Why nobody can tell you last month’s house prices

No publisher of achieved prices anywhere can tell you what happened last month, and the reason is not that they are slow. A sale enters a public register only after it has completed and been registered, and that sequence takes weeks to months. Every achieved-price figure you read describes a period that has already closed, and the most recent part of it is always incomplete.

What is the registration lag made of?

Four separate delays stack, and only the last one belongs to the publisher.

• Offer to completion. Searches, surveys, mortgage offers and chains. Months, and highly variable.
• Completion to registration. The conveyancer lodges the transfer and the registry processes it. Weeks to months, and longer for new builds, where the plot may not exist as a title yet.
• Registration to publication. The register’s own release cycle. Monthly in some countries, twice a year in others.
• Publication to the page. Whatever the site you are reading takes to rebuild. This is the only one anybody can shorten, and it is the smallest.

How long does one sale take to appear?

Follow a single ordinary sale through that chain. The offer is accepted and the conveyancing begins; months pass before completion, and the price is fixed early in that stretch, so the register’s date already trails the moment the market actually spoke. After completion the conveyancer lodges the transfer, the registry works through its queue, and the sale becomes a public row weeks to months later still. Then it waits for the next release, and then for the next rebuild of whatever site you are reading. By the time the row is readable, the price it carries describes an agreement struck perhaps half a year earlier. Nothing in that chain is anyone being slow. It is what a register of completed, verified transfers costs, and it is the same chain in every country, differing only in the length of each link.

How far behind is each market today?

This network runs three registers and their vintages are months apart from each other at any given moment, because the release cycles differ. Today England & Wales is anchored at August 2026, Singapore at September 2026 and France at December 2025.

That is not three degrees of diligence. It is one register that publishes monthly, one that publishes monthly with a shorter conveyancing tail, and one that publishes twice a year. A French figure cannot be as current as a Singaporean one no matter who is publishing it, and a site that showed all three as though they described the same week would be inventing a currency it does not have.

Why does the newest period always look like a fall?

This is the practical consequence and it catches professionals as well as readers. The most recent month or quarter in any achieved-price series is missing the sales that have not been registered yet, and those arrive over the following weeks. So the latest point starts low, on a thin sample, and revises upward as it fills.

Read naively, that produces a market downturn every single month, in every country, for ever. Publishers handle it in one of three ways: mark the period as partial, exclude it from any growth calculation, or wait to publish it at all. This network marks it and excludes it, which is why a chart here can end with a lighter final bar and why a growth figure never uses it.

How does each register handle revisions?

Differently, and the difference decides how a published figure can change under you. The England & Wales register publishes additions, corrections and deletions every month, each row carrying a status, so a past month keeps filling in and an old row can be amended years after the sale. Singapore’s resale dataset is updated in place within days of registration, so its recent months fill quickly and settle quickly. France republishes the whole file at each half-yearly release, each one superseding the last, so a French figure is perfectly stable between releases and can move all at once when a new file lands. Three registers, three revision shapes, and none of them is a fault: they are three answers to the same question of when a register lets itself be corrected.

How do you read a period that has not finished arriving?

• Find the anchor. Every honest publisher states the last date its data covers. If you cannot find one, you do not know what period you are looking at.
• Distrust the final point. Not because it is wrong, but because it is incomplete, and incompleteness has a direction.
• Do not compare across vintages. Two markets anchored months apart are describing different weather, not different climates.
• Prefer a rate to a change, over short horizons. A level built on a twelve-month window is robust; a month-on-month change built on the newest, thinnest data is mostly registration timing.

The lag is also why asking-price measures exist and why they are worth reading despite being hopes rather than outcomes: they are the only ones that can be current, and they buy that currency by measuring something else entirely. Every market page on RealScout, IndexProp SG and IndexProp FR prints the date its own window closes, which is the only honest way to answer “is this current?”.