piccini papers
Papers / PIC:2610.04 Evidence reviewed through 15 October 2026

Social protection · Child mortality · Evidence synthesis

Does Bolsa Família Save Children's Lives — and Does the Answer Apply to the Program Brazil Pays Today?

Brazil's flagship cash transfer is credited with averting hundreds of thousands of deaths. Convergence across municipal panels and individual-level cohorts now puts the effect on under-five mortality near a 13–17% reduction. But almost every mortality evaluation stops observing outcomes in 2019 — before the pandemic suspension, the Auxílio Brasil interlude, the 2023 rebuild, and the revaluation that took effect this month.

Apollo · AI research assistant · Prepared for Luiz Piccini

Independent synthesis Version 1.1 Not peer reviewed Open access

Abstract

Bolsa Família is one of the largest conditional cash transfers in the world and the most studied. Convergence across municipal panels and an individual-level cohort puts its effect on under-five mortality in the range of a 13–17% reduction during the program's first two decades, and a 2026 working paper using the 2012 generosity reform finds a 14% fall in all-cause mortality among the extremely poor — from a research group independent of the Brazilian panels. But every quantitative mortality evaluation measures exposure and outcomes that end in 2019 or earlier. The program has since been suspended in its conditionality enforcement, renamed and restructured as Auxílio Brasil, rebuilt under a new law, and — for the first time since 2023 — revalued by 15% in September 2026. The evidence says the pre-2020 programme probably saved children's lives. It does not yet say the same about the programme Brazil pays today.

Central finding

Converging evidence — now including an independent 2026 study of the 2012 reform — suggests Bolsa Família reduced under-five mortality by roughly 13–17% under the pre-2020 programme, but no evaluation observes a single outcome year after 2019, so the effect of the current 2023–26 version is unmeasured.

The question, and the two ledgers it needs

In September 2026 Brazil raised the value of Bolsa Família by 15%, the first revaluation since 2023. The national cash-transfer programme now reaches about 19.3 million households — roughly 50 million people — at an average benefit near R$678 a month before the increase and about R$777 after. It is one of the largest conditional cash transfers (CCTs) in the world and one of the most evaluated social programmes anywhere.

The natural question is whether it keeps children alive. That question sounds simple, but answering it honestly requires two ledgers that are usually kept apart. The first is a policy ledger: Bolsa Família has not been one programme for the last six years. It was suspended in its conditionality enforcement during the pandemic, replaced by Auxílio Brasil in 2021, rebuilt under a new legal design in 2023, and just revalued. The second is an outcome ledger: for each study, what is the latest year in which a death could actually have been observed? Vital statistics in Brazil now run through 2024. The evidence base, as this brief shows, stops years earlier.

The puzzle is the collision between those two ledgers: the confidence in public commentary far exceeds the recency of the measurements it rests on. This brief states what is measured, what is inferred, and where the two separate.

What Brazil pays, version by version

Version In force What changed Child-mortality evidence
Original Bolsa Família (Lei 10.836/2004) 2004 – Aug 2021 Fixed benefit schedule; eligibility lines at R$89 (extreme poverty) and R$178 per person as of 2020; health and school conditionality enforced Yes — all quantitative evaluations described below
Pandemic-era original programme Mar 2020 – Aug 2021 Conditionality non-compliance effects and blocking suspended by Portaria/MC 335 (20 Mar 2020, 120 days), extended by Portaria 443 (17 Jul 2020, 180 days); school-attendance data for Feb–Mar 2020 not collected; emergency aid (Lei 13.982/2020) could replace Bolsa Família where more advantageous; a residual top-up brought beneficiary families to a R$300 floor from Sept 2020 None directly — evaluation windows end earlier
Auxílio Brasil (MP 1.061 of 9 Aug 2021, converted to Lei 14.284/2021; payments from 17 Nov 2021) Nov 2021 – 2023 Revoked the Bolsa Família statute, reset the benefit structure, raised lines to R$105/R$210, later a R$400 minimum; conditionality enforcement still unsettled; a transitional compensation benefit cushioned families whose value fell None for mortality
Rebuilt Bolsa Família (Lei 14.601/2023, 19 Jun 2023) 2023 – present Restored health and education commitments; per-person R$142, a floor taking families to R$600, R$150 for children 0–7, a R$50 variable benefit; eligibility line R$218 per person; a protection rule keeps half the benefit for up to a year for families just above the line (2.26 million families in mid-2026) None for mortality
Revalued programme (Decreto 13.120, 17 Sep 2026) Payroll from Oct 2026 +15.04% (INPC Mar 2023–Aug 2026): per-person R$164, child R$173, variable R$58, floor R$691; average benefit R$777; never a reduction. The Supreme Court (MI 7300, 18 Sep 2026) recognized compliance None — too recent

This ledger is not pedantry. Each row changes the dose, the conditionality, or the counterfactual, and a health effect estimated on one row does not automatically transfer to another.

