France’s €211bn business-aid figure distorts the budget debate

On 10 September 2026, the French newspaper L’Humanité presented €211 billion as money the state pays businesses each year without receiving anything in return. The article introduces a forthcoming book by Fabien Gay, the Senate inquiry’s rapporteur and the newspaper’s director. ^[1]

The figure comes from the Senate’s broad estimate of business support for 2023. But its components include repayable finance, guarantees, tax measures, social-security contribution relief and grants. The same report also gives a narrower estimate of €108 billion. Neither total measures money simply given away by central government. ^[2]

The €211 billion figure combines forms of public support measured in different ways. It estimates neither a consistent net public cost nor the savings from withdrawing the programmes. Those distinctions matter when the number enters a budget debate.

France’s next presidential election is due in 2027. In that setting, presenting the aggregate as an annual giveaway risks turning an inventory into an apparent source of funding. A candidate may reasonably propose ending a tax break or cutting a grant. The proposal should identify the measure, its net fiscal yield and its consequences. A large, undefined total avoids those choices rather than settling them. ^[13]

Four categories behind the total

The Senate’s arithmetic is straightforward: €41 billion of Bpifrance interventions, €88 billion of broadly defined tax expenditure, €75 billion of social-security contribution relief and €7 billion in selected grants outside Bpifrance and excluding public-service compensation. These figures are rounded to the nearest billion. ^[2]

They add up. Their meanings do not line up so neatly. A loan normally creates a claim for repayment. A non-repayable grant is a different kind of transaction, subject to any clawback provisions. Tax relief measures a departure from a chosen tax benchmark. A guarantee concerns a contingent risk.

The graphic keeps the four categories apart. Expressing them all in euros does not make them interchangeable measures of taxpayer cost.

The Senate’s own aggregation raises a measurement problem. Describing it as an annual cost assumes a comparability that has not been established. A broad inventory can identify the reach of public intervention. It cannot turn the face value of financing into a fiscal expense simply by including it in a larger total. ^[2][4]

The Senate also points to omissions, including local and European support. Those gaps justify further work. They do not establish €211 billion as a lower bound for net public cost. Incomplete coverage and incompatible valuation are different problems; adding missing items does not resolve the latter. ^[2]

Public lending creates an asset as well as a cash outflow

When a public bank lends money, cash leaves its account. It also acquires a claim. An equity investment buys securities. A guarantee requires a separate assessment of the exposure, fees, calls and recoveries. Financing mobilised, the benefit to the recipient and the eventual public loss are distinct quantities. ^[4]

Bpifrance’s own activity report makes this concrete. For 2023, it records €9.24 billion in short-term credit authorisations. That does not establish how much borrowers actually drew. Its guarantee business reports €8.65 billion of activity, while separately identifying €4.33 billion of risk taken. Neither number is a realised loss to the public sector. ^[3]

These figures illustrate the underlying data; they are not a certified line-by-line reconstruction of the Senate’s €41 billion. The detailed reconciliation has not been obtained. That block also includes genuine grants, notably for innovation. Describing all €41 billion as repayable would therefore be another error. ^[2][3]

Consider a wholly hypothetical one-year loan of €100 million, with principal and interest paid at maturity. The public lender charges 3%, against a 5% benchmark assumed appropriate for a comparable loan. The borrower pays €3 million in interest instead of €5 million. The benefit is €2 million at maturity, worth roughly €1.90 million at inception when discounted at 5%. It is not a €100 million grant.

This is the idea behind a grant equivalent: putting a monetary value on the advantage provided by favourable terms. The European Commission expressly distinguishes the aid element from the nominal amount granted. Its legal state-aid perimeter is not, however, an exhaustive definition of all economic support in France. ^[4]

The illustration does not estimate a lender’s expected losses or the fiscal cost of an actual programme. It assumes repayment and leaves fees aside. In a real assessment, the benchmark must reflect credit risk and security, among other factors. Public lending can be costly even when repaid; its profitability also depends on pricing, risk and costs. Counting the entire principal as a gift prevents that assessment from even beginning.

