Why property markets resist comparison

The obvious question about a network of property datasets is which country is more expensive. It is a harder question than it looks, and the honest answer today is that IndexProp does not publish one. Here is precisely what stands in the way, because the obstacles are more useful than a number that ignores them.

This piece was written when the network covered two markets and promised to say what a third made comparable. France has since launched, and the answer is instructive: it removed two of the five obstacles and left the other three standing.

1. Are the three registers even about the same thing?

Singapore’s national figure is public housing: HDB flats, sold on 99-year leases under eligibility rules, in a market where public housing is where most of Singapore’s residents live. England & Wales and France both publish the whole market, houses and flats, every tenure. So two of the three now describe the same KIND of thing, and the Singapore figure still does not: setting it against either compares a public-housing resale market with an entire private one. This is the obstacle France removed, and it removed it only for one of the two pairs.

2. Is a square metre the same square metre?

A per-square-metre price is a ratio, and its denominator is defined differently in every market: an Energy Performance Certificate’s measured area in England & Wales, the housing authority’s own recorded area in Singapore, and the cadastre’s recorded built area in France. Three definitions, three measuring conventions, three sets of things counted and not counted. Adding a third market added a third denominator rather than a tiebreak, and two figures built on two definitions are not made comparable by sharing a unit.

3. What does the exchange rate add?

Converting one market’s prices into the other’s currency introduces an exchange rate with its own date, its own volatility and its own choice of spot-versus- average. A converted gap can move on a day when no home changed hands in either country, which is a poor property of a property statistic.

4. Are the mix adjustments comparable?

Two of the three markets correct for the fact that what sells is not a random sample of what exists, and they do it differently: one weights property types by their share of past sales, the other weights flat types by the national housing stock. France applies no mix adjustment at all, because the estimator it would need has not been shown to hold at commune grain. Both corrections are defensible and so is declining to make one. None of the three is interchangeable with another, and a headline that ranked them would be ranking three estimators as much as three markets.

5. Which questions have no data behind them at all?

Two questions get asked more than any others: where rental yields are highest, and where housing is least affordable against local incomes. Neither can be answered across these three datasets. England & Wales publishes no rent figures at all, deliberately: the available national rent series is a mean, and dividing a mean rent by a median price produces a number with no clean interpretation. Singapore publishes gross yields but no income data at this geography. France publishes a modelled ASKING rent, which is a different object from a contracted one, and refuses to divide it by transaction prices for exactly that reason. Three markets, three different reasons, one unanswerable question.

What would have to change?

A third market has landed, so this piece owes the account it promised.

What France made comparable. England & Wales and France now publish the same kind of statistic over the same kind of population: a median price per unit of floor area, over the whole market, from a national transfer register, on a twelve-month window. Obstacle 1 falls for that pair, and obstacle 5 partly falls with it, because both markets can at least be asked the same question about coverage.

What it did not. The denominators are still defined by three different authorities for three different purposes. The vintages still differ: one register updates monthly and the other twice a year, so on any given day the two figures describe windows that end months apart. And a currency conversion still sits between them, with its own date and its own volatility, so a gap could move on a day when no home changed hands in either country.

The remaining obstacles are not clerical. They would each need a decision that trades accuracy for a headline, and this network has not made one. What would change it is a rent register on the England & Wales side, an income series at a matching geography in Singapore, or an agreed floor-area definition that all three authorities happen to share. Until then, each market answers the questions its own data supports, and the gaps are published rather than papered over. What each one can answer is on its own site: RealScout for England and Wales, IndexProp SG for Singapore, and IndexProp FR for France.