Essays in Multidimensional Measurement: Welfare, Poverty, and Robustness

Abstract

This dissertation consists of three different projects concerning three different aspects of measuring social welfare and poverty, when the measurement is based on more than one attribute or dimension of well-being. The first project develops a class of multidimensional social welfare indices that is sensitive to two distinct forms of inequality across the population. One form of inequality is concerned with the dispersion of the distributions of achievements and the other is concerned with the association across the attributes. This project also applies these newly developed indices to the Indian context analyzing how the consideration for the sensitivity to inequality may alter the relative performance of the Indian states. The second project is concerned with the composite indices that evaluate the overall performance of societies by taking a weighted average of their performance in different attributes and rank the societies accordingly. The selection of weights is crucial because a choice of different weights other than the one initially used often alters the ranking and thus resulting in ambiguous or non-robust comparisons. This project develops a technique that can gauge the extent of robustness of each pair-wise comparison using a 0-100 percent scale. It is further shown that the prevalence of robustness is affected by the association across the attributes. The third project is concerned with the evaluation of multidimensional poverty in the context of India when the attributes are mostly dichotomous or categorical. The project uses the poverty index proposed by Alkire and Foster (2008), which is appropriate for this purpose, and shows how the multidimensional evaluations are different from the purely income based evaluations.