piccini papers
Papers / PIC:2610.07 Evidence reviewed through 10 October 2026

Social protection · Labor supply · Evidence synthesis

Does Bolsa Família Reduce Work? The Evidence, Sorted by Which Version of the Program Was Tested

The most-studied cash transfer in the developing world was long found to leave work essentially untouched, and the real argument was always about formality, not effort. But the program has since been multiplied in value and reshaped three times, and the newest evaluations — including a clean causal test of the 2022 increase — are reading a different program.

Apollo · AI research assistant · Prepared for Luiz Piccini

Independent synthesis Version 1.0 Not peer reviewed Open access

Abstract

The pre-2019 evidence on Brazil's Bolsa Família was remarkably consistent: the transfer did not reduce aggregate labor supply, and the live question was whether means-testing on formal income nudged beneficiaries toward informal work. Since 2019 the benefit's real value has roughly quintupled and the program has been renamed, rebuilt, and revalued. This brief sorts the literature by the program version each design actually tests. The strongest causal test of the modern program — the August 2022 increase of the floor from R$400 to R$600 — finds a small reduction in participation (2.2–4.7 points), no effect on hours, and no effect on informality, with the exit concentrated among people leaving precarious or informal jobs and women with caregiving duties. A rival reading of the 2023 expansion finds larger declines concentrated in men. The honest conclusion is not that the old consensus was wrong; it is that a much more generous, differently designed program is now being measured, and the two readings disagree about who adjusts and by how much.

Central finding

The pre-2019 evidence that Bolsa Família did not reduce work was about a smaller program; the cleanest causal test of the modern benefit increase (2022, R$400 to R$600) finds only a small participation drop, no hours effect, and no informality effect — but a rival reading of the 2023 expansion finds larger declines in men, so the modern program's labor effect remains genuinely contested.

The question, and why it keeps returning

Bolsa Família is the largest means-tested cash transfer in the Americas and one of the most evaluated social programs anywhere. The charge most often made against it is simple: that a guaranteed transfer dulls the incentive to work. For two decades the empirical answer was an unusually firm no. A wave of studies through the 2010s found that the program did not reduce aggregate labor force participation, and several found a small positive effect among women. In August 2025, a widely covered economics study declared that this had changed — that "for every two families receiving the benefit, one leaves the labor force." The claim is striking, but it is stated without a version: it does not say which program was tested.

That omission is the whole puzzle. Bolsa Família has not been one program. Between 2019 and 2026 its real value roughly quintupled, it was suspended in its conditionality during the pandemic, renamed Auxílio Brasil, rebuilt under a new law, and revalued by 15% in September 2026. A labor-supply estimate is only interpretable against the dose, the eligibility lines, and the withdrawal rules in force when it was measured. This brief does what the public argument rarely does: it sorts the evidence by the program version each design actually tests, separates observation from causal estimate from extrapolation, and states where the newest studies genuinely conflict.

What Brazil pays, version by version

Version In force What changed for work incentives
Original Bolsa Família (Lei 10.836/2004) 2004 – 2021 Eligibility means-tested on declared per-capita income, in practice read through formal-sector records; a transfer withdrawn as formal income rises. Almost all the classic labor-supply evidence is on this version.
"Regra de Permanência" (Portaria GM/MDS 617/2010) 2010 onward A family whose per-capita income rises above the line but stays under half a minimum wage could keep the benefit for up to two years. This is the first explicit attempt to soften the withdrawal cliff.
Pandemic program Mar 2020 – 2021 Conditionality penalties suspended; emergency aid (Lei 13.982/2020) could substitute for the transfer; a floor brought beneficiary families toward R$300.
Auxílio Brasil (MP 1.061/2021 → Lei 14.284/2021) Nov 2021 – 2023 Statute replaced; benefit floor raised first to R$400 and then, in August 2022, to R$600 — the single cleanest modern "more money" shock.
Rebuilt Bolsa Família (Lei 14.601/2023) 2023 – present R$600 family floor, R$142 per person, R$150 per child aged 0–7, a R$50 variable benefit, eligibility line R$218 per person. Adds the Regra de Proteção (retain 50% of the benefit above the line) and Retorno Garantido (automatic return on income loss).
Tightened protection (MDS, July 2025) Payroll from Jul 2025 For new entrants, protection shortened from 24 to 12 months and the ceiling cut from R$759 to R$706 per person; only two months where income is stable (pension, BPC).
Revalued program (Decreto 13.120/2026) Payroll from Oct 2026 +15.04%; average benefit near R$777; ~19.3 million households.

Two facts in that ledger matter for everything below. First, the modern benefit is far larger in real terms than the one most studies tested. Second, the withdrawal rules have oscillated: the 2023 rebuild softened the cliff, and the 2025 tightening partly re-steepened it. Any statement that "Bolsa Família reduces work" must therefore specify which rule and which value.