What the evidence shows

The quantitative literature is unusually broad in design for social policy. The studies below were all verified in full this run.

Study Design and window Key estimate
Shei 2013, Health Affairs Municipal panel, programme rollout ~2003–2008 Infant mortality fell ~9.3% during the programme's first five years; larger postneonatal than neonatal effects; larger where primary-care coverage was higher
Guanais 2013, AJPH 4,583 municipalities, 1998–2010 Post-neonatal mortality falls most where Bolsa Família and the Family Health Strategy (ESF) co-expand; the primary-care effect grows with cash-transfer coverage
Rasella 2013, The Lancet 2,853 municipalities, 2004–2009, fixed effects Under-five mortality rate ratio 0.94 (intermediate coverage) and 0.88 (high); consolidated 0.83 (0.79–0.88); malnutrition 0.35 (0.24–0.50); diarrhoea 0.47 (0.37–0.61); external causes (negative control) null
Silva & Paes 2019, Ciência & Saúde Coletiva 1,133 semi-arid municipalities, 2004–2010 Bolsa Família and the ESF each independently associated with lower child mortality
Ramos 2021, PLOS Medicine Individual-level: 6.3 million children registered 2006–2015, 4.86 million beneficiaries, 14,897 linked deaths Weighted odds ratio 0.83 (0.79–0.88); preterm 0.78, mothers recorded as Black 0.74, poorest quintile 0.72, best-managed municipalities 0.76; null in the richest quintile
Aransiola 2023, Lancet Regional Health — Americas Municipal panel, 2004–2019 CCT coverage: infant rate ratio 0.868, child 0.874; modestly stronger in poorer municipalities; projections to 2030
Aransiola et al. 2024, JAMA Network Open 2,548 municipalities, 2004–2019, microsimulation to 2030 The three programmes together (Bolsa Família, the social pension BPC, and the ESF) associated with all-cause rate ratio 0.95 and under-five 0.87; an estimated 1.46 million deaths averted over 2004–2019
Cavalcanti et al. 2025, Lancet Public Health 3,671 municipalities, 2000–2019, plus modelling to 2030 High coverage: overall mortality rate ratio 0.824; high benefit adequacy 0.849; coverage and adequacy together, under-five rate ratio 0.67 (0.65–0.69); 713,083 deaths and 8.2 million hospitalisations averted 2000–2019

Three features carry weight. First, dose–response: mortality falls more as coverage rises, and more again when benefits are adequate — the pattern an effective mechanism should produce. Second, cause specificity: the strongest associations are for malnutrition and diarrhoea, the pathways a cash transfer plausibly moves, while external causes show no association. Third, subgroup coherence: effects concentrate among poorer, preterm and better-governed municipalities. The largest, most recent study extends this to overall mortality and finds the same gradient.

An independent design — and the limit of what it shows

Every evaluation above comes from one research lineage. A 2026 working paper by Best, Lobel and Pinho Neto breaks from it. Exploiting the 2012 reform that raised Bolsa Família's generosity for the extremely poor, they find the reform raised employment by 5% and — the result that matters here — cut hospitalisations by 8% and all-cause mortality by 14%, roughly 1,000 lives. The authors are an independent economics team, and the finding challenges the common view that cash transfers reduce work.

Two limits bound it. First, the outcome is all-cause mortality among the extremely poor, not child mortality, so it supports the general claim that the transfer saves lives without measuring children specifically. Second, it is an unreviewed working paper, and its central argument concerns labour supply, with the mortality result secondary. Read as a robustness check, it is welcome: it is the first mortality evidence in this literature from outside the Bahia/ISGlobal research group. Read as confirmation of the child-mortality magnitude, it is silent.

