← All papers

Camille Landais, Thomas Piketty, Emmanuel Saez · Travail, genre et sociétés n° 27 (2012/1): 177-181

Réponse de Camille Landais, Thomas Piketty et Emmanuel Saez

8 problems found: 1 major · 3 moderate · 4 minor

What the paper claims

Every problem found

1majorconfirmed · newly identified here · external inconsistency

The reply says giving the RSA minimum income to each spouse individually would add only 'a few thousand' people. On the authors' own data and rules, between about half a million and three million people would become eligible, most of them women at home. The cost would be roughly EUR 1-8 billion a year, and a government estimate put it near EUR 9 billion. The cost side of the authors' argument is therefore wrong by a factor of hundreds.

Technical detail and evidence

Why it matters, in detail. Under an RSA assessed on own resources with 2010 parameters, newly entitled adults in couples number: - 3.11 M on own resources only; - 1.45-2.17 M with a 3-6% notional return on own net wealth, i.e. the reply's own design; - 0.50 M in the most restrictive variant. The annual cost is EUR 0.9-8.2 bn (central EUR 3.5-4.8 bn), or 9-83% of the QC's pure schedule value (EUR 9.9 bn adjudicator, 10.2 bn Auditor A, about 10 bn in Allègre, Périvier & Pucci 2021). Against 'quelques milliers' (about 2,000-9,000), the understatement is a factor of at least about 50 even in the most restrictive case, and about 150-1,500 for the other variants. It survives partial take-up and is corroborated by INSEE's 2.1 M and the government's about EUR 9 bn costing. Most new recipients are spouses of below-median earners. The reform would therefore not 'display openly' the QC's redistribution, which rises with income up to a bound of about EUR 13,400; it would be a different and much larger transfer.

Evidence. Reply p.180 §13; the text is identical in the 2015 and 2022 Wayback copies and the 2026 publisher text. Adjudicator scripts/rsa_individualised.py (output/rsa_individualised_log.txt) uses the authors' parameters from Annexe simulateur §C.4: 2010 base EUR 460.09/month, 12% forfait logement, 38% taper, ages 25-64. Newly entitled adults aged 25-64 in couples whose foyer social gets no RSA today: - own resources only: 3.11 M (EUR 8.2 bn/yr); - plus own imputed rent: 2.28 M (EUR 5.1 bn); - with a 3%/4%/6% notional return on own net wealth: 2.17/1.88/1.45 M (EUR 4.8/4.1/3.5 bn); - strictest variant of the book's p.114 idea (gross imputed rent + rent paid by the household + 4% of wealth): 0.76 M (EUR 1.7 bn); - with each spouse at 75% of the single base: 0.50-2.30 M (EUR 0.9-4.9 bn). 86-91% are women. At one-third take-up, the most restrictive case still leaves about 0.17 M. In the 4% variant, of 1.44 M newly entitled married spouses, 0.81 M have a partner earning below the median and 0.008 M a partner in the top 1%. These figures reproduce Auditor A's v0/v1 results exactly (3.11 M / EUR 8.2 bn; 2.28 M / EUR 5.1 bn). External benchmarks: - INSEE Première n° 1463 (Aug. 2013), p.1: 2.1 M 'femmes au foyer' aged 20-59 in 2011; - government answer to Sénat written question n° 21106 (JO Sénat 01/04/2021, p.2232): 'le coût d'une individualisation totale du RSA avait été estimé à près de 9 milliards d'euros en 2016'. The authors' own commented-out individual 'revenu universel' sketch implies 7.0 M recipients aged 25-64 and EUR 30.3 bn (scripts/revenu_universel_sketch.py).

Why this severity. The reply's case for extending individualisation to transfers rests on the cost being a few thousand extra recipients. Every defensible design gives hundreds of thousands to millions, at roughly EUR 1-8 bn a year (10-80% of the QC's own value), so the cost side of the trade-off is wrong by two to three orders of magnitude. The defence of individual income taxation itself is untouched, hence major rather than critical.

The authors' response. None located. The sentence is hedged ('il est bien possible'). The companion reply in Revue de l'OFCE 122 (2012), p.149, only proposes 'intégrant le RSA au barème de l'impôt sur le revenu unique (issu de la fusion IR et CSG) et individualisé', with no count. The authors' best case is that the figure was meant to describe only spouses of very rich earners who would remain eligible after a wealth test: about 8,000 under a 4% notional return in their own files.

