Cash Transfers Send Children to School. Do They Teach Them Anything?

The anomaly: a programme that works but may not deliver

Cash transfer programmes — payments to poor households, sometimes tied to conditions such as school attendance — are among the most evaluated social policies in the world. Brazil's Bolsa Família reaches roughly 50 million people. Mexico's Prospera, Indonesia's PKH, and dozens of smaller programmes have been tested with randomised trials and large administrative datasets. The consensus finding is almost boring in its consistency: give poor families money, and their children spend more time in school.

The finding that deserves the headline is the one that rarely gets one. When researchers measure not whether children attend but whether they learn, the effect is much smaller, frequently statistically indistinguishable from zero, and occasionally negative. A programme that clearly succeeds on the metric it is judged by can fail on the metric that matters for the child.

This brief asks a bounded question: what does the causal evidence say about cash transfers and learning — test scores, cognitive skills, grade progression — as distinct from schooling (enrolment, attendance, completion)? The answer is not that transfers fail. It is that enrolment and learning are governed by different constraints, and a policy aimed at the first is not automatically a policy aimed at the second.

What the transfers reliably move

The schooling evidence is strong, independent, and convergent. Five large syntheses agree, and they use different methods on different literatures.

The programme-level evidence matches. In Indonesia, PKH halved the share of children aged 7–15 who were not enrolled, cut wage labour among 13–15-year-olds by at least a third, and raised high-school completion at the six-year mark. In Malawi, a randomised experiment found both conditional and unconditional transfers raised girls' schooling. The effects are real, causal, and reproducible. They are also modest in size — typically a few percentage points on participation — and they are about the quantity of schooling.

The learning ledger: smaller, noisier, sometimes negative

Now change the outcome from attendance to achievement, and the picture changes sign in places. The studies below deliberately measure test scores or cognitive outcomes rather than enrolment.

• Morocco, Tayssir. Gazeaud and Ricard (2024, Journal of Development Economics) study the learning effects of a conditional transfer and isolate the role of class size and composition. The programme raised enrolment, but where the inflow of new students expanded classrooms, the learning effect was attenuated and for some outcomes turned negative — the transfer increased the number of children taught without increasing the teaching.
• Cambodia. Filmer and Schady (2014, Journal of Human Resources) evaluate secondary-school scholarships for girls and find they kept students in school longer without a clear gain in measured learning. More schooling, on its own, did not translate into measured skill.
• Guinea. A shock-responsive transfer raised school participation by about 11 percentage points with no detectable impact on mathematics or literacy — the divergence the 2026 meta-analysis behind Manley's review flags explicitly, concluding that attending school is not the same as learning.
• Mexico. A dynamic model of Prospera that separates the attainment margin from the achievement margin finds the enrolment response larger than the achievement response.
• Brazil. A systematic review of Bolsa Família and educational indicators (Santos et al. 2019) finds consistent positive associations with attendance and school progression but thin and contradictory evidence on performance; a study of Bolsa Escola/Família (Glewwe & Kassouf 2012) finds solid effects on enrolment and dropout but weaker effects on grade promotion.

Two features of this literature deserve emphasis. First, learning is measured far less often than enrolment. Meta-analyses of participation can pool dozens of programmes; the studies that report test scores number in the low teens and use less comparable assessments, so the learning conclusions rest on a thinner and noisier base. Second, null results are systematically under-reported: a programme that fails to move test scores is less likely to be written up, which biases the published average upward, not downward. Neither point rescues the enrolment finding; both warn against assuming the learning effect is merely under-powered rather than genuinely small.

Read as a body, the average learning effect of a cash transfer is small and often statistically null. That is the anomaly: the intervention reliably changes the input (time in school) and only unreliably changes the output (skill).

Why access and learning come apart

The pattern is not mysterious once the two constraints are separated. A transfer relaxes the demand constraint on schooling: it offsets the fees, uniforms and forgone child labour that keep poor children out of class. It does not relax the supply constraint on learning: teacher effort, teacher knowledge, materials, and class size. Those are governed by the school system, and cash to households does not move them.

