In many developing countries the standard approach to analyse poverty and inequality is still to adjust household income or consumption expenditure by the number of household members (per capita approach). However, when we want to identify the poor and target them with specific policies it becomes apparent how the per capita approach tends to under-estimate poverty among small households and over-estimate it in large ones. Indeed, households incorrectly excluded from social assistance benefits because considered better-off are more likely to make complaints and the exclusion of small households usually is also noted by social workers. Moreover, when social protection policies are considered, it is very important to determine the specific needs of subgroups of the population who tend to have extra needs. In particular this applies to people with disabilities who otherwise tend to be excluded from certain types of social assistance.

This paper explores these issues by analysing rich household survey data and administrative targeting data for social assistance from two different countries: Moldova and Mongolia. It estimates equivalence scales using different methods: expert opinions, use of subjective assessments and the living standard approach. In particular we measure equivalence scales for people with disabilities and we find substantial extra costs that should be accounted for when assessing their living conditions. Although every estimate represents a simplification, we assess the average performance of such estimates against subjective assessments of household living conditions made by social agents, who visited households in their homes to determine their eligibility to social assistance, and compare the relative performance of the proposed equivalence scales against the per capita approach.

Results show that it is extremely important to use equivalence scales and economies of size when calculating welfare living conditions and that the simplicity of the per capita approach should not justify its use since it generates large targeting mistakes.