Official figures and this site's calculations
How much public money goes to companies, and what is known of its effect?
How much public support companies receive under each official count, who receives it, how long each tax break lasts, and what is known of its effect.
- of public support to companies in 2023 by the Senate's broad count, by the narrowest official one.
- of the research tax credit went to large firms in 2023.
- of the 2024 cost of tax expenditures comes from measures with no end date.
Estimated: a model, an imputation or a surveyUnmarked figures are published as observed.
Companies
How much public money goes to companies
It depends on what counts as support. Five official estimates for 2023, from the Senate and from the Haut-Commissariat à la Stratégie et au Plan (HCSP), formerly France Stratégie, run from to . They differ on whether to count low-wage contribution cuts, reduced VAT rates and repayable loans.
- HCSP, perimeter 1Direct and targeted aid only
- Senate, strictWithout Bpifrance and VAT
- HCSP, 2025 noteWith VAT, without contribution cuts
- HCSP, perimeter 2With contribution cuts and VAT
- Senate, broadLoans and guarantees at face value
Show the data as a table
| Item | Amount | Year |
|---|---|---|
| HCSP, perimeter 1 | €82bn | 2023 |
| Senate, strict | €108bn | 2023 |
| HCSP, 2025 note | €111.9bn | 2023 |
| HCSP, perimeter 2 | €187bn | 2023 |
| Senate, broad | €211bn | 2023 |
The €108bn often quoted as a “Sénat/HCSP” figure is the Senate's strict estimate alone. The HCSP's own number for 2023, , is close by coincidence: it leaves out the contribution cuts the Senate keeps, and counts VAT the Senate leaves out.
What the Senate counted
- Tax expenditures41.7%
- Contribution cuts35.5%
- Bpifrance loans and equity19.4%
- Subsidies3.3%
Show the data as a table
| Component | Amount | Share |
|---|---|---|
| Tax expenditures | €88bn | 41.7% |
| Contribution cuts | €75bn | 35.5% |
| Bpifrance loans and equity | €41bn | 19.4% |
| Subsidies | €7bn | 3.3% |
| Total | €211bn | 100% |
The Senate's contribution cuts, for 2023, are its own count. The general reductions alone cost in 2024 across all compulsory schemes, of it in the basic social-security schemes, by the social security budget's annex on exemptions, which the register of public money on the evaluations page also reads.
The largest schemes
- General contribution reductionsAll compulsory schemes
- Reduced VAT ratesCost to the State after transfers
- Research tax credit (CIR)Largest tax expenditure
- Hiring aid for apprenticesEstimate of the Observatoire français des conjonctures économiques (OFCE)
Show the data as a table
| Item | Amount | Year |
|---|---|---|
| General contribution reductions | €77.6bn | 2024 |
| Reduced VAT rates | €10.9bn | 2024 |
| Research tax credit (CIR) | €7.8bn | 2024 |
| Hiring aid for apprentices | €4.1bn | 2024 |
How much of it changes what firms do?
No evaluation covers the total, and the HCSP writes that most schemes have never been evaluated. The broadest count, the Senate's, includes of Bpifrance loans, guarantees and equity at face value, not at their cost; its rapporteur calls it a floor because regional, local and EU aid are left out.
Companies
Follow one company's money
The Senate inquiry made the largest groups state, under oath, the public support they received in 2023. Three of the groups whose executives it heard are shown here as examples, not as a sample: those for which both a total stated at the hearing and the group's annual filing for 2024 were read for this page. They are ordered by the support they received, beside what they paid their shareholders the next year:
- Public support, 2023
- Paid to shareholders, 2024
LVMH
Dividends , buybacks . Taxes paid in France €3.8bn.
Sanofi
Dividends , buybacks . French R&D €2.5bn.
Michelin
Dividends , buybacks . French R&D budget €400m.
Show the data as a table
| Group | Public support, 2023 | Paid to shareholders, 2024 |
|---|---|---|
| LVMH | ||
| Sanofi | ||
| Michelin |
Public support received in 2023, beside what was paid to shareholders in 2024: different flows and different years, not meant to be set against each other.
Much of the support rewards activity in France that dividends do not measure. The Senate report's own figures contain errors (dollars reported as euros, dividends that include buybacks), so the payouts here come from the filings. Most of the filings read for this page do not disclose the research tax credit at all.
