This is the first part of three papers in which we revisit issues surrounding poverty calculations in India during the 1980s and 1990s. A number of recent papers have put forward or endorsed poverty calculations based on poverty lines computed using Unit Values (expenditure divided by quantity) of food, and fuel and light items in the National Sample Survey Consumer Expenditure Surveys, which they suggest are more plausible than those produced by the Indian Planning Commission. Others have criticised a growing gap between money-metric poverty based on poverty lines and food poverty based on normative calorie consumption levels.
In this first paper we explore the use of Unit Values to compute Consumer Price Indexes (UV CPIs). It is shown that (i) the UVs calculated from the National Sample Survey (NSS) Consumer Expenditure Surveys (CES) are multi-modal corresponding perhaps to different prices being paid by different population groups, for different qualities of produce, or at different times or places; (ii) UVs vary within states (specifically by the National Sample Survey Regions (NSSR)), by expenditure group, and by town size within the urban sector; (iii) UVs do not always correspond well with prices used by the PC for its poverty line calculations; (iv) the differences between rural and urban CPIs that are reported both by the PC and by other researchers are not soundly based; (v) neither alternative methods of computing UVs nor alternative methods of computing CPIs from UVs overcome the problems identified. We argue that UVs computed from household expenditure surveys can be a useful check on prices obtained from markets and by quotation that are generally used in computing price indexes. But they are no substitute for proper price data required for CPI calculation. However, current practices of price collection and CPI calculation in India need to be thoroughly overhauled, in part because they are based on long out of date sampling schema which cannot now assess the differences in costs of living in different domains and for different social groups.
Artha Vijnana (2005) XLVII (3-4) 223-258