Proxy means test (PMT) poverty targeting tools have become common tools for
beneficiary targeting and poverty assessment where full means tests are costly.
Currently popular estimation procedures for generating these tools prioritize
minimization of in-sample prediction errors; however, the objective in generating such
tools is out-of-sample prediction. In this paper, we present evidence that application of
machine learning algorithms to PMT development can substantially improve the out-ofsample
performance of these targeting tools. In particular, we show that stochastic
ensemble methods can improve out-of-sample performance by 2 to 18 percent over
current methods. While we take the USAID poverty assessment tool and base data for
demonstration of these methods, the methods applied in this paper should be considered
for PMT and other poverty targeting tool development more broadly.