What are the characteristics of poor households and how are they different across regions and time? This paper proposes and operationalises a method for estimating the composition of poverty.  

The paper takes two measures of poverty – child mortality and primary school non-completion – from the internationally comparable, and nationally representative, Demographic and Health Surveys (DHS) to produce estimates of the composition of health and education poverty.

The approach is operationalised in 33 countries and estimates of the composition of ‘poverty’ by the two indicators are made for sub-Saharan Africa, South Asia and Southeast Asia for 1998 and 2007 based on a total sample of over a million households.

Those estimates generated are, in 1998 and 2007, largely consistent across the two measures used in terms of the characteristics of poor households. The estimates suggest ‘poverty’ is overwhelmingly concentrated in: (i) rural households; (ii) in households where the head of the household has ‘no education’ or ‘incomplete primary education’; (iii) in households where the head of the household is ‘not in work’ or ‘working in agriculture’. However, there are some differences across regions. The strengths and weaknesses of the approach are also discussed.