2moderateconfirmed · previously known · overclaiming

Today, the stay-at-home wife of a man on an ordinary income gets no RSA, so taking a job costs her no benefit. If every spouse had their own RSA, she would receive it at home and then lose 38 cents of it for every euro she earned. For one to two million mostly female spouses, the hidden tax on going to work would go up, the opposite of what the reply promises.

Technical detail and evidence

Why it matters, in detail. With q = 0.38 unchanged, individualisation extends the taper to every low-earning spouse below F/q (about EUR 1,065-1,210 net/month). In the authors' data, 1.2-2.1 M lower earners (about 92% women) move from a 0% to a 38% RSA withdrawal rate. Against that, 0.5-0.7 M spouses, mostly primary earners in RSA couples, see their rate fall from 38% to 0. For an inactive spouse in a couple not on the RSA, the participation tax rate from the RSA alone goes from 0 to 38%, on top of social contributions and the new income tax. The net effect on second-earner incentives is negative, not 'une forte baisse'. Only a much lower taper (raising F1's cost further) or a non-withdrawn individual benefit could deliver the claimed gain.

Evidence. Reply p.180 §13. - RSA rule: CNAF, l'e-ssentiel n° 90 (Sept. 2009), p.1: RSA = montant forfaitaire + 62% of the foyer's activity income - resources. The authors' simulator sets $tx_rsa = 0.38 (Annexe simulateur §C.4). - Algebra (Auditor B, codex/report.md I1): if the joint benefit max(1.5F - q(w1+w2), 0) is positive and w2 <= w1, then w2 < 0.75F/q < F/q, so the lower earner's individual RSA is also being withdrawn at q. - Population-weighted check on the authors' TAXIPP files (scripts/rsa_marginal_rates.py; 9.21 M married/PACS couples, both spouses aged 25-64): lower-earning spouses facing the 38% withdrawal rise from 0.77 M today to 2.81 M (own resources) or 1.77 M (with 4% notional wealth income). Of these, 1.20-2.12 M face it for the first time; 92-93% are women and most have no earnings today. The rate falls for only 0.08-0.20 M lower earners and 0.41-0.51 M higher earners. - A newly entitled spouse taking a half-time or full-time net-SMIC job loses 38% of those earnings in RSA (about EUR 2,400 or 4,800 a year). - The authors' own commented-out individual 'revenu universel' uses a 50% taper on earnings (SimulationsNouvTransferts.xls).

Why this severity. The benefit side of the RSA trade-off has the wrong sign for the group the reply cares about, second earners deciding whether to work. Together with F1, this leaves the RSA extension (C6) unsupported. It is graded moderate because it concerns a hedged, undeveloped extension: a redesigned schedule (e.g. a lower taper or integration into the tax) could change it, at extra cost, and the income-tax argument is untouched.

The authors' response. Not addressed in the reply. The companion OFCE reply (Revue de l'OFCE 122, 2012, p.149) proposes integrating the RSA into the individual income-tax schedule but specifies no taper or design.

3moderateconfirmed · previously known · overclaiming

The authors say their computer simulations prove that one-earner families would not be the big losers. Their own published tables show such couples are about three times more likely than two-earner couples to pay more, and most middle-income one-earner couples without children would pay a little more. The amounts are usually small, around 1% of income, except for high earners. So the reassurance about size is broadly right, but 'prove' overstates it and hides who loses.

Technical detail and evidence

Why it matters, in detail. Incidence: - the probability of a loss is 41% for single-earner vs 13% for dual-earner couples (52% vs 21% after Auditor A's rescaling to the xls vintage); - unequal couples' winner share is 24 points below equal couples' (65% vs 89%), and 45 points below among full-time workers. Magnitude: - below EUR 5,000/month of earner income, single-earner couples' mean change lies between -2.2% and +1.1% of income, so 'grands perdants' is not supported for the middle class; - losses become material above about EUR 7,000/month (+3.6% to +13% of income), partly through the reform's broader taxation of capital income. The reply's second reason is weak: for single-earner couples the QC is worth about 4-8% of gross pay from about the median wage up, which is not small 'until a very high level of income'. It is largely offset by the new schedule's lower rates.

