995" as a real answer, not an error.
The trailing months are always incomplete
A sale completes on one day and is registered when the buyer's conveyancer lodges it, typically weeks later and sometimes months. So the latest month in the file is a fraction of what it will become. Count sales in the most recent quarter and you will conclude the market has collapsed; wait six months and the same quarter looks ordinary.
Every MarketCode series that reads Price Paid carries a marker for this. The volume series returns a provisional_from month; the house price index returns a last settled period. The market pages leave the filling months out of their twelve-month totals rather than quoting a number that will rise. The registration lag entry in the glossary is the short version.
From a file to a property
The address in the file is the address as registered, which is not always the address as written on a letter or shown on a portal. Joining a sale to a property means resolving both to the same UPRN, the identifier every address in Great Britain carries. Where the join can be made with confidence, the sale is attached to the property and appears in its registered sales and in the comparables behind a valuation. Where it cannot, the sale stays in the area statistics and is attached to nothing. That conservatism is why the cube's sale count runs below the registry's own: it counts matched sales, and matching is deliberately strict.
Three questions and the right tool for each
- How many homes sold in SE23 each month, and at what median? market_volume_series, district by district, from 1995.
- What did this flat sell for before? property_history for the readable story, transactions_by_uprn for the rows.
- What do two-bedroom flats in Lambeth sell for? Not Price Paid directly: the market cube, which is built from it and grades each cell by how many sales sit behind it.
How the data is loaded
The registry publishes a monthly change file rather than a fresh dump. MarketCode loads it as additions, changes and deletions, so a corrected registration updates the existing row instead of duplicating it. The source page carries the licence (Open Government Licence v3.0, with attribution) and the refresh cadence, and lists every tool that reads the file.
Do it today
Ask your assistant, with MarketCode connected: "How many homes sold in E5 in each of the last 24 months, and from which month is the count provisional?" The answer will name the month, because the tool returns it. The Hackney market page shows the authority's picture; its postcode districts each carry the monthly sale counts drawn from this file.