Reading a Housing Market Before You Trust a Number

A housing market analysis is only as good as the data behind it. Before comparing two neighborhoods, two years or two countries, a reader has to identify who produced the figure, which year it covers, what geographic area it includes and which definition was used. Skipping those four checks turns a useful statistic into a misleading one, and the same discipline applies whether the market is in New York, Paris or anywhere else.

Start With the Source, Not the Headline

Every published price, rent or sales count comes from an institution with a method. In France, the main reference for recorded sales is the DVF database, built from notarized transactions and published by the state. Price indices come from the Notaires-Insee partnership, average mortgage rates from the Banque de France, construction starts from Sitadel, energy ratings from the DPE, and address-level hazard data from Georisques. Each of these has a defined scope and a defined lag. A reader who wants to understand how those French sources fit together, and where each one stops being reliable, can look at the work published by Observatoire immobilier, which is a French-language educational outlet on reading public housing data. The point is not the country. The point is that a number without a named producer is not evidence.

In New York City, the equivalent habit means checking whether a figure comes from the Department of Finance sales file, from a broker's own closed deals, or from a listing platform's asking prices. Those three are not the same measurement. Asking prices describe what sellers hope to receive. Recorded sales describe what buyers actually paid. A market report that mixes them without saying so is not comparable to one that does not.

What Does the Year on a Figure Actually Mean?

A year label is not a single date. A sales figure for 2024 may cover deals signed in 2024, deals closed in 2024, or deals recorded by the city in 2024. Those three sets differ, sometimes by several months, because a contract signed in November often closes in January or later. A price index for a quarter may be provisional, revised once, then revised again.

Rent data has the same problem. A median rent for a neighborhood can be computed from new leases only, from all occupied units, or from listings that were advertised online. New leases move faster than the whole stock, so a median built from new leases will swing more than one built from all units. Neither is wrong. They answer different questions.

A practical rule: when two sources disagree, compare their definitions before comparing their numbers. If one counts closed sales and the other counts signed contracts, the gap is a definition gap, not a market signal.

Compare Territories at Constant Scope

Administrative boundaries change. Neighborhood definitions change. A district that gained or lost blocks between two years cannot be compared to itself without adjustment. This is why serious analysts rebuild past data on the current boundary before drawing a trend line.

The same applies to price per square meter. In France, the surface used in a transaction is the habitable area under the Carrez definition for co-owned units, which excludes certain low-ceiling spaces, cellars and balconies. In New York, listing platforms often quote a different interior measurement, and some quote gross area. A price per square foot computed on one definition cannot be placed next to a price per square meter computed on another without conversion and a note about the difference.

Medians and averages also behave differently. A median splits the market in half and resists distortion by a handful of very large sales. An average does not. In a neighborhood with a few penthouse closings, the average can sit well above the median, and a report that quotes only the average will describe a market that most buyers never see.

Which Numbers Should a Buyer or Tenant Check First?

For a buyer, the first useful figure is the count of recorded sales in the specific building type and area over the last four quarters. Volume tells you whether a price is supported by many transactions or by two. The second is the median price per unit of surface, not the average. The third is the time on market, which is usually measured from listing to contract, not listing to closing.

For a tenant, the first figure is the median rent excluding charges, because charges vary by building and by lease. The second is the scope of the sample: new leases only, or all leases. The third is whether the area has rent regulation, since a controlled unit and a free-market unit in the same block can carry very different rents.

For both, the financing side matters. An average mortgage rate published by a central bank describes the market as a whole. The rate a specific borrower receives depends on the loan term, the down payment, the borrower's profile and the fees included in the annual percentage rate. A rate quoted without saying whether fees are included is not comparable to one that includes them.

Where Definitions and Vintages Break Down

Some figures are published with a long lag. Construction permits and housing starts can take a year or more to appear in a national series, and they are revised. Vacancy counts depend on a census or a tax record and may be several years old. Energy performance ratings depend on when the assessment was made and which version of the method was in force.

Hazard data is a separate case. A flood or subsidence risk attached to an address is a regulatory statement about that parcel, not a market opinion. It can affect insurance cost and resale, and it should be read at the address level rather than at the city level.

When a figure is old, the honest move is to label it as old. A 2021 vacancy rate can still be useful for context, but it cannot describe the market in 2025.

A Short Checklist Before Quoting Any Number

Name the producer. Name the year or quarter. Name the geographic scope. Name the definition, especially for surface, rent and charges. Say whether the figure is provisional or final. Say whether it is a median or an average. Say what it excludes.

A reader who runs those seven checks will find that most disagreements about a housing market are disagreements about method, not about reality. The market did one thing. The numbers describe it from different angles, and the angles have to be stated before the description can be trusted.

That habit travels. A New Yorker reading a French price index and a Parisian reading a Brooklyn sales file are doing the same work: finding the producer, the year, the scope and the definition, then deciding whether the comparison in front of them is fair.

A housing market page tracks prices, permits and turnover, but a building also runs on equipment that must be documented. Owners and supers who keep a binder for boilers, elevators and sprinklers should treat emergency power the same way. Engine brand files, Deep Sea and ComAp controllers, ATS cabinets, load testing and safe operation all belong in one place, ready for an inspector or a contractor. For small sites, the same habit applies: write down what was tested, when, and by whom. The page on standby generator records explains how that operating file is organized.