Girls' education is among the most researched interventions in development economics, and the findings have been consistent enough to shape international policy for decades.

Worth being specific about what the evidence supports, since the topic attracts both overclaiming and dismissal.

The individual returns

Estimates of returns to schooling — the increase in earnings associated with each additional year of education — have been calculated extensively.

Reviews of this literature have generally found returns that are positive, substantial, and in many analyses somewhat higher for women than for men.

Returns tend to be higher in lower-income countries and at lower levels of education, though the pattern varies.

Methodological caveats matter. Isolating the causal effect of education from the characteristics of people who obtain more of it is difficult, and studies using natural experiments — policy changes affecting some cohorts and not others — generally find effects, sometimes smaller than simple comparisons suggest.

The wider effects

What makes this literature distinctive is the range of outcomes associated with female education beyond earnings.

Child health and survival. Maternal education is consistently associated with lower child mortality across a large body of research. Proposed mechanisms include health knowledge, healthcare utilisation, resources and household bargaining power.

Child nutrition and education. Children of more educated mothers show better outcomes on both, with effects persisting across generations.

Fertility. Female education is associated with later first birth and lower total fertility, with effects observed across many settings.

Age at marriage. Later marriage, with associated effects on health and autonomy.

Health knowledge and behaviour. Including vaccination uptake and use of antenatal care.

Causal interpretation requires care throughout — education correlates with other advantages — and studies exploiting policy changes have found effects on several of these outcomes, which strengthens the causal case.

What actually increases attendance

The more practically useful literature concerns which interventions work, and the findings are frequently counterintuitive.

Reducing costs. Eliminating fees, providing uniforms, and cash transfers conditional on attendance all show effects. Cost is a substantial barrier and reducing it works.

Reducing distance. Building schools closer to communities has shown large effects, particularly for girls, where distance carries safety concerns.

Sanitation facilities. Separate and functioning toilets are associated with attendance, particularly for adolescent girls. Evidence on menstrual product provision specifically is more mixed than commonly claimed, with several rigorous trials finding smaller effects than expected.

Health interventions. Deworming programmes have shown effects on attendance, though the magnitude of long-term benefits has been debated extensively.

Information about returns. Studies providing information about the earnings associated with education have found effects on enrolment, suggesting that expectations matter.

Female teachers. Associated with girls' attendance and attainment in several studies.

Attendance is not learning

An important qualification that has reshaped thinking in this area.

Enrolment has increased dramatically in many countries. Learning outcomes have not followed proportionally, with large-scale assessments finding that substantial proportions of children in school are not achieving basic literacy and numeracy.

This has been described as a learning crisis, and it means that measuring success by enrolment overstates progress.

Interventions targeting learning specifically — teaching at the appropriate level rather than the curricular level, structured pedagogy, and targeted instruction — have shown stronger effects on learning than interventions targeting access alone.

Where girls remain out of school

The remaining gaps are concentrated and identifiable.

Conflict-affected areas, where disruption affects girls disproportionately.

Secondary rather than primary level, where gaps are larger.

Contexts with high rates of early marriage.

Rural and remote areas, where distance and cost compound.

And among children with disabilities, where the intersection produces the largest exclusion.

What to take from it

The evidence supports girls' education as an intervention with returns extending beyond the individual, and the mechanisms are reasonably well understood.

It also supports being specific. Enrolment is not learning. Interventions vary enormously in effectiveness. And the remaining gaps are concentrated in contexts where the general interventions are hardest to deliver.

Which suggests that the useful contribution now is less about the general case, which is established, and more about delivery in the specific circumstances where children are still being left out.

The disability intersection

Worth stating separately because it is where exclusion is largest and attention smallest.

Children with disabilities are substantially less likely to be in school across most contexts, and girls with disabilities face compounded exclusion.

The barriers are practical as much as attitudinal: physical inaccessibility, absence of trained teachers, lack of assistive materials, and transport that does not accommodate.

Inclusive education has an evidence base and requires investment in specific provisions rather than only in enrolment campaigns. It is also the area where general interventions perform worst, which means it needs to be addressed deliberately rather than expected to follow from broader progress.

Cost-effectiveness comparisons

One further contribution of this literature is comparative. Where budgets are finite, the question is not whether an intervention works but how it compares with alternatives per unit of spending.

Analyses ranking education interventions by cost per additional year of schooling, or per unit of learning gained, have consistently found large variation — with some widely funded approaches performing considerably worse than cheaper alternatives.

Providing information about the returns to education, for instance, has appeared very cost-effective in several analyses, largely because it is inexpensive. Building schools is effective and expensive.

That ranking exercise is more useful than advocacy for any single approach, and it is the kind of comparison that rarely appears in campaigning material.