Detached, semi, terraced or flat: price per square metre by property type

Divide the price by the floor area and the four property types stop looking like one market. In the 727 districts that publish a median for both, a flat sold for less per square metre than a terraced house in 90% of them. Detached houses sat above terraced houses in 93% of the 1,007 districts publishing both, by 22% at the median. That order is the reverse of what most people expect, and the reason it surprises them is the reason to read the by-type figure at all.

The expectation comes from cash prices. A flat is cheaper than a house, so a flat must be the cheap way to buy, so per square metre it must be the dear one, because you get less space. The first two are true and the third does not follow. This article is about why, with the numbers read live from the current dataset, 12 months to August 2026.

Which property type costs most per square metre?

It depends on whether you read across the country or within a district, and the two answers disagree. Read across the country first:

The middle column ranks places as much as types. Flats are where the cities are, and in a city every type costs more, so the middle flat district sits high in the table even though inside almost any one district the flat is the cheapest type per square metre. That is a composition effect, and it is the same trap the median-and-mean article describes from the other side: a national figure for one type is a figure about where that type is built. It is not a figure about the type.

So the honest comparison is within a district, one type against another over the same streets and the same year. On that basis the order is stable enough to state. Detached above semi-detached in 90% of the 1,142 districts publishing both, semi-detached above terraced in 85% of 1,293, and terraced above flats in 90% of 727. At the median, a terraced house sits 14% above a flat on the same streets; a detached house sits 22% above a terraced one.

Why do flats come out cheaper per square metre inside a district?

Four reasons, and the first is the measure itself. The floor area on an EPC is the area inside the home’s own walls. For a flat that excludes the shared stair, the lift, the landing and the bin store, all of which the buyer is paying for through the price and none of which appears in the denominator. A house sells with a garden, a drive, a garage and the land under it, and those enter the price without entering the square metres either. The rate rewards the house for what it does not count.

The second is tenure. Flats are almost always leasehold, and a leasehold sale is a sale of a diminishing term with a service charge and often a ground rent attached. The tenure article shows how far that alone moves the rate. The third is stock: a large share of the flats in the record are conversions and estate blocks of the kind that sell at a discount to the houses around them, and the type code cannot tell a converted Victorian villa from a walk-up. The fourth is that the exceptions cluster. Where flats do sell above terraced houses per square metre, it tends to be where the flats are new or riverside and the terraces are not, which is a fact about the local stock rather than about flats.

How big is the gap, district by district?

Big enough that pricing a flat against a district’s all-types median gets the wrong answer. Six districts that publish all four types, chosen for volume rather than for drama:

Read along any row and the four figures step down from left to right more often than not, with the odd pair swapping places where one type is thin or the local stock is unusual. Read down the flat column and the level is set by the place: a flat in Hove is not cheap, it is cheaper than a Hove house per square metre. The two readings are the whole argument. Only 457 districts publish all four medians at once, because each needs thirty matched sales of its own type before the site will print it, so most district pages show two or three types and say which.

The gap is not uniform across the country. In the South East, a flat came in below a terraced house in 98% of the 167 districts publishing both, and in London in 95% of 167. In the North West it was 76% of 63, and in Yorkshire and The Humber 72% of 39. Where terraced housing is old, dense and cheap, a flat has less to be cheaper than.

What does the match rate do to the comparison?

It tilts it, and the direction is known. A rate can only be computed for a sale whose address found an EPC, and flats find one far more reliably than detached houses: 89.3% of flat sales in the whole register carry a floor area against 69.2% of detached sales, because a detached house is more likely to be known by a name than a number and names match badly. So the detached median in any district rests on a smaller share of its own market than the flat median does, and the missing third of detached sales is not a random third. The coverage article measures how far that moves the figures. The short version is that the by-type gap is real and its exact size is less certain than the table’s tidy columns suggest.

How should you use the by-type figure?

1. Price a flat against flats. Every district page splits its median by type where the sample allows. If the flat row is there, that is the row; the all-types median above it is a median of houses with some flats mixed in.
2. Read the count under each figure. A type median on thirty-one sales is a type median on thirty-one sales. The site will not print one on fewer, and the number beside it tells you how much further it could move.
3. Convert before you compare. Listings quote square feet more often than square metres, and the by-type figures here are all per square metre; the unit article gives the factor and the trap in rounding it.
4. Compare the same type across areas, not different types in one. The question “is a terraced metre in Sale dearer than in Oadby?” has a clean answer in the table above. The question “are flats overpriced in Sale?” does not, because there is no version of a flat that is a house.

What will the by-type figure still not tell you?

The type is the one the register recorded at the sale, in four letters, and it carries no bedrooms, no floor, no age and no condition. A first-floor conversion and a new-build penthouse are both an F. A bungalow and a six-bedroom house are both a D. The figure is also silent on the fifth type, “other”, which the site excludes from every median because it mixes garages, plots and mixed-use lots that are not homes. Within those limits, the by-type rate is the closest thing the public record offers to comparing like with like, and it is on every district page that has the sales to support it.

Every figure on this page is read from the current dataset as the page is built, currently August 2026, so a rate here and the district page it links to cannot disagree. Everything is rebuilt monthly from HM Land Registry sold prices matched to EPC floor areas; the glossary defines every term used here.