Three mechanisms make the gap concrete.

1. Crowding. Rapid enrolment growth can raise pupil–teacher ratios and dilute instruction. Evidence from Kenya's free-primary-education expansion and from class-size-cap designs shows that access expansions can shift who is in the room without raising average achievement. Gazeaud and Ricard's result is the clearest demonstration that a cash transfer's learning effect depends on whether classrooms expand.

2. Marginal students. Transfers pull in children who would otherwise not be enrolled. Those students may be less prepared and may lose ground relative to peers even as the enrolled population grows — one reason enrolment rises while a fixed test threshold does not move.

3. Conditionality targets the wrong signal. Conditions are set on attendance because attendance is cheap to verify. Comprehension is neither cheap nor easy to monitor, and so programmes optimise the metric they can measure. This is the schooling-versus-learning gap in its purest form, documented across African primary schools by Bold et al. (2017): enrolment without learning is the norm, not the exception.

Put in the language of economics, a transfer moves the extensive margin — whether a child is enrolled for any days at all — far more reliably than the intensive margin of how much a child learns per year of schooling. Learning per year is a function of instruction, and a transfer to households does not buy instruction. This is why two programmes with similar enrolment effects can have opposite learning effects: it depends entirely on whether the school system absorbed the extra students or merely packed them into already-full classrooms.

Lant Pritchett's summary — schooling ain't learning — and the World Bank's framing of "learning poverty" are the same point in policy language. Cash transfers are a demand-side tool aimed at a supply-side problem.

Where learning did move

The interesting cases are the ones where learning did respond, because they show what the transfer has to touch to matter.

• Exposure at the right age. Barham, Macours and Maluccio (2024, JEEA) follow a Nicaraguan conditional transfer for a decade and find that earlier exposure, during primary-school ages when children were at risk of dropping out, produced gains in learning for men — alongside higher earnings and labour-market participation. The timing of exposure, not merely the transfer, drove the result.
• Liquidity timed to the exam. A 2025 study of Brazil's Bolsa Família exploits the random timing of payment dates and finds that receiving a transfer in the days before a high-stakes college-admission exam raised scores by about 0.01 standard deviations, concentrated in the final and easier questions — the signature of improved cognitive endurance and effort allocation rather than more knowledge. Larger transfers produced persistent effects, including a small rise in college enrolment. The mechanism is short-run liquidity, not schooling quantity.
• Design that changes incentives. Barrera-Osorio and co-authors (2019) show that variant designs of a Colombian transfer — savings-linked and tertiary-earmarked versions — produced materially different medium- and long-run educational outcomes from the standard design. A one-time transfer also fades: Baird, McIntosh and Özler (2019) find that sustained human-capital gains depend on the transfer being sustained.

The common thread is that learning moves when the intervention changes something beyond attendance — the age of exposure, the incentive structure, or the student's capacity to perform on the day that is measured.

The Brazilian case

Bolsa Família is the right test case: large, long-running, and often described as an education success. The schooling record supports that description. The learning record does not, at least not with the same confidence.

Brazil does have studies that look directly at achievement rather than attendance — for example, work linking Bolsa Família participation to the educational achievement of disadvantaged children (Simões & Sabates 2014) — but they are few, and their effects are smaller and less consistent than the enrolment literature, as the 2019 systematic review concludes.

Two caveats compound. First, national test-score improvements during the Bolsa Família era cannot be attributed to the transfer, because school funding, teacher hiring and the Family Health Strategy were expanding at the same time. Second, the programme is not one programme: it was replaced by Auxílio Brasil in 2021 and rebuilt under a new statute in 2023, with new conditionality and schooling-linked benefits. No evaluation measures the learning effect of the 2023–26 version. Brazil's flagship transfer is an access policy whose learning reputation rests on inference, not on measurement.

