Stop arguing ideology; run the experiment

Banerjee and Duflo open by rejecting the decades-long macro-debate over whether foreign aid works or whether free markets alone will lift poor countries out of poverty — a debate they see as unresolvable at that scale because it's essentially unfalsifiable, with each side able to point to selective success stories.

Their alternative is to shrink the question down to something testable: not "does aid work" but "does this specific program — say, free deworming pills for schoolchildren — actually improve outcomes, measured against a comparable group that didn't get it." This is the logic of the randomized controlled trial (RCT), borrowed from medicine, applied to poverty interventions for the first time at scale.

The payoff of thinking this small is that it's actually answerable, and answerable questions accumulate into real knowledge, while unanswerable ideological ones just accumulate into more shouting. Test the specific policy, not the ideology it's attached to.