Tax expenditure needs a benchmark

The €88 billion tax block raises a different issue. The Senate includes both officially recognised tax expenditures and provisions that the administration has removed from that category. Some of these reclassified provisions are treated as part of the ordinary rules for calculating tax. ^[2][5]

France’s tax-consolidation regime is a useful example. The tax authority explains that a group’s taxable result combines its companies’ results with the required adjustments. Losses at one company may therefore enter the calculation alongside profits elsewhere in the group. The apparent concession depends on the alternative tax system chosen for comparison. ^[6]

A different benchmark can be defended. Taxing entities without the group rules, or changing the taxation of distributions within a group, would be a fiscal policy choice. But the difference from that hypothetical system is not money the government first collected and then handed back.

The converse matters too. Tax relief can confer a real advantage and reduce revenue even when it is not delivered as a grant. Some tax credits are refundable. The analysis needs to show how each instrument reduces revenue or creates expenditure. ^[5]

Nor does the identity of the taxpayer establish who ultimately benefits. Reduced VAT can lower consumer prices, raise margins or produce a mixture of effects. Research by France’s Institute for Public Policy on the 2009 restaurant VAT cut found that owners captured a substantial part of the benefit. That challenges the claim that reduced VAT necessarily goes to consumers. It does not justify attributing every reduced VAT rate in 2023 entirely to businesses. ^[7]

How the narrower estimate reaches €108 billion

The Senate’s narrower calculation retains the €75 billion social-contribution block and the €7 billion grant block. It removes Bpifrance interventions and reduces the tax component from €88 billion to €26 billion, excluding reclassified provisions and VAT measures together. The bridge is 211 − 41 − 62 = 108. ^[2]

That subtraction explains a classification change. It does not measure a saving. The €62 billion difference is not a demonstrated double count or a fabricated sum. Removing all Bpifrance interventions also removes the grants included within them. ^[2]

Replacing the original claim with a supposedly definitive €108 billion cost would therefore be premature. First decide what is being measured: budget spending, recipient benefit, financing volume or public exposure. Then value the relevant operations on a consistent basis.

Even a well-constructed cost figure cannot establish whether a policy works. Support can be expensive and worthwhile, inexpensive and ineffective, or pay firms for investments they would have made anyway. The size of a programme does not supply its counterfactual.

Contribution relief deserves a real debate

The €75 billion block should not vanish merely because contributions are forgone rather than collected and refunded. Relative to a contribution schedule, relief changes both the levy and collective financing. Its effectiveness is a legitimate subject of debate. The employer receiving relief directly must still be distinguished from its ultimate distribution through profits, wages, prices or employment. ^[2][12]

In his 1 June 2026 submission to the consultation led by the Haut-commissariat à la stratégie et au plan, France’s public policy analysis body, Fabien Gay explicitly argues for including general contribution relief. He has a valid point here: a decision to forgo a levy can constitute public support even when no cheque changes hands. ^[8]

The same submission nevertheless connects repayable loans and advances with terms more favourable than the market’s. It also distinguishes eligibility rules from additional conditions. Following tax and social support does not require flattening every financial instrument and every obligation into a single category. ^[8]

The criticism of €211 billion consequently does not establish that all contribution relief should be retained. It establishes that an ill-defined aggregate cannot tell us the revenue from removing it.

Uneven checks and evaluation

The accessible opening of L’Humanité also presents a general absence of conditions and evaluation. That wording erases important distinctions. ^[1]

The former CICE competitiveness and employment tax credit was evaluated, including in a France Stratégie synthesis published in September 2020. The research tax credit, or CIR, was examined in a national innovation-policy evaluation report in June 2021. Both are public. They disprove a blanket claim that evaluations do not exist; they do not establish that every current programme has been adequately assessed. ^[9][10]

The Senate report itself describes compliance checks and France 2030 disbursements linked to contractual milestones. It also criticises weaknesses in monitoring and evaluation. Its findings are not an account of an entirely rule-free system. ^[2]

Eligibility determines which firms or expenditures qualify. Compliance checks test those rules. A requirement to maintain employment adds a particular obligation. Evaluation asks whether the policy changes outcomes compared with a situation without it. These are different tests.

