Targeting when Poverty is Multidimensional

The view of poverty as a multidimensional phenomenon has swiftly become mainstream. However, the debate remains open as to how such ?multidimensional poverty? should be assessed in practical settings, particularly when identifying the beneficiaries of poverty alleviation programmes. This paper develops a novel empirical approach that explicitly takes into account the goals and needs of the policy-maker. In particular, the paper takes up the case of a government official running a budget constrained programme to alleviate a few dimensions of poverty, and translates her concerns into a set of desiderata which the multidimensional measure should meet. The policy-maker targeting ability and aversion to the risk of leakages play crucial roles in setting the desired properties. We illustrate our methodology in the context of a CCT programme in Peru, and show that it improves expected leaking and undercoverage relative to alternative Alkire-Foster based approaches.