Édouard Philippe’s positions: what we were able to cost

Effects on households

Eight model households, annual disposable income, calculated with OpenFisca-France for only the measures the engine can simulate (1 measure). A zero difference means that these measures do not affect this household, not that the programme changes nothing for it. The detail, household by household, and the comparison with the other candidates.

The public accounts

How to read these amounts. These are static costs and revenues, with no feedback effect, over a full year: nothing on growth, employment or the revenue that economic activity would gain or lose. A higher net cost does not on its own make a worse policy, nor net savings a better one: each programme rests on an economic model (demand-led stimulus, lower compulsory levies, consolidation of the public accounts…) that the cost schedule does not settle. GDP itself measures a volume of activity, not well-being or the usefulness of what is produced. Economic models and feedback effects.

Line by line

• LeverPositionEffect on the public accounts, per year
• Numbers and pay of civil servantsOne retirement in three not replaced (750,000 retirements expected)* Outside the range measured by the source: linear extrapolation (measured range: -150,000 to 150,000 staff). Assumption by the editors. Partial costing. Coefficient calculated by the editors from Sénat, « L'essentiel » sur le PLF 2026, programme « Fonction publique » de la mission « Transformation et fonctions publiques » (page 2): the position is marked as an assumption by the editors. Reading: one retirement in three not replaced out of the 750,000 expected during the five-year term, i.e. around 250,000 fewer staff at the end of the term, costed at the average cost of a public employee (bottom of the range at the cost of a local authority employee); beyond 150,000 staff (domain measured by the Cour des comptes), linear extrapolation. The campaign website (priority « Pour une France plus prospère ») writes « Supprimer un tiers des fonctionnaires partant à la retraite, soit plus de 250 000 postes sur le quinquennat » [“Cut one third of civil servants retiring, i.e. more than 250,000 posts over the five-year term”]: costed at 250,000, the most cautious value (partial costing). The reorganisation of State agencies is not costed.+€11bn^{* (outside the range measured by the source: linear extrapolation)}saving, +€7bn to +€16bn
• Teachers' pay+20% in average pay over the five-year termAssumption by the editors. Coefficient calculated by the editors from Sénat, rapport spécial PLF 2026, Enseignement scolaire: the position is marked as an assumption by the editors. In the same source, the candidate says he will fund this increase and the reduction in class sizes through the fall in the number of pupils (« près d'un million d'élèves en moins » [“almost a million fewer pupils”]), « sans dégrader les finances publiques » [“without damaging the public finances”]. This counterpart, with no amount, is not costed: the cost shown is gross.−€7.1bnnew spending, −€12bn to −€7.1bn
• New nuclear reactors8 EPRs on top of the 6 already planned, i.e. 14* Outside the range measured by the source: linear extrapolation (measured range: 0 to 6 reactors). Assumption by the editors. Coefficient calculated by the editors from EDF, communiqué du 18 décembre 2025, « EDF shares its forecasted cost estimate of the EPR2 programme for €72.8bn » (copie d'archive du 13 octobre 2026, le site renvoyant une erreur 403): the position is marked as an assumption by the editors. Central value brought down to the smallest amount: 2032 financing need acknowledged as overestimated beyond the six EPR2s of the PPE3, disbursed later than the coefficient assumes (2027 to 2038); lower bound: EDF's December 2025 estimate, 60% financed by a State loan paid out over 12 years (€0.61bn per reactor). Fourteen EPR2s, i.e. eight more than the six in the PPE3 (multi-year energy programme). Beyond the six EPR2s of the PPE3, the cost-schedule coefficient (disbursements spread from 2027 to 2038) is extrapolated linearly: it probably overestimates the 2032 financing need, since the subsequent reactors are disbursed later. State loan: weighs on the debt, shown separately from the balance.−€4.9bn^{* (outside the range measured by the source: linear extrapolation)}additional State lending, on the debt, −€9.6bn to −€4.9bn
• Legal retirement age65, graduallyAssumption by the editors. Partial costing. Timetable given by the same source: first restore the 2023 reform (64), then reach 65 « en moins de dix ans » [“in less than ten years”], i.e. the Cour des comptes’ current-pace scenario (65 for the 1968 to 1972 generations: €4.7bn in 2035). The Court costs only the year between 64 and 65, in 2035 and not in 2032; the quarter between 63 years and 9 months (the law in 2032) and 64 is not counted. This gain and that of the contribution period (costed at 176 quarters, the costing horizon) are added together, minus the share common to both increases, removed from the smaller of the two gains (overlap declared in the cost schedule, measured from Cnav data: 8% of the gain from the age, 23% of that from the contribution period, shares that apply to two increases at the pace of one quarter per generation). Overlap with “Contribution period”: the estimated common part (€0.4bn) is withheld from “Contribution period”; this one is counted in full.+€4.7bnsaving
• Unemployment, max. benefit duration12 months maximum for under-50s, « comme l’Allemagne » [“like Germany”]* Outside the range measured by the source: linear extrapolation (measured range: 18 to 24 months). Assumption by the editors. Partial costing. Coefficient calculated by the editors from Unédic, Réforme 2023 : premiers effets de la réforme de contracyclicité: the position is marked as an assumption by the editors. Central value brought down to the smallest amount: gain acknowledged as overestimated, since 50 to 54-year-olds, whom the candidate excludes, are counted; for this parameter, the lower bound of the coefficient (€0.7bn per month) is also its central value. 12 months, read as the maximum compensation period before 55 under current law (cost-schedule parameter); below 18 months (domain measured by Unédic, 18 to 24 months), linear extrapolation. The candidate’s age threshold (50) differs from that of current law (55): 50 to 54-year-olds, counted here, would not be affected, and the gain is overestimated accordingly. Claimants after a negotiated termination (rupture conventionnelle), already compensated for at most 15 months before 55 since 1 September 2026 (law no. 2026-470 of 11 June 2026), are also counted as if they were compensated for 18 months: the gain is further overestimated on this side; Unédic only costs the amendment as a whole (€0.72bn to €0.945bn, seniors included), which does not make it possible to isolate this share, which is therefore not removed. The central coefficient (€0.7bn per month) is already the lower bound of the range: no more cautious amount is published.+€4.2bn^{* (outside the range measured by the source: linear extrapolation)}saving, +€4.2bn to +€4.5bn