The counterevidence that will not go away

The picture is not unanimous. The clearest contrary finding comes from Dyngeland, Oldekop and Evans (2020, PNAS), who built a rural-municipality panel for 2004–2013 and applied covariate-balancing weights. Using the poorer-population administrative health system (SIAB), they found that higher Bolsa Família investment was associated with higher infant mortality (coefficient 0.16 ± 0.05, p = 0.002; mean predicted change +59.4%, about 725 additional deaths per 100,000 live births) and with more child malnutrition (+67.7%). Their census-based estimate was null (0.05 ± 0.22). Critically, they report that the positive mortality effect "was not detectable" once municipalities with lower-quality data were excluded — the effect is fragile to data quality.

A second strand is nutritional. Labrecque and colleagues (2018, Social Science & Medicine), analysing 1,703 Bolsa Família–eligible children from the 2004 Pelotas birth cohort with doubly robust estimation, found that the programme was associated with lower length-for-age and weight-for-age at 24 months: length-for-age −0.14 (low exposure) and −0.20 (high), weight-for-age −0.18 (high). Only a strong unmeasured confounder could explain it away, the authors argue — but it runs opposite to the most-cited mechanism. A 2024 scoping review of 94 studies of Brazilian cash transfers found consistent positive impacts on poverty, employment, child and adult mortality, schooling and violence outcomes, with the Dyngeland mortality result standing as a rare dissent; the review's own conclusion is that more studies are needed to confirm several results.

The nutritional evidence is more mixed than either side implies, and one large new study cuts both ways. Falcão and colleagues (2025, BMJ Global Health) followed 3,116,138 children born 2008–2015 in a CadÚnico-linked birth cohort and found Bolsa Família participation associated with 17% lower stunting (OR 0.83, 95% CI 0.81–0.85) — strongest among children of less-educated mothers and in rural areas — but also 19% higher wasting (OR 1.19, 1.16–1.22) and, in urban or better-educated subgroups, higher overweight/obesity. The dual result fits Brazil's rapid nutrition transition and the "double burden" of malnutrition. It does not vindicate the simple more cash, better growth story, and it does not support Labrecque's blanket negative either.

There are also structural reasons for caution. Nearly all the mortality evidence is ecological: municipalities are the unit, and coverage is credited with co-moving mortality while time-varying confounders are addressed only by fixed effects. The Family Health Strategy is the largest of these: cash transfers and primary care expanded together, and Guanais's own conclusion is that the primary-care effect depends on cash-transfer coverage. The individual-level study is stronger in design but is confined to CadÚnico registrants — a poorer population than Brazil as a whole — and its authors report a striking sensitivity: in municipalities where no nameless death records were lost to linkage, the association changed direction, albeit with wide intervals.

Why the evidence base is thinner than it looks

The strongest evaluations are not independent tests. Several of the most prominent — Rasella 2013, Aransiola 2023 and 2024, Cavalcanti 2025 — share authorship, municipal-panel construction, and the same group's modelling conventions (Federal University of Bahia / ISGlobal / CIDACS). They extend one another's windows, which is useful, but they do not supply independent replication. The largest magnitudes (a 33% under-five reduction in the 2025 study) come from models with explicit, and reasonable, structural assumptions. Read as a family, they are a convergent research programme rather than a pile of separate confirmations. That is worth stating plainly on both sides: it inflates neither the certainty nor the risk.

The 2026 Best–Lobel–Pinho Neto paper is the one mortality result here from outside that family, which is why it carries more weight than its page count suggests — even though it measures all-cause, not child, mortality. It eases the lineage worry; it does not remove it. One independent test is not the several this question still needs.

What no study covers: the version boundary

Here is the fact this brief was written to establish. Every quantitative mortality evaluation found this run observes outcomes that end in 2019 or earlier. Rasella 2004–2009; Shei ~2003–2008; Guanais 1998–2010; Silva and Paes 2004–2010; Ramos registrations 2006–2015; Aransiola 2004–2019; the JAMA and Lancet Public Health studies 2000/2004–2019; even the independent 2026 study measures the 2012 reform, not the current programme. None covers Auxílio Brasil, none covers the 2023 rebuild, and none covers the September 2026 revaluation.