Evidence. Reply p.179 §12; repeated verbatim in Revue de l'OFCE 122, p.148. Authors' published gagnants.xls: - sheet '18+_couples_inegaux': equal couples 89.1% winners (H5) vs unequal couples 65.1% (H24); note A48: '90% des couples egaux gagnent avec notre reforme mais seulement 65% des couples inegaux'; - sheet '18-65_workers_couples_inegaux': 88.7% (H5) vs 43.6% (H24), and 33.9%/24.9%/31.2% winners for P30-P60 (H28-H30). Authors' FAQ (revolution-fiscale.fr, 2011): winners are 'nettement plus faible dans les couples inégaux au dessus de P30'. Adjudicator scripts/single_earner_incidence.py on indiv_nouvel_ir_2010.dta (spouse = next record, linkage verified): - 3.12 M single-earner couples (24% of 12.76 M married/PACS couples): 40.8% pay more, vs 12.7% of 5.65 M dual-earner couples; - 27.3% vs 6.9% lose more than 1% of income; - single-earner couples bear 50% of married couples' losses (57% below EUR 10,000/month). By the earner's gross monthly earnings: - EUR 2,000-2,500: 63% lose (88% of childless couples, mean +EUR 331, +1.1% of income); - EUR 2,500-3,000: 59% (77%); - EUR 5,000-7,000: 60% (+1.1% of income); - EUR 7,000-10,000: 84% (+3.6%); - EUR 10,000 and above: 98% (+13.4%). In net terms single-earner couples pay EUR 4.0 bn more overall, but gain EUR 1.6 bn net below EUR 10,000/month. Dual-earner couples gain 2-4% of income at EUR 3,000-10,000/month. These results reproduce Auditor A's exactly; Auditor B's childless subgroup (97% losing at EUR 2,000-3,000, mean EUR 38/month) is consistent. Typical case (scripts/typical_case.py; authors' cas-type parameters): a childless single-earner couple under the book's schedule (p.79, 10% at EUR 2,200; params.xls uses 10% at EUR 2,170) loses EUR 96-469/yr at EUR 1,850-3,050 gross/month, and again from EUR 5,480 (EUR 1,548 at EUR 8,000; EUR 4,912 at EUR 10,000). Under SimulCasTypes.xls 'v0' (8% at EUR 2,200) it loses only above EUR 5,480, which is the FAQ's 'about EUR 5,000'. For such couples the QC itself is worth about 4-8% of gross pay between EUR 1,900 and 6,000/month (4.6% at EUR 2,000-2,500, 5.9% at EUR 4,000, 7.7% at EUR 6,000). The QC is worth about EUR 9.9 bn in total (scripts/qc_value.py; Auditor A EUR 10.2 bn; Allègre, Périvier & Pucci 2021 about EUR 10 bn), 72% of it below the top decile of couples.

Why this severity. The magnitude reassurance survives for most middle-income single-earner couples: net changes are within about ±1% of income below EUR 5,000 a month. But the reply presents as 'proven' a conclusion its own simulations contradict on incidence, and omits that unequal couples are the main losing group among couples below the top. That is a noticeable qualification of a core claim, not a reversal.

The authors' response. §12 is itself the authors' answer to Sterdyniak, repeated verbatim in the OFCE reply (Revue de l'OFCE 122, p.148). Their site FAQ (2011) acknowledges that the share of winners is 'nettement plus faible dans les couples inégaux au dessus de P30'.

4moderateconfirmed · newly identified here · misrepresentation of source

The reply says almost every developed country taxes spouses separately, and cites an article in the same issue as saying so. That article actually says most EU and OECD countries still tax couples jointly or in equivalent ways; the United States, Germany and Spain, for example, allow joint filing. The point is not central to the authors' case, but the source is misreported.

Technical detail and evidence

Why it matters, in detail. By country count, separate taxation is the majority regime (19/30 in Jepsen's European count; about 20-24 of 30-34 OECD members), so 'commonly accepted' is defensible but 'quasi-totalité' is not. About a third of OECD countries, with 45-48% of the OECD population in 2010, kept joint or optional joint taxation, and Jepsen stresses that joint 'equivalents' (transferable allowances, credits) are widespread. The correction removes the picture of France as an isolated outlier on the tax unit. It does not touch Jepsen's and OECD PF1.4's finding that France, Germany and partly Slovakia stand out in favouring an unequal division of paid work.