• Contribution period45 years, graduallyAssumption by the editors. Partial costing. Coefficient calculated by the editors from Service-public.fr, fiche F35063 (durée d'assurance par génération): the position is marked as an assumption by the editors. Central value brought down to the smallest amount: gain acknowledged as overestimated, since the coefficient for the extension is that of 2035 (Cour des comptes), which overstates the 2032 gain; lower bound of the coefficient: €1bn per quarter, an effect reached as early as 2030 according to the Court. 45 years, i.e. 180 quarters, reached « progressivement » [“gradually”]; the same source gives the timetable of the reform, « en moins de dix ans » [“in less than ten years”] (video of 29 September 2026). At the pace of one quarter per generation, that of the Cour des comptes, only the first four quarters (176, 1969 generation) have an effect within the costing horizon: the position is costed as the Court’s « 176 trimestres » scenario (table no. 7 and table no. 12, page 93: 1969 to 1973 generations), €5.2bn in 2035, i.e. €4bn to €5.2bn; quarters 177 to 180, reached beyond the horizon, are not costed (partial costing). This gain and that of the age of 65 are added together, minus the share common to both increases, removed from the smaller of the two gains (overlap declared in the cost schedule, measured from Cnav data). Of which €0.4bn not added, common base with “Legal retirement age”. For an increase in both age and contribution period, the Cnav (Prisme model) costs, for 2030, in current euros, the age raised by two years at 3 months per generation with the period accelerated by one quarter per generation (€17.7bn), the accelerated period alone (€4.7bn) and the age raised by three years at 4 months per generation (€18.8bn): report annexed to the amending social security financing bill for 2023, table 5, page 27. The age alone raised by two years at 3 months per generation is not costed there: in 2030, it has reached three-quarters of the increase at 4 months per generation, hence 18.8 × 3/4 = €14.1bn. Common part: 14.1 + 4.7 − 17.7 ≈ €1.1bn, i.e. 8% of the gain from age (1.1 / 14.1) and 23% of that from the period (1.1 / 4.7). Editors' calculations.+€3.6bnsaving, +€3.6bn to +€4.8bn; €0.4bn more, costed, not added
• Income tax scaleNo income tax for 3 years after a birth (middle classes), €3,000 above thatAssumption by the editors. Coefficient calculated by the editors from DGFiP, Impôt sur le revenu : répartition de l'impôt moyen par dixièmes de RFR par part, onglet a2024: the position is marked as an assumption by the editors. Benefit per household: min(tax, €3,000) during the three years following a birth. Editors' calculation: average net tax of the deciles of RFR (reference taxable income) per share (DGFiP, 2024 income), capped at €3,000, i.e. €877 per household at most, applied to 1.93 million households at most (three times the 643,905 births of 2025, Insee), assuming that these households are distributed like all households: €1.7bn a year at most. The candidate announces €3bn (campaign website): the gap comes from the distribution assumption, which only a microsimulation (OpenFisca, LexImpact) could settle; « familles qui travaillent » [“working families”] is not defined (not taken into account).−€1.7bnrevenue cut
• Adaptation to climate changeSingle climate adaptation fund of €2bn a year (Barnier fund and Fonds vert)Assumption by the editors. Candidate’s amount, taken as is. Reading adopted: the single fund of €2bn a year brings together existing appropriations; the new spending is the difference from those of the PLF 2027, in payment appropriations: Barnier fund €238.8m and Fonds vert €981.4m, i.e. 2 − 1.22 = €0.78bn a year (candidate's amount used). Other reading: in commitment authorisations (€300m and €1bn), €0.7bn.−€0.8bnnew spending
• Funded shareCollective funded pensions for all employees, €500 from the State at first jobAssumption by the editors. Partial costing. Coefficient calculated by the editors from Céreq Bref n° 487-488, « S'insérer dans un monde instable : le diplôme reste une valeur sûre » (enquête 2024 auprès de la Génération 2021): the position is marked as an assumption by the editors. Reading adopted: €500 paid once by the State to each young person at their first job, the annual flow of first jobs being taken as equal to that of first-time leavers from the education system (773,000 in 2021, Céreq), i.e. about €0.39bn a year. Not costed: the extension of collective funded pensions to all employees, with no rate or funding method (contributions diverted from the pay-as-you-go system or new savings). The payment is counted as spending; an account remaining the property of a public fund could be a financial transaction.−€0.4bnnew spending
• Right of asylumSafe countries voted yearly, end of asylum for life; suspended in Mayotte and GuyaneAssumption by the editors. Partial costing. Coefficient calculated by the editors from Sénat, commission des finances, note de présentation PLF 2026, mission « Immigration, asile et intégration »: the position is marked as an assumption by the editors. Central value brought down to the smallest amount: gain acknowledged as overestimated: shifts to ordinary emergency accommodation and State medical aid are not deducted; the central value is the lower bound of the cost per application (rule common to positions costed through the gain per asylum application). Only the suspension of the registration of applications in Mayotte and Guyane is costed: 10,490 applications in 2025 (Guyane 8,897, Mayotte 1,593, Ofpra), at the national average cost of an application, measured in mainland France (material support is lower overseas: the gain is probably overestimated). The annual list of safe countries and the end of « asile à vie » [“asylum for life”] have no costed effect. The knock-on effects on emergency accommodation and state medical aid are not counted: the gain is overestimated. Compliance with international and European law not assessed.+€0.02bnsaving, +€0.02bn to +€0.09bn
• Corporate income taxNo tax increase for businesses, « qui payent déjà trop » [“which already pay too much”]€0bnzero cost