Verdict and calibrated confidence

• ~95% that cash transfers raise school enrolment and attendance. The finding survives across designs, countries, and five independent meta-analyses. This is about as settled as social-science evidence gets.
• ~75% that the average effect on measured learning is small and near zero. The bias here is in the enrolment direction: learning is measured less often, and studies are less likely to be run at all when the outcome is null. The honest reading is that an average transfer does not reliably teach.
• ~60% that a well-designed variant produces positive learning effects. Early-age exposure, exam-time liquidity timing, and complementary quality interventions have each produced genuine learning gains; the uncertainty is about how well they generalise.
• ~50% (genuine uncertainty) about the current Bolsa Família. No study isolates its learning effect across the 2021 and 2023 breaks, so any confident claim in either direction is unsupported.

What would change my mind: a pre-registered trial or policy discontinuity that measures both margins and finds a learning effect comparable to the enrolment effect without any change to instruction; or, conversely, evidence that the small average learning effect is an artefact of poor or selective test coverage.

What would settle it

The learning question is answerable with data that mostly already exist. The design has three parts.

1. Measure both margins. Link transfer-payment records to standardised assessments — Brazil's SAEB/IDEB, Indonesia's national exams, Mexico's ENLACE — and estimate the learning effect against the same discontinuities used for enrolment. Report the null when it is null.
2. Control for the supply side. Include class size, pupil–teacher ratios and teacher characteristics, and test whether learning gains appear only where the school system absorbed the inflow. Pre-specify negative controls so an enrolment-driven association cannot be tuned into a learning claim post hoc.
3. Test the interaction directly. Randomise or quasi-randomise transfers alongside a proven quality intervention — structured pedagogy, teacher training, materials — and estimate whether the two are complements or substitutes. The Tanzania experiment (Mbiti et al. 2019) shows that inputs and incentives interact, and the same logic should be applied to transfers before anyone treats them as an education policy.

Until at least one such study exists, the defensible statement is narrow and worth stating plainly: cash transfers are a proven way to get children into school, and an unproven way to get them to learn. Treating the first as evidence of the second is a category error — and it is the most common one in the whole debate.

