The India Human Development Survey (2005) is a nationally representative survey of 41,554 households. The sample encompasses 33 states and union territories of India with the exception of small populations living in the island states of Andaman & Nicobar and Lakshadweep. The sample comprises 1504 villages and 970 urban neighborhoods in 383 (of the 602) districts in India. Of the total, 26,734 households reside in rural areas and 14,820 in towns and cities. The IHDS sample design combines two separate samples: a re-interview sample (N=13,900 households) of households first contacted in 1994-5 and a new sample (N=27,654 households).
The IHDS aimed to re-interview about half the rural households that had been in the earlier Human Development Profile of India, 1994-5 (HDPI) (N=33,230). Households were selected randomly within the stratified 1994 sampling design. Split households that remained within the village were re-interviewed. New households were added in four distinct components:
Sufficient records were not always available from the original HDPI villages to re-locate those households for IHDS. To maintain a representative sample, an equal number of replacement households were drawn from those HDPI villages with missing records.
Two new villages in each of the HDPI districts were randomly drawn with probability proportional to population size. A sample of 20 households was selected from a village listing in each of the two new villages.
IHDS was conducted in 10 states and union territories that were not part of the earlier HDPI survey. A three-stage sample (districts, villages, households) was randomly drawn in each of these new states or territories (N=1080 households). Sample sizes varied from 60 to 400 depending on the size of the state.
The 1994 HDPI included only rural areas (although 277 households in 19 villages had been re-classified as towns by the 2001 census). Therefore, IHDS drew a new sample of households (N= 14,543) for urban areas. Within each state or union territory, towns or cities were selected with probability proportional to population, and three or more census neighborhoods were randomly chosen within each town. Fifteen households were randomly selected from each neighborhood following a complete household listing.