It might seem obvious to some of you reading this that it might be possible to learn something from a book. But as a recent review for RISE by Paul Glewwe and Karthik Muralidharan found, researchers have actually so far failed to show rigorously that there is any improvement in test scores in developing countries after handing out textbooks to schools. There have now been four different Randomized Controlled Trials showing no improvement (and for four different reasons).

So when might books “work”? A new paper from the World Bank proclaims to answer just this question: “When Do In-service Teacher Training and Books Improve Student Achievement? Experimental Evidence from Mongolia.”

Their answer, somewhat disappointingly, seems to be “when it happens in Mongolia”. More constructively, though books alone seem to work, they work better when combined with teacher training, reinforcing Glewwe and Muralidharan’s conclusion about the importance of complementarities.

But that “there are complementarities” isn’t a very satisfying conclusion by itself. The more comprehensive hypothesis being developed for RISE by Lant Pritchett is: “when there is an accountability framework which is coherent for learning” – that is, when all of the relevant actors are held accountable for common goals through clear delegation of those goals, and have the resources to accomplish them. We’re hoping that this accountability coherency system diagnostic can be a useful tool for thinking through systematically what it is about specific contexts that mean that interventions work in some places and not others. What is it about Mongolia which means that providing books alone can be enough, in contrast to those other studies in Kenya and Sierra Leone? It just might be a coherent accountability system.