Development Experiments: Ethical? Feasible? Useful?

Bill Easterly weighs in on randomized controlled trials for development projects. I have mixed feelings - on one hand, we need this research to begin to reveal honest, data-driven best practices in development. On the other hand, is it ethical to deny something you think will help? I think that it's more dangerous to package something that's uncertain as a foolproof development strategy (which is our current practice)