Targeting Performance and Poverty Effects of Proxy Means-Tested Transfers: Tade-offs and Challenges

In the absence of reliable and exhaustive income data, Proxy Means Tests (PMTs) are frequently employed as a cost-eective way to identify income-poor beneciaries of targeted anti-poverty programs. However, their usefulness depends on whether proxies accurately identify the income poor. Based on Receiver Operating Characteristics (ROC)-analysis, we nd that PMTs perform poorly in terms of identifying poor households in Bolivian data when transfers are targeted narrowly to the poor but that the true positive rate is highly responsive to increases in the proportion of beneciaries. Using non-parametric regression- techniques, we show that the resulting leakage can largely be conned to the non-poor close to the poverty line. However, simulating the eect on poverty measures of a uniform transfer to beneciaries across inclusion rates suggests that the largest poverty eect is attained with very narrow targeting. Hence, we nd a trade-o between targeting accuracy and poverty eect.