What IndexProp measures, market by market

IndexProp is not one dataset. It is one method applied to each country’s own public transaction registers, and those registers differ enough that the differences are worth stating before any figure is read.

What does Singapore measure?

Four registers, never pooled: HDB resale registrations, HDB renting-out approvals, URA private sale caveats and URA private lease contracts. HDB publishes the price and the floor area on the same record, so there is no address matching step and no match rate to report. The current build carries 987,909 resales on record back to 1990, of which 24,659 fall in the last twelve months, across 55 districts, with 4,237 named roads resolving to one of them by search, plus 131,793 private sale caveats.

The headline is a median price per square foot, which is how the market itself quotes it. Public and private housing are published side by side and never averaged: they are different products with different buyers, different eligibility rules and different tenure. Open Singapore.

What does England and Wales measure?

Two sources joined on address: HM Land Registry Price Paid Data for the price, and Energy Performance Certificates for the floor area. The join is the defining constraint. A sale without a certificate is counted but cannot be priced per square metre, which is why that market publishes a match rate and a matching-bias figure that Singapore has no need for. The current build carries 609,912 sales in the last twelve months, of which 482,172 are matched to a floor area, across 2,279 postcode districts. The match rate is not decoration: it is the market’s statement of how much of the register its rate can see, and it moves with the certificate register as much as with sales. It is published as RealScout on its own domain. Open RealScout UK.

What does France measure?

One register, and a legal condition attached to it. Demandes de valeurs foncières is the tax administration’s record of transfers, published as open data with the surface and the price on the same row, so France has no address-matching step either. The current build carries 808,552 sales in the last twelve months, across 4,251 published communes in 97 départements.

Two constraints shape what may be published. The register excludes Alsace-Moselle and Mayotte, which are governed by a different land-registration regime, so those départements cannot be covered from this source at all. And the reuse conditions forbid letting the underlying data be indexed by search engines, which is why that market publishes derived statistics on its indexable pages and serves its data endpoints noindex. Ouvrir la France.

How do the three registers differ in shape?

Strip away the vocabulary and three structural properties decide how each market’s figures behave: what one record is, whether price and size arrive on the same record, and how the market itself quotes a rate. The table is the network’s method in miniature, and every difference in it is a difference a reader will meet on the pages.

What follows from the three being different?

The same word means different things in the three markets. “Floor area” is a certificate’s measurement in one, a register’s own field in the second and a statutorily defined measurement in the third. “Transaction” is a registered transfer in one, a resale registration or a lodged caveat in the second, and a recorded mutation in the third. Those are not rounding differences, and they are the reason IndexProp publishes each market on its own terms rather than flattening three registers into a league table.

The coverage differs as sharply as the vocabulary. One market can price four fifths of its sales per unit of area, one can price effectively all of them, and one cannot cover three départements at all. A cross-country table would have to hide every one of those facts to produce a single number. Why that is harder than it sounds is the companion to this piece, and what it rules out is set out on the methodology page.

How should you read a figure from any of the three?

The same way in all three markets, in the same order. First find the register behind the figure, because the register decides what “a sale” even is. Then find the date the window closes, which every page on this network states, because three release cycles mean three different definitions of “now”. Then check the denominator, because a rate is only as comparable as its definition of a square metre, and no two countries measure one the same way. Last, look at the count published beside the figure, because a median over a handful of sales is an anecdote wearing a statistic’s clothes. Every one of those four checks is answerable on the page in front of you in each market, which is the property this network is built to protect.