# Trends: sources

The long series behind the question the site leads with: are day-to-day lives
in France getting better or worse, and for whom. `build/extract.exe` reads the
files below and writes `series.csv`, one row per published cell with its file
and cell; `data/trends.ml` reads it, declares each indicator's kind (flow or
stock), deflator and breaks, and builds the claims of the answer:

    dune exec data/trends/build/extract.exe -- data
    dune exec data/trends/build/extract.exe -- data --claims

The files of the sections below were downloaded on 2026-10-19 (times UTC),
those of the section "Insee, prices, wages and incomes by group" on
2026-10-20, and each was checked against a second download from its
publisher; the SHA-256 is that of the committed file. Every reading of a workbook checks the row label and the unit line it
relies on, so a republished file whose rows moved fails the extraction.

## Insee

Licence Ouverte 2.0.

- `../income/raw/reve-niv-vie-decile-individu.xlsx` (already in the tree,
  read by `data/income/build/living_standards.exe` too): "Distribution des
  niveaux de vie, données annuelles de 1996 à 2024", published 9 July 2026,
  <https://www.insee.fr/fr/statistiques/fichier/2416808/reve-niv-vie-decile-individu.xlsx>,
  SHA-256 `1e485ca348684f272e781de8a32631a1aa20369dc5509ab2465d39feadbcf2b4`
  (identical to the file fetched at 16:36). Sheet "Données", rows D1 to D9
  and C95, every year: indicator `erfs_threshold`.
- `../income/raw/reve-niv-vie-decile.xlsx`: "Niveau de vie moyen par
  décile, données annuelles de 1996 à 2024", published 9 July 2026,
  <https://www.insee.fr/fr/statistiques/fichier/2417897/reve-niv-vie-decile.xlsx>,
  SHA-256 `8bcdecd0705ddb902ffebafd206cc2436363397a7691bce1cc7c7d199f34e3ab`.
  Sheet "Données", the mean within each tenth and overall: `erfs_mean`.

  Both are in constant euros of 2024, deflated by the consumer price index
  (all households), metropolitan France, people in ordinary housing whose
  household's declared income is positive or zero and whose reference person
  is not a student. Three breaks, each published twice: 2010 (financial
  income from the Patrimoine 2010 survey), 2012 (recomputed on 2013's
  methods, Patrimoine 2014-2015), 2020 (renovated ERFS on the new Labour
  Force Survey). A column whose year repeats the previous one opens a new
  basis, so the column `basis` is `erfs-1996`, `erfs-2010`, `erfs-2012` or
  `erfs-2020`, and a change across a break is chained through the year
  published on both. Insee calls the 2020 points fragile (flag `fragile`).
- `raw/insee_ip2117.xlsx`: Insee Première n° 2117, "Niveau de vie et
  pauvreté en 2024", data workbook, published 9 July 2026,
  <https://www.insee.fr/fr/statistiques/fichier/9019316/ip2117.xlsx> (page
  <https://www.insee.fr/fr/statistiques/9019316>), fetched 16:03, SHA-256
  `df80d57ee7449f1a766ca8fc2ba59721bc735a41a18ee63e2d807b43f6dad026`.
  Insee's series 1996-2024 retropolated across the breaks of 2010, 2012 and
  2020 (its note: "Les données de 1996 à 2020 sont rétropolées pour
  permettre une comparaison temporelle"); `basis` is `retropolated`.
  - "Tableau complémentaire 1", rows 5-13 and 15-23: the decile thresholds
    after and before taxes and benefits, index 2008 = 100 in constant euros
    (`decile_index_after`, `decile_index_before`).
  - "Tableau complémentaire 2", rows 6-8: the median, D1 and D9 in euros of
    2024 a year (`living_standard`); rows 14-18: S20 to S90
    (`income_share`); rows 20 and 26: (100-S80)/S20 after and before
    (`quintile_ratio`); rows 23 and 27: the Gini index (`gini`).
  - "Tableau complémentaire 3", rows 5-9 (60% of the median) and 11-15
    (50%): number of poor, poverty rate, poverty line in euros of 2024 a
    month, median living standard of the poor, intensity.
- `raw/insee_t_2102_fr.xlsx`: "Revenu et pouvoir d'achat des ménages en
  2025", national accounts base 2020, table 2.102, published 29 May 2026,
  <https://www.insee.fr/fr/statistiques/fichier/8988851/t_2102_fr.xlsx>
  (page <https://www.insee.fr/fr/statistiques/8988851>), fetched 16:31,
  SHA-256 `1d9ad81ea39f2deb6b8608be592cf1540a443db0c1986ccf84bab6a235c802cc`.
  Sheet "T_2102", France, households including sole proprietors, 1959-2025:
  rows 7-9, gross disposable income (B6G), pre-committed consumption
  (CPDPE) and arbitrable income (RARBI) in billions of current euros
  (`accounts_level`); rows 16, 17, 21, 22, the change on the year before of
  the purchasing power of disposable income (_PAM) and of arbitrable
  income (_PAA), in total and per consumption unit (`accounts_growth`).
  Disposable income is deflated by the price of household final
  consumption, arbitrable income by the price of consumption other than
  pre-committed. Insee marks 2025 provisional and 2024 semi-final (flags `p`
  and `sd`). Pre-committed spending (Quinet 2008): housing, water, energy,
  insurance net of claims, telecoms, television subscriptions, canteens and
  financial services. The same purchasing-power series are in
  <https://www.insee.fr/fr/statistiques/2830244> and the arbitrable one in
  <https://www.insee.fr/fr/statistiques/2385829>; both agree with this
  table in every year they share (checked on 2026-10-19).