What the evidence shows

Study Design and program version Key estimate
Barbosa & Corseuil 2013 Original program; occupational-choice analysis Little to no distortion of the formal/informal choice
de Brauw, Gilligan, Hoddinott & Roy 2015 Original program; EDCC No aggregate labor-supply disincentive; sizeable hour reallocation from formal to informal work (urban), and from women to men (rural), plausibly driven by means-testing on formal income
Fruttero, Leichsenring & Paiva 2020 (IMF) Original program; unannounced eligibility-rule change Positive effect on formal employment, stronger for younger beneficiaries
Gerard, Naritomi & Silva 2021 (World Bank) Original program; 2009 expansion across municipalities Increased local formal employment, concentrated in non-tradables; no wage change holding composition fixed; cost ≈ US$9,799 per formal job-year
Gerard & Gonzaga 2021 (AEJ) Original program; UI interaction Formal-sector informality responds to program design; the formal/informal margin is the efficiency-relevant one
Vidigal 2026 (J. Dev. Effectiveness) Original program; PNAD 2001–2009 fixed effects Raises the extensive margin, lowers the intensive margin; informal participation +4.2 pp, formal −3.4 pp
Best, Lobel & Pinho Neto 2026 (NBER) 2012 extreme-poverty top-up; regression discontinuity Employment +≈5%, earnings +5%, hospitalizations −8%, mortality −14% — "productive inclusion"
Vale & Soares 2025 (Ipea NT 78) Aug 2022 floor increase R$400→R$600; PSM + fixed effects Participation −2.2 to −4.7 pp; hours 0 to −1/week (n.s.); no informality effect; exits concentrated among prior informal/precarious workers and women with caregiving duties
Hecksher 2025 (Ipea Radar 78) 2019–2023 expansion; PNAD-C descriptive Despite real value multiplying ~five-fold, formal employment and real labor income rose most among the poorest; consistent with a labor-demand expansion dominating a modest supply retreat
Duque 2024/2025 (FGV Ibre) 2023 eligibility-line change R$210→R$218; eligible vs near-ineligible Participation −11%, employment −12%, formal employment −13% among the newly eligible; concentrated in men, North and Northeast
Silva 2026 (Estudos Econômicos) Rebuilt program; PNAD-C 2023 Q2 causal forest Beneficiaries less likely employed, especially formally; fewer hours; effects largest for women, age 30+, urban
Fietz, Morgandi, Lyrio de Oliveira & Superti 2024 Exit dynamics Local employment growth accelerates exits from the program
Nazareno & Galvão 2023 COVID-19 period Cash transfers cut poverty and inequality without evidence of reduced employment

The mechanism that is actually at stake: work versus formality

The strongest old result was never that beneficiaries work the same amount; it was that they work the same amount in a different place. Because eligibility is means-tested on income that is observable to the state — in practice formal-sector wages — a transfer creates an implicit tax on formal employment, not on work as such. de Brauw and colleagues made this explicit: they found no aggregate labor-supply disincentive, but a clear reallocation of hours from formal to informal work, and the reallocation was largest where the means test bites hardest. Vidigal 2026 reaches a compatible conclusion from a different data window (informal participation up, formal down). Gerard and Gonzaga 2021 give the design principle its cleanest economic statement: under a program keyed to formal contracts, the relevant distortion is the formal/informal wedge.

This reframing matters because the public argument usually claims the wrong thing. The credible version of the "disincentive" concern is not that the poor stop working; it is that a means test read through formal payroll creates a reason to prefer informal arrangements, which are exactly the arrangements with no severance, no unemployment insurance, and no pension. That is a real efficiency cost, and it is the one worth debating.

The modern evidence complicates the mechanism rather than confirming it. Vale and Soares (2025), using the August 2022 increase from R$400 to R$600 as a shock, find the participation reduction that the anti-transfer camp predicts — but find no informality effect at all. Where people left the labor force, they left precarious informal jobs or unemployment, and the leavers were disproportionately women with caregiving responsibilities. On this reading, a more generous benefit let people exit bad work, not work in general, and did not push anyone toward informality. This directly undercuts the formality-distortion story for the modern, larger program.

Where the evidence genuinely conflicts

The honest state of the field is not a consensus. It is two serious, well-identified studies reaching different magnitudes and different subgroups.

  • The Ipea reading (small, benign): the 2022 increase cut participation by 2.2–4.7 points, did nothing to hours, and nothing to informality; the adjustment was women leaving bad jobs. Hecksher's descriptive work adds that over 2019–2023 formal employment and real labor income rose fastest among the poorest even as the benefit's real value multiplied — consistent with labor demand expanding faster than any supply retreat.
  • The FGV Ibre reading (larger, maligned): using the 2023 eligibility expansion (the line rose from R$210 to R$218), Duque finds participation down 11%, employment down 12%, and formal employment down 13% among the newly eligible, concentrated in men in the North and Northeast. If correct, this is a material negative effect.

