OneDpe study
French energy certificates: at the “thermal sieve” line, 7 in 10 fall on the favourable side
In England, the same pattern appears only where the rule applies, and the national certificate register makes it measurable. In France, according to the Court of Audit, fewer than one certificate in a thousand was checked in 2023.
At the line that puts a home into band F, the French “thermal sieve” category, 7 energy performance certificates (DPE) in 10 fall on the favourable side, against 1 in 2 elsewhere on the scale. Across 9.8 million certificates issued between 2022 and 2025, OneDpe estimates the excess at 33,449 certificates, or +35%. England shows the same pattern, only in the stock where the rule applies, and its national certificate database makes it measurable.
Study published on · Figures frozen on · Source: OneDpe.fr
- Press release (PDF, French)
- Press release (PDF)
- The 5 graphics, French (ZIP)
- The 5 graphics (ZIP)
- Method note (PDF, French)
- Method note (PDF)
Press contact: Steven Annonziata · presse@onedpe.fr
Key figures
- Counted, no model7 in 10
- certificates one unit from the line fall on the favourable side, against 1 in 2 elsewhere on the scale
- 43,882 homes show 69 kg CO₂/m²/yr (the last value of band E), 17,299 show 70 (the first value of band F): 72%, against 51% on average at other values. None of the 131 control values exceeds 56%. Homes over 40 m².
- Estimated, interval published+35%
- more certificates than expected just short of band F: 33,449 out of 9.8 million
- Band F emissions line (70 kg CO₂/m²/yr), certificates based on a site visit, mainland France, 2022-2025. 95% confidence interval: +22% to +43%. Positive in all 948 calculation variants tested.
- Estimated, England and Wales+39.8%
- more certificates just above the minimum letting standard in English and Welsh private rentals, against −1.1% in social housing, which the rule does not cover
- 39 SAP-point threshold, England and Wales register 2012-2026: +39.8% on 4,395,103 private rented certificates, −1.1% on 4,229,524 social housing certificates. The two confidence intervals do not overlap.
- French Court of Audit, 2023Under 1 in 1,000 in 2023
- French certificates checked in 2023, according to the Court of Audit; no official figure is published for 2024 and 2025
- 0.08% according to the French Court of Audit (report of 3 June 2025): 3,819 surveillance operations, 937 of them on site. The two countries do not check the same object (the assessor in France, the certificate across the Channel); reaching the volume of the English floor of 2% of certificates would take 80,448 operations a year.
« The problem is not the assessor, it is the system. We have made the energy certificate the key to the rental market without giving ourselves the means to check that it is right. The British have given themselves the means to measure their own reliability. In France, none of these indicators exists. »
What the study shows
Just short of band F, 7 certificates in 10 on the favourable side
At the emissions line that puts a home into band F, 70 kg of CO₂ per m² per year, certificates are not spread at random. Among homes over 40 m², 43,882 show 69 kg, the last value of band E, and only 17,299 show 70: 72% fall on the favourable side, against 51% on average everywhere else on the scale. Compared with what the curve predicts, there are 33,449 more certificates than expected near the line: +35%. This excess is also visible, more weakly, at the band E line (+21%). The French Council of Economic Analysis (CAE) measured this kind of threshold effect in 2024 (Focus no. 105), on the energy label and up to the end of 2023, with a rise at the F/G line in 2023; the climate (emissions) label, which this study measures, was not covered there.
When the line moves, the bunching follows
On 1 July 2024, the band lines changed for homes of 40 m² or less, which now have a line specific to their floor area. They did not change for other homes. The bunching followed: among small homes, the share on the favourable side of their new line rose from 50% to 74%, while at the old 70 kg line it fell to 47%, from 71% before. Among larger homes, whose line did not move, nothing changed: 73%, then 70%. Nor is it a liking for round numbers: for 98.9% of these small homes, the new line is not a multiple of ten.
In England, the anomaly exists only where the rule applies
In the private rented sector of England and Wales, a home rated below 39 SAP points, in band F or G, has been barred from new tenancies since 2018, and from all tenancies since 2020. Social housing is not subject to this rule. In the same register, at the same threshold, the excess reaches +39.8% in private rentals and is nil in social housing (−1.1%). This comparison is only possible because the British register records whether a home is rented or owner-occupied. The French database does not, since July 2021 no longer distinguishes sales from lettings, and no longer records who issued the certificate.
