Statistical revisions

Revisions published by the producers

• Producer correctionINSEE · 2021-2025 (housing), 2024-2025 (failures)

  Housing starts and business failures

  Value previously served	Housing starts: 412.3 thousand in 2021 · 388.4 in 2022 · 296.2 in 2023 · 259.0 in 2024 · 264.2 in 2025. Failures: 66,133 in 2024 · 68,560 in 2025
  Value served today	Housing starts: 414.2 thousand in 2021 · 389.5 in 2022 · 296.5 in 2023 · 259.6 in 2024 · 263.4 in 2025. Failures: 66,123 in 2024 · 68,540 in 2025
  Reference publication	INSEE — Macroeconomic database (BDM), series 001718270 and 001656101
  Publication date	2026-09-01
  Applied by France ÉcoScan on	2026-09-09

  Re-query of the BDM by the canonical ingestion pipeline, with no manual entry. Both series are MONTHLY at the producer and aggregated into annual figures here: their latest points are provisional by construction and are consolidated as returns come in. The gaps range from 0.1 to 1.9 thousand housing units and from 10 to 20 failures — the order of magnitude of an ordinary consolidation, not a change of method. France ÉcoScan does not state what INSEE consolidated: the publication does not detail it, and deducing it from an order of magnitude would be an inference, not an observation.

  What this changes on screen — Five annual figures for housing starts and two for failures move at the margin. No curve changes shape: the 2023-2024 trough in construction remains the trough, and the continuous rise in failures since 2021 remains continuous. No editorial conclusion rests on the revised decimals.

• Methodological revisionEurostat · 1975-2025 (productivity), 1988-2025 (confidence)

  Hourly labour productivity and consumer confidence

  Value previously served	Hourly productivity: index 45.626 in 1975 · 101.654 in 2025. Consumer confidence: −4.8833 in 1988 · −16.3 in 2025
  Value served today	Hourly productivity: index 45.49 in 1975 · 101.678 in 2025. Consumer confidence: −4.875 in 1988 · −16.2917 in 2025
  Reference publication	Eurostat — Labour productivity and unit labour costs (nama_10_lp_ulc) and Consumer opinion surveys (ei_bsco_m)
  Publication date	2026-08-18
  Applied by France ÉcoScan on	2026-09-09

  Re-ingestion by the canonical Eurostat pipeline, with no manual entry. The two series involve two different mechanisms, which must be told apart. Hourly productivity is an INDEX whose base has been recalculated: the gaps reach 0.3% on the older years and shrink towards the present — the signature of a rebasing, not a one-off correction. Consumer confidence is an annual average of MONTHLY balances, and the gaps there are worth less than a hundredth of a point: they are rounding differences in the average, revealed because the producer re-released its monthly figures. France ÉcoScan does not state what Eurostat rebased: the publication does not detail it, and deducing it from the shape of the gaps would be an inference, not an observation.

  What this changes on screen — No curve changes shape, and no editorial conclusion rests on the revised decimals. Hourly productivity keeps its 2010s plateau and its recent recovery; consumer confidence keeps its troughs of 2008, 2022 and 2025. The gaps are smaller than the thickness of the curve.

• Methodological revisionEurostat · 2003-2023

  Number of ILO unemployed, aged 15-74 (millions)

  Value previously served	2.364 M in 2003 · 2.365 M in 2021 · 2.263 M in 2023
  Value served today	2.368 M in 2003 · 2.408 M in 2021 · 2.300 M in 2023 — 2024 and 2025 unchanged
  Reference publication	Eurostat — Unemployment by sex and age, annual data (une_rt_a)
  Publication date	2026-09-10
  Applied by France ÉcoScan on	2026-09-13

  Re-ingestion by the canonical Eurostat pipeline, with no manual entry. The twenty-one annual figures from 2003 to 2023 are raised by 4,000 to 43,000 people, the last two do not move, and the series-break flag the producer carried on 2024 disappears. This shape — a history raised until it meets the last two years, with no remaining break — is that of a back-calculation; it is consistent with the extension of the France coverage to Mayotte that INSEE applied to the Labour Force Survey on 7 August 2026 (methodological note on the ILO rate in the public truth table). The Eurostat publication does not detail it: France ÉcoScan relates the two, it does not assert it.

  What this changes on screen — The shape of the curve does not change: a rise from 2008 to 2015, a decline until 2022, an upturn in 2024-2025. Levels before 2024 are slightly higher, by at most 1.8%, which reduces the apparent upturn of 2024 by as much. No analysis rests on the gap between these years.

• Methodological revisionEurostat · 2019 and 2024

  Car industry production (index)

  Value previously served	136.5 in 2019 · 101.7 in 2024
  Value served today	136.6 in 2019 · 101.8 in 2024
  Reference publication	Eurostat — Production in industry, annual data (sts_inpr_a)
  Publication date	2026-09-11
  Applied by France ÉcoScan on	2026-09-13

  Re-ingestion by the canonical Eurostat pipeline, with no manual entry. Two annual figures raised by a tenth of an index point, in a re-release of 11 September 2026; the other thirty-four are unchanged. A gap of this size is that of a monthly-average rounding re-released by the producer, not of a change of base.

