How France works
The research tax credit, against a synthetic France
Whether the research tax credit, the largest tax expenditure, raised business research in France.
The research tax credit (CIR) returns to firms part of what they spend on research and development: in 2024. Every firm qualifies, so this page compares France with a synthetic France built from OECD countries that left their own incentives unchanged. From 2008 to 2012 French business R&D differed from it by , against had each extra euro of credit become research. Ranked among the donor countries, the data can reject neither one euro of research per euro of credit nor none. Firm-level studies for France Stratégie found about one euro per euro among firms already claiming the credit, significant for small and medium firms, not for larger ones.
Evidence
The research tax credit (CIR) costs a year. In 2008 it became a credit of 30% of all R&D spending up to €100m, and tax relief for business R&D rose from to . Firms are not assigned to it, they all qualify, so no French firm can serve as a control. The comparison is between countries: a synthetic France, the weighted average of OECD countries that did not change their own R&D tax incentives, chosen to track French business R&D from 1995 to 2007.
- France
- Synthetic France
- Euro for euro
Why believe these series: · ·
Show the data as a table
| Year | France | Synthetic France | Euro for euro |
|---|---|---|---|
| 1995 | 1.4% | 1.3% | – |
| 1996 | 1.4% | 1.3% | – |
| 1997 | 1.3% | 1.3% | – |
| 1998 | 1.3% | 1.3% | – |
| 1999 | 1.3% | 1.4% | – |
| 2000 | 1.3% | 1.4% | – |
| 2001 | 1.4% | 1.4% | – |
| 2002 | 1.4% | 1.3% | – |
| 2003 | 1.3% | 1.4% | – |
| 2004 | 1.3% | 1.3% | – |
| 2005 | 1.3% | 1.3% | – |
| 2006 | 1.3% | 1.3% | – |
| 2007 | 1.3% | 1.3% | 1.3% |
| 2008 | 1.3% | 1.4% | 1.5% |
| 2009 | 1.4% | 1.4% | 1.6% |
| 2010 | 1.4% | 1.3% | 1.5% |
| 2011 | 1.4% | 1.4% | 1.5% |
| 2012 | 1.4% | 1.4% | 1.5% |
| 2013 | 1.4% | 1.4% | 1.6% |
| 2014 | 1.4% | 1.4% | 1.6% |
| 2015 | 1.4% | 1.5% | 1.7% |
| 2016 | 1.4% | 1.6% | 1.8% |
| 2017 | 1.4% | 1.6% | 1.8% |
| 2018 | 1.4% | 1.6% | 1.8% |
| 2019 | 1.4% | 1.7% | 1.9% |
Synthetic France is Norway 50%, United States 40%, Germany 11%. From 2008 to 2012, France's business R&D differed from it by ; had every extra euro of credit become R&D (the "euro for euro" line), the gap would have been . By 2019 the gap was , as the countries that make up its synthetic counterpart raised their R&D faster. The same method given a fake start in 2001 finds , and of countries, each treated in turn as if it had been reformed, show a gap as large as France's.
The direct test is whether these data can tell one euro of R&D per euro of credit from none. Tested directly, by taking the euro-for-euro path off France's series and ranking what is left among the donor countries treated in turn, the one-for-one effect gets a p-value of , half a euro per euro . No test ranked among so few countries can give less than , so the aggregate data can reject neither one euro per euro nor none. The gap also depends on which countries make up the synthetic: leaving out one at a time moves it from to , and fitting from 2001, after two of the four breaks in France's series, gives .
Firm-level studies address this directly. For the evaluation commission at France Stratégie (CNEPI, 2021), the 2008 reform led firms that already claimed the credit to raise their R&D by about as much as the extra credit, roughly one euro per euro. The effects are positive and significant for small and medium firms, and not statistically significant for mid-sized and large firms, which receive most of the credit. Not significant does not mean no effect: the studies cannot say how much of large firms' credit paid for R&D they would have done anyway. The OECD work the commission cites finds €0.4 of R&D per euro of relief in firms of 250 employees or more, €1 in firms of 50 to 249 and over €1.4 below 50.
