The data used in the analysis are from the two rounds of the National Family Health
Survey of India (NFHS), conducted in 1992/3 and in 1998/9, respectively. Although
this survey was primarily concerned with reproductive and child health, the household
questionnaire of the survey contains information on schooling for every individual in
the surveyed household. The survey covered the 26 main states of India, interviewing
88,563 households in 1992 and 92486 in 1998. In 1992/3, 69 per cent and, in 1998/9,
66 per cent of households reported living in rural areas. For rural households, we have
merged in information on relevant infrastructure indicators that is available from a
village questionnaire. In 1992, 485 villages were surveyed and, in 1998, 622. An
advantage of household survey data over administrative data is that the latter often
exaggerate school enrolment, possibly because this reflects well on school
administrators and district officials, and because public expenditure allocations to
schools and districts are often based upon the number of enrolled students (e.g. World
Bank 2004).