A government can therefore demand stronger employment conditions while acknowledging that tax checks exist. A beneficiary can comply fully with the rules of an ineffective programme. Turning weak conditionality into no obligations whatsoever, or patchy evaluation into no evaluation, converts a defensible criticism into an inaccurate claim.

The savings an election promise would need

The fiscal implication is where the headline becomes especially damaging. The article’s accessible opening places €211 billion alongside a €173 billion deficit. The latter matches, to rounding, France’s central-government budget deficit in 2023. The support aggregate includes social contributions and public-bank interventions; it does not share that budget perimeter. Putting the numbers next to each other does not establish that removing one would eliminate the other. ^[1][11]

A reform also changes the situation from which a tax expenditure was estimated. Recipients may adjust activity, investment or prices, or make greater use of another provision. Conversely, ending an ineffective subsidy can generate a genuine saving. The scale and direction require analysis. Neither follows from the headline total.

The source budget document gives an unusually direct warning. Page 35 of Volume II of the 2025 draft budget’s revenue assessment explains that the estimates exclude behavioural effects and that measures interact. It consequently cautions that adding their costs lacks a meaningful interpretation. ^[5]

That warning is not an assertion that every tax break costs nothing, nor an argument against inventories. It is a reason not to present separate estimates as the revenue obtainable from a combined reform.

Lending requires similar discipline. Stopping future loans changes disbursements, but also the financial assets and repayments that would otherwise accompany them. Timing matters, as does the accounting measure. A cash saving, a smaller deficit and an improvement in public net worth are not interchangeable results.

The revised definitions proposed in July 2026

The Haut-commissariat’s July 2026 report proposes illustrative perimeters of approximately €82 billion and €187 billion. It includes a grant-equivalent treatment of certain financial interventions. The report sets out reference definitions while recording continuing disagreements. ^[12]

Its methodological note is crucial: the figures were not updated for the new publication. They largely reuse 2023 data and, for some financial support, a European source describing 2022. €187 billion is not evidence that business aid has fallen since 2023. It reflects a different classification. ^[12]

Competing totals do not entitle critics to choose the largest or defenders to choose the smallest. Their definitions must be made explicit before they are compared.

The political responsibility is to show the trade-offs

As France approaches its 2027 presidential election, portraying €211 billion as unconditional annual handouts is a false account of the underlying transactions. Loans and equity investments carry financial counterparts, the aggregate is not a grant-disbursement flow, and some programmes have been evaluated. These differences warrant correcting the description, without presuming the writer’s intent. ^{[2][4][9][10][13]}

The risk is that an apparent fiscal resource enters the debate before its availability has been established. Spending can then be promised without identifying the extra levies, withdrawn financing or specific grants that would pay for it. A headline number conceals the political cost of choosing.

It would be equally unsound to use this rebuttal to excuse every existing subsidy. The Senate documents the lack of a consolidated picture. Demands for transparent records, verifiable objectives and usable evaluation remain justified. ^[2]

The editorial rule should be straightforward. €211 billion can be cited as the outcome of a broad, heterogeneous inventory for 2023, with its components and limitations visible. It should not be presented as an annual gift or a saving ready to fund a manifesto. That would let the debate focus on specific reforms, policy effectiveness and the consequences of each choice.

---

Method and limits. Sources checked as of 10 September 2026. The criticism of L’Humanité concerns the accessible introduction and opening paragraphs, not the unavailable remainder or the forthcoming book. The Senate’s underlying worksheet was not obtained; no amount of double counting is claimed. Source figures and documents were inspected. The loan example is hypothetical. No interviews or responses from the parties are claimed.

Further reading on public accounts: Senegal’s government arrears and the suppliers left waiting for payment. It distinguishes payment obligations, cash flow and budget recording.