That boundary falls exactly where the programme changed. Brazil suspended conditionality penalties in March 2020 and kept suspending them through 2021; school-attendance data for the first pandemic months were never collected; emergency aid could substitute for the transfer; the statute was replaced in 2021 and then again in 2023; and the benefit is only now being revalued. The programme measured by the evidence is not a different programme by accident — it is a different programme by law. The strong inference most commentary makes — that Bolsa Família as it exists today is protecting children — is an extrapolation from a version that no longer exists.

What the national numbers can and cannot say

Brazil's own vital statistics now reach 2024. A June 2026 Fiocruz/CIDACS report puts infant mortality at 12.3 per 1,000 live births in 2024, down from 28.1 in 2000, with life expectancy rising from 71.1 to 76.6. That is a genuine, large national improvement over a quarter-century. It cannot be attributed to the 2023–26 programme: the decline is mostly a long-run trend that predates both the pandemic and the recent redesign, and a separate analysis of 2020–2021 found under-five deaths fell during the suspension years while stillbirths rose — a pattern that reflects disrupted care and reporting as much as any programme effect. National aggregate data are, at best, a sanity check. They cannot separate the programme from everything moving alongside it.

Two figures also make clear that the current version differs materially in dose: an average benefit near R$777 after the 2026 revaluation, reach of roughly 50 million people, and a protection rule for families just above the line. Whether more money per family, with conditionality restored, translates into more survival than the pre-2020 schedule did is an empirical question. Nobody has answered it.

Verdict and calibrated confidence

On the pre-2020 programme: ~90% that it reduced under-five mortality. The four strongest supports are convergence across two designs, a coverage/adequacy dose–response, cause-specificity consistent with the proposed mechanism, and subgroup coherence (poorer and better-managed municipalities benefit more), now reinforced by an independent 2026 mortality result. The discount reflects the still-shared research lineage of the child-specific studies, the ecological design of most of them, the ESF entanglement, and the fact that one well-executed study found the opposite sign.

On the current programme: unmeasured, not safe. My subjective ~75% that a well-powered evaluation of 2023–2026 exposure and 2024 outcomes would find a protective association of the same sign is inference, not evidence, and I label it as such. Several things could make it smaller today — a different benefit formula, restored conditions, a higher income floor nationally — or larger, given the revaluation's real-terms increase.

What would change my mind: (1) an individual-level study replicating Ramos's design with enrolment cohorts from 2019 on and deaths through 2024, spanning the Auxílio Brasil switch; (2) a municipal panel extended past 2019 that treats the November 2021 restructuring and the September 2026 revaluation as dose shocks; (3) a documented, pre-registered national evaluation rather than a retrospective one. Until at least one exists, "we know Bolsa Família saves children's lives" should be spoken in the past tense.

The study that would settle it

The design is available and cheap relative to the programme's size: link CadÚnico/household-payment records to the mortality information system for cohorts paid under each version — original, Auxílio Brasil, rebuilt, revalued — and estimate mortality against the November 2021 and September 2026 discontinuities. Two safeguards matter. First, pre-register the analysis and the negative-control outcomes (external causes) so that the ecological association cannot be tuned post hoc. Second, report linkage-loss sensitivity as Ramos did, because nameless death records are exactly where this kind of study can quietly flip. Brazil already pays for the programme and collects the data. It has not yet joined them across the boundary that matters most.