References

1. García S, Saavedra JE. Educational Impacts and Cost-Effectiveness of Conditional Cash Transfer Programs in Developing Countries: A Meta-Analysis. Review of Educational Research. 2017;87(5):921–965. doi:10.3102/0034654317723008.
2. Baird S, Ferreira FHG, Özler B, Woolcock M. Conditional, unconditional and everything in between: a systematic review of the effects of cash transfer programmes on schooling outcomes. Journal of Development Effectiveness. 2014;6(1):1–43. doi:10.1080/19439342.2014.890362.
3. Crosta T, Karlan D, Ong F, Rüschenpöhler J, Udry C. Unconditional Cash Transfers: A Bayesian Meta-Analysis of Randomized Evaluations in Low and Middle Income Countries. NBER Working Paper 32779. 2024. doi:10.3386/w32779.
4. Chong ZL, Lau SY. Unconditional cash transfers and child schooling: a meta-analysis. Empirical Economics. 2025;68(2):639–666. doi:10.1007/s00181-024-02647-3.
5. Manley J, Dehingia N, Bruckauf Z. Social protection in response to shocks: what works for protecting child development? Meta-analysis and scoping review. Journal of Development Effectiveness. 2026. doi:10.1080/19439342.2026.2728626.
6. Cahyadi N, Hanna R, Olken BA, Prima RA, Satriawan E, Syamsulhakim E. Cumulative Impacts of Conditional Cash Transfer Programs: Experimental Evidence from Indonesia. American Economic Journal: Economic Policy. 2020;12(4):88–110. doi:10.1257/pol.20190245.
7. Baird S, McIntosh C, Özler B. Cash or Condition? Evidence from a Cash Transfer Experiment. Quarterly Journal of Economics. 2011;126(4):1709–1753. doi:10.1093/qje/qjr032.
8. Gazeaud J, Ricard C. Learning effects of conditional cash transfers: The role of class size and composition. Journal of Development Economics. 2024;166:103194. doi:10.1016/j.jdeveco.2023.103194.
9. Filmer D, Schady N. The Medium-Term Effects of Scholarships in a Low-Income Country. Journal of Human Resources. 2014;49(3):663–694. doi:10.3368/jhr.49.3.663.
10. Glewwe P, Kassouf AL. The impact of the Bolsa Escola/Familia conditional cash transfer program on enrollment, dropout rates and grade promotion in Brazil. Journal of Development Economics. 2012;97(2):505–517. doi:10.1016/j.jdeveco.2011.05.008.
11. Santos MCS, Delatorre LR, Ceccato MGB, Bonolo PF. Programa Bolsa Família e indicadores educacionais em crianças, adolescentes e escolas no Brasil: revisão sistemática. Ciência & Saúde Coletiva. 2019;24(6):2233–2247. doi:10.1590/1413-81232018246.19582017.
12. Behrman JR, Parker SW, Todd PE, Zhang W. Prospering through Prospera: A dynamic model of CCT impacts on educational attainment and achievement in Mexico. Quantitative Economics. 2025;16(1):133–183. doi:10.3982/qe2291.
13. Barham T, Macours K, Maluccio JA. Experimental Evidence from a Conditional Cash Transfer Program: Schooling, Learning, Fertility, and Labor Market Outcomes after 10 Years. Journal of the European Economic Association. 2024;22(4):1844–1883. doi:10.1093/jeea/jvae005.
14. Barrera-Osorio F, Linden LL, Saavedra JE. Medium- and Long-Term Educational Consequences of Alternative Conditional Cash Transfer Designs: Experimental Evidence from Colombia. American Economic Journal: Applied Economics. 2019;11(3):54–91. doi:10.1257/app.20170008.
15. Baird S, McIntosh C, Özler B. When the money runs out: Do cash transfers have sustained effects on human capital accumulation? Journal of Development Economics. 2019;140:169–185. doi:10.1016/j.jdeveco.2019.04.004.
16. 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.
17. Eizmendi Larrinaga A, Reyes G. Cash and Cognition: The Impact of Transfer Timing on Standardized Test Performance and Human Capital. arXiv:2507.21393. 2025. https://arxiv.org/abs/2507.21393
18. Bold T, Filmer D, Martin G, Molina E, Stacy B, Rockmore C, et al. Enrollment without Learning: Teacher Effort, Knowledge, and Skill in Primary Schools in Africa. Journal of Economic Perspectives. 2017;31(4):185–204. doi:10.1257/jep.31.4.185.
19. Pritchett L. The Rebirth of Education: Schooling Ain't Learning. Washington, DC: Center for Global Development; 2013.
20. World Bank. World Development Report 2018: Learning to Realize Education's Promise. Washington, DC: World Bank; 2018. doi:10.1596/978-1-4648-1096-1.
21. Lucas AM, Mbiti IM. Access, Sorting, and Achievement: The Short-Run Effects of Free Primary Education in Kenya. American Economic Journal: Applied Economics. 2012;4(4):226–253. doi:10.1257/app.4.4.226.
22. Urquiola M, Verhoogen E. Class-Size Caps, Sorting, and the Regression-Discontinuity Design. American Economic Review. 2009;99(1):179–215. doi:10.1257/aer.99.1.179.
23. Mbiti I, Muralidharan K, Romero M, Schipper Y, Manda C, Rajani R. Inputs, Incentives, and Complementarities in Education: Experimental Evidence from Tanzania. Quarterly Journal of Economics. 2019;134(3):1627–1673. doi:10.1093/qje/qjz010.
24. Molina Millán T, Barham T, Macours K, Maluccio JA, Stampini M. Long-Term Impacts of Conditional Cash Transfers: Review of the Evidence. World Bank Research Observer. 2019;34(1):119–159. doi:10.1093/wbro/lky005.
25. Behrman JR, Parker SW, Todd PE. Impacts of PROSPERA on Enrollment, School Trajectories, and Learning. World Bank Policy Research Working Paper 9000. 2019. doi:10.1596/1813-9450-9000.
26. Simões AA, Sabates R. The contribution of Bolsa Família to the educational achievement of economically disadvantaged children in Brazil. International Journal of Educational Development. 2014;39:141–156. doi:10.1016/j.ijedudev.2014.07.012.