Targeting The Poor In Mexico: An Evaluation Of The Selection Of Households For Progresa

This report reevaluates PROGRESA’s targeting methods since the program began adding beneficiary households through a process called “densification.” The authors first evaluate PROGRESA’s accuracy in targeting both at the community and household levels. Second, they evaluate the targeting in terms of its impact on poverty alleviation relative to other feasible methods assuming the same total budget. The first step is accomplished by comparing PROGRESA’s method to an alternative selection method based on household consumption. For the second task, the costs associated with different schemes are compared for their effects on the budget available for poverty alleviation. Schemes considered include uniform transfers that involve no targeting, targeting based on consumption, and targeting at the locality rather than the household. The authors find that PROGRESA’s marginality index performs quite well when contrasted to a consumption-based model of locality selection. The consumption-based model results in a more precise categorization of poverty, which implies that geographic targeting based on the marginality index is more likely to result in leakage rather than undercoverage. The fit between the two methods is particularly tight for the low and very high marginality categories, and is more diffuse in the middle categories. This suggests that the PROGRESA marginality index loses its power of distinction between medium-marginality localities. The analysis also showed that PROGRESA’s targeting appears to accurately identify extremely poor households, but makes more errors identifying moderately poor households. This finding raises serious concerns about PROGRESA’s current targeting method for the less-poor areas. However, in comparison to other schemes, PROGRESA’s targeting appears to miss fewer extremely poor households. In addition, households that are included erroneously in the list of beneficiaries appear to be closer to the poverty line than households that are included incorrectly by other methods. In short, the errors of exclusion and inclusion occurring with PROGRESA’s targeting are less serious than those occurring with other feasible target and transfer schemes. Finally, the authors performed a number of simulations comparing PROGRESA’s targeting in terms of its impact on poverty alleviation relative to other schemes assuming the same total budget. For the social objectives of reducing the depth of poverty, PROGRESA’s targeting appears to be the second most effective scheme to “perfect” targeting based on consumption. However, it should be noted that the reduction in the higher-order measures of poverty accomplished by household targeting over and above those accomplished by simply including all the households in the locality are relatively small.