Randomised controlled trials have revolutionised development policy. But do the interventions that work in the short run have a benefit 10 or 20 years later? Ted Miguel tells Tim Phillips how he and his colleagues aim to find out.