What would it cost to end extreme poverty? To obtain realistic and actionable cost estimates, we study poverty targeting it as a statistical learning problem. In contrast to existing estimates of the global poverty gap, which assume that policymakers can perfectly target the minimum transfer required to lift each household above the poverty line, our estimates assume that policymakers have imperfect information and imperfect ability to target. In a sample of 13 countries that collectively account for 35% of the world's poor, the cost of reducing the population-weighted poverty gap index to 1% via a cash-transfer policy that targets based on observable characteristics is $67.6B nominal per year. This cost is three times that of an oracle policy that perfectly targets the minimum transfer required to lift households above the poverty line but 28% of the cost of universal basic income at the level of the poverty line.
The talk is based on joint work with Joshua Blumenstock, Paul Niehaus, Leo Selker, and Stefan Wager.