Proxy means tests don’t work

A review of the effectiveness of proxy means tests by Brown, Ravallion and van de Walle finds that they are not an effective way of concentrating resources on the poor. The process is simply not accurate enough.

“Standard PMTs help filter out the non-poor, but exclude many poor people, thus diminishing the impact on poverty. … The prevailing methods are particularly deficient in reaching the poorest.  … The most widely-used form of PMT in practice does only slightly better on average than an untargeted universal basic income scheme, in which everyone gets the same transfer, whatever their characteristics. Even under seemingly ideal conditions, the “high-tech” solutions to the targeting problem with imperfect information do not do much better than age-old methods using state-contingent transfers or even simpler basic income schemes.”

Proxy means tests are being used because poor countries just don’t have the quality of information to make fuller assessments work.  As many critics of means testing have pointed out, richer countries don’t have the capacity to do it either.  People’s incomes fluctuate, boundary problems are intrinsic, people don’t understand what should be included and what should not be, and take-up is consistently poor.

There is however one large reservation to make about this study’s findings.  There have to be doubts as to whether any country, rich or poor, really has the capacity to produce the kind of information that detailed quantitative studies of this kind call for.  This study points to the difficulty that any test has in determining whether or not specific individuals are poor.  The standard they use to verify the connections, household consumption, is not absolute proof of poverty; it’s an indicator.  It’s probably more valid than some other indicators, but it isn’t perfect and it is just as difficult to collect as income.  I happen to agree with the paper’s conclusions about proxy means tests, because they happily coincide with my own judgment, but nothing can be supposed to be proved beyond doubt; the core information that the analysis is built on is not good enough, and it cannot be.

Leave a Reply