A sharper test would follow, firm by firm, the companies that declared more than €100 million of research in 2023, where each further euro earns a 5% credit instead of 30%: it needs the ministry's file of credit declarations (GECIR) and firms' tax returns, open to researchers only through the CASD after the comité du secret, so this site names the design rather than running it; it would show whether the largest spenders, whose response the published evaluations cannot tell from none, cut research at the margin or spread it across subsidiaries, which a synthetic France built from national totals cannot see.
Tax credits or programmes
Governments support business R&D in two ways: tax relief, which follows what firms choose to do, and direct funding, grants and contracts for work the state chooses, as the US defence agency DARPA does. Since 2007 France, like the United Kingdom and the United States, has moved towards tax relief; Japan, Korea and the Netherlands moved the other way, and Germany still gives almost all of its support directly.
- 2007
- Latest year (2022-2024)
France
United Kingdom
United States
Japan
Korea
Netherlands
Italy
Germany
Show the data as a table
| Group | 2007 | Latest year (2022-2024) |
|---|---|---|
| France | ||
| United Kingdom | ||
| United States | ||
| Japan | ||
| Korea | ||
| Netherlands | ||
| Italy | ||
| Germany |
Moretti, Steinwender and Van Reenen, using defence R&D as a source of public R&D that does not follow private plans, find that public R&D draws in private R&D:
a 10% increase in government-financed R&D generates a 5% to 6% additional increase in privately funded R&DThe Intellectual Spoils of War? (NBER w26483, REStat 2025)
By their estimates, public R&D spending « generates twice as much overall R&D compared to R&D tax credits ». Tax relief is of France's public support for business R&D.
OECD subsidy rates use headline rates and ignore the €100m ceiling. Business R&D in France has methodological breaks in 1997, 2001, 2004 and 2006. Donors are countries whose implied subsidy rates moved by less than 0.05 over 2003-2019; with all OECD countries as donors neither benchmark is rejected either. The tests follow Firpo and Possebom (2018). The CNEPI findings are from its 2021 opinion, pages 3, 12 and 13. The code is analysis/cir.ml and the results are in results.csv, benchmark.csv and loo.csv.
Its own aims
Measured against what it was meant to do
Each aim as the scheme's own texts state it, quoted, then every outcome this site measured against it, with the same design as above: the matched comparison, or the synthetic control. Positive, negative and null results are listed alike. Where no open dataset covers an aim, the dataset it would need is named, with why it is not open. Before them, the article of law that founds the scheme, quoted from Légifrance or the parliamentary record with the dates its wording was in force; after them, the indicators the State set for the programme, as the State reports them, with their source: this site shows them but has not checked them.
In the law, or the nearest official text
Each aim as the founding text states it, with the article and the dates that wording was in force. Where an article states no aim, a line says so, and the aim is quoted from the nearest official text that states one: the bill's explanatory memorandum or impact study, a government instruction, or the programme's performance report.
Code général des impôts, art. 244 quater B, en vigueur depuis le 01/01/2026 (modifié par loi n° 2026-103 du 19 février 2026, art. 42) states no aim: the article grants a credit on research spending. The article states no aim (no afin de, objectif or inciter anywhere in it); the aims come from art. 67 VII of the loi de finances pour 1983, the loi n° 82-610 it cites, the exposé des motifs of the 2008 reform and the Voies et moyens fiche.
Count the credit towards the national R&D effort set by the 1982 research programming law
« VII - Le montant annuel des pertes de recettes résultant pour le Trésor public des dispositions ci-dessus sera pris en compte chaque année comme une composante de l'effort budgétaire à consentir, tel qu'il est prévu à l'article 2 de la loi d'orientation et de programmation pour la recherche et le développement technologique de la France (n° 82-610 du 15 juillet 1982). »Loi n° 82-1126 du 29 décembre 1982 de finances pour 1983, art. 67 VII, en vigueur du 01/01/1983 au 01/09/2007 (périmé)
The founding article states no aim of its own and ties the credit to the R&D target of loi n° 82-610, art. 2. The credit was a Government amendment (art. 52 bis nouveau at the Assemblée), so no exposé des motifs exists for it.