## Eurostat

CC BY 4.0, © European Union. EU-SILC for France; the survey year N measures
incomes of N-1, and housing costs and deprivation at the interview of N.
Each file is the API's JSON-stat answer
(`https://ec.europa.eu/eurostat/api/dissemination/statistics/1.0/data/DATASET?QUERY&lang=EN`),
with its flattening beside it (`.csv`: a column per dimension of size above
one, then the flag and the value). The extraction reads the JSON, keeps both
sexes together where a set gives each sex, and names every dimension but geo,
freq and time in the column `group`, so a row's group is the filter that
selects its cell. Flags as published: `b` break (2001, 2004, 2008, 2020 and
2022 in these files), `p` provisional, `u` low reliability, `e` estimated. The
certificate chain of ec.europa.eu was verified against GlobalSign Root R46
from the system's root store.

| File | Query | Updated | Fetched | SHA-256 |
|---|---|---|---|---|
| `eurostat_ilc_li02_FR.json` | `geo=FR&unit=PC&sex=T&age=TOTAL&statinfo=MEAN_EI&statinfo=MED_EI&rskpovth=B_40&rskpovth=B_50&rskpovth=B_60&rskpovth=B_70` | 2026-10-15 | 16:29:56 | `5e0cabf96e6d3e132417941f16fd3c0ff2939c056321bc4f499d3a6bb71d2438` |
| `eurostat_ilc_li22b_FR.json` | `geo=FR&unit=PC&statinfo=MED_EI&rskpovth=B_60&age=TOTAL&age=Y18-64&age=Y_GE65&age=Y_LT18` | 2026-10-05 | 16:29:58 | `27aa0866f7619c247bb1126c2165f1322af34f548bd23dca8da15aff604b2986` |
| `eurostat_ilc_li22_FR.json` | the same, anchored in 2019 | 2026-10-05 | 16:29:57 | `587cfd747a3cb2f1e0ed250147229c4982963769e06eb81fdc6b845550dccf9a` |
| `eurostat_ilc_di11_FR.json` | `geo=FR&age=TOTAL&sex=T&unit=RAT` | 2026-10-05 | 16:29:55 | `f61d43122f13faac7292d83c3f63e499e78a5123c2fcd5795017cf6d0c38484f` |
| `eurostat_ilc_mdsd02_FR.json` | `geo=FR&unit=PC&hhcomp=TOTAL` | 2026-09-17 | 16:29:59 | `03c3a1144765da38af8d931321d2b1ff39490d3f319213c50e3d780379a34e0d` |
| `eurostat_ilc_mdsd07_FR.json` | `geo=FR&unit=PC&age=TOTAL&sex=T` | 2026-09-17 | 16:29:59 | `850be7641267f0913b4e19a4472fbb6c7ebd7f112a6b441beed8bc3b7a43e8e2` |
| `eurostat_ilc_mdsd11_FR.json` | `geo=FR&unit=PC&age=TOTAL&sex=T` | 2026-09-28 | 16:29:59 | `027ec64ecf1c13ed393522c2490924eca26019f686898552c60667b33907a65a` |
| `eurostat_ilc_mdsd13_FR.json` | `geo=FR&unit=PC&hhcomp=TOTAL` | 2026-09-17 | 16:30:00 | `77a1c066476edc59698fada61ad263ce67e771fafe8a1a497c8870192da7c814` |
| `eurostat_ilc_mddd11_FR.json` | `geo=FR&unit=PC&age=TOTAL&sex=T` | 2025-10-28 | 16:29:58 | `eff46e1f8d570db1bec7233f07736e87027a3be362f1f8fe907c3928ae644dec` |