The two designs are not variants of one test. Vale and Soares study a big change in the value of the benefit; Duque studies a small change in the eligibility line that sharply expanded coverage. A value shock and a coverage shock can move behavior differently, and both are legitimate. But it is also fair to note the asymmetry in what each is being asked to carry: Duque's headline ("one in two families") is a big causal claim built on a two-real-per-person line change, and one of his own institute's labor economists publicly cautioned that the earlier program had been unfairly attacked on the same grounds. The Ipea result, meanwhile, is a small, precisely bounded effect — the kind that is easy to under-report because it is unexciting.

A third body of work points the other way entirely. Best, Lobel and Pinho Neto (2026) isolate the 2012 top-up for the extremely poor with a regression discontinuity and find a positive labor effect: employment up about 5%, earnings up about 5%, plus large health gains. Gerard, Naritomi and Silva (2021) find that the original program's municipal expansion raised local formal employment. So across three distinguishable research programs — original program, 2012 generosity reform, 2022 value increase — the modal finding is "little harm, sometimes help," and the outlying negative estimate comes from one coverage-expansion design whose subgroup signal (men, North/Northeast) does not obviously match the mechanism (caregiving women) that the value-shock design identifies.

The design fix that already exists

The interesting policy fact is that Brazil has already instrumented against the formality wedge. The Regra de Proteção lets a family whose income rises above the line keep half the benefit for a defined period, and Retorno Garantido restores full benefits automatically if income falls. In incentive terms, both are attempts to lower the effective marginal tax at the eligibility threshold — the precise margin that de Brauw and Gerard and Gonzaga identify as the costly one. Whether they work is, again, an empirical question, and the 2025 tightening (24 months down to 12, ceiling down from R$759 to R$706) reduced the generosity of that buffer for new entrants while leaving the earlier cohort on the old rules. The result is a program running two withdrawal regimes at once, which is itself a natural experiment nobody has yet used.

What the national numbers can and cannot say

Brazil's aggregate data are genuinely encouraging: through 2024–2026 the unemployment rate fell to historic lows and formal employment grew even as the transfer's reach stayed near 19–20 million households. Descriptive decompositions — including the government's own tabulations of formal-job registrations — show that a large share of net formal hirings went to Cadastro Único registrants, and a majority to program beneficiaries. These are real facts. They are also not causal. The beneficiaries of a hot labor market and the beneficiaries of a cash transfer overlap heavily, and a high share of formal hires being registrants is at least as consistent with strong labor demand and active formalization as with the transfer changing anyone's behavior. Aggregate national series can falsify the strongest anti-transfer claim — a program that emptied the labor force would show up — but they cannot separate the program from the labor market it sits inside.

Verdict and calibrated confidence

On the pre-2019, smaller program: ~85% that it did not reduce aggregate labor supply, and ~75% that it did shift some hours toward informality. The old consensus looks robust to design and data window for the program as it then existed.

On the modern, much larger program: genuinely unsettled. I put ~65% on the proposition that the modern benefit produces at most a small (order of a few percentage points) participation reduction concentrated in secondary earners, and ~35% on a materially larger negative effect of the kind the FGV Ibre reading implies. I put only ~25% on the stronger claim that the modern program measurably increases informality: the one clean causal test of a large value increase found no informality effect, which is weak evidence against, though it tests a value shock rather than a coverage expansion.

What would change my mind: (1) a pre-registered replication of the 2022 R$400→R$600 increase extended to outcomes through 2024, with formal/informal status measured directly rather than inferred; (2) an evaluation that uses the two coexisting Regra de Proteção regimes (pre- and post-July-2025 cohorts) as a withdrawal-cliff experiment, the cleanest available test of whether the formality wedge responds to design; (3) convergence between the value-shock and coverage-shock designs on the same subgroup. Until at least one lands, the correct verb tense is the one the field keeps avoiding: the small program probably did not discourage work, and the large one is still being measured.

The study that would settle it

The design is available, bounded, and cheap relative to the program. Brazil already links Cadastro Único and payment records to formal-employment registers and to PNAD Contínua. Two moves would resolve most of the disagreement. First, estimate labor outcomes around the August 2022 value increase and the 2023 eligibility expansion in one framework, so the value shock and the coverage shock cannot be smuggled into each other's conclusions. Second, treat the July 2025 shortening of the protection rule — which left an earlier cohort on a more generous withdrawal schedule — as a discontinuity in the effective tax rate, and ask whether formality responds. Report informal status directly, pre-register the subgroup cuts, and state the estimator's sensitivity to the labor-market cycle, because the deepest problem with this literature is not a missing effect; it is that everyone is measuring a different program and calling it the same name.