Evidence. Reply p.177 §3. Jepsen (2012), « Vers une individualisation de nos systèmes fiscaux ? », Travail, genre et sociétés 27: 171-176 (doi:10.3917/tgs.027.0171): - §1: 'onze pays maintiennent encore le principe d'une imposition conjointe, alors que dix-neuf autres ont adopté une imposition séparée'; - §6: 'la plupart des pays de l'Union européenne et de l'ocde appliquent l'imposition conjointe ou des systèmes équivalents'; - §8: 'il n'existe pas dans l'Union européenne de systèmes d'imposition totalement individualisés'; - §11: 'le ménage continue d'être considéré comme l'unité d'imposition appropriée'. CESifo DICE Report 1/2011, 'Neutrality of tax systems', Table 1 (p.49; source OECD Taxing Wages 2007-2008) lists the Czech Republic (couples with children), France, Germany, Ireland, Luxembourg, Norway (optional), Poland, Spain (optional), Switzerland and the US. With World Bank 2010 populations, these countries hold 45-48% of the OECD-34 population (9-12 countries depending on the Czech Republic, Portugal and Belgium; adjudicator recomputation of Auditor A's figure). The authors' own book is more careful: 'comme cela se pratique déjà dans la plupart des pays européens' (p.64).

Why this severity. The international claim frames the reply but is not needed for its economic argument; the book itself calls it 'pas le plus déterminant' (p.65). The misattribution is nonetheless clear-cut and inverts the cited source, so it materially misstates a secondary claim.

5minorconfirmed · methodology

The reply says joint family taxation is a small corner of the French system because the income tax raises only 3% of national income, against 6% for the CSG and 10% for some social contributions. The numbers are right. But how much a tax raises does not show how much would change if couples were taxed separately, and part of the CSG also depends on the household. This is a matter of framing and changes little in the debate.

Technical detail and evidence

Why it matters, in detail. Standalone CSG is 5.0% rather than 6% of national income, and about 18% of CSG revenue (on replacement incomes) has a familialised rate schedule. More substantively, joint versus individual assessment matters only for progressive or means-tested instruments, which are precisely the familialised IR, PPE, RSA and housing benefits. Revenue shares of proportional individual levies therefore cannot measure how much individualisation changes. The relevant magnitude is the QC's value: about EUR 10 bn, roughly 0.6% of national income and 18% of couples' schedule IR. That is modest relative to total levies, so 'relatively marginal' is defensible on that measure, but the argument the reply gives does not establish it.

Evidence. Reply p.177 §4 (also Revue de l'OFCE 122, pp.145-146). - Book p.43 table: IRPP 52, CSG 94 and 'cotisations maladie, famille, formation, etc.' 164 of EUR 1,680 bn, i.e. 3.10%, 5.60% and 9.76%. - Book appendix p.6: 'nous avons inclus les recettes du prélèvement libératoire avec les recettes de la CSG-CRDS (89.1+4.8=93.9 milliards)'. The IRPP figure is the 2011 assessment (51.6) rather than 2010 receipts (47.4, i.e. 2.8%). - CCSS (Sept. 2011), p.39: CSG 2010 EUR 84,006 M, i.e. 5.0% of EUR 1,680 bn (Auditor B; verified). - Authors' do_cotsoc.do, phase 2: CSG on pensions and unemployment benefits is exempt if rfr_irpp <= a threshold rising with nbp, and at the reduced rate if irpp_net = 0. - TAXIPP: CSG on replacement income is EUR 17.0 bn (pensions 15.8, unemployment 1.2). - QC value about EUR 9.9 bn (scripts/qc_value.py). - Book p.67.

Why this severity. The figures are correct to rounding under a disclosed convention, and the 'continuity' framing is a rhetorical judgement the critics contest. Correcting the labels and the inference changes no substantive conclusion.

The authors' response. The reply's §4 is itself the response to Sterdyniak and Monnier. The CSG grouping and the IR vintage are disclosed in the book appendix (p.6).