| `eurostat_ilc_mddd13_FR.json` | `geo=FR&unit=PC&hhcomp=TOTAL` | 2026-08-13 | 16:29:59 | `23d71c12a63b1708df5b9e9e3252f13f2f1d14a631a29298e8428e41d087ebd8` |
| `eurostat_ilc_lvho07a_FR.json` | `geo=FR&unit=PC&age=TOTAL&sex=T` | 2026-10-16 | 16:29:58 | `fdfa5425b126d7800439228bddbe9cead147fa659a08bc8a29f32b28f7390c0c` |
| `eurostat_ilc_lvho07b_FR.json` | `geo=FR&unit=PC` | 2026-10-16 | 16:29:58 | `811d32dd068356093da88a47e15584529f32cf65728f264f3de05f89c62ca236` |
| `eurostat_ilc_mded01_FR.json` | `geo=FR&unit=PC&hhcomp=TOTAL` | 2026-10-16 | 16:29:59 | `91516caad2754e6bf04b375f0da5a743ac9ee700745fb0d9970a11d72dd7fc55` |
| `eurostat_ilc_pw01_FR.json` | `geo=FR&statinfo=AVG&unit=RTG&isced11=TOTAL&life_sat=LIFE&age=Y_GE16` | 2026-09-17 | 16:30:01 | `d90556c3fd4f295881a24017db3f79324a0fb7786e726c7b9a47cef93eded8dd` |
| `eurostat_ilc_pw02_FR.json` | `geo=FR&statinfo=AVG&unit=RTG&deg_urb=TOTAL&life_sat=LIFE&hhcomp=TOTAL` | 2026-07-06 | 16:30:03 | `5b3da1ba6cd3c5ed3a953f0eeb7dacf144b2889f9cea9775a667ef92ea4746fc` |
| `eurostat_hlth_hlye_FR.json` | `geo=FR` | 2026-07-17 | 16:29:54 | `ccccd824edeabe7e96b2cf96c87caaf0dbbaf9b1a2e72f2f6c351ea8c12019e7` |
| `eurostat_demo_mlexpec_FR.json` | `geo=FR&unit=YR&sex=F&sex=M&sex=T&age=Y_LT1&age=Y65` | 2026-09-25 | 16:29:54 | `2d050db2e372cfef84299b8e7dd19c6388abb67cb1a22f6a1be2816a87ce3f61` |

Notes. `ilc_li22b` holds a line fixed at 60% of the 2008 median, deflated
by the harmonised index of consumer prices: it is the absolute line of the
measures. The severe material deprivation series (`ilc_mddd*`, 9 items, 4
lacking) stops in 2020; the material and social deprivation rate
(`ilc_mdsd*`, 13 items, 5 lacking, 7 for the severe rate) runs from 2014 and
is the one the EU's 2030 target uses; the two are not one series. France has
no housing-cost row in 2021. Life satisfaction is flagged `u` in every year.
Healthy life years changed question in 2008 and are provisional for 2022
to 2024; life expectancy is provisional for 2023 and 2024.

## World Inequality Database

CC BY 4.0. Bulk download `wid_all_data.zip`
(<https://wid.world/bulk_download/wid_all_data.zip>, 886,945,845 bytes, last
modified 15 October 2026). The archive is not kept: its members
`WID_data_FR.csv` (SHA-256
`64429f6bfd5fe834688e7a5051a2ae6dc01ad4c8bbc4ba5e1f59205070cbd5ff`) and
`WID_metadata_FR.csv` (SHA-256
`3779cad426d95b7eefd1854c5529f3f10ab77183c7c172d909846e7ce7ba6990`) were
read by HTTP range requests on 2026-10-19 and the rows below kept.