References

  1. Rasella D, Aquino R, Santos CAT, Paes-Sousa R, Barreto ML. Effect of a conditional cash transfer programme on childhood mortality: a nationwide analysis of Brazilian municipalities. The Lancet. 2013;382(9886):57–64. doi:10.1016/S0140-6736(13)60715-1.
  2. Shei A. Brazil's Conditional Cash Transfer Program Associated With Declines In Infant Mortality Rates. Health Affairs. 2013;32(7):1274–1281. doi:10.1377/hlthaff.2012.0827. PMID 23836744.
  3. Guanais FC. The Combined Effects of the Expansion of Primary Health Care and Conditional Cash Transfers on Infant Mortality in Brazil, 1998–2010. American Journal of Public Health. 2013;103(11):2000–2006. doi:10.2105/AJPH.2013.301452. PMID 24028257.
  4. Silva ES da, Paes NA. [Effects of the Bolsa Família program and the Family Health Strategy on child mortality in the semiarid region of Brazil]. Ciência & Saúde Coletiva. 2019;24(2):623–630. doi:10.1590/1413-81232018242.04782017. PMID 30726394.
  5. Ramos DO, et al. Conditional cash transfer program and child mortality: a cross-sectional analysis nested within the 100 Million Brazilian Cohort. PLOS Medicine. 2021;18(9):e1003509. doi:10.1371/journal.pmed.1003509.
  6. Aransiola TJ, Ordoñez JA, Cavalcanti DM, et al. The combined effect of social pensions and cash transfers on child mortality: evaluating the last two decades in Brazil and projecting their mitigating effect during the global economic crisis. Lancet Regional Health — Americas. 2023;27:100618. doi:10.1016/j.lana.2023.100618. PMID 38029069.
  7. Aransiola TJ, Cavalcanti D, Ordoñez JA, et al. Current and Projected Mortality and Hospitalization Rates Associated With Conditional Cash Transfer, Social Pension, and Primary Health Care Programs in Brazil, 2000–2030. JAMA Network Open. 2024;7(4):e247519. doi:10.1001/jamanetworkopen.2024.7519. PMID 38648059.
  8. Cavalcanti DM, Ordoñez JA, da Silva AF, et al. Health effects of the Brazilian Conditional Cash Transfer programme over 20 years and projections to 2030. The Lancet Public Health. 2025;10(7):e548–e558. doi:10.1016/S2468-2667(25)00091-X. PMID 40451230.
  9. Dyngeland C, Oldekop JA, Evans KL. Assessing multidimensional sustainability: Lessons from Brazil's social protection programs. PNAS. 2020;117(34):20511–20519. doi:10.1073/pnas.1920998117. PMID 32788353.
  10. Labrecque JA, Kaufman JS, Balzer LB, et al. Effect of a conditional cash transfer program on length-for-age and weight-for-age in Brazilian infants at 24 months using doubly-robust, targeted estimation. Social Science & Medicine. 2018;211:9–15. doi:10.1016/j.socscimed.2018.05.040. PMID 29879565.
  11. Magalhães J, Ziebold C, Evans-Lacko S, Matijasevich A, Paula CS. Health, economic and social impacts of the Brazilian cash transfer program on the lives of its beneficiaries: a scoping review. BMC Public Health. 2024;24(1):2818. doi:10.1186/s12889-024-20046-2.
  12. Ministério da Cidadania. Portaria nº 335, de 20 de março de 2020 (suspension of conditionality effects, 120 days). Diário Oficial da União.
  13. Ministério da Cidadania. Portaria nº 443, de 17 de julho de 2020 (extension, 180 days).
  14. Ministério da Cidadania / SAGI. Portaria nº 682, de 6 de outubro de 2021 (further suspension, 180 days).
  15. Brasil. Lei nº 13.982, de 2 de abril de 2020 (emergency aid, Lei 13.982/2020 art. 2).
  16. Brasil. Medida Provisória nº 1.061, de 9 de agosto de 2021 (creation of Auxílio Brasil); converted into Lei nº 14.284/2021.
  17. Brasil. Lei nº 14.601, de 19 de junho de 2023 (rebuilt Bolsa Família).
  18. Brasil. Decreto nº 13.120, de 17 de setembro de 2026 (benefit revaluation of 15.04%); see also Supremo Tribunal Federal, MI 7300 decision, 18 September 2026.
  19. Ministério do Desenvolvimento e Assistência Social. Informes Bolsa Família (coverage, 2026); Cenário da Infância e Adolescência no Brasil 2026 (report, 2026).
  20. Centro de Integração de Dados e Conhecimentos para Saúde (CIDACS/Fiocruz). Efeitos do Programa Bolsa Família na Saúde da População Brasileira. Technical report, 2 June 2026. https://cidacs.bahia.fiocruz.br/wp-content/uploads/2026/06/PBF_ptbr_02JUN26_ptbr-finalizado.pdf
  21. Best MC, Lobel F, Pinho Neto V. Cash Transfers and Productive Inclusion: Evidence from Bolsa Família. NBER Working Paper 35006. March 2026. doi:10.3386/w35006. https://www.nber.org/papers/w35006
  22. Falcão IR, et al. Brazil's Bolsa Família conditional cash transfer and child malnutrition: a nationwide birth cohort study. BMJ Global Health. 2025;10(7):e018431. doi:10.1136/bmjgh-2024-018431. PMID 40685158.