Answer in brief
France’s 29 September construction release shows starts improving as permits soften. Both annual totals remain below their five-year averages, and August estimates carry an explicit revision warning.
The pipeline moves in two directions
French housing construction produced a mixed August reading in the SDES release published on 29 September 2026. Housing authorisations fell 1.1% from July to 29,749, while starts were estimated to have risen 4.1% to 27,663. The monthly series are adjusted for seasonal patterns and working days. They measure different stages of supply, so their opposite movements do not make the report internally contradictory.
A permit creates the possibility of building; a start moves an earlier project into physical activity. The projects behind August starts need not be the ones receiving permission during August. Treating the two counts as a monthly conversion rate would therefore misrepresent the process. The useful signal is that current building activity can improve while the earlier supply pipeline becomes less vigorous.
Annual totals need the right baseline
Across September 2025 to August 2026, SDES reports 369,723 authorised dwellings and 300,244 starts. These totals were respectively 9.7% and 9.5% below the average of the preceding five years. The comparison is with a multi-year benchmark, rather than simply with the immediately preceding twelve months. That distinction changes what readers can legitimately infer from the negative percentages in the report.
The same release says average monthly starts during the first eight months of 2026 were 21.2% above the average for the whole of 2025. That is compatible with the shortfall against the five-year baseline. A recovery from a weak recent level does not require a return to the longer-term norm. Keeping both comparisons makes the uneven improvement visible without labelling it a full recovery.
| Measure | Latest value | Comparison |
|---|---|---|
| August authorisations | 29,749 dwellings | −1.1% versus July |
| August estimated starts | 27,663 dwellings | +4.1% versus July |
| Twelve-month authorisations | 369,723 dwellings | −9.7% versus five-year average |
| Twelve-month starts | 300,244 dwellings | −9.5% versus five-year average |
August contains a larger estimated component
SDES explicitly warns that August starts are more fragile than other months because fewer records are collected, increasing the share supplied by estimation. It says subsequent months may bring larger revisions. The 4.1% monthly increase is consequently an initial statistical estimate with a specific reporting limitation. It should not be presented as an exact, final count of observed work at every construction site.
Insee’s methodological description explains the underlying delay: information arrives through administrative forms, while declarations of starts and completions depend on applicants reporting them. That description is background about the reporting system, not independent confirmation of August activity. It helps explain why a physical event and its appearance in an official database can be separated in time and why revisions are a normal part of interpretation.
The database has also changed
The release notes that Sitadel3 replaced Sitadel2 for new-housing statistics from March 2026, with slight revisions to published data. This is a separate issue from August’s thin reporting. A change in the collection system and late declarations can both affect the series, but they should not be collapsed into one explanation for every movement in the monthly figures or their subsequent revisions.
For comparisons, readers should retain the definitions and latest historical series supplied with the new release. The national monthly data exclude Mayotte, despite the broad France-wide wording of the publication title. The available detail also distinguishes detached, grouped individual, collective and residence accommodation. National totals describe a broad construction system rather than a single type of home available to a specific household.
What the split means for future availability
Our interpretation is that the improvement in starts gives contractors evidence of projects moving forward, while weaker permissions temper confidence in a sustained replenishment of the pipeline. That is a reading of the two stages, not a forecast of prices or future completions. A development can still encounter financing, scheduling or delivery problems after its start has entered the statistical record.
The distinction is especially useful for someone comparing an existing home with a purchase in a development under construction. A national starts figure cannot establish that a particular apartment will be ready on its advertised date. Project documentation, the actual construction stage and the agreed delivery arrangements address that question. The statistical release supplies market context but does not certify any developer’s timetable or the quality of the eventual dwelling.
Watch revisions alongside the next month
A sound follow-up will examine both the next month and the revised August estimate. If later declarations change August substantially, part of the apparent new movement may belong to earlier activity. Looking only at successive headlines would miss that adjustment. Following authorisations and starts separately will also show whether the present improvement in building activity is accompanied by a stronger supply of newly authorised projects.
For now, the evidence supports a qualified improvement in starts alongside softer permits and annual totals below their five-year benchmarks. The important uncertainty is stated by the producer itself. Preserving it makes the report more informative: readers can distinguish a possible strengthening in construction from a final count, and a recovery against a weak year from a return to the longer-run level.
Questions and answers
Are the 27,663 starts completed homes?
No. Starts describe entry into construction, not handover or occupation. A development still needs to pass through the remaining building process before it supplies usable homes.
Does the 9.7% permit shortfall compare with last year?
No. The comparison is between the latest twelve-month total and the average of the preceding five years. It is not a conventional year-on-year percentage.
Why might August be revised more heavily?
SDES says fewer records are collected in August, leaving a larger estimated component in the starts series. Later declarations can therefore change the initial result more substantially.