- `raw/wid_FR_trends.csv` (SHA-256
  `28b85d7b99d026e1b365a945f7f369fdc9926253cacd34e3efad31b50e5090da`): years
  1970 onwards, groups `p0p50` (bottom half), `p50p90` (middle 40%),
  `p90p100` (top tenth), `p99p100` (top 1%) and `p0p100`, of the variables
  `ahwealj992` and `shwealj992` (average and share of net personal wealth),
  `adiincj992` and `sdiincj992` (post-tax disposable income, from 1980),
  `aptincj992` and `sptincj992` (pre-tax national income), adults aged 20
  and over, split equally between spouses; plus `inyixxi999` (national
  income price index) and `npopuli992`. The extraction reads all but
  `sptincj992` and `npopuli992`, which stay for cross-checks.
- `raw/wid_FR_trends_metadata.csv` (SHA-256
  `26415d31cd18cb90499e9224d295b4eeb0548ff4b6bfefa3b5159f98c8cc2ec3`): the
  matching metadata rows (units, sources, method).

Prices: WID's amounts are in constant euros of its latest year, deflated by
the national income price index, which is 1 in 2025; so these are euros of
2025. Wealth is a stock at the end of the year (Garbinti, Goupille-Lebret
and Piketty, from estate-tax, capitalised-income and survey sources that
change over the span); incomes are flows of the year scaled to the national
accounts. This file supersedes nothing: `data/wealth/raw/wid_FR_wealth.csv`
(2015 onwards, bulk data of 9 September 2026) stays the wealth page's input.

## IGEDD (Jacques Friggit)

Reuse with the source named: "Les séries peuvent être utilisées librement,
sous réserve d'en mentionner explicitement la source, y compris leur source
première", as the workbook states (row 12). Source: IGEDD d'après Insee,
base de données notariale, indices Notaires-Insee, Banque de France.

- `raw/igedd_friggit_valeur-immobilier-depuis-1800.xls`: "Valeurs
  immobilières, économiques et financières de 1800 à 2025", last saved 16
  October 2026, from the page
  <https://www.igedd.developpement-durable.gouv.fr/prix-immobilier-evolution-a-long-terme-a1048.html>,
  file
  <https://www.igedd.developpement-durable.gouv.fr/IMG/xls/valeur-immobilier-depuis-1800_cle0a1c56.xls>,
  fetched 16:32, SHA-256
  `269bcbcc8c9793da4b58afdf11bf8e4f3f25c44651ce4ad83c6181b8af323e3a`
  (identical to the copy of 14:30). Sheet "Series", header row 16 (years in
  columns D to HU, 1800-2025), rows 19 (consumer price index, base 2000),
  21 (household disposable income, EUR bn), 25 (households, thousands), 36
  (house price index, France, base 2000, from 1936) and 39 (rent index,
  base 2000, from 1949), the years 1936 onwards: indicator `friggit`. The
  sheet prints provisional values in red, which the extraction cannot see;
  2025 is flagged `p`.

House prices against income are Friggit's ratio: the price index over
household disposable income per household, so no deflator is needed. The
same ratio is the "tunnel de Friggit" of his presentation (1.53 at the second
quarter of 2026 on a quarterly basis, 1.56 here for the annual 2025).

The site draws the ratio from 1949, not from 1936, on what Friggit's own
notes say, read on 2026-11-12 (not committed):

- "Sources of the long run series" (October 2026), the document the
  workbook's "Cliquer ici pour accéder aux sources" opens,
  <https://www.igedd.developpement-durable.gouv.fr/IMG/doc/sources-valeur-immobilier-depuis-1800_cle26c473.doc>,
  SHA-256 `4c102056b87f5e51753f75e10386229e923ca21944dc23202c9d7d9db560cb58`:
  household disposable income "1896-1938: after Villa P. [...] From 1938:
  Insee (missing years interpolated)"; households "1861-1982: after Insee
  (missing values extrapolated [...])"; home price index "1840-1999: as
  described in Comparing Four Secular Home Price Indices [...] From 1999:
  indices Notaires-Insee".
- "Comparing Four Secular Home Price Indices" (June 2008),
  <https://www.igedd.developpement-durable.gouv.fr/IMG/doc/house-price-index-Paris-and-others-secular_cle7fed11.doc>,
  SHA-256 `e8652b7c999978e304c1f26402f44caa8935ec44d35017b901d342c1d9feaea2`,
  on the France index built from the previous sale recorded in the
  notaries' databases (sales of 1993 to 1999): "Prior to the 1970's, when
  the previous transaction year gets older, the number of records decreases
  sharply, and so does the precision of the index. We could go back only to
  1950 for Paris (1936 for France)", the records falling "by a factor 10
  when the previous transaction is older by 10 years"; and "concordant
  reports of considerable price dissimulation during and after WWII to
  evade transaction taxes (which amounted to 17% to 36% of the price during
  the Occupation and were still very high in the following years)".