  What this changes on screen — Smaller than the thickness of the curve: no reading changes.

• Change of baseINSEE · 1991-2025

  Cumulative consumer prices (base 100 in 1990)

  Value previously served	179.45 in 2025 (series 001764363, base 2015, discontinued by INSEE)
  Value served today	179.56 in 2025 (series 011814639, base 2025, linked to 001764363 for 1990-1995)
  Reference publication	INSEE — Macroeconomic database (BDM), series 011814639 (annual consumer price index, base 2025), linked to 001764363 (base 2015, discontinued)
  Publication date	2026-02-27
  Applied by France ÉcoScan on	2026-09-13

  The base-2015 series is marked “Séries arrêtées” (discontinued series) by the producer: the 2026 figure would never arrive there. The base-2025 series prevails wherever it publishes (1996-2025); the discontinued base only fills 1990-1995, multiplied by a chain-linking factor measured on 1996 (0.826302) and checked against the thirty overlapping years (maximum gap 0.11%). The whole is rebased to 100 in 1990 as before: the values move by less than a tenth of an index point.

  What this changes on screen — Smaller than the thickness of the curve: no reading changes; the 2026 figure can now be served.

• Methodological revisionINSEE — annual national accounts (base 2020) · 1949-2025

  Margin rate of non-financial corporations (%)

  Value previously served	Eurostat nasa_10_ki (B2G_B3G_RAT_S11), latest year 2024 = 32.16% — the figure transmitted in 2025, two decimals
  Value served today	INSEE, annual national accounts base 2020, table 4.101 of the non-financial corporations accounts (29 May 2026): 2024 revised to 32.7% (semi-final), 2025 = 32.2% (provisional), 1949-2023 identical after rounding — served at the published decimal
  Reference publication	Les comptes de la Nation en 2025 (national accounts 2025) — non-financial corporations, table 4.101 “Principaux ratios des comptes des sociétés non financières (S11)”, T_4101_fr.xlsx (sha256 32f8ed52d085ecf8ac21bd394f22613ef45329130c6e1f6884a41d1dd9cddc4b); Insee Première No. 2105, figure 5
  Publication date	2026-05-29
  Applied by France ÉcoScan on	2026-09-13

  The same concept on both sides — gross operating surplus relative to the gross value added of non-financial corporations (S11), where mixed income is zero —, but Eurostat only receives the French transmission at T+9 months: in September 2026 it was still serving the 2024 figure of May 2025 when INSEE had published a revised 2024 and a 2025. The most recent official figure is the national producer's; the series is brought back to INSEE (M0D.1), with a cross-check at ingestion against BDM 011793882.

  What this changes on screen — The 2024 margin rate goes from 32.16% to 32.7% and the 2025 figure (32.2%) appears; the curve is served at the published decimal instead of two decimals, with no other change in value.

• Methodological revisionINSEE — annual national accounts (base 2020) · 2025

  Public debt (€bn) and public deficit (€bn) — 2025 figure; derived 2025 debt per inhabitant

  Value previously served	Maastricht debt at end-2025 = €3,460.5bn; 2025 net borrowing = €152.5bn; 2025 debt per inhabitant = €50,093 — INSEE tables of 29 May 2026 (Insee Première No. 2106), retrieved 10 August 2026
  Value served today	debt at end-2025 = €3,460.1bn; 2025 net borrowing = €151.6bn (−151.592 published); 2025 debt per inhabitant = €50,087 (€3,460.1bn ÷ 69,081,996 inhabitants on 1 January 2026) — ratios of 115.7% and 5.1% of GDP unchanged; 44 other years identical
  Reference publication	Debt of general government under the Maastricht definition by sub-sector (econ-fin-pub-dette-adm-pub.xlsx, sha256 923b35afe7fe72929a87452b8a1a3042292967302c680f16d2e0eb10e4e74bf1) and Net lending or borrowing of general government by sub-sector (econ-fin-pub-deficit-secteur.xlsx, sha256 a0302dca10488db73864d218d6b61443a92d97ee0a59f7a78772aac53f3d22a1), tables “Paru le 28/08/2026” (published 28/08/2026)
  Publication date	2026-08-28
  Applied by France ÉcoScan on	2026-09-15

  On 28 August 2026 the producer updated the two annual general government tables, together with the detailed quarterly accounts for the second quarter of 2026: only the provisional 2025 figure moves — by €0.4bn on the debt and €0.9bn on net borrowing. Re-ingestion by the canonical connector (SHA-256 fingerprints re-pinned), with no manual entry. The quarterly Maastricht series (Informations rapides No. 158 of 25 June 2026, BDM 010777616) still carries €3,460.5bn for Q4 2025: it is an earlier vintage of the same stock, which INSEE will realign when it publishes Q2 2026 on 29 September 2026.

