Identifying the poor: a multiple indicator approach

The standard approach to the study of poverty assumes the existence of an ideal variable that captures the extent of deprivation. In this paper we postulate that poverty is involved with many dimensions. We use a latent variable framework to predict the extent of an individual's hardship as a function ?i =ax1i + bx2i +..., where the x's are indicators of i's income status, yi, and the latter variable is not observed.