Asking price, valuation, or achieved price?

Almost every property price you will read is one of three different objects. It is what a seller advertised, or what a model estimated, or what a buyer actually paid and a public authority recorded. They are produced by different people for different purposes, they disagree with each other by a lot, and a great many published figures do not say which one they are. Working that out is the first thing to do with any property statistic.

What is an asking price measuring?

What the property was advertised at. It exists before any negotiation, it is set by the seller on advice, and it is available the moment a listing goes up, which is why asking-price measures are always the most current.

Its weakness is that it is a hope rather than an outcome. In a falling market asking prices lead achieved prices down and the gap widens; in a rising one the reverse. A listing that is reduced three times and finally sells contributes its original number to the asking-price series and its final number to nothing at all, unless a register picks it up.

What is a valuation measuring?

A model’s estimate of what a property would fetch. Lender valuations, agents’ online estimates and automated valuation models all sit here. They are useful because they exist for properties that are not for sale, which is most of them, and because they can be produced instantly.

Their weakness is that they are outputs rather than observations. A valuation is only as good as the comparable evidence and the model behind it, and neither is usually published. When a valuation and an achieved price disagree, nothing in the valuation tells you which of its assumptions failed.

What is an achieved price measuring?

What was actually paid, recorded by a public authority as part of transferring ownership or collecting tax. This is the number that is hardest to argue with, because the transaction happened and somebody official wrote it down.

It has two costs and they are structural. It is late: a sale enters the record only once it has been registered, so the most recent months are always incomplete. And it is sparse: it exists only for homes that changed hands, so it says nothing directly about the ninety-odd per cent of the stock that did not.

Which of the three does this network publish?

Every figure on this network is the third kind. In the last twelve months that is 609,912 registered sales in England & Wales, 24,659 HDB resale registrations in Singapore and 808,552 recorded transfers in France. No listings, no model output, and no valuations of any kind.

That is a deliberate limit rather than a claim of superiority. An achieved-price publisher cannot tell you what your house is worth, and should not pretend to. What it can do is tell you what comparable homes actually sold for, with the sample size attached, which is the evidence a valuation ought to rest on anyway.

How do you tell which number you are reading?

Four questions settle it, and any publisher worth quoting answers all four somewhere.

• At what stage of the sale? Listing, offer, mortgage approval, exchange, completion or registration. Each is weeks to months from the next.
• Whose properties? One portal’s listings, one lender’s borrowers, or every transfer in the country. A lender measure excludes cash buyers entirely, and cash buyers are neither rare nor typical.
• Observed or modelled? If the figure exists for a property that did not sell, it is modelled. There is no third possibility.
• How many observations? A publisher that will not tell you the sample size behind a local figure is telling you something.

Why are cross-country tables the worst case?

International comparisons are where these three get mixed most freely, because assembling one number for eighty countries means taking whatever each country makes easy. The result is often a table in which some rows are registry data, some are listings and some are survey estimates, presented in one column with one unit.

This network publishes three countries rather than eighty for that reason, and does not combine them even so: five specific obstacles stand between three honest national datasets and one honest cross-country number, and three of them are still standing. What each register actually records is the companion to this piece. All three of this network’s markets publish the third number and only the third: RealScout, IndexProp SG and IndexProp FR.