In `series.csv` income reads to the million from 1949 (9.196 bn), the
first year of Insee's annual accounts, and as chained decimals before it,
where Friggit interpolates; the years before 1949 also carry the
understated war prices. Households are interpolated between censuses in
every year to 1982, a smooth term that moves the ratio little. The index's
lower precision before the 1970s is stated in the series' caveat and under
the chart.

## Insee, prices, wages and incomes by group (2026-10-20)

Licence Ouverte 2.0. Fetched on 2026-10-20 between 15:00 and 15:30 UTC and
checked against a second download at 15:29 (workbooks of incomes and
wages) or 15:38 (workbooks of prices): every SHA-256 below is that of the
committed file, and each second copy was identical.

- `raw/insee_ip2117.xlsx` (above) also gives, read since 2026-10-20:
  "Figure 4", rows 6-26 (by activity status and household type) and
  "Tableau complémentaire 4", rows 5-10 (by density of commune), columns B-E
  (2023) and F-I (2024): share of the population (`profile_share`), median
  living standard in euros of 2024 a year (`profile_median`), poor in
  thousands (`profile_poor`) and poverty rate (`profile_poverty`); groups
  `status=...`, `household=...`, `density=...`, `all`.
- `raw/insee_reve-niv-vie-individu-age.xlsx`: "Niveau de vie selon l'âge",
  published 9 July 2026,
  <https://www.insee.fr/fr/statistiques/fichier/2416878/reve-niv-vie-individu-age.xlsx>,
  SHA-256 `662eb4e4495677f32e0eea18415f56ef78a1d77d2c44d286f66f682257cf1de8`.
  Sheet "Figure 1", rows 5-12, the median by age band in euros of 2024,
  1996-2024, on the ERFS bases as above (`age_median`).
- `raw/insee_reve-pauv-seuil-age.xlsx`: "Pauvreté selon l'âge et le seuil",
  published 9 July 2026,
  <https://www.insee.fr/fr/statistiques/fichier/3565548/reve-pauv-seuil-age.xlsx>,
  SHA-256 `fa00f7e38c79df5420e68667604f125cbb4842e8a1f60a690b15709b05283ae6`.
  Sheet "Figure 1", rows 5-12, poverty rate at 60% by age band
  (`age_poverty`); its footnote mark for the fragile 2020 is 4, not 3.
- `raw/insee_reve-niv-vie-pouv-achat-trim.xlsx`: "Évolution du revenu
  disponible brut et du pouvoir d'achat", quarterly accounts base 2020,
  published 28 August 2026 (detailed results of 2026-T2),
  <https://www.insee.fr/fr/statistiques/fichier/2830166/reve-niv-vie-pouv-achat-trim.xlsx>,
  SHA-256 `b0444dc5ddeadfc12c76c01d29a93edafc1cb6cc78abfd0f198586d1eed09eff`.
  Sheet "Données", a row per quarter (newest first, 2026-T2 to 1949-T2),
  columns D, E, F: purchasing power of disposable income, the same per
  consumption unit (changes on the quarter before, seasonally and
  working-day adjusted) and the saving rate (`accounts_quarter`, groups
  `pp`, `pp_uc`, `saving_rate` with `quarter=T1`...). Row 5, the carry-over
  for 2026, is not read. Insee publishes the arbitrable variant yearly only.
- `raw/insee_marc-salair-smic.xlsx`: "Smic, données annuelles de 1980 à
  2026", published 9 September 2026,
  <https://www.insee.fr/fr/statistiques/fichier/1375188/marc-salair-smic.xlsx>,
  SHA-256 `9371a8fb45317f79f0bb4a994dcac90bcc75b0451710f7c316269bd8cca7382e`.
  Sheet "Figure 1", column D, each hourly gross SMIC from its date in force
  in column C (Excel dates; `smic_hourly`, groups `from=YYYY-MM-DD`). Sheet
  "Figure 4", rows 5-20 (2025 to 2010), the yearly change of the purchasing
  power of the hourly gross SMIC (C, `smic_real`), of Dares' hourly base
  wage of workers and employees (E, SHBOE from the Acemo survey,
  `shboe_real`; 2025 is "nd") and of the CPI with tobacco (F,
  `cpi_annual`).
- `raw/insee_ECRT2026-E3.xlsx`: Insee Références "Emploi, chômage, revenus
  du travail", édition 2026, éclairage 3, "Entre 2019 et 2024, des
  évolutions salariales contrastées selon les métiers", published 2 July
  2026 (title corrected on 7 July), page
  <https://www.insee.fr/fr/statistiques/8733117>, file
  <https://www.insee.fr/fr/statistiques/fichier/8733117/ECRT2026-E3.xlsx>,
  SHA-256 `a34768fae1d75ee7903bc45646c8802f7edb5992c7dcad7f455c8ec5f795dd60`.
  Sheet "Figure 1", rows 4-24: the real change of the mean net wage in
  full-time equivalent from 2019 to 2024 by socio-professional group (B,
  `pcs_real_change`) and the group's share of full-time equivalents in
  2024 (C, `pcs_share`); private sector outside agriculture, bases Tous
  salariés. The text names no deflator, only "la hausse des prix (+14,9 %
  sur cette période)". The count of 121 professions of 343 is in the PDF
  only (below).
- `raw/insee_econ-gen-conso-menage-fonction-2.xlsx`: "Consommation
  effective des ménages par fonction", national accounts base 2020,
  published 29 May 2026,
  <https://www.insee.fr/fr/statistiques/fichier/2830276/econ-gen-conso-menage-fonction-2.xlsx>,
  SHA-256 `3b483b6b8685ddd53c9cb1716132283e5a35683983d1dc56dbffaf5505a83e67`.
  Sheet "Données", rows 6-18, 2025 (provisional): EUR bn (E,
  `consumption_value`) and the share of household consumption expenditure
  (G, `consumption_share`; imputed rents included, so housing is 28%).
  Column F, the share of actual consumption, which counts public spending
  in kind, is not read.

