What Actually Saves the Children in a Cash Transfer?

Two claims, one programme

The sentence "cash transfers reduce child mortality" is now conventional. So is its opposite twin: "cash transfers only work because of the conditions and the health workers." Both are asserted with more confidence than the underlying literature can carry, and they are not the same claim.

The two statements also imply different budgets. If the money is doing the work, the policy levers are transfer size and coverage, and administrative conditionality is an expensive monitoring apparatus with a political function. If the conditions and the clinics are doing the work, the money is a delivery vehicle, and cutting monitoring or primary care would be the expensive mistake. Most programmes are sold as a single package precisely because this question is left unresolved.

This brief asks a narrow question: of the money, the strings, and the clinics, which part of the package has the best claim to the child-survival effect? The answer is not symmetric. The money has the clearest evidence and the clearest dose-response. Conditionality has never been shown to save lives on its own, and the only direct comparison finds conditional and unconditional programmes performing similarly on mortality. Complementary health services matter a great deal for the outcomes money cannot buy — growth, anaemia, service use — but they are not what most of the mortality estimates are measuring.

There is a second, less comfortable answer underneath. The mortality evidence is not global. It is Latin American, from programmes that bundled cash with health-system expansion, and the policy claim has travelled further than the measurements have.

The Latin American record is a package, not a pill

Start with the magnitudes, because they are genuinely large and they repeat.

Brazil's Bolsa Família, created in 2003, is the most studied. A municipal fixed-effects panel covering 2004–2009 and 2,853 municipalities with adequate vital statistics found under-five mortality falling steadily as programme coverage rose: rate ratios of 0.94 for intermediate coverage, 0.88 for high coverage, and 0.83 for consolidated coverage (≥100% of the target population for at least four years). For deaths from malnutrition the consolidated-coverage estimate was 0.35, and for diarrhoea 0.47 — a cause pattern that matches the proposed mechanism far better than a generic time trend would (Rasella et al., 2013).

A different design, an earlier window, and a smaller estimate reach the same sign. Analysing Bolsa Família's first five years, Shei (2013) found a 9.3% reduction in infant mortality, with larger declines in postneonatal than neonatal deaths and larger effects where coverage of the Family Health Program was higher — the first hint that the clinic is part of the story rather than scenery.

Mexico's Progresa, evaluated through its phased rollout across rural municipalities, produced a 17% decline in infant mortality among treated communities, no detectable average effect on neonatal mortality, and a benefit-cost ratio between 1.3 and 3.6. Effects were larger where pre-programme mortality was above the median, where illiteracy was higher, and where access to electricity was lower (Barham, 2011).

Ecuador's Bono de Desarrollo Humano adds a coverage gradient of its own: for each percentage point of county coverage, under-five mortality from malnutrition fell by about 3% (rate ratio 0.971, 95% CI 0.953–0.989), with a smaller effect on respiratory deaths (Moncayo et al., 2019). A three-country municipal panel covering Brazil, Ecuador and Mexico from 2000 to 2019 estimated an under-five rate ratio of 0.76 for conditional cash transfer coverage, and attributed roughly 739,000 child deaths averted to the programmes (Cavalcanti et al., 2023).

Two features of this record matter for the question. First, every one of these programmes attached health conditions and expected beneficiaries to use clinics; several also supplied nutritional supplements or required growth monitoring. The treatment was never cash alone. Second, the designs are ecological: municipalities or counties are the unit, coverage is the exposure, and the counterfactual is a fixed-effects construction. That is a reasonable design for a national programme, but it cannot separate cash from the primary-care expansion happening at the same time, and the authors know it — Rasella's models adjust for the Family Health Program rather than isolating it.

The money has a dose-response

If the money is an active ingredient, more money should do more. On the best evidence, it does.

Brazil's 2025 twenty-year evaluation is the cleanest demonstration because it treats coverage and adequacy — average transfer per family — as separate exposures. High coverage was associated with an age-standardised mortality rate ratio of 0.824 (95% CI 0.807–0.842); high adequacy with 0.849 (0.833–0.866). The combination of high coverage and high adequacy produced the largest under-five effect, a rate ratio of 0.67 (0.65–0.69), and the models attributed 713,083 deaths averted between 2000 and 2019 (Cavalcanti et al., 2025). Adequacy doing independent work is the important part: coverage without adequate benefit value is a mailing list, not an income shock.

The same gradient appears inside a single programme. In Mexico's Oportunidades, a doubling of the cash transfer — holding the health and nutrition components constant — was associated with a height-for-age Z-score 0.20 higher (95% CI 0.09–0.30), a 10-percentage-point lower prevalence of stunting, and better motor, cognitive and language scores (Fernald, Gertler and Neufeld, 2008). That is a within-programme comparison, which makes it unusually good evidence that the money itself, not just the programme's existence, changes child outcomes.