Fewer than one certificate in a thousand checked in 2023
In 2023, according to the French Court of Audit, 0.08% of certificates were checked: 3,819 surveillance operations, 937 of them on site; no official figure is published for 2024 and 2025. In England and Wales, at least 2% of certificates are audited each year. The two systems do not check the same object, the assessor in France, the certificate across the Channel; reaching that volume would nonetheless take 21 times more inspection operations, about 80,000 a year. The contrast also exists within France: works funded by energy savings certificates must be checked on site in 10% to 15% of cases depending on the year, a hundred to a hundred and ninety times more than the certificate that triggers them (a works operation on one side, an assessor on the other: a comparison of instrumentation, not of probity). Across the Channel, an audit fails beyond a set margin, 5 SAP points, and the certificate is replaced. In France, no numerical tolerance is published, and no rule requires a faulty certificate to be withdrawn. France does have two rules England lacks: three of the six checks on an assessor take place on site, and since October 2025 ADEME automatically flags anomalies to the certification bodies.
Four measures to find out what a French energy certificate is worth
The EU Energy Performance of Buildings Directive (EU 2024/1275, art. 27 and Annex VI) requires an independent control system ensuring that at least 90% of the energy performance certificates issued are valid, at 95% statistical confidence, measured by random sampling, together with publication of the number of certificates assessed. It was due to be transposed by 29 May 2026; to our knowledge these provisions have not been, and none of these indicators is published in France. OneDpe is not calling for more sanctions: what is missing first is the means to measure. Four measures would cost little; the directive requires the first, and the other three already exist across the Channel:
- Publish every year the number of certificates checked, the inspection rate and the failure rate.
- Set and publish a numerical tolerance beyond which a check fails, like the British 5 SAP points.
- Record in the ADEME database whether a home is rented or owner-occupied, as the English register does.
- If a cost-based exemption is introduced (the French Senate adopted one in July 2026), keep it in a public register, like the England and Wales exemptions register.
What the study does not say
- +35% does not mean that 35% of certificates are wrong: it is 33,449 more certificates than expected just short of band F, out of 9.8 million.
- The study does not say that these certificates are disguised energy sieves, nor which class these homes should have. And “7 in 10” concerns certificates one unit from the line, not all certificates.
- It points the finger at no professional: it describes the shape of a file of several million rows, never an individual case. No statistic by assessor, firm or software is produced, on either side of the Channel.
- It cannot decide between two explanations: input data adjusted near the line, or unfavourable certificates disappearing more often from the published database. Deactivated certificates are not in it, and the reason for a deactivation is recorded nowhere.
- It does not say that English certificates are more reliable: neither country publishes the failure rate of its checks.
- It says nothing about any particular home. The published figures are floors: several methodological choices all push towards underestimation.
Copy to publish, method to check
The full analysis does not cut well from the bottom. Rather than let it be trimmed badly, the copy is here — free to reuse, including commercially, with the credit “Source: OneDpe.fr”. For anyone who would rather check the count than reuse it, the method note gives the perimeter, the keys, the conventions and the limits in full.
- Press release (French)
The press release of 3 November 2026, one page, in French: the four findings, the four proposals, the quotes and the caveats.
- Press release (English)
The same press release in English.
- Full method note (French)
Everything needed to redo the calculation without us, in French: corpora, filters, boundary conventions, the two measures and their checks, the France / England-and-Wales non-comparability table, the exact statements with their counts and intervals, and the log of withdrawn statements.
- Full method note (English)
The same document in English.
Infographics
Free to reuse, including commercially, with the credit “Source: OneDpe.fr”. Click a thumbnail to enlarge and download — PNG 3200 px, editable SVG, greyscale proof.
Just short of band F, certificates bunch up
When the line moves, the bunching follows
England: the anomaly only where the rule applies
Fewer than one certificate in a thousand checked
What each country is able to measure
The other figures in the file
- Legal timetable4 times in 11 years
- France raises the minimum standard for a rented home; England has not raised it since 2018Mainland France: under 450 kWh of final energy per m² per year from 1 January 2023, band F from 1 January 2025, E in 2028, D in 2034 (Law 89-462, art. 6; Decrees 2021-19 and 2023-796). England and Wales: band E for new tenancies since 1 April 2018, for all tenancies since 1 April 2020; the increase announced on 21 January 2026 is not yet law.
