The INE median and the asking price measure different things

A listing price is a unilateral proposal: a person decided what they want to sell for, and wrote it down. Nobody validated it, nobody paid it, and it can change tomorrow without anything happening to the property. The INE median is a different thing: it is the central value of sales actually registered in a territory, drawn from completed and declared transactions.

Comparing the two is useful, but it is not comparing two measurements of the same object. It is comparing an intention with a track record. The distance between them is not an error in either: it is the margin between what the market asks and what the market closed, and that margin always exists.

So the right question is never “which of the two is correct”. It is “how far does this listing sit from what has already been registered here, and what justifies that distance”. A large gap can be an exceptional property, or a price that has not come down yet.

• Asking price. Chosen by the seller, published in the listing, changeable at any moment.
• Registered sale. The value of a completed transaction, drawn from tax and statistical sources.
• Median. The central value of a set, not the average. Half the sales fell below, half above.

The series this site uses publishes the median sale value per square metre of family dwellings, by municipality, computed over the twelve months ending each quarter. In this period — 1.º Trimestre de 2026 — the national median is €2,168/m², and 304 municipalities have a published value.

The twelve-month window is what confuses people most. The figure you read today is not “this quarter’s price”: it is the central value of everything registered over the previous four quarters. That makes the series stable, but also slow — a market that turned two months ago does not show up here yet.

The series holds 12 quarters. In the first of them, 2.º Trimestre de 2023, the national median was €1,541/m²; today it is €2,168/m², which is +40.7% across the series. That is an observed change between two published periods. It is not an annual rate, not a projection, and says nothing about the next quarter.

Family dwellings covers flats and houses, new and second-hand, of every size and condition, in the same figure.

When a listing sits away from the municipal median, six explanations have nothing to do with the price being right or wrong. They are worth ruling out before concluding anything.

The first is the mix: the median puts flats and houses, studios and five-bedroom homes, new builds and renovation projects into one number. If the municipality is largely older stock and the listing is a new building, the gap is about product, not price.

The second is geography: the median is municipal, and a municipality can hold a waterfront and an interior. The third is time, covered in the next section. The fourth is the denominator — floor area — covered below. The fifth is the nature of the value: asked against registered. The sixth is sample size: where transactions are few, the central value moves on very little.

• Product mix. Different sizes, ages and conditions inside one figure.
• Geographic scale. A whole municipality in one value, including areas that do not compete with each other.
• Time lag. Twelve months of registrations, published some months after the period.
• Area definition. The listing’s area and the register’s area are rarely the same thing.
• Asked against registered. An intention compared with completed transactions.
• Sample size. Municipalities with few sales give unstable medians.

The 1.º Trimestre de 2026 figure aggregates registrations over the twelve months ending in that quarter, and is only published some weeks after the quarter closes. Put together, the series’ centre of gravity sits more than half a year behind the day you read it.

In a stable market that is irrelevant. In a market moving quickly, in either direction, it is the difference between reading the map and reading the ground. The series is not wrong: it is answering a question about the past, which is the only question registered data can answer.

The practical consequence: use the median to frame the order of magnitude and to compare municipalities, and use the listing’s own history to see what is happening now. They are instruments with different clocks, which is why the report shows both.

A value per square metre is a division, and a division has two sides. The numerator — the price — is usually reliable. The denominator rarely is: the area in a listing may be gross private area, with or without dependent area, or usable area, or a rounded number nobody checked.

The area on the caderneta predial is the one the tax register recognises, and it is the closest to the denominator official statistics use. When the two differ — and they often do — the listing’s €/m² and the series’ €/m² are not measuring the same surface.

Before concluding that a listing is expensive per square metre, confirm the area on the caderneta predial. A ten per cent difference in area moves the €/m² by ten per cent, with the price unchanged.

The caderneta predial urbana is free to the owner on the tax portal, and a seller can supply it to a serious buyer.

The median is good for three things and fails at two. It is good for learning the order of magnitude of a municipality you do not know; for comparing municipalities on one methodology; and for spotting a listing that sits outside the price universe of the territory it is in.

It is not good for valuing a specific property, because it does not know that property. And it is not good for forecasting anything: this page does not forecast prices, and an observed change between two published quarters is not a projectable trend.

An honest way to use the number: if a listing sits +47.6% above the median, ask what this property has that the municipal median does not — view, year of construction, condition, floor, lift, parking, energy rating. If the answer is convincing, the gap explains itself. If there is no answer, you have a question for the seller.

• Good for. Framing an unfamiliar area, comparing municipalities, spotting listings outside the local universe.
• Not good for. Valuing a specific property or anticipating future values.
• Pairs with. The listing’s own history, which runs on a much faster clock.

Some municipalities have no published value in this period: of Portugal’s 308 municipalities, the series publishes 304. A missing row means a median was not published, usually because the number of transactions is too small for a statistically robust figure. It does not mean there is no market.

Some municipality names also repeat in Portugal, and a series that identifies municipalities by name alone does not disambiguate them. Where we cannot be sure which municipality is meant, we do not run the municipal comparison — we show the national median and say that is what we are showing.

And some properties do not belong to the distribution: a palace, a farm estate, a derelict building for full reconstruction. Comparing them with a municipality’s family-dwelling median is a valid arithmetic operation and an invalid conclusion.