Reading a housing market without guessing

What does a housing market analysis actually measure?

Most published housing numbers fall into three families, and mixing them up is the most common error.

The first family describes transactions: how many sales closed, at what price, in which territory. In France the reference for this is the DVF database, built from recorded notarial deeds. It reports actual sale prices rather than asking prices, which makes it more reliable than listing data, but it also lags: a sale appears only once the deed is registered, so the most recent months are always incomplete.

The second family describes prices through indices, such as the Notaires-Insee index. An index tracks the change in value of a comparable set of dwellings over time. It answers a different question than a median price does. A median tells you what a typical buyer paid in a given area; an index tells you whether the same kind of home became more or less expensive. A market can show a rising median while the index is flat, simply because larger or better-located homes changed hands that quarter.

The third family describes financing conditions: average mortgage rates, annual percentage rates and borrowing capacity. These come from banking supervisors and central bank surveys rather than from property records, and they move faster than prices do. A rate shift of a few tenths of a point changes monthly payments enough to alter what a household can bid, which is why financing data often leads price data rather than following it.

Readers who want to verify definitions, vintages and sources before interpreting a figure can consult the French housing data reference, which is built specifically around that verification step.

Why do two sources give different prices for the same area?

Disagreement between two price figures for the same town is normal, and it usually comes down to four causes.

The first is the field of observation. A figure covering only apartments will not match one covering houses and apartments together. A figure covering a whole department will not match one covering a single commune inside it. Comparisons are only meaningful at constant scope, meaning the same property types and the same geographic boundaries on both sides.

The second is the moment of recording. A price agreed in March and recorded in June belongs to different periods depending on whether the source dates it by agreement or by registration. Two sources can both be correct and still differ by a full quarter.

The third is the treatment of outliers. Very high or very low prices, unusual properties, bulk sales between institutions, and sales including furniture or commercial premises all distort averages. Some sources exclude them, some do not, and the exclusion rules are rarely stated on the front page of a report.

The fourth is the difference between asking prices and transaction prices. Listing platforms publish what sellers hope to receive. Notarial records publish what buyers actually paid. The gap between the two is not a measurement error; it is a measurement of something else entirely.

A useful habit is to read the methodology note before the headline. If a source does not state its scope, its dating convention and its exclusions, treat the number as indicative rather than comparable.

How do mortgage rates change what buyers can afford?

Borrowing capacity is arithmetic, and it responds to rates more sharply than most buyers expect.

Capacity depends on three inputs: the rate, the term, and the share of income a lender will accept as a monthly payment. When the rate rises, the affordable loan amount falls even if income is unchanged. A household that could borrow a given sum at one rate may qualify for roughly ten to fifteen percent less at a rate one point higher, depending on the term. That reduction feeds directly into what the household can offer, which in turn feeds into observed prices with a delay.

The annual percentage rate matters more than the nominal rate for comparing offers. It folds in insurance, arrangement fees and other costs, so two loans advertised at the same nominal rate can differ meaningfully once everything is included. The total cost over the life of the loan is the figure that actually compares two offers.

Rental figures follow a separate logic. Rents excluding charges are tracked by local observatories, and in some areas an administrative cap applies. A rent figure quoted with charges included is not comparable to one quoted without them, and the two are frequently printed side by side in the same table.

What do vacancy, energy ratings and risk maps add?

Price and transaction data describe what changed hands. Three other datasets describe the condition of the stock itself.

Construction data, collected through the Sitadel system, reports building permits and housing starts. It is a leading indicator of future supply: units authorized today reach the market years later, so a drop in permits signals tighter supply well before prices reflect it.

Vacancy data counts dwellings with no usual occupant. A high vacancy rate in a growing market usually points to a mismatch between the stock and demand, for instance obsolete or poorly located units, rather than to a simple surplus of housing.

Energy performance ratings classify dwellings by consumption. Because ratings affect both running costs and, increasingly, financing and rental conditions, they split a single market into segments that behave differently. Two identical apartments in the same building can trade at different prices if their ratings differ.

Risk mapping adds a further layer. Flood zones, subsidence exposure and similar designations are recorded per address, and they affect insurability and therefore value. These are address-level facts, not market-level averages, and they cannot be read from a commune-wide price table.

How should a reader compare two territories?

Comparison is where most analysis breaks down, because the two sides are rarely measured the same way.

Start by fixing the scope. Compare apartments with apartments, houses with houses, and use boundaries that nest properly, such as communes within the same department, rather than arbitrary radiuses around a point.

Then fix the period. Quarterly data compared against annual data will produce a difference that reflects the calendar, not the market. Where a source publishes rolling twelve-month figures, use those on both sides.

Then check the volume. A median computed from forty sales is far less stable than one computed from four thousand. Small territories produce large swings that reverse the following period, and those swings are often reported as trends.

Finally, separate the level from the change. A territory can have high prices and flat growth, or low prices and rapid growth. Both facts matter, and they answer different questions: affordability on one side, momentum on the other.

What a market analysis cannot tell you

Published data describes the past, and it describes it at a level of aggregation that rarely matches a single property.

A commune-level median says nothing about a specific street, floor or orientation. A transaction count says nothing about the condition of the units sold. An index says nothing about whether a particular buyer will find a particular home at a particular moment.

The honest use of housing data is to establish context: what the range of prices is, how financing conditions have moved, how much stock is being built, and which risks attach to an address. The decision itself rests on facts the data does not contain, including the state of the building, the terms of the loan and the time horizon of the buyer.

Reading the definitions first, and the headline second, is what keeps a market analysis from turning into a guess.