- Law in force£3,500
- the spending cap available to a landlord in England and Wales; in France, no cap exempts from the letting banCap per home including VAT, in force since 1 April 2019; beyond it, the landlord registers a five-year exemption on the public exemptions register. The increase to £10,000 is announced, not in force. Law as at 19 August 2026.
- Counted27%
- of certificates are issued from the whole-building certificate, not from a visit to the home: they serve as a control group4,436,634 certificates out of 16,273,573 (27.26%). For them, the value is computed from the building: crossing a line is not decided at the level of the home. At the band E line they show no excess (−0.5%), against +21.4% for certificates based on a visit.
The data
What each country records, and therefore what it can measure
| Data | England and Wales | France, since July 2021 |
|---|---|---|
| Is the home rented or owner-occupied? | Recorded | Absent |
| Sale or letting? | Recorded | Absent since July 2021 |
| Who issued the certificate? | Recorded | Anonymised since July 2021 |
| Why was a certificate deactivated? | Not documented by the study | The reason exists nowhere |
| Inspection floor | 2% of certificates a year, plus one audit per assessor every 12 months | No floor expressed as a share of certificates; 0.08% checked in 2023 |
| Failure threshold of a check | 5 SAP points; the certificate is replaced | No numerical tolerance published |
| Failure rate of checks published every year | No | No (18% of non-compliance cited once by the Court of Audit, for 2023) |
| Register of exemptions from the letting ban | Opened on 1 October 2017, public since 5 May 2026 | None |
| Who enforces energy standards in tenancies | The local authority, with a duty to act | No authority |
Law and registers as at 19 August 2026. The audit failure rate is published in neither country.
France: the share on the favourable side, line by line, including where it is inconvenient
| Value (kg CO₂/m²/yr) | Share on the favourable side | Certificates counted | What this value is |
|---|---|---|---|
| 70 | 72% | 61,181 | entry into band F |
| 50 | 64% | 121,943 | entry into band E |
| 11 | 64% | 430,294 | B/C boundary, no direct regulatory stake |
| 100 | 61% | 12,965 | entry into band G |
| 131 control values | 51% median, 56% at most | — | values more than two units from a line |
Certificates based on a visit, mainland France, 2022-2025, homes over 40 m². The B/C boundary, which carries no legal stake, also shows a step: the study publishes it and draws no estimate from it.
The French internal control: certificates derived from a whole-building certificate
| Threshold and population | Excess | 95% confidence interval | Certificates in the window |
|---|---|---|---|
| Band E (50 kg) — certificates based on a visit | +21.4% | +14.0% to +24.8% | 1,871,687 |
| Band E (50 kg) — certificates derived from a building | −0.5% | −5.0% to +4.4% | 743,625 |
For a certificate derived from a building, crossing a line is not decided at the level of the home: at the band E line, the excess is nil. The band F rows are in the method note. These certificates are not a perfect control (flats, more frequent shared heating).
When the line moves, the bunching follows: homes of 40 m² or less
| Population | Period | Share on the favourable side of the applicable line | Share on the favourable side of the former 70 kg line |
|---|---|---|---|
| 40 m² or less | before 1 July 2024 | 50% (3,476) | 71% (5,785) |
| 40 m² or less | after 1 July 2024 | 74% (3,207) | 47% (3,942) |
| Over 40 m² (control) | before 1 July 2024 | 73% (41,617) | — |
| Over 40 m² (control) | after 1 July 2024 | 70% (19,594) | — |
In brackets, the number of certificates counted. Emissions line between bands E and F, certificates based on a visit, mainland France, 2022-2025. Exact intervals (Clopper-Pearson) in the method note.
England and Wales: two housing stocks, two rules, one register
| Stock | Certificates | Excess at the 39 SAP-point threshold | 95% confidence interval |
|---|---|---|---|
| Whole register, excluding new homes | 20,716,467 | +13.94% | +9.4% to +18.7% |
| Private rented (covered by the rule) | 4,395,103 | +39.8% | +28.1% to +52.4% |
| Social housing (not covered) | 4,229,524 | −1.1% | −7.9% to +6.9% |
England and Wales certificate register, 2012-2026, excluding new homes. Tenure is missing on 14.5% of rows. The French and English scales share neither unit nor band width: these percentages cannot be compared directly with the French figures.