6minorconfirmed · newly identified here · arithmetic or transcription

A few references and one quotation are slightly wrong or incomplete. This is sloppy, but it does not change what the reply argues.

Technical detail and evidence

Why it matters, in detail. No effect on any claim. The elided Sterdyniak clause concerns the authors' acceptance of small gaps between assistance incomes and earnings, i.e. the other half of the inconsistency he alleged, which the reply's elasticity argument does not address.

Evidence. - 'Raj J. Chetty et al. [2011]' (p.178) vs the bibliography 'CHETTY Raj, FRIEDMAN John, OLSEN Tore et PISTAFERRI Luigi, 2009' (p.181, working-paper title). - 'Pollack' (p.179) vs 'POLLAK' in the bibliography. - Blundell, Bozio & Laroque (2011) is listed (p.180) but never cited. - The note-2 call is inside the quotation ('n'avaient pas² conscience', p.178). - §8 quotes 'par contre' where Sterdyniak's TGS text (§11) has 'en revanche'. Its '[…]' omits his clause on reducing the gap between social minima and earnings 'puisque les individus préfèrent toujours vivre d'un emploi'. - 'impôt sur les revenu' (p.177).

Why this severity. Presentation slips; none affects any conclusion.

7minorconfirmed · newly identified here · version instability

The authors posted their simulation files online, which is to their credit, but the different files do not match one another. A reader cannot tell which version 'proves' the reply's claims, and in one case the choice of version decides whether middle-income one-earner couples win or lose.

Technical detail and evidence

Why it matters, in detail. Whether a childless single-earner couple at EUR 1,850-3,050 gross/month loses (book schedule, by up to EUR 469/yr) or gains (v0 schedule) depends on which posted schedule is used. The winners share differs by 10 points between the posted micro-file and the published table. The individualised-RSA claim has no executable simulation behind it.

Evidence. - scripts/single_earner_incidence.py: 83.45% winners, ir/y 0.1313 and ira/y 0.1442, vs gagnants.xls '18+' G5 = 73.35%, E5 = 0.1297, F5 = 0.1304. - scripts/check_new_ir.py: the posted ir reproduces the params.xls 2010 schedule (max relative error 1.1e-7). - SimulCasTypes.xls, 'Méthodes' A1 '(11-1-2012) (TP)'; sheet v0 F27:G32. BaremesNouvelIR.xls BAR-V0: 10% at EUR 2,170. Book p.79: 10% at EUR 2,200. - SimulationsMTR_PTR.xls, sheet PTR, row 7 (O7 = 0.0108, P7 = 303,314, R7 = 3,906); book p.100. - do_nouv_transferts.do lines 45 (rsan=rsa) and 51-57 (commented out). - The authors' notice of 2 April 2012 (revolution-fiscale.fr/presentation/99, via Auditor B) says the public simulator stayed on 2010 data.

Why this severity. This is a documentation and versioning problem: no single published version reproduces the reply's 'simulations'. It does not change the direction of F3, and it explains why the authors' FAQ and their micro-files disagree.

8minorlikely · previously known · overclaiming

The authors say every study finds that wives' decisions to work react strongly to taxes. Research does find wives more responsive than husbands, but the effects are moderate and have been shrinking, and a review co-authored by one of the cited authors finds similar figures for men. The reply overstates the evidence without reversing the argument.

Technical detail and evidence

Why it matters, in detail. Quasi-experimental extensive-margin elasticities for married women of about 0.15-0.4 (CGMW; Blau-Kahn's 2000 estimates about 0.36-0.4, mostly extensive) imply a real but modest response, not 'fortement'. In the authors' own simulation, removing the QC lowers married women's IR participation tax by 1.5-5.3 points. Scaling that simulation (women's average elasticity 0.59, giving +1.9% female employment; SimulationsMTR_PTR.xls, PTR rows 7-9) linearly to elasticities of 0.15-0.4 gives about +0.5-1.3%. SSG supports the claim that top earners' measured responses are mainly avoidance rather than labour supply, so note 2 is a fair citation for the labour-supply comparison.

