Is Randomized Testing the Key to Better Aid Programs?

Just telling adolescent girls in Kenya that the older the man the more likely he is to be HIV infected reduced the incidence of pregnancy (and presumably of HIV infection) - that's one conclusion of a randomized trial among schoolgirls in Kenya, reports Celia Dugger in the NYTimes. That single result is likely to influence programs - at least in Kenya - and gives heart to the right-wing proponents of abstinence as the best AIDS prevention and the left-wing proponents of information, education - and randomized trials as the key to more effective aid programs.But are randomized trials the key to better aid programs - or do they amount to "ring-fencing" as Nick Stern, Ian Goldin and others of the World Bank assert, by "randomists" (the word Angus Deaton uses for single-minded proponent of RCTs ("Randomized Controlled Experiments"). Do randomized trials risk substituting a feckless faith in "international best practice" for a developing country's own decisions and "ownership" of policy and programs - as implied by Robert Bates? Read about "bureaucratic aversion" to impact evaluation as thoughtfully summarized by Ruth Levine, and why it matters for aid effectiveness and development practice as set out by Banerjee, Stern, Deaton, Bates, Bhagwati and others in the July/August issue of the Boston Review.And for an excellent review of the statistical (as opposed to random trial) literature see Clemens, Radelet and Bhavnani.