In France, the certificate is checked far less than the works it triggers
| What is checked | Year | Floor or rate |
|---|---|---|
| Works funded by energy savings certificates: required on-site checks | 2023 | 10% |
| Same | 2024 | 12.5% |
| Same | 2025 | 15% |
| Comprehensive renovation bonus, pipe insulation | — | 100% |
| Home certificates checked (Court of Audit) | 2023 | 0.08% |
| Same number of checks, over certificates in the open data | 2023 | 0.097% |
Sources: annual reports of the French energy and climate directorate (DGEC); Court of Audit, report of 3 June 2025. The two floors do not cover the same object (a works operation on one side, an assessor on the other): this compares instrumentation, not probity.
Methodology
Published in full so anyone can check, replicate or challenge it.
Scope, definition, sources and limits
Scope
France: 9,843,867 certificates for existing homes based on a visit (excluding those derived from a whole-building certificate), mainland France, issued 2022 to 2025. The 72% share (43,882 and 17,299 certificates) excludes homes of 40 m² or less, whose band lines depend on floor area. The 2021 vintage and overseas territories are excluded. England and Wales: 20,716,467 certificates excluding new homes, 2012 to mid-2026; Scotland and Northern Ireland keep separate registers.
What the figure means
Two measures, to be read in this order. The share: at each band line, we count certificates at the last value on the favourable side and at the first value on the unfavourable side, with no model at all; elsewhere on the scale it averages 51%. The excess: we compare the number of certificates near the line with what a curve fitted to the rest of the distribution predicts; it depends on calculation choices, hence the 948 variants tested, all positive.
Sources
- ADEME, energy performance certificates for existing homes (2021 method), snapshot of 11 June 2026: 16,273,573 certificates. Licence Ouverte 2.0.
- England and Wales energy performance certificate register: 23,552,693 certificates, 2012 to mid-2026. Contains public sector information licensed under the Open Government Licence v3.0.
- French Court of Audit, report of 3 June 2025; French energy and climate directorate (DGEC), energy savings certificate reports; Scheme Operating Requirements (England and Wales), 2025 edition.
- Excess estimation method: Chetty, Friedman, Olsen and Pistaferri (2011); Kleven and Waseem. Validity of the share on a discrete score: Cattaneo, Titiunik and Vazquez-Bare (2017).
- Study produced by OneDpe (RIANN SAS), publisher of certificate analysis tools: the commercial interest is declared. Right of reply: droit-de-reponse@onedpe.fr, under article 6 IV of French Law 2004-575 of 21 June 2004.
Limits — what this study does NOT say
- Two mechanisms are consistent with the excess, and the data cannot separate them: input data adjusted near the line, or selective disappearance of certificates from the published database. Over 2021-2026, at least 1,428,641 certificates were removed from the database (8.07% of the original cohort), with no recorded reason.
- The excess is neither proof of intent nor a statement about any case or professional. No statistic by assessor, firm, software or body is produced, on either side of the Channel.
- The published figures are floors: the excluded window, dilution by homes whose label does not change at the line (22.71% of the window), removed certificates and the boundary regime used all push towards underestimation.
- The two countries do not use the same scale: a cost score in SAP points on one side, two physical quantities on the other, the French label taking the worse of the two bands. The study compares the two countries only with aligned methods, never through raw rates.
- Inspection rules say what each system requires, not what it delivers. The French check covers an assessor, the English audit a certificate: their rates do not convert into one another.
- The French database holds no certificate issued before 2 July 2021: the pre-reform period cannot be measured. The study publishes no inspection rate for 2024 or 2025, for lack of a published source.
Quotes
Attributable and ready to publish. For an interview, see the contact below.
« What we are proposing costs almost nothing: publish an inspection rate, set a tolerance, record whether a home is rented. England has had the last two for years. »
About OneDpe
OneDpe.fr is an independent platform of analysis tools for the French energy performance certificate (DPE). It offers free tools — certificate consistency checks, discount estimation, 3CL simulation, simulation of the rating under the 1.7 factor for certificates issued in mainland France from 1 January 2027, renovation works estimation — built on the analysis of several million certificates from the ADEME open data, cross-referenced with French property transaction data (DVF). OneDpe is published by RIANN SAS (Paris).
Going further
Journalist? Detailed figures, interviews and custom breakdowns: presse@onedpe.fr — reply within 24 hours.
Steven Annonziata, co-founder of OneDpe (RIANN SAS) and author of the study — available to comment, in French or English, within 24 hours.