References

  1. de Brauw A, Gilligan DO, Hoddinott J, Roy S. Bolsa Família and Household Labor Supply. Economic Development and Cultural Change. 2015;63(3):423–457. doi:10.1086/680092.
  2. Vidigal VG. The Impact of Conditional Cash Transfers on Labour Supply: New Evidence from Brazil's Bolsa Família. Journal of Development Effectiveness. 2026;18(1):86–103. doi:10.1080/19439342.2025.2611036.
  3. Barbosa ALNH, Corseuil CHL. Bolsa Família, Occupational Choice and Informality in Brazil. IPC-IG One Pager No. 226. Brasília: Ipea/UNDP; 2013. https://repositorio.ipea.gov.br/handle/11058/14908.
  4. Fruttero A, Ribeiro Leichsenring A, Henrique Paiva L. Social Programs and Formal Employment: Evidence from the Brazilian Bolsa Família Program. IMF Working Paper 2020/099. 2020. doi:10.5089/9781513547701.001.
  5. Gerard F, Naritomi J, Silva J. Cash Transfers and Formal Labor Markets: Evidence from Brazil. World Bank Policy Research Working Paper 9778. 2021. doi:10.1596/1813-9450-9778.
  6. Gerard F, Gonzaga G. Informal Labor and the Efficiency Cost of Social Programs: Evidence from Unemployment Insurance in Brazil. American Economic Journal: Economic Policy. 2021;13(3):167–206. doi:10.1257/pol.20180072.
  7. Best M, Lobel F, Pinho Neto V. Cash Transfers and Productive Inclusion: Evidence from Bolsa Família. NBER Working Paper 35006. 2026. doi:10.3386/w35006.
  8. Silva RR da. Os efeitos do Programa Bolsa Família (PBF) no mercado de trabalho: uma análise com Causal Forest. Estudos Econômicos (São Paulo). 2026;56(2). doi:10.1590/1980-53575625rrs.
  9. Fietz K, Morgandi M, Lyrio de Oliveira G, Superti LH. Does Local Employment Growth Accelerate Exits from Social Assistance? Evidence from Brazil's Conditional Cash Transfer Bolsa Família. World Bank Policy Research Working Paper 10676. 2024. doi:10.1596/1813-9450-10676.
  10. Nazareno L, de Castro Galvão J. The Impact of Conditional Cash Transfers on Poverty, Inequality, and Employment During COVID-19: A Case Study from Brazil. Population Research and Policy Review. 2023;42(2):22. doi:10.1007/s11113-023-09749-3.
  11. Vale RC, Soares FV. O Efeito do aumento no valor das transferências de renda sobre a inserção dos beneficiários no mercado de trabalho: uma análise com dados em painel da PNAD-Contínua. Ipea, Nota Técnica Dinte No. 78 (Publicação Expressa). Brasília; set. 2025. https://repositorio.ipea.gov.br/handle/11058/18706.
  12. Hecksher M. Notes on Bolsa Família Expansion, Employment, and Poverty. Ipea/IPC-IG Policy Research Brief. 2025. https://repositorio.ipea.gov.br/entities/publication/b993db65-d3a6-411e-91be-cfa21fdca1d1.
  13. Hecksher M. Observações sobre emprego, Bolsa Família e metas de redução da pobreza. Ipea, Radar No. 78. 2025. https://repositorio.ipea.gov.br/items/87e50fed-7e84-48d4-8e83-a3d26f52d038.
  14. Duque D. Expansão do Bolsa Família desencoraja participação no mercado de trabalho de vulneráveis. Blog do IBRE/FGV. 26 August 2024. https://blogdoibre.fgv.br/posts/expansao-do-bolsa-familia-desencoraja-participacao-no-mercado-de-trabalho-de-vulneraveis.
  15. Schymura LG (ed.). Bolsa Família ampliado reduz oferta de trabalho, mas pode ser aperfeiçoado. Carta do IBRE/FGV. August 2025. https://portalibre.fgv.br/revista-conjuntura-economica/carta-da-conjuntura/bolsa-familia-ampliado-reduz-oferta-de-trabalho.
  16. Brasil. Lei nº 14.601, de 19 de junho de 2023 (rebuilt Bolsa Família; Regra de Proteção and Retorno Garantido).
  17. Brasil. Portaria GM/MDS nº 617, de 11 de agosto de 2010 (Regra de Permanência); Ministério do Desenvolvimento e Assistência Social, July 2025 revision of the protection rule (12 months, R$706 ceiling).
  18. Brasil. Decreto nº 13.120, de 17 de setembro de 2026 (15.04% benefit revaluation).