IF A girl in a poor country goes to school, she will probably have a more comfortable life than if she stays at home. She will be less likely to marry while still a child, and therefore less likely to die in childbirth. So, not surprisingly, there is an Indian charity that tries to get girls into school and ensure they learn something, and there are Western philanthropists willing to pay for its work. What is noteworthy is how they have gone about this transaction.

On July 13th the Brookings Institution, a think-tank, presents the results of the world’s first large development-impact bond, which paid for girls’ education in the northern Indian state of Rajasthan. In this novel way of funding charitable work, a financial institution gives money to a charity, which tries to achieve various specified outcomes. If a neutral arbiter rules that it has succeeded, a donor or philanthropist repays the investor, plus a bonus. If it fails, the investor loses some or all of its money.

This is more convoluted than the usual way of funding charitable projects, in which a donor gives money to a charity, which spends it according to a pre-agreed plan. The donor tries to ensure the money is not wasted by keeping track of inputs—the number of solar panels installed or vaccinations given, say. Often, no one knows whether the intervention did much good.

In this case, the more complicated approach did achieve something. Educate Girls, the charity, identified 837 out-of-school girls aged 7-14 in the villages where it was active, and enrolled 768 of them. By using volunteers to teach both boys and girls in village schools for a few hours a week, it managed to raise test scores substantially relative to a control group. So the investor, UBS Optimus Foundation, will be repaid by the Children’s Investment Fund Foundation.

For Safeena Husain, who runs Educate Girls, the process was as satisfying as the results. Instead of having to send tedious reports to a donor about how she was spending money, she concentrated on solving problems. Educate Girls found, for example, that many pupils could not do long division because they did not understand the concept of place value. So its workers taught remedial classes. IDinsight, the independent assessor, found that the main boost to children’s test scores came in the third year of the programme, when Educate Girls hit its stride.

Creating the development-impact bond was also complicated and time-consuming. Staff from several organisations spent months pinning down what Educate Girls would aim to achieve, how progress would be measured and what would be repaid. Outside experts were drafted in. The randomised controlled trial that IDinsight used to assess the teaching was, like many such trials, neither simple nor cheap.

More development-impact bonds are now under way or under discussion, some involving big donors like the World Bank, USAID and DfID (America’s and Britain’s aid agencies). But they will probably remain infrequent oddities in the aid landscape. Not only are they complex, ponderous and costly; they also offer small returns to investors. And, as Emily Gustafsson-Wright of Brookings points out, no one can yet say for certain that they are better than other ways of delivering aid.

They are useful, even so. The problem with much aid (and social spending in general) is that inputs are scrutinised more closely than results. Experimenting during a project is hard or impossible. It would be good if development-impact bonds teach donors to give charities freer rein and to focus on outcomes. Rajasthani girls are not the only people with lessons to learn.