Overlaps

Amounts announced by the candidate

The amounts that Édouard Philippe puts forward, with the direction given to them (“+”: gain for the public accounts, “−”: cost). They are never added to the totals or to each other. The gap with our calculation is only given when both cover the same scope: not for a position that is not costed, costed in part, based on an assumption by the editors, or whose amount is the candidate’s own, taken as is; and only when the announcement falls outside the range of our calculation (otherwise: “within the range”).

• LeverWhat the candidate saysAnnounced, per year
• Income tax scaleCampaign website, « Trois principales annonces d'Édouard Philippe à l'Heure de vérité » (8 October 2026): « Concrètement, tous les foyers qui paient moins de 3 000 euros d’impôt sur le revenu seront exonérés d’impôt pendant trois ans, et ceux qui en paient plus bénéficieront d’une réduction d’impôt de 3 000 euros. Ce dispositif, qui coûte 3 Mds € par an […] » [“In concrete terms, all households paying less than 3,000 euros in income tax will be exempt from tax for three years, and those paying more will benefit from a tax reduction of 3,000 euros. This scheme, which costs €3bn a year […]”]Our calculation: −€1.7bn per year. No gap calculated: costing based on an assumption by the editors.−€3bncost to the public accounts
• Production taxes« 50 Md€ de baisses d'impôts de production contre 50 Md€ d'aides publiques improductives en moins » [“€50bn of cuts in production taxes in exchange for €50bn less in unproductive public subsidies”]: two equal amounts, without saying whether they apply per year or over the five-year term.Position not costed. Quoted, not counted.−€50bnrevenue cut
• Adaptation to climate change« doté de 2 milliards d'euros par an » [“endowed with 2 billion euros a year”]; « Deux milliards, c'est une somme importante. Mais rapportée aux 5 milliards d'euros que coûtent chaque année les catastrophes naturelles, c'est de l'argent bien investi » [“Two billion is a large sum. But compared with the 5 billion euros that natural disasters cost every year, it is money well invested”] (Les Ponts-de-Cé speech, 17 September 2026, reported by AFP).Our calculation: −€0.8bn per year. Candidate’s amount, taken as is.−€2bncost to the public accounts
• Public aid to businesses« 50 Md€ d'aides publiques improductives en moins » [“€50bn less in unproductive public subsidies”], in exchange for « 50 Md€ de baisses d'impôts de production » [“€50bn of cuts in production taxes”]: period (per year or over the five-year term) not specified.Position not costed. Quoted, not counted.+€50bnsaving

What this costing does not measure

• Growth and employmentThe effect of the measures on activity and employment, and the revenue gained or lost as a result; it depends on the economic model used: models and feedback effects.
• Education, health, researchLong-term effects on human capital, productivity and innovation.
• InequalityThe distribution of gains and losses across the whole population, beyond the eight model households.
• Public servicesWhat new spending or savings change in the quality of the services provided and in access to them.
• Environment and climateEffects on emissions, natural resources and adaptation to climate change.
• Sovereignty and external dependenceEffects on the country’s dependence on other countries: energy, industry, defence, supplies.
• Interest rates on the debtThe effect of a programme on the terms on which the State borrows, which also depend on the make-up of the holders of its debt.