A Bayesian Approach to the Measurement of Poverty in India

Abstract

For efficient implementation of various anti-poverty measures devised by Indian planners, a suitable definition of poverty is essential. One of the many definitions of poverty is in terms of a cut-off point for the minimum calorie intake level, MCIL, fixed exogenously by the nutritionists. A basic problem of economic interest is, however, to determine a threshold for the consumption in monetary terms, which corresponds to MCIL.

In this paper, we depart from the traditional approach which uses conditional measures to estimate the consumption threshold, such as conditional means, median etc., and formulate poverty line as follows: If y is the per capita consumption of a household and x is the corresponding per capita calorie intake, the poverty line is defined as the minimal (y _*), which guarantees the availability of the fixed MCIL of (x _*) cals with probability arbitrarily close to one. We assume a simple linear relationship between suitably transformed x and y and use the Bayesian approach in estimating the poverty line so defined. Extensive computations using the official National Sample Survey data for the Indian State of Karnataka reveal that poverty is under-estimated by the orthodox methods.