Raise R&D spending to 2.5% of GDP by 1985
« Pour atteindre l'objectif retenu par le plan intérimaire tendant à porter à 2,5 % en 1985 la part du produit intérieur brut consacrée aux dépenses de recherche et de développement technologique, les crédits inscrits au budget civil de recherche et de développement technologique progresseront à un rythme moyen annuel de 17,8 % en volume d'ici 1985, et les effectifs employés dans la recherche publique croîtront au rythme moyen annuel de 4,5 %. »Loi n° 82-610 du 15 juillet 1982 d'orientation et de programmation pour la recherche et le développement technologique de la France, art. 2, en vigueur depuis le 16/07/1982; abrogée par l'ordonnance n° 2004-545 du 11 juin 2004, art. 6, avec effet différé (art. 7 II)
The orientation law that the founding article 67 VII of the CIR cites by article number.
Reward firms that raise their research effort
« un amendement à la loi de finances prévoit la mise en place d'un crédit d'impôts en faveur des entreprises développant leur effort de recherche scientifique et technologique. »Sénat, avis n° 96 (1982-1983), tome VII « Recherche scientifique et technique » (M. Adrien Gouteyron), audition du ministre de la Recherche et de l'Industrie, 27 octobre 1982, compte rendu de l'audition, 22 novembre 1982
Only contemporary statement of purpose found for the 1983 creation: the minister's words as reported by a Senate committee, not a legal text.
Raise firms' research spending (Lisbon strategy)
« Cette réforme, qui s’inscrit dans le cadre de la stratégie de Lisbonne, devrait inciter les entreprises à accroître leurs dépenses de recherche. »Projet de loi de finances pour 2008, n° 189 (Assemblée nationale, XIIIe législature), article 39, adopté comme article 69 de la loi n° 2007-1822 du 24 décembre 2007, exposé des motifs de l'article 39, déposé le 26 septembre 2007
Art. 69 of the loi de finances pour 2008 (JORFARTI000017853890) states no aim; the 2008 reform that made the credit volume-based states its aim in the bill's exposé des motifs.
Innovation, and the location in France of innovative firms and researchers
« Cette disposition favorise l'innovation dans les entreprises françaises et la localisation en France des entreprises innovantes et des chercheurs. »Projet de loi de finances pour 2008, n° 189 (Assemblée nationale, XIIIe législature), exposé général des motifs, déposé le 26 septembre 2007
Same reform; the sentence before it reads « Afin d’encourager la contribution des entreprises à l’amélioration de la croissance française de long terme, une majoration significative du crédit d’impôt recherche est proposée. »
Raise research spending
« Finalité : Augmenter les dépenses de recherche »Évaluation des voies et moyens, annexe au PLF 2026, tome II « Dépenses fiscales », fiche 200302, PLF 2026 (octobre 2025)
The State's official purpose for tax expenditure 200302, which names programme 172 and its indicator P172-2.2 (PDF p. 44).
In the programme's texts
Raise firms' research and development
« pour inciter les entreprises à accroître leur effort de recherche et développement »Ministère de l'Enseignement supérieur et de la Recherche, Bilan du crédit d'impôt recherche au titre de l'année 2006, Introduction, partie 1, p. 1 (copie : data/literature/aims/mesr-cir-bilan-2006-p1-2.txt), 2007
- Business R&D, % of GDP: (rank p-value 0.46).
- Patent families invented in France: (rank p-value 0.41).
Raise the employment of researchers
« un processus d’évaluation de son impact sur les dépenses de R&D et l’emploi des chercheurs par les entreprises »Ministère de l'Enseignement supérieur et de la Recherche, Bilan du crédit d'impôt recherche au titre de l'année 2006, Introduction, p. 1 (copie : data/literature/aims/mesr-cir-bilan-2006-p1-2.txt), 2007
- Researchers per 1,000 in employment: (rank p-value 0.10).