Evidence. Reply p.178 §7-8 and note 2. - Blau & Kahn (2007), Journal of Labor Economics 25: 393-438 (Jan. 2007 manuscript, p.2): own-wage elasticity 'fell by 50 to 56 percent'. - Chetty, Friedman, Olsen & Pistaferri (2011), QJE 126: 749-804: married women bunch more than single men (1.79 vs 0.25); intensive margin. - Chetty, Guren, Manoli & Weber (2011), NBER WP 16729, Table 1 panel A: Juhn-Murphy-Topel 0.13 (men), Devereux 0.17 (married women), Liebman-Saez 0.15 (women married to high-income men), Eissa-Liebman 0.30 (single mothers), mean 0.25. Text p.17: the most elastic groups are single mothers and older workers, and extensive-margin values 'above 1 are not' reasonable. - Saez, Slemrod & Giertz (2009), NBER WP 15012, p.3: elasticity 'close to zero for prime-age males, although for married women the responsiveness of labor force participation appears to be significant'. - Jepsen (2012) §4: 'les recherches n'ont pas permis d'aboutir à des résultats concluants'. - Authors' SimulationsMTR_PTR.xls, sheet PTR, columns K-L: the reform lowers married women's IR participation tax rates by 1.5-5.3 points.

Why this severity. The direction of the claim, that married women respond more than primary earners at the participation margin, is mainstream and supported by the cited sources. Only the universality and the strength are overstated, and no conclusion of the reply turns on the difference.

The authors' response. The reply's §7-8 is itself the response to Sterdyniak's charge of inconsistency.

Claims examined and rejected

Problems one auditor raised that did not survive checking.

Audit notes

What reproduced. Reproduced: - the 3/6/10% shares from the book's table (3.10/5.60/9.76%); - the 38% RSA taper (CNAF 2009; simulator $tx_rsa); - the posted new-tax variable from the params.xls schedule (max relative error 1.1e-7); - the second-earner wedge and Sterdyniak's EUR 120/month QC example (Auditor B); - Auditor A's typical case for a childless single-earner couple (loss of EUR 96-469/yr at EUR 1,850-3,050 gross/month under the book schedule; none below EUR 5,480 under SimulCasTypes v0); - Auditor A's couple-level incidence, exactly (40.8% vs 12.7% losers; 50% of losses); - Auditor A's individualised-RSA counts, exactly where definitions coincide (3.11 M / EUR 8.2 bn; 2.28 M / EUR 5.1 bn), plus new variants down to 0.50 M / EUR 0.9 bn; - the scale of the authors' 'revenu universel' sketch (7.0 M, EUR 30.3 bn); - the QC's pure schedule value (EUR 9.9 bn vs Auditor A's 10.2 bn and Allègre et al.'s about 10 bn; joint schedule tax equal to the authors' irpp_bar0, EUR 55.0 bn); - OECD population shares under joint taxation (45-48%); - every quoted cell of gagnants.xls, SimulCasTypes.xls, BaremesNouvelIR.xls and SimulationsMTR_PTR.xls. Not reproduced: - the reply's 'quelques milliers': no calculation exists in the published materials, and every recomputation contradicts it; - the published gagnants.xls headline (73.35% winners) from the posted micro-file (83.45%), a vintage mismatch; - the FAQ's EUR 5,000 break-even under the book's own schedule. The English translation could not be compared: its body was not retrievable.

Limits of this audit. The reproductions use the authors' TAXIPP 0.0 synthetic files for 2010, built by statistical matching and scaled to national accounts. Capital income and wealth are split 50/50 within couples, incomes and wealth are NA-scaled (which makes the notional-wealth RSA variants conservative), AAH and ASS are not modelled, take-up is full and there are no behavioural responses. The posted micro-file is not the vintage behind the authors' published tables; the direction of every result holds across the published table and the micro-file. The individualised-RSA design is not specified in the reply, so the individual base, notional return and treatment of children and housing are adjudicator assumptions, reported as a range from 0.50 to 3.11 M. The government's 'about EUR 9 bn (2016)' figure is second-hand; its underlying study was not located. Typical cases use the authors' simplified cas-type assumptions: 12% employee contributions, wage income only, no niches. The Cairn-int English translation could not be read: it returns 403 live, and archived captures contain only bibliographic metadata. The historical claim attributed to Delalande (2011) (C8) and Jepsen et al. (1997) were not verified. The attribution of greater bargaining power to the primary earner under the QC is an untested conjecture and was not graded.