  What this changes on screen — End-2025 debt reads €3,460.1bn instead of €3,460.5bn, the 2025 deficit €151.6bn instead of €152.5bn, debt per inhabitant €50,087 instead of €50,093 — below the thickness of each curve; no ratio, no ranking, no editorial conclusion changes.

• Methodological revisionINSEE — quarterly Maastricht debt of general government (base 2020) · 2025-Q4, 2026-Q1

  Quarterly Maastricht debt (€bn) — Q4 2025 and Q1 2026

  Value previously served	Q4 2025 = €3,460.5bn; Q1 2026 = €3,536.1bn — Informations rapides No. 158 of 25 June 2026 (BDM 010777616)
  Value served today	Q4 2025 = €3,460.1bn; Q1 2026 = €3,535.9bn (OBS_REV=1); Q2 2026 = €3,595.5bn, 119.0% of GDP, added — Q4 2025 and Q1 2026 ratios unchanged (115.7% and 117.5%)
  Reference publication	Informations rapides No. 239 — “À la fin du deuxième trimestre 2026, le ratio de dette publique s'établit à 119,0 % du PIB” (at the end of Q2 2026, the public debt ratio stands at 119.0% of GDP); BDM series 010777616 (LAST_UPDATE 2026-09-29)
  Publication date	2026-09-29
  Applied by France ÉcoScan on	2026-09-29

  The second-quarter 2026 release realigns Q4 2025 on the annual table of 28 August 2026 (REV-2026-016) and revises Q1 2026 by €0.2bn. It becomes the official anchor of the debt counter, whose pace is no longer a twelve-month moving average but the slope of the nowcast (M0U).

  What this changes on screen — The outstanding amount displayed goes from €3,536.1bn (Q1 2026) to €3,595.5bn (Q2 2026) and the ratio from 117.5% to 119.0%. The counter restarts from €3,595.5bn on 30 June 2026; its pace, dated by its last recalibration, is displayed alongside.

• Producer correctionINSEE · 2022, 2023, 2025 (housing), 2024-2025 (failures)

  Housing starts and business failures

  Value previously served	Housing starts: 389.5 thousand in 2022 · 296.5 in 2023 · 263.4 in 2025. Failures: 66,123 in 2024 · 68,540 in 2025
  Value served today	Housing starts: 389.6 thousand in 2022 · 296.3 in 2023 · 263.0 in 2025. Failures: 66,115 in 2024 · 68,524 in 2025
  Reference publication	INSEE — Macroeconomic database (BDM), series 001718270 (LAST_UPDATE 2026-09-29) and 001656101 (LAST_UPDATE 2026-09-18)
  Publication date	2026-09-29
  Applied by France ÉcoScan on	2026-09-29

  The same mechanism as REV-2026-009: both series are monthly at the producer, and their latest points are consolidated as returns come in (housing starts “in actual date” are revised over eighteen months). Picked up by the confrontation with primary producers of 29/09/2026, absorbed by the canonical ingestion pipeline, with no manual entry.

  What this changes on screen — Three annual figures for housing starts move by 0.1 to 0.4 thousand, two for failures by 8 to 16 units. No curve changes shape; no editorial conclusion rests on these decimals.

France ÉcoScan source corrections

These gaps come from no institute: France ÉcoScan was serving a reconstructed value, a wrong unit or a weaker source, and now serves the official publication. The responsibility is entirely ours.

• France ÉcoScan source correctionFrance ÉcoScan (export correction) — producers: ECB, Eurostat, INSEE, official texts · 1981-2025

  Public CSV export — seven series realigned on the served register

  Value previously served	values from the bootstrap CSV still present in public/data/france-ecoscan-indicators-annual.csv for the France-Germany spread (60 bp in 1990), the fertility rate (1.75 in 1998), housing starts (410 thousand in 2000), the saving rate (18% in 1981), the margin rate, the CSG and the social levies on investment income
  Value served today	the values the Explorer serves from the R4/R5/TRUST-R2 authorities: 122 bp in 1990 (ECB), 1.78 in 1998 (Eurostat demo_find), 337.6 thousand in 2000 (INSEE BDM 001718270), 16.87% in 1981 (Eurostat nasa_10_ki)…
  Reference publication	scripts/fes-public-export-csv.ts — projection of the served register (fes-catalog-manifest-verify fails if the export diverges)
  Publication date	not established — left blank rather than approximated
  Applied by France ÉcoScan on	2026-09-13

  The public CSV export had only been realigned on some of the authorities: seven series still carried the bootstrap CSV values there while the Explorer was already serving the producers' series. The export is now regenerated from the served register and can no longer diverge.

  What this changes on screen — Readers of the CSV file saw, for these seven series, unsourced values — including an EMS crisis missing from the France-Germany spread and a saving rate one point too high in the 1980s. The Explorer, for its part, was correct.

Status of the series served

A provisional figure will be revised by its producer; a final figure will not move again, except for a change of base or method. A dated snapshot is a value taken at one moment, not updated continuously — no data on this site is real-time. A year in progress is never complete: its values carry the period actually covered.