And the individual-level Brazilian cohort study, linking 6.3 million children in the unified registry to death records, found participation associated with a weighted odds ratio of 0.83 (0.79–0.88) for mortality at ages one to four, with stronger associations for preterm children and children of Black mothers (Ramos et al., 2021). The subgroup pattern is what a money mechanism predicts: more effect where baseline risk is higher.

The multi-country study reinforces the pattern. Richterman and colleagues, using individual-level longitudinal mortality data from many low- and middle-income countries, found significant reductions in under-five and female adult mortality, with larger effects for programmes covering a larger share of the population and paying larger amounts — and larger effects in countries with lower health expenditure, lower baseline life expectancy, and higher perceived regulatory quality (Richterman et al., 2023).

Note what that last set of modifiers implies. Effects were larger where health spending was lower. If the clinic were the active ingredient, the opposite would be expected. That does not prove cash works without clinics; it suggests the marginal value of income is highest where the public system is thinnest.

The conditionality test is weaker than either side admits

Conditionality is the feature everyone argues about, and it is the one with the thinnest mortality evidence.

The only direct mortality comparison comes from Richterman's secondary heterogeneity analysis: effects were similar for conditional and unconditional programmes. That is a program-level comparison, not a randomised one, and programme type is confounded with country, donor, and administrative capacity. But it is the best available test, and it fails to find the advantage conditionality is supposed to deliver.

Randomised tests of conditionality exist, but they measure behaviour, not death. In Malawi, adolescent girls were randomised to an unconditional transfer, a transfer conditional on school attendance, or a control group. The conditional arm cut dropout substantially more, and did better on reading comprehension. The unconditional arm, however, produced substantially lower pregnancy and marriage rates — entirely among girls who had dropped out of school (Baird, McIntosh and Özler, 2011). Conditionality changed which behaviour improved, not whether the programme helped.

Morocco's labelled cash transfer sharpens the point. A small transfer to fathers, not conditional on attendance but explicitly labelled as education support, produced large gains in school participation. Adding conditionality and targeting mothers "made almost no difference" in that context (Benhassine et al., 2015). The label and the money did the work; the enforcement mechanism did not.

For health, the Cochrane review is the canonical caution. It found that conditional cash transfers improved use of health services, nutritional status and some self-reported illness outcomes, but concluded that "it is hard to attribute these positive effects to the cash incentives specifically because other components may also contribute" (Lagarde, Haines and Palmer, 2009). That was 2009 and the synthesis was qualitative, but the conceptual problem has not been solved by later study designs: conditions are almost always bundled with supply-side investment, so their independent contribution is unidentified.

The steelman for conditionality deserves its hearing. Conditions can be the only mechanism that gets a child into a clinic in a place where attendance is otherwise rare; they generate the monitoring data that makes a programme auditable; and they buy political durability by framing transfers as investment rather than charity. Those are real benefits. None of them is a demonstrated mortality effect, and the distinction between "conditions improve service use" and "conditions save lives" is exactly where the policy argument usually slides.

The clinic is not a detail

If conditionality is not the active ingredient for survival, the complementary services are still not optional — they are just doing a different job.

The clearest evidence is for outcomes money alone does not fix. A 2025 meta-analysis of 26 estimates found that cash transfers had no significant effect on haemoglobin (0.065 d/L, 95% CI −0.054 to 0.184) or anaemia prevalence (−0.092, 95% CI −1.227 to 1.042). Programmes combining cash with behaviour-change communication or nutritional supplements did better (Manley and Alderman, 2025). Argentina's universal child allowance, evaluated with matching on the 2019/2020 national survey, likewise found no significant increase in health-care use for affiliated children and no significant reduction in stunting or wasting, though secondary-school enrolment rose among boys (Nikoloski, Zapata and Mossialos, 2025).

Those nulls are informative rather than damning. Argentina's allowance was modest and its health system strained; the Mexican programme that did improve growth supplied supplements and required growth monitoring alongside the cash. A review of 17 quantitative studies put the tension plainly: effects on mortality were mixed, and "cash transfers alone or the use of conditions may not be able to mitigate poverty and health inequalities in the presence of poor health services" (Cruz, Moura and Soares Neto, 2017).

The interaction runs both ways. Shei's Brazilian estimates were larger where Family Health Program coverage was higher, and the author's policy conclusion was that governments should check whether basic services can absorb the demand that cash creates. That is the correct reading of complementarity: cash raises demand for care, and the health system determines whether that demand becomes survival or queues.

This is also where the multi-country panel evidence needs care. The 46-country study found social protection coverage associated with lower under-five mortality (incidence rate ratio 0.71, 95% CI 0.54–0.96), lower stunting (0.75) and lower wasting (0.87), and estimated 3.05 million under-five deaths averted over 2000–2021 (Landin Basterra et al., 2025). But its exposure is "at least one social protection benefit" — a category that includes pensions, school feeding and disability grants, not cash transfers alone — and coverage at the country level is entangled with state capacity. It is the best global-scale estimate available and it does not isolate cash.