Consumer prices. Insee's monthly CPI series are in its BDM, served from
bdm.insee.fr, which the build's network refuses (below). The anchors are
read instead from the tables of the releases:

- `raw/insee_ipc_2017_03_indices.xls` and `raw/insee_ipc_2017_06_indices.xls`:
  "Indices détaillés", members of the zips
  <https://www.insee.fr/fr/statistiques/fichier/2832611/Donnees_complementaires_Mars_2017.zip>
  (SHA-256 `4134eba6ac486b6068e21ab8f6e435ab3702abdbe3f26f3328d7c95c97026db0`)
  and
  <https://www.insee.fr/fr/statistiques/fichier/2900793/Donnees_complementaires_Juin_2017.zip>
  (`2efb9d2bf59cd2afd9108c004287d02fd385682b015de0db69a5a01b49178f0a`),
  final CPI of March and June 2017; SHA-256
  `8a353d4a94b65e1fae57ae68edbeff22014cb80dbe20483ef00dddc87602b330` and
  `3055c667c3d52cfffde80bb60220212362df987571727c798e2a14c3b7402126`.
  Row 9 (ENSEMBLE), columns E-I, base 2015: March 2016, December 2016 and
  January to June 2017 (`cpi_level`, basis `base-2015`, groups
  `month=MM`). March 2017 is in both and read once, its two values equal.
- `raw/insee_ipc_annuel_2025.xls`: "Données complémentaires" of
  Informations rapides n° 8, 15 January 2026, updated 21 January,
  <https://www.insee.fr/fr/statistiques/fichier/8726461/Donnees%20complementaires%202025_maj_21012026.xls>,
  SHA-256 `5e7800edc0f4e7ceb7b5a69faf0911f613af7ca9ba93d48ac3e86474b3e65969`.
  Sheet "Moyennes annuelles", row 6 (ENSEMBLE), the averages of 2023 to
  2025 on base 2015, France excluding Mayotte (`cpi_level`, group `year`).
  Its sheet "Regroupements conjoncturels" still holds the 2023-2024 table.
- `raw/insee_ipc_2026_09_indices.xls`: "Tableau Données Détaillées" of the
  final CPI of September 2026, Informations rapides n° 247, 15 October
  2026, member of
  <https://www.insee.fr/fr/statistiques/fichier/9080209/Donn%C3%A9es%20Compl%C3%A9mentaires%20IR%20IPC%20septembre%202026.zip>
  (SHA-256 `7953e9303cc0828526e12997e1ea549865ed9eb2e4e0592eba68f6c4eb7b34f8`);
  SHA-256 `ead618ece08104a085169acd9fd1864b1f36a5f2a4e42df0c868c062d2b37220`.
  Sheet "Feuille1", row 2 (Ensemble), columns D-H: September 2025 and June
  to September 2026 on base 2025, France (`cpi_level`, basis `base-2025`).

