Girls' education is one of the few areas of social policy with a large body of randomised evidence, because it has been the subject of sustained trial-based research over three decades.
The returns to girls' education are among the most replicated findings in development economics, with associations documented for child health, fertility, household income and intergenerational outcomes. The causal claims vary in strength across these, but the association is consistent.
What is more useful, and less frequently summarised, is what the trials found about which interventions actually increase attendance and completion.
The pattern in the results
Across a large number of trials, interventions that removed a specific identifiable obstacle outperformed interventions that attempted to change attitudes or increase motivation.
This is the same pattern found in the domestic literature on subject choice and university access: information and encouragement produce small effects; removing a concrete barrier produces larger ones.
Interventions with strong results
Reducing the direct cost of attendance. Trials providing uniforms, eliminating fees, or providing cash transfers conditional on attendance have generally found substantial effects on enrolment and attendance.
Uniform provision in particular has produced large effects relative to cost in several trials, which is notable because the cost of a uniform is small relative to most educational spending.
Reducing the cost of travel. Distance to school is among the strongest predictors of girls' attendance in many settings, and interventions reducing travel time or cost — including the widely cited bicycle distribution programmes — have produced measurable effects.
The gender asymmetry here is specific: distance affects girls' attendance more than boys' in most studies, and safety concerns during travel are the commonly cited reason.
Health interventions. Deworming and other health programmes have been found in several studies to increase attendance, through a straightforward mechanism — sick children do not attend.
Sanitation facilities. Provision of separate and private toilet facilities has been associated with improved attendance for adolescent girls in several studies. This recurs across contexts and is among the more consistent findings.
Interventions with weaker or mixed results
Information campaigns about the value of education. Generally small effects, with the notable exception of information about the returns to education where parents' beliefs were substantially inaccurate — in those cases, correcting the belief produced effects.
General teacher training. Mixed results, with specific structured programmes performing better than general professional development.
Unconditional transfers versus conditional. The comparison is genuinely contested. Conditional transfers generally produce larger attendance effects; unconditional transfers have advantages on other measures and lower administrative cost. The evidence does not clearly favour one across all contexts.
The menstruation question
The claim that lack of menstrual products causes substantial school absence has been widely repeated and is more contested than its prevalence suggests.
Trials providing menstrual products have produced mixed results, with several finding smaller effects on attendance than expected.
What appears to matter more, in the studies that have separated the components, is facilities — private toilets, water, disposal — rather than products alone.
This is a useful example of a plausible mechanism producing a specific intervention that turned out to address the wrong part of the problem, and it is worth citing whenever an obvious-seeming solution is proposed without testing.
Marriage and the age question
Early marriage is a substantial cause of girls' school leaving in a number of contexts, and interventions addressing it have been studied.
Conditional transfers that continue only while a girl remains unmarried and in school have produced effects on both marriage timing and attendance in several trials.
Legal minimum age provisions exist widely and enforcement varies enormously; studies of legal change alone generally find smaller effects than economic interventions, which suggests the constraint is frequently economic rather than legal.
What generalises to high-income contexts
The specific interventions do not transfer — uniform cost is not the binding constraint in most high-income systems.
The methodological lesson does. It is that the barrier should be identified empirically before an intervention is designed, that plausible mechanisms frequently turn out to be wrong, and that removing a concrete obstacle generally outperforms changing an attitude.
Applied domestically, that logic points at the findings elsewhere on this site: changing facilities affecting sport participation, subject availability affecting physics take-up, form completion affecting university applications, and scheduling affecting who can attend training.
All of them are obstacles rather than attitudes, and all of them are more tractable than the things that receive more attention.