Poor proxies for poverty: The problems of rationing benefits through poverty-targeting.

Poor proxies for poverty: The problems of rationing benefits through poverty-targeting

Proxy means testing

When considering social assistance schemes, the inevitable debate over financing takes centre stage: how to reach (target) the poorest, while keeping costs down? Means testing based on income is difficult in low- and middle-income countries as many people work in the informal economy.

To address this, proxy means testing (PMT) is a popular method to identify poor households. Proxies are derived from national household surveys, where statistical methods are used to determine the relationships between observable household characteristics and consumption. These proxies are meant to correlate with poverty and are based on factors such as asset ownership, housing, and demographic characteristics.

How accurate is proxy means testing?

PMT proponents argue it’s an accurate method for predicting the welfare of different households. Two recent reports challenge this popular narrative.

A World Bank report analyzes various PMT methods in nine Sub-Saharan African countries, concluding these ‘methods are particularly deficient in reaching the poorest’. When assessing the most common PMT method, on average almost half (48%) of those identified as poor are non-poor. Alarmingly, 81% are incorrectly identified as non-poor.

The second report assesses the sensitivity of a PMT tool (Progress out of Poverty Index (PPI)) for a Catholic Relief Services’ project. PPI data is compared with two years of financial diaries, and in-depth wealth/wellbeing rankings for 255 households in Zambia. Findings show that although the PPI score found a difference of just 7 percentage points between households categorized as ‘managing’ vs. ‘poor’, ‘managing’ households actually spent 160% more than ‘poor’ households. While PPI can identify poor households, it cannot identify who the poorest households are.

Unintended negative consequences…

PMT may also lead to unintended negative consequences. Another World Bank report presents evidence that increased village incomes from a conditional cash transfer programme in the Philippines (using PMT), likely led to price increases of perishable protein-rich foods. The indirect effect was increased stunting among non-beneficiary children, while stunting reduced for beneficiary children. Considering the inaccuracies of PMT, many excluded households were likely poor, and unable to absorb price increases. The report concludes that in areas where poverty is prevalent, benefits shouldn’t be rationed, but provided to all households.

Investing in inclusion

While PMT is advocated to governments as an effective mechanism, the evidence shows it excludes the majority of the poorest households, cannot accurately identify the poorest households, and can negatively impact children.

While providing universal access to certain support involves greater initial investment and political buy-in, it can still be affordable and is the most effective way to invest in the poorest in society.

Save the Children is building evidence on how inclusive social protection systems are essential to ensure social assistance is accessible to all who live in poverty. For example, in Myanmar, where two in five rural households experience poverty and one in five children from the wealthiest households are malnourished, we are supporting the government to phase-in a universal approach to cover all pregnant women and children under two with income support and nutrition education.

Mathew Tasker, Social Protection Advisor, Save the Children UK

Leave a Reply


  • Richard Morgan

    Hi Mat – thanks for the great blog on this vital issue. I would add just a couple of thoughts which to my mind further reinforce the case for categorical targeting: it seems inappropriate to rely on “point in time” (and seldom, expensively updated) measures of who is poor, when we know that incomes among the lowest-earning households are extremely fluid and descents into extreme poverty can be very rapid. Secondly, the costs to individuals (not least to children) and to wider society of inadvertent exclusion by blunt targeting methods are likely to far outweigh the costs of inadvertent inclusion. “Errors” should err on the side of inclusiveness.