Sources and documents

1. L’Humanité · 10 September 2026. Article on Fabien Gay’s book and the €211bn estimate. Hélène May. Title, introduction and opening paragraphs read through Exa; direct access unavailable. Neither the complete article nor the book was reviewed. The criticism concerns only the accessible wording.
2. Sénat · Report no. 808, volume I, submitted 1 July 2025. Senate inquiry into transparency and evaluation of business support. Method and charts: pp. 185–187; summary p. 21 and sections VII–VIII on checks and evaluation. Charts visually inspected. The narrower chart on p. 187 says 2019, whereas the text and summary place it in 2023: this labelling inconsistency is disclosed, not silently repaired. Rounded figures. Underlying worksheet not obtained. HTML version.
3. Bpifrance · 29 February 2024 · 2023 activity. 2023 activity review. Financing and guarantee sections: €9.24bn of short-term credit authorisations; €8.65bn of guarantee activity and €4.33bn of risk taken. The institution’s account of its own activity; not a certified reconciliation to the Senate’s €41bn block.
4. European Commission · DG Competition · Methodology accessed 10 September 2026. State aid Scoreboard: data and methodology. The aid element measures economic advantage rather than the nominal amount. EU coverage excludes general measures among others; it is not an exhaustive definition of French support. The loan illustration is an l0g teaching calculation, not a regulatory assessment of an actual transaction.
5. French budget ministry · National Assembly · 2025 draft budget · document published in 2024. Revenue assessment, volume II: tax expenditure. Pp. 7–10: tax benchmark and reclassification; p. 35: behaviour, interactions and aggregation limits. Page 35 inspected as an image. The 2023, 2024 and 2025 columns are not interchangeable. Since the 2024 draft budget, VAT costs shown concern the central-government share after transfers, not automatically the full general-government revenue effect.
6. DGFiP · BOFiP · Version dated 15 April 2020. Tax consolidation: calculation of a group’s taxable result. BOI-IS-GPE-20, paragraphs 1–20. Describes the regime and adjustments; does not estimate the aggregate 2023 benefit or settle the policy choice of a different benchmark.
7. Institut des politiques publiques · 23 May 2018. Brief no. 32: beneficiaries of the restaurant VAT reduction. Youssef Benzarti and Dorian Carloni. Study of the 2009 reform; its findings are not extrapolated to all reduced rates or to 2023.
8. Fabien Gay · submission published by HCSP · 1 June 2026. Submission to the third consultation session. P. 5: definition and loans on favourable terms; pp. 6–7: general contribution relief; p. 10: eligibility versus additional conditions. The rapporteur’s own argument; an interested political source, not independent validation of the total.
9. France Stratégie · archive HCSP · September 2020. CICE evaluation: synthesis of further research. Evaluation of the former CICE. Cited to establish that research exists, not to extrapolate employment effects or validate all current programmes.
10. CNEPI · France Stratégie · archive HCSP · June 2021. Evaluation of France’s research tax credit. Report by Mohamed Harfi and Rémi Lallement. A counterexample to a blanket claim of no evaluation, not an exhaustive audit of 2023 recipients or a universal endorsement.
11. French Senate · Finance Committee · 12 June 2024 · 2023 budget outturn. Report no. 685 on the deterioration in public finances. Summary and central-government budget section: 2023 deficit rounded to €173.0bn. Not the general-government deficit.
12. Haut-commissariat à la stratégie et au plan · July 2026. Business support: definition, perimeters, monitoring and evaluation. P. 25: incidence; pp. 27–29: 82/187 perimeters, methodological notes and grant equivalents. Printed p. 27 is PDF page 29, visually checked. Figures were not updated and mainly combine 2023 with a financial-support source for 2022; no time-series decline is inferred.
13. French Interior Ministry · Page accessed 10 September 2026. France’s presidential election. The 2027 election. The article refers to the approaching vote and pre-election debate; it does not claim that the official campaign period has begun.