The State's own indicators
The indicators the State reports for the programme, with the latest value and the target as it prints them, the report and page they come from, and the open data that could measure each at this site's unit.
| Indicator | Latest value | Target | This site's measure | Source | Open data that could measure it |
|---|---|---|---|---|---|
| P172-2-2, programme 172Mesures de l'impact du crédit d'impôt recherche (CIR) : Dépenses de R&D privée supplémentaires par euro de crédit d'impôt recherche | 12025 | >12025 | € of business R&D per € of tax support2008-2012 | RAP 2025 (PLRG 2025), data.economie.gouv.frRAP 2025 dataset, programme 172, indicator P172-2-2, sous-indicateur « Dépenses de R&D privée supplémentaires par euro de crédit d'impôt recherche »; cible 2026 (PAP 2026, carried in the RAP 2025 dataset): >1 | OECD MSTI (business R&D, BERD) and OECD RDTAX (tax support for BERD) by country, which this site already reads: the additional BERD per euro of tax support can be estimated at the country level, as this site's synthetic control does; the RAP reports 1 for every year from 2022 to 2025, which reads as a fixed estimate rather than a yearly measurement. |
| P172-2-2, programme 172Mesures de l'impact du crédit d'impôt recherche (CIR) : Indice de rotation des entreprises bénéficiant du CIR | 18,42025 | 202025 | not measured | RAP 2025 (PLRG 2025), data.economie.gouv.frRAP 2025 dataset, programme 172, indicator P172-2-2, sous-indicateur « Indice de rotation des entreprises bénéficiant du CIR »; cible 2026 (PAP 2026, carried in the RAP 2025 dataset): 20 | No open data measures it: it needs the list of firms claiming the CIR each year (the ministry's GECIR files), which is not published; no country-level counterpart exists. |
| MRA-4-1, Mission Recherche et enseignement supérieurEffort de la recherche de la France : Part de la dépense intérieure de R&D (DIRD) dans le produit intérieur brut (PIB) | 2,182024 | 2,52025 | % of GDP2024 | RAP 2025 (PLRG 2025), data.economie.gouv.fr, indicateur de missionRAP 2025 dataset, mission record MRA-4-1, sous-indicateur « Part de la dépense intérieure de R&D (DIRD) dans le produit intérieur brut (PIB) »; objectif MRA-4 « Améliorer la part de la R&D des entreprises françaises dans l'effort national de recherche »; exec 2025: non déterminé; cible 2026 (PAP 2026, carried in the RAP 2025 dataset): 2,3 | OECD MSTI (GERD as a share of GDP) and Eurostat rd_e_gerdtot, by country, which this site already reads. |
| MRA-4-1, Mission Recherche et enseignement supérieurEffort de la recherche de la France : Part des dépenses intérieures en R&D des entreprises (DIRDE) dans la dépense intérieure de R&D (DIRD) | 662024 | 66,32025 | % of GERD2024 | RAP 2025 (PLRG 2025), data.economie.gouv.fr, indicateur de missionRAP 2025 dataset, mission record MRA-4-1, sous-indicateur « Part des dépenses intérieures en R&D des entreprises (DIRDE) dans la dépense intérieure de R&D (DIRD) »; exec 2025: non déterminé; cible 2026 (PAP 2026, carried in the RAP 2025 dataset): 66,4 | OECD MSTI (BERD and GERD), by country, which this site already reads. |
The State reports one euro of additional private R&D per euro of research tax credit (P172-2-2) in every year from 2022 to 2025: a fixed estimate carried from one report to the next, not a measurement.
This site's synthetic control finds euro of business R&D per euro of tax support over 2008-2012, the first years of the 2008 reform, at the country level; the one-sided p-value against one euro per euro is , and no test ranked among so few countries can give less than .
The census outcomes compare 2021 with 2015, so they serve only designs that start in 2015-2017, and no placebo can be run on them; land taken comes from CEREMA's land-registry series, 2011-2024. The aims and their sources are in aims.csv, the texts of law in law.csv, the State's indicators in indicators.csv, and this site's measures of them in measured.csv and, year by year, measured_series.csv.