Instruments of Development

How can we reduce global poverty? How do we know what works, and what doesn't? There is great dissatisfaction among development economists about the lack of convincing evidence of policy effectiveness, and with an associated failure to learn from past development assistance and projects. Econometric methods are seen as part of the problem, not the solution, and there has been a movement to discard econometrics in favour of a rigorous and comprehensive programme of randomized controlled trials from which, at last, we will learn what works. I review recent work, both experimental and econometric, and argue that, while much can be learned from the new methods, they are unlikely to deliver all that is promised, and that it would be unwise to abandon econometric analysis. Our failures, such as they are, come from our recent neglect of models of behaviour and of development, and the absence of such models from econometric analysis.