Measuring Changes in Multidimensional Inequality - An Empirical Application

During the past decade there has been a growing opinion of including more than an income perspective in the examination of inequality. As a result a broad theoretical literature on the subject of multidimensional inequality is present. This can mainly be divided into three different parts: item-by-item, non-aggregative and aggregative approach. However, there is hitherto no agreement over the measurement of inequality when each individual or household is characterized by a variety of attributes of wellbeing. In addition, there are less empirical examinations applying a multidimensional perspective to inequality. We apply three existing techniques, one from each of the mentioned strands of the theoretical literature, to the particular question of whether multidimensional inequality increased or decreased in Zambia between 1998 and 2004 using household indicators on expenditures, educational level, health status and land holdings. The purpose is to assess strengths and weaknesses of these theoretical methods in an empirical context and accordingly review their usefulness for measurement and policy analysis. Our examination points to that inequality comparisons taking interrelations between attributes into account repeatedly are at odds with comparisons of independent distributions. Consequently, if employing the item-by-item approach, at minimum, one should check the correlations between welfare distributions. The assessment of the aggregative approach show evidence of that different dimensions of wellbeing compensate and reinforce each other with respect to inequality in an empirical context. However, a majority of the results are very sensitive to the degree of substitution between attributes chosen. Sensitivity analyses and explicitness should thus accompany examinations of this kind. In applying a non-aggregative approach few combinations fulfill the required dominance conditions. Accordingly, generality and less imposed structure come at a cost. We conclude that the empirical usefulness of these existing techniques is reasonable as long as we stay aware of intrinsic weaknesses. Clearly, careful interpretations and analyzes involving more than one technique is constructive to portray multidimensional inequality.