The research tax credit of fifteen groups
Annex 12 of the Senate report sets out, for fifteen of the groups heard, the research tax credit received and the spending on research and development that their executives gave at the hearings, for 2023:
| Group | Research tax credit | R&D spending | Credit as a share of R&D |
|---|---|---|---|
| Michelin | |||
| Renault | |||
| ArcelorMittal | |||
| Sanofi | |||
| Stellantis | |||
| Safran | |||
| STMicroelectronics | |||
| Vinci | |||
| Air Liquide | |||
| EDF | |||
| Orange | |||
| Thales | |||
| Danone | |||
| SNCF | |||
| Accor |
The annex is a selection of the figures executives gave in public, which the report calls a sample of them; it is not a sample of the firms that claim the credit. The amounts are the groups' own, not the tax administration's, and the annex does not say whether the R&D is spent in France or worldwide: Michelin's and Sanofi's executives gave theirs for France. The table is in tome I, p. 430, and the report introduces it on p. 251.
Evidence
How long tax expenditures last
France has tax expenditures, taxes it chooses not to collect, costing in 2024. of that cost comes from measures created before 2000, and from measures with no end date: they last until a law repeals them; the evaluation rule applies only to measures created since 2024.
- Before 1950
- 1950s
- 1960s
- 1970s
- 1980s
- 1990s
- 2000s
- 2010s
- 2020s
Show the data as a table
| Item | Amount | Year |
|---|---|---|
| Before 1950 | €7bn | 2024 |
| 1950s | €1.1bn | 2024 |
| 1960s | €3.6bn | 2024 |
| 1970s | €11.6bn | 2024 |
| 1980s | €14.3bn | 2024 |
| 1990s | €11.9bn | 2024 |
| 2000s | €23.3bn | 2024 |
| 2010s | €11.6bn | 2024 |
| 2020s | €4.9bn | 2024 |
The five largest
- Research tax creditCreated 1982, no end date
- Tax credit for employing help at homeCreated 2006, no end date
- Pacte DutreilCreated 1999, no end date
- 10% allowance on pensionsCreated 1977, no end date
- Profit-sharing and employee savings exemptionCreated 1973, no end date
Show the data as a table
| Item | Amount | Year |
|---|---|---|
| Research tax credit | €7.8bn | 2024 |
| Tax credit for employing help at home | €6.4bn | 2024 |
| Pacte Dutreil | €5bn | 2024 |
| 10% allowance on pensions | €4.8bn | 2024 |
| Profit-sharing and employee savings exemption | €2.9bn | 2024 |
The law already says they should expire
Since 2024, the public-finance programming law limits every new tax expenditure to three years, renewable only after an evaluation:
Les dépenses fiscales instituées par une loi promulguée après le 1er janvier 2024 sont applicables pour une durée qui est précisée par la loi les instituant et qui ne peut excéder trois ans. Les dépenses fiscales ne peuvent être prorogées que pour une période maximale de trois ans et à la condition d'avoir fait l'objet d'une évaluation, présentée par le Gouvernement au Parlement, des principales caractéristiques des bénéficiaires des mesures, qui précise l'efficacité et le coût de celles-ci.Loi n° 2023-1195, article 7
The Cour des comptes finds the rule has not been applied. Under the previous law's four-year limit, of the 82 measures created in 2018-2022 had no end date. Of the 35 evaluations the government scheduled for 2020-2023, were carried out; of the 11 the current law requires, by the end of 2025. The Cour des comptes:
La LPFP 2023-2027 a pour l'heure échoué, comme les lois de programmation qui l'ont précédée, à imposer de manière effective un plafonnement du coût et de la durée d'application à trois ans des nouvelles dépenses fiscales.
Every measure, with its creation year, end date, 2024 cost and beneficiaries, is in tax-expenditures.csv, parsed from the budget annex.
Evidence
Is this public euro working?
The same six questions for each large scheme, answered only where an evaluation answers them. Click a cell for the finding, the quote and the report. A question mark means no adequate evaluation was found.
| Scheme | Cost | Additionality | Permanence | Leakage | Spillovers | Who gets it | Strategic value |
|---|---|---|---|---|---|---|---|
| Contribution reductions on low wages | 2024 | ||||||
| Research tax credit (CIR) | 2024 | ||||||
| Hiring aid for apprentices | 2024 | ||||||
| CICE (2013-2018) | 2016 |
- Positive
- Mixed
- Negative
- Suspected: raised by evaluators, not measured
- Measured: a fact about who receives it
- Not evaluated: no adequate evaluation found
Contribution reductions on low wages
- Evidence quality
- Medium: strong evidence for the 1990s; current effects rest on elasticities chosen for macro models.
- What would change the assessment
- An evaluation of the 2026 single reduction, firm-level incidence estimates (wages, margins, prices), and wage-mobility studies around the thresholds.
Research tax credit (CIR)
- Evidence quality
- Medium: a quasi-experimental study of the 2008 reform, covering 2008-2012 only; the credit is universal, which makes a counterfactual hard to build.
