How much of the housing market the sold-price record can measure

A price per square metre needs two numbers: what a home sold for, and how big it is. The first is public for every sale in England & Wales. The second is not. Across the whole register, 29,580,544 sales have been recorded and 23,005,376 of them could be given a rate, which leaves 6,575,168 that could not. This article is about those.

Why does a sale have no floor area?

Floor areas here come from Energy Performance Certificates, matched to sales by address. A sale goes unmatched when no certificate exists for that address, when the address is written differently in the two registers, or when the certificate records a floor area that fails a basic sanity check. Certificates became compulsory on sale in 2008, so a home that last changed hands before then and has not been marketed since may never have had one.

This is not a defect peculiar to this site. Every service that publishes a price per square metre for England & Wales is doing the same join and dropping the same sales. The difference is whether it tells you.

Which homes are missing?

Not a random fifth. The match rate runs from flats at 88.18% against detached houses at 65.19%, a gap of 23.0 percentage points.

The pattern is consistent and it has a plain explanation: the more individual a building is, the harder it is to match one register’s address to another’s. A flat has a number, a block and a postcode, and those three strings agree across databases. A detached house may be a name rather than a number, may have been renamed, and may sit on a lane that two registers spell differently.

The consequence is worth stating plainly, because it runs against intuition: the price per square metre statistics on this site describe flats and terraces better than they describe detached houses. On a district of Victorian terraces the rate rests on most of the market. On a district of large detached homes it rests on two thirds of it. Both are published; only one of them is nearly the whole picture.

Has coverage improved?

Yes, and the shape of the improvement is informative. Matching runs from 74.15% of sales in 2011 to 83.49% in 2023, with the current overall figure at 78.15%.

Two things move this series. It climbs after 2012 because the certificate register keeps growing: a home assessed in 2020 can be matched to its sale in 1998, so old years keep gaining coverage long after the fact. And it dips around 2009 to 2011 because those years were thin, transactions collapsed, and the homes that did sell were disproportionately the ones that had not been marketed in the certificate era.

Are the missing homes different?

This is the question that decides whether the gap matters, and it has an answer.

Over the last five years the homes that do carry a rate had a median sale price of £285,000, and the homes that do not had a median of £290,000. The ratio is 0.983: the priced subset is 1.7% cheaper at the median than the market it is drawn from. That is a small number, and publishing it is the point. It is the difference between a statistic whose selection you can check and one you cannot.

A ratio near 1.0 means the priced subset is close to price-representative of the whole market. It does not mean the two groups are identical in every respect, and the property type table above shows one respect in which they clearly are not. What it does mean is that the missing fifth is not systematically the expensive end or the cheap end, so a median rate computed on the priced subset is not being pulled by selection in the way it could be.

When was the floor area measured?

A certificate records a floor area on the day it was issued, and a sale happened on some other day. Match a 2024 certificate to a 2001 sale and you have priced the 2001 transaction using the home as it stands now, extension included. The error runs one way: homes grow more often than they shrink, so an undated join makes historic prices per square metre look too low, and therefore makes growth look too fast.

There is a way to measure it. 3,873,329 addresses in the certificate register have been assessed more than once, a median of 7.73 years apart, which is the same dwelling measured twice. Between the two assessments 55.07% of them record exactly the same floor area. Most homes are not quietly growing, and the ones that are can be bounded rather than guessed at.

This site publishes a date-bounded series alongside the unbounded one for exactly this reason, and the methodology page sets out the five-year rule it uses. The bias is small, and it is in the same direction as the renovation bias in the repeat-sales index, which is worth remembering whenever the two series agree.

What should you do with this?

Three practical points. First, read the matched count printed beside every rate on this site: it is that figure’s own denominator, and it is not the number of sales in the area. Second, treat a rate in a detached-heavy area as describing a smaller share of its market than the same rate in a flat-heavy one. Third, when comparing two areas, check both denominators before you compare the rates; that is the first step in comparing areas honestly, and it is the one most often skipped.

Every figure here is read live from the published dataset and is current to June 2026. The underlying coverage tables are in the same file, and what an EPC floor area measures is its own article.