A month of 2026 is carried to base 2015 by 2025's average on base 2015
(120.95): base 2025 is the 2025 average. August 2026 is then 125.00, 23.4%
above June 2017. The two bases' fields differ slightly (France excluding
Mayotte, then France).

- `raw/insee_ip1958.xlsx`: Insee Première n° 1958, "En 2022, les prix
  restent plus élevés dans les DOM qu'en France métropolitaine", 11 July
  2023,
  <https://www.insee.fr/fr/statistiques/fichier/7648939/donnees_insee_premiere_n1958.xlsx>,
  SHA-256 `c824a18d8ee0c38bd78e57b2579f360823b5b4ac59d25d991a907ed7076e2736`.
  Sheet "Figure 2", rows 4-8, the Fisher price gap of each overseas
  département to metropolitan France in 2010, 2015 and 2022
  (`dom_price_gap`); "Figure 1", columns C and D, the 2022 gaps on the
  metropolitan and the local basket. Prices surveyed in March-April 2022;
  consumption excluding fuel oil, town gas and rail.
- `raw/insee_ip2002.xlsx`: Insee Première n° 2002, "La consommation des
  ménages en 2023", 4 July 2024,
  <https://www.insee.fr/fr/statistiques/fichier/8210847/ip2002.xlsx>,
  SHA-256 `7f01819893e04d08cbe82ac1e741fd7ac367f341fd4e603da59a5fe444311665`.
  Sheet "Figure 1", rows 5 and 6, volume (B-E) and price (F-I) changes of
  food, 2020-2023 (`food_volume`, `food_price`). Food fell 3.1% in volume
  in 2023 in this vintage; the accounts of May 2026 (Insee Première
  n° 2105) give -3.3%.
- `raw/insee_ip2110.xlsx`: Insee Première n° 2110, "La consommation des
  ménages en 2025", 16 June 2026,
  <https://www.insee.fr/fr/statistiques/fichier/9006503/IP2110.xlsx>,
  SHA-256 `b4c85a192100babd650014abdb1a59e16dda1914ae332ce1a224ab512c8358f2`.
  Sheet "Figure 1", rows 5 and 6: volume changes of 2024 and 2025 (B, C),
  from 2019 to 2025 (D, `food_volume_since_2019`), price changes (F, G).

## Transcribed values

`raw/transcribed.csv` holds the values printed only in a PDF or a web
page, one row each: indicator, group, year, value, flag, the document, the
page and the line of its text the value was read from (`pdftotext -layout`,
spaces collapsed). `extract.exe` refuses a row whose line does not hold its
value. The documents are not kept:

| Document | URL | Date | SHA-256 | Read |
|---|---|---|---|---|
| `insee_ir_ipc_2025.pdf`, Informations rapides n° 8, "En 2025, l'inflation ralentit nettement" | <https://www.insee.fr/fr/statistiques/fichier/version-html/8726461/IR08_IPCan-2025.pdf> | 15 Jan 2026 | `f3c44fbb6a31b7959779879a52b0455cdf28e028ba47c0cca287233f3a2320fe` | p. 3, the yearly CPI by household category, 2023-2025 (`cpi_category`) |
| `insee_ir_ipc_2023.pdf`, Informations rapides n° 7 | <https://www.insee.fr/fr/statistiques/fichier/version-html/7750173/IR7-IPC.Annuel-2023.pdf> | 12 Jan 2024 (erratum Dec 2024) | `c06872f88bf82f2e8bed990fabfaaa3bc085b493dcd8bf5c0bb1e88ecfb87952` | p. 3, the same table, 2021 and 2022; its 2023 column equals the 2025 release's in every row |
| `insee_ndc_2022_06_eclairage_2.pdf`, Note de conjoncture, June 2022, éclairage on inflation by household category | <https://www.insee.fr/fr/statistiques/fichier/6464607/ndc-juin-2022-eclairage-2.pdf> | June 2022 | `a8836bbf499aecac74de9327d5a3316840d46d4c1c5356142704780f82256e27` | p. 26, figure 3: average inflation over the year to April 2022 (`cpi_yoy_2022_04`) and each category's gap (`cpi_gap_2022_04`) |
| `ofce_pbrief_112.pdf`, OFCE Policy brief n° 112 (Madec, Plane, Sampognaro) | <https://www.ofce.sciences-po.fr/pdf/pbrief/2023/OFCEpbrief112.pdf> | 22 Feb 2023 | `fb0b2e6f967c6e709308d727855c8b3bc9ebb7140a8ae2f94d53a7ec6b372899` | p. 15, tableau 6: 2022 inflation within each tenth, first quartile, median, third quartile (`ofce_inflation_2022`) |
| `insee_ndc_2026_09.pdf`, Note de conjoncture, "Vigilance orange sur la croissance" | <https://www.insee.fr/fr/statistiques/fichier/9050757/ndc-sept-2026.pdf> | 10 Sep 2026 | `e19538438c662703ce046933fe4266206e3a0ef0341c0c741e8fb910b2fd359d` | p. 69, table 8: purchasing power per consumption unit by quarter, 2024-2026, and by year; 2026-T3, 2026-T4 and 2026 are forecasts (flag `f`, `ndc_pp_uc`) |
| `insee_ecrt2026_e3.pdf`, the éclairage above | <https://www.insee.fr/fr/statistiques/fichier/version-html/8733117/ECRT2026-E3.pdf> | 2 Jul 2026 | `62cf5ac5be4328d62736be87fbf25e84830a8564a7def71e0ce8c8f3b652ed93` | p. 1: 121 of 343 professions, 37.8% of full-time equivalents (`pcs_count`) |
| `dgafp_point_indice.html`, DGAFP, "Connaître le point d'indice" | <https://www.fonction-publique.gouv.fr/etre-agent-public/ma-remuneration/connaitre-le-point-dindice> | updated 5 Dec 2023 | `c6c6f84fc139dace465924de9d73594c4c2ba7aa8c40c1811767eac2b6e2c9e2` | the table "Évolution de l'indice 100 depuis 2002", values from 2010 (`point_indice`, groups `from=YYYY-MM-DD`) |