- What would change the assessment
- A causal evaluation after 2012 and of the 2025 reform, group-level data linking the credit to where R&D is located, and a measured spillover estimate.
Hiring aid for apprentices
- Evidence quality
- Low for the aid itself: descriptive trends and a model estimate.
- What would change the assessment
- A causal evaluation using the 2023, 2025 and 2026 cuts by diploma level and firm size, and a cost per additional apprentice.
CICE (2013-2018)
- Evidence quality
- High for the firm-level job estimate (two independent teams, a public committee); low for the macro estimates.
- What would change the assessment
- An evaluation of the 2019 conversion beyond 2020, with group-level data on wages, margins and dividends.
Additionality
Contribution reductions on low wages
Does it cause what would not have happened anyway?
Mixed
Strong job effects in the 1990s; smaller today, and no counterfactual exists for the current scheme.
La politique de réduction des cotisations employeur sur les bas salaires a eu un impact significatif sur l'emploi lors de sa mise en place dans les années 1990. [...] l'amélioration de la situation du marché du travail et le niveau important des réductions actuelles de cotisations employeur rendent son efficacité en termes de création d'emploi moindre que dans les années 1990.
Permanence
Contribution reductions on low wages
Do the effects last?
Mixed
Jobs depend on the reduction continuing; long-run effects such as low-wage traps cannot be measured with current methods.
La difficulté de mettre en évidence l'impact causal de la pente des allègements sur l'existence de trappes à bas salaires provient de son caractère graduel dont les effets s'accumulent décennie après décennie ce qui rend les méthodologies impuissantes à les mesurer.
Source: Cour des comptes, RALFSS 2025, ch. III, allègements généraux de cotisations sociales (2025)
Leakage
Contribution reductions on low wages
Does the money end up with those it does not target: margins, dividends, wages abroad or prices?
Not evaluated
No study measures how much passed into margins, dividends or prices.
No source found.
Spillovers
Contribution reductions on low wages
Does it benefit others beyond the recipient?
Suspected
Possible over-specialisation in low-wage sectors, raised by the evaluators and not measured.
les politiques d'allègements de cotisation peuvent avoir conduit à une sur-spécialisation dans certains types d'emplois ou de secteurs à bas salaires, au détriment d'autres à plus forte valeur ajoutée voire exportateurs qui auraient ainsi pu être pénalisés. [...] Au cours de nos discussions, nous avons été frappés par un déficit d'études sur ces points importants.
Who gets it
Contribution reductions on low wages
Which firms receive it?
Measured
Small firms get the highest rate; firms of 250+ employees receive a third of the total.
en montant, les entreprises de 250 salariés et plus perçoivent tout de même un tiers des cotisations exonérées (pour 44,7 % de la masse salariale).
Strategic value
Contribution reductions on low wages
Does it build or keep a capability France needs?
Not evaluated
No evaluation of what capability it builds or keeps.
No source found.
Additionality
Research tax credit (CIR)
Does it cause what would not have happened anyway?
Mixed
It raises R&D in small firms; no significant effect is found for large and mid-sized ones.
Le CIR est un dispositif qui stimule les activités de R & D des entreprises, mais ses effets sont perceptibles surtout sur les PME. En ce qui concerne les entreprises entrées dans le dispositif avant 2008, les travaux sur la période 2008-2012 n'identifient pas d'impact significatif du CIR pour les grandes entreprises et les ETI.
Source: France Stratégie, Évaluation du crédit d'impôt recherche, avis de la CNEPI (2021)
Permanence
Research tax credit (CIR)
Do the effects last?
Not evaluated
Causal estimates cover 2008-2012 only; no study measures whether effects persist.
Toutefois, l'étude IPP-PSE estime que ces gains sont caractéristiques d'entreprises très jeunes et, ne pouvant mener une analyse causale, elle ne peut confirmer que ces gains soient imputables au CIR.
Source: France Stratégie, Évaluation du crédit d'impôt recherche, avis de la CNEPI (2021)
Leakage
Research tax credit (CIR)
Does the money end up with those it does not target: margins, dividends, wages abroad or prices?
Mixed
No significant effect on patents or value added; for larger firms it mainly eased financing.
l'accès à un CIR plus généreux après 2008 semble donc surtout avoir desserré les contraintes financières des bénéficiaires dans une période de crise économique, leur permettant ainsi de développer leur activité sans nécessairement privilégier une stratégie d'innovation
Source: France Stratégie, Évaluation du crédit d'impôt recherche, avis de la CNEPI (2021)
Spillovers
Research tax credit (CIR)
Does it benefit others beyond the recipient?
