How Best to Target the Poor? An operational targeting of the poor using indicator-based proxy means tests

This paper seeks to answer an operational development question: how best to target the poor? In their endeavor, policy makers, program managers, and development practitioners face the daily challenge of targeting policies, projects, and services at the poorer strata of the population. This is also the case for microfinance institutions that seek to estimate the poverty outreach among their clients. This paper addresses these challenges. Using household survey data from Uganda, we estimate four alternative models for improving the identification of the poor in the country. Furthermore, we analyze the model sensitivity to different poverty lines and test their validity using bootstrapped simulation methods. While there is bound to be some errors, no indicator being perfectly correlated with poverty, the models developed achieve fairly accurate out-of-sample predictions of absolute poverty. Furthermore, findings suggest that the estimation method is not relevant for developing a fairly accurate model for targeting the poor. The models developed are potentially useful tools for the development community in Uganda. This research can also be applied in other developing countries.