What the geography hides

Here is the sentence that should accompany every citation of this literature. Richterman's own framing: existing evidence was "limited to a few specific conditional cash transfer programmes, primarily in Latin America." Their study was designed to escape that limit, and largely did — but the escape is thinner than the resulting headlines.

The Latin American mortality literature is three countries deep in its quantitative core (Brazil, Mexico, Ecuador), with Argentina contributing outcome evidence that is null on nutrition, and with a shared research lineage: several of the largest Brazilian estimates come from overlapping author groups, panel constructions and modelling conventions. That is a convergent research programme, not a set of independent replications.

Outside Latin America, the evidence is mostly about intermediate outcomes. The sub-Saharan African systematic review found that cash transfers improve social determinants — nutrition, dietary diversity, child deprivation, teen pregnancy, birth registration — with moderate evidence on health and quality of life, and concluded that its findings were largely consistent with Latin America (Owusu-Addo, Renzaho and Smith, 2018). Consistency on determinants is not the same as measured mortality reduction, and the review does not claim otherwise.

So the accurate statement is narrower than the slogan: cash transfers reduce child mortality where they have been large, broad, and embedded in health systems that expanded alongside them, and in one multi-country individual-level study that reports the direction but whose country list and effect sizes this brief could not verify because the full text is paywalled. Anyone asserting a global mortality effect is extrapolating from a Latin American base rate.

The counterevidence and why it survives

The picture is not unanimous, and the dissent is not fringe.

A national analysis of Brazil's Zero Hunger programme found that summed investment did not alleviate child malnutrition or infant mortality, and that Bolsa Família was mainly associated with non-beneficial impacts — with substantial variation across subprogrammes (Dyngeland, Oldekop and Evans, 2020). The result sits awkwardly against the municipal panels, and the usual reconciliation is data quality: the mortality signal in that study was sensitive to which municipalities and which data source were used. Both things can be true — an association that is fragile to data quality is not a robust association, and a null that depends on inclusion choices is not a clean refutation.

There are structural reasons for caution that no single study resolves:

• Ecological exposure. Coverage is measured at the municipality or county level, not assigned to households. Migration, informal receipt and misclassification all blur the dose.
• Concurrent policy. Cash transfers expanded alongside primary care, conditionalities, school feeding and macroeconomic boom. Fixed effects handle stable differences between places, not simultaneous national trends.
• Publication and salience. The programmes are politically salient and heavily studied by groups with a stake in the answer. The literature has produced large positive estimates and at least one credible null on the same programme.
• Outcome measurement. Vital registration quality varies; the strongest Brazilian panels deliberately exclude municipalities with inadequate death statistics, which is defensible and also a selection on the outcome's denominator.
• No mortality trial. No randomised trial has child mortality as a pre-registered primary endpoint. The randomised literature is about schooling, consumption, service use and nutrition.

None of these points overturns the Latin American estimates. Together they explain why the field's confidence should be described as high on direction, moderate on magnitude, and low on generalisation.

What would settle it

Four designs would move this from a plausible package effect to a decomposed one.

1. Randomised or regression-discontinuity assignment of coverage. Where a programme is oversubscribed, a lottery for enrolment with mortality follow-up would identify the money's effect cleanly. Eligibility thresholds with credible running variables would do the same at lower cost.
2. Randomised conditionality with health outcomes. The Malawi design generalised to health: unconditional versus conditional transfers, same amount, same population, with clinic utilisation and child survival as pre-registered outcomes.
3. Cash versus cash-plus. Hold the transfer constant and randomise the supply-side component — a supplemented food ration, a community health worker visit schedule, a transport voucher. That isolates the clinic's contribution from the money's.
4. Dose shocks with mortality endpoints. Brazil's 2012 benefit reform and the 2023 rebuild and 2026 revaluation are natural experiments in adequacy. Linking payment records to death records across those discontinuities, with negative-control outcomes such as external causes, would test the dose-response outside the ecological frame.

The field also needs the unglamorous infrastructure: pre-registration, published nulls, and standard reporting of linkage loss, because nameless death records are exactly where an association can quietly change sign.

Verdict

On the direction: ~85% that large, well-implemented cash transfer programmes reduced under-five mortality in the settings where they have been evaluated. The convergence across designs and countries, the cause-specific pattern, the coverage and adequacy gradients, and the individual-level cohort result are collectively strong. The discount is for shared lineages, ecological designs, and one credible null.

On the mechanism: ~70% that the money is the largest single contributor to the survival effect, with complementary services contributing more to growth, nutrition and service use than to survival itself. The evidence is the dose-response on amount and coverage, the within-programme doubling result in Mexico, and the failure of the only direct comparison to find an advantage for conditionality.