Not evaluated
Knowledge spillovers are the rationale and are assumed in the macro model used, but never measured for the CIR.
Source: France Stratégie, Évaluation du crédit d'impôt recherche, avis de la CNEPI (2021)
Who gets it
Research tax credit (CIR)
Which firms receive it?
Measured
Large firms, 3.5% of beneficiaries, receive 41% of the research credit.
Elles bénéficient d'une créance de 3,0 Md€ qui représentent 41 % de la créance au titre de la recherche en 2023.
Source: MESR, Le crédit d'impôt recherche (CIR) en 2023 (2025)
Strategic value
Research tax credit (CIR)
Does it build or keep a capability France needs?
Negative
It did little to keep multinationals' R&D in France.
le CIR n'a guère contrecarré la détérioration de l'attractivité du site France pour la localisation de la R & D des entreprises multinationales
Source: France Stratégie, Évaluation du crédit d'impôt recherche, avis de la CNEPI (2021)
Additionality
Hiring aid for apprentices
Does it cause what would not have happened anyway?
Mixed
Apprenticeships surged in 2020-2022 after the aid widened; no cost-benefit analysis has been published.
aucune analyse coûts-bénéfices des aides à l'embauche d'apprentis n'a été publiée à ce jour par le ministère, malgré leur coût pour les finances publiques.
Permanence
Hiring aid for apprentices
Do the effects last?
Mixed
Helps early careers, strongly at vocational level; long-run effects are not established.
l'apprentissage pourrait favoriser l'emploi en début de carrière, mais le faible nombre d'études scientifiques sur le sujet ne permet pas d'être conclusif quant à ses effets à plus long terme.
Source: DG Trésor, Trésor-Éco n° 376, « L'apprentissage en France : quel bilan ? » (2025)
Leakage
Hiring aid for apprentices
Does the money end up with those it does not target: margins, dividends, wages abroad or prices?
Not evaluated
No study of whether the aid passed into pay, margins or training prices.
No source found.
Spillovers
Hiring aid for apprentices
Does it benefit others beyond the recipient?
Not evaluated
No study of skills spillovers or displaced hires.
No source found.
Who gets it
Hiring aid for apprentices
Which firms receive it?
Negative
Most of it goes to the young people who, on average, have the least difficulty finding work.
La majorité du financement public de l'apprentissage bénéficie aujourd'hui aux jeunes rencontrant en moyenne le moins de difficultés d'accès à l'emploi.
Strategic value
Hiring aid for apprentices
Does it build or keep a capability France needs?
Not evaluated
No evaluation of strategic value.
No source found.
Additionality
CICE (2013-2018)
Does it cause what would not have happened anyway?
Mixed
About 100,000 jobs, which its evaluators call small for the cost.
L'effet total reste estimé à 100 000 emplois environ, ce qui est faible, rapporté au coût du CICE – de l'ordre de 18 milliards d'euros en 2016.
Source: France Stratégie, Évaluation du CICE. Synthèse des travaux d'approfondissement (2020)
Permanence
CICE (2013-2018)
Do the effects last?
Not evaluated
The evaluation of the credit stops at 2017. The IPP's ex post evaluation of its 2019 conversion into a contribution cut detects no effect on employment, value added, turnover or investment in 2019 and 2020; whether the credit's own effects lasted is not measured.
aucun effet n'est détecté sur les variables d'emploi, ni de valeur ajoutée, de chiffre d'affaires ou d'investissements, pour les années 2019 et 2020
Leakage
CICE (2013-2018)
Does the money end up with those it does not target: margins, dividends, wages abroad or prices?
Not evaluated
No study measures how much went into margins, dividends or prices; the share passed into wages is under “Who gets it”.
No source found.
Spillovers
CICE (2013-2018)
Does it benefit others beyond the recipient?
Not evaluated
No effect on exports detected; no spillover study.
Malgouyres et Mayer (2018) réalisent une telle estimation pour le CICE mais ne détectent aucun effet significatif sur les exportations jusqu'en 2015.
Who gets it
CICE (2013-2018)
Which firms receive it?
Measured
Evaluators estimate at least half was passed into wages, to the employees of the firms the credit targets: who ends up with it, recorded as a fact, not as leakage.
L'incidence du CICE est ainsi estimée avoir été transmise dans les salaires pour au moins 50 %.
Strategic value
CICE (2013-2018)
Does it build or keep a capability France needs?
Not evaluated
No evaluation of strategic value.
No source found.
Large firms receive of the research tax credit, and firms of 250 employees or more of exempted contributions. No source breaks the whole of company support down by firm size.
Every line of public money, and the studies this site runs itself, are on Does it work?.