Insee's documents are under Licence Ouverte 2.0 and the DGAFP's page under
the Licence Ouverte 2.0 (Etalab), as its footer says. The OFCE states no
licence: only its figures are taken, with the source named. Each document
was checked against a second download, except the OFCE's: its host,
www.ofce.fr, to which www.ofce.sciences-po.fr redirects, is refused by the
build's network, and the copy was saved by a fetch outside it. The DGAFP
page is rebuilt at each request, so its bytes differ between downloads
(the second copy's SHA-256 is
`0b7ca2688f98c02d6d1f24721515bb39be0818d2a3377e598c6e8f2d97b72afd`); its
table is identical in both. The DGAFP's note of 8 July 2026 on the "rendez-vous
salarial"
(<https://www.fonction-publique.gouv.fr/toutes-les-actualites/retour-sur-le-rendez-vous-salarial-de-la-fonction-publique>,
saved with SHA-256
`8b4e3cd0f86fd0b347b7b6bbe97fce3dda5d90b2f48f6247bc3ab83bdbed7e75`) says of
2027: "il n'est pas prévu de mesure générale sur le point d'indice, ni de
nouvelle mesure catégorielle", subject to the 2027 budget bill.

Insee's annual releases of 2017 to 2022 give the inflation of household
categories in their text only, without a table; the table by tenth starts
with the release of January 2024. The categories are tenths of living
standard ("tranche", "décile" in the text), on metropolitan France.

## Hosts and what was not fetched

- www.cgedd.fr (Friggit's xlsx variants `valeur-immobilier-france.xlsx`,
  the PDF of the presentation) and secure.globalsign.com are not reachable
  from the build's network; the IGEDD host serves the same workbook.
- On 2026-10-20 the build's network refused bdm.insee.fr and api.insee.fr
  (Insee's series database and its API: the monthly CPI, the CPI by
  household category before 2021 and its seasonally adjusted series),
  api-diffusion-catalogue-donnees-externe.insee.fr and
  catalogue-donnees.insee.fr (Insee's Melodi datasets of the CPI, among
  them "indices catégoriels"), api.db.nomics.world, www.ofce.fr and
  www.lemonde.fr. dares.travail-emploi.gouv.fr answers with a bot check
  only, so Dares' quarterly index of the base monthly wage (SMB) is not in
  this set; the SMIC workbook's SHBOE is the wage index read instead.
  www.legifrance.gouv.fr and www.economie.gouv.fr answer 403.
- Not in this set: Insee's ERF points of 1970-1990 (relative poverty before
  1996), the Enquêtes Logement (taux d'effort by quartile, 1996-2020), the
  Histoire de vie et Patrimoine waves as a series, life expectancy by
  living standard over time (Insee gives one period, 2012-2016), the CPI
  by tenth before 2021 and by month, and Dares' SMB. `data/trends.ml` and
  the answer say where these gaps leave the question open.