On conditionality: ~75% that conditions are not a necessary ingredient for mortality reduction. That is not the same as saying they are useless — they plausibly improve targeting, monitoring and political durability — but the mortality claim attached to them is an inference, not a measurement.

On global generalisation: ~40% that the Latin American magnitude would replicate in low-income settings with weak health systems. The direction probably travels; the size probably does not, and the multi-country evidence that would settle it remains partly opaque.

The practical implication is unromantic. If a government wants child-survival returns from cash, the best-evidenced levers are how many people are covered and how much each receives. Conditions should be judged as an administrative and political technology, not a proven medical one. And the health system that conditions point at has to exist, because cash can fill a family's refrigerator and a clinic's waiting room, but it cannot stock the shelves.

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.
3. Barham T. A healthier start: the effect of conditional cash transfers on neonatal and infant mortality in rural Mexico. Journal of Development Economics. 2011;94(1):74–85. doi:10.1016/j.jdeveco.2010.01.003.
4. Moncayo AL, Granizo G, Grijalva MJ, Rasella D. Strong effect of Ecuador's conditional cash transfer program on childhood mortality from poverty-related diseases: a nationwide analysis. BMC Public Health. 2019;19(1):1132. doi:10.1186/s12889-019-7457-y.
5. Cavalcanti DM, Ordoñez JA, Aransiola TJ, et al. Evaluation and forecasting analysis of the association of conditional cash transfer with child mortality in Latin America, 2000–2030. JAMA Network Open. 2023;6(7):e2323489. doi:10.1001/jamanetworkopen.2023.23489.
6. 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: a retrospective analysis and modelling study. The Lancet Public Health. 2025;10(7):e548–e558. doi:10.1016/S2468-2667(25)00091-X.
7. Fernald LCH, Gertler PJ, Neufeld LM. Role of cash in conditional cash transfer programmes for child health, growth, and development: an analysis of Mexico's Oportunidades. The Lancet. 2008;371(9615):828–837. doi:10.1016/S0140-6736(08)60382-7.
8. Ramos D, da Silva NB, Ichihara MY, 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.
9. Richterman A, Millien C, Bair EF, Jerome G, Suffrin JCD, Behrman JR, Thirumurthy H. The effects of cash transfers on adult and child mortality in low- and middle-income countries. Nature. 2023;618(7965):575–582. doi:10.1038/s41586-023-06116-2.
10. Baird S, McIntosh C, Özler B. Cash or condition? Evidence from a cash transfer experiment. The Quarterly Journal of Economics. 2011;126(4):1709–1753. doi:10.1093/qje/qjr032.
11. Benhassine N, Devoto F, Duflo E, Dupas P, Pouliquen V. Turning a shove into a nudge? A "labeled cash transfer" for education. American Economic Journal: Economic Policy. 2015;7(3):86–125. doi:10.1257/pol.20130225.
12. Lagarde M, Haines A, Palmer N. The impact of conditional cash transfers on health outcomes and use of health services in low and middle income countries. Cochrane Database of Systematic Reviews. 2009;(4):CD008137. doi:10.1002/14651858.CD008137.
13. Manley J, Alderman H. Do cash transfer programmes affect child anaemia? Results from a meta-analysis. Maternal & Child Nutrition. 2025;e70026. doi:10.1111/mcn.70026.
14. Nikoloski Z, Zapata ME, Mossialos E. Impact of conditional cash transfer programs on health outcomes in Argentina: a retrospective, observational analysis based on MICS 2019/2020. The Lancet Regional Health — Americas. 2025;101011. doi:10.1016/j.lana.2025.101011.
15. Cruz RCS, Moura LBA, Soares Neto JJ. Conditional cash transfers and the creation of equal opportunities of health for children in low and middle-income countries: a literature review. International Journal for Equity in Health. 2017;16:161. doi:10.1186/s12939-017-0647-2.
16. Landin Basterra E, Gentilini U, Cavalcanti DM, et al. Impact of social protection on child malnutrition and mortality across 46 LMICs: a longitudinal study over two decades with insights from the COVID-19 pandemic. eClinicalMedicine. 2025;87:103414. doi:10.1016/j.eclinm.2025.103414.
17. Owusu-Addo E, Renzaho AMN, Smith BJ. The impact of cash transfers on social determinants of health and health inequalities in sub-Saharan Africa: a systematic review. Health Policy and Planning. 2018;33(5):675–696. doi:10.1093/heapol/czy020.
18. Dyngeland C, Oldekop JA, Evans KL. Assessing multidimensional sustainability: lessons from Brazil's social protection programs. Proceedings of the National Academy of Sciences. 2020;117(34):20511–20519. doi:10.1073/pnas.1920998117.