This paper is the result of a global review of the effectiveness of different methods of selecting social protection recipients, both universal and poverty-targeted social protection schemes. The work, supported by the Church of Sweden, considered the effectiveness of 38 programmes across 23 low- and middle-income countries, including means-tested schemes and schemes using proxy means testing, community-based targeting, self-targeted and pension testing. The research sought to answer both how effective the different types of targeting mechanism are in reaching their intended recipients, and their effectiveness in reaching those living in extreme poverty specifically. The findings are a damning indictment of advocates for poverty-targeted social protection, with only one of the programmes using this selection method reaching over half of the poorest 20 per cent of the households it is targeted at. Some means-tests performed well when compared to other income testing mechanisms. This included the simple means-test used in Brazil’s Bolsa Família programme, though this is likely due to its use of quotas in each municipality; and South Africa’s means-test, though this sets out to exclude the more affluent rather than targeting those living in extreme poverty. However, universal schemes were the most effective in reaching both their intended recipients and the poorest 20 per cent of people within their intended categories.