Optimising Anti-Poverty Transfers With Quantile Regressions

Anti-poverty transfer schemes are one of the main way of fightingpoverty. Under perfect observation of incomes, designing such scheme boilsdown to solving an optimisation program under constraints, which can beachieved with well-defined methods. In contrast, when incomes cannot beperfectly observed, the schemes are usually based on predictions of livingstandards using ancillary regressions and household survey data to predict theunobserved living standards of households. In this paper, we study the povertyminimisation program under imperfect information. We show why usingpredictions of living standards helps to deal approximately with an otherwiseintractable problem. Then, we propose a new approach to the practicaloptimisation procedure based on improved predictions of living standards interms of the targeting problem to be solved. Our new empirical methodology totarget direct transfers against poverty is based on observable correlates and onestimation methods that can focus on the poor: the quantile regressions. Weillustrate our results using data from Tunisia.