The DVF database (Demandes de Valeurs Foncières) remains the only official source that records transactions for consideration registered by land publicity. To know real estate prices by neighborhood with a usable level of reliability, one must know how to query this database correctly and then correct its structural biases.
Biases of the DVF: what average prices by neighborhood do not reveal
An average price per square meter calculated over an entire neighborhood aggregates properties with heterogeneous characteristics. A studio sold on the ground floor facing a courtyard and a T4 apartment crossing on the fifth floor with a continuous balcony fall within the same geographical perimeter, but not the same market.
Read also : The latest trends and must-know news in the real estate market to discover
The DVF does not provide information on the condition of the property, the floor, or the orientation. It provides a transfer price, a Carrez surface area, a type of property, and an address. Any average drawn from this data without reprocessing produces a raw indicator, not a reliable estimate. To better understand how to determine real estate prices by neighborhood, it is useful to know the limitations of this database before using it.
We observe three recurring biases in the use of the DVF at the neighborhood level:
You may also like : How to Optimize Your Savings While Shopping in Pas de la Case
- The composition bias: a neighborhood where transactions mainly involve small properties displays a mechanically higher price per square meter, without the larger properties being actually more expensive.
- The temporal bias: the DVF covers transactions since 2014, but recent transactions (less than six months) have not yet been published. The delay between signing and publication can exceed three months.
- The volume bias: in low-density neighborhoods, a few atypical sales (family sale, discount for heavy work) are enough to distort the median over an entire year.
To reduce these biases, we recommend working on comparable transaction lots, filtered by surface area, type, and period, rather than on a raw average by postal code or cadastral section.
DVF on impots.gouv.fr: an underutilized research tool
Most content directs users to the Etalab application or to third-party platforms like Immo-DVF or Immo-Data. The tool “Search for real estate transactions” accessible from the personal space on impots.gouv.fr (formerly Patrim) offers a different level of detail.
This access, free and official, was designed for tax purposes: declaration of IFI, inheritance, valuation control. It allows users to consult comparable sales around a specific address, with filters for surface area, number of rooms, and date of transfer.
The main interest lies in the geographical proximity of the results. While public platforms aggregate by municipality or by IRIS, the impots.gouv.fr tool centers the search on a radius around the entered address. For an apartment on a specific street, this granularity changes the reading of the local market.
Access requires FranceConnect authentication or a personal tax account. The trade-off: the number of consultations is limited, and the use is strictly personal.
DVF analysis by micro-neighborhood: building a usable price series
The DVF has been continuously recording transactions for consideration since 2014. This depth covers more than one recent market cycle (increase, plateau, corrections post-2022), allowing for a deeper understanding than just the observation of “price per square meter over the last twelve months.”
To produce a reliable analysis at the micro-neighborhood or street level, we structure the work in three steps.
Filter by strict typology
Isolate T2-T3 apartments on one side, houses on the other. Mixing typologies in the same comparison lot renders the median unusable. Each surface segment must contain at least ten transactions for the statistics to make sense.
Trace annual evolution
Comparing the median price per square meter year by year, since 2014, reveals the dynamics of the neighborhood. A neighborhood where the median progresses regularly does not have the same profile as an area where prices stagnate despite an overall municipal increase. This divergence often signals a deficit of attractiveness that the city average masks.
Cross-reference with local data
The DVF alone is not sufficient. Local housing observatories and ADIL publish analyses by sector that incorporate variables absent from the raw database: vacancy rates, share of social housing, ongoing development projects. In Paris, notarial data (BIEN database) complements the DVF with detailed information on the floor and condition of the property.
Real estate estimation: why online tools diverge so much
MeilleursAgents, Orpi, SeLoger, and other estimation platforms use the DVF as a base but add layers of proprietary modeling. The discrepancies between two estimates for the same property often exceed several hundred euros per square meter.
These discrepancies arise from different methodological choices:
- Temporal weighting: some models overweight recent transactions, while others smooth over three years.
- The reference geographical perimeter: an algorithm that reasons at the IRIS level does not produce the same result as a model based on a metric radius around the address.
- Qualitative corrections: floor, exposure, proximity to transport. These variables are injected by statistical models, not by DVF data.
No online tool replaces an analysis of comparable transactions filtered manually. Automated estimates provide a ballpark figure. For a precise neighborhood price, the combination of raw DVF, the impots.gouv.fr tool, and local knowledge remains the most reliable method.
The French real estate market today benefits from unprecedented data transparency thanks to open data DVF. Exploiting this transparency requires going beyond the averages displayed by platforms and working on filtered, comparable, and contextually informed transaction lots.