Poverty has many dimensions, which, in practice, are often binary or ordinal in nature. A number of multidimensional measures of poverty have recently been proposed that respect this ordinal nature. These measures agree that the consideration of inequality across the poor is important, which is typically captured by adjusting the poverty measure to be sensitive to inequality. This, however, comes at the cost of sacrificing certain policy-relevant properties, such as not being able to break down the measure across dimensions to understand their contributions to overall poverty.
In addition, compounding inequality into a poverty measure does not necessarily create an appropriate framework for capturing disparity in poverty across population subgroups, which is crucial for effective policy. This paper proposes using a separate decomposable inequality measure – a positive multiple of variance – to capture inequality in deprivation counts among the poor and decompose across population subgroups. The study provides two illustrations using Demographic Health Survey datasets to demonstrate how this inequality measure adds important information to the adjusted headcount ratio poverty measure in the Alkire-Foster class of measures.
Key findings:
- There have been recent developments in both theory and practice in the measurement of multidimensional poverty based on counting approaches. The categorical or binary nature of dimensions and the fact that the counting measures of poverty are based on direct deprivations make the use the counting approaches more amicable. Even in counting approaches, however, it is important that all three „I?s of poverty – incidence, intensity, and inequality among the poor – are incorporated. If the object of a policy maker is to reduce only the incidence of poverty, then only the marginally poor people would be driven out of poverty ignoring the poorest of the poor completely. If the objective is to reduce both the incidence and intensity of poverty, then while the policy maker has no reason to focus on the marginally poor instead of the poorest of the poor, the policy maker has no strong incentive to assist the poorest of the poor either.
- This paper suggests the use of a separate inequality measure to capture inequality among the poor and disparity across population subgroups. Then the question is, which inequality measure is to be used? The choice of inequality measure is determined by certain desirable properties, in addition to the standard properties of inequality measures, that this study requires it to satisfy. The first requirement is that the inequality measure is additively decomposable so that it can be expressed as a sum of total within-group inequality and between-group inequality. Moreover, the total within-group inequality should not change as long as the inequality within each population subgroup does not change. Second, the study requires that inequality across deprivation scores to remain unchanged when the deprivation score increases by the same amount rather than by the same factor. In other words, it requires that inequality should be perceived through absolute distances between deprivation scores rather than relative levels. The only inequality measure that satisfies our requirements is a positive multiple of variance.
- The paper provides two illustrations to show the application of the inequality measure to analyze inequality among the poor and disparity across population subgroups. The first illustration used the DHS of 23 developing countries around the world. Although we find that the level of inequality among the poor and the level of poverty in terms of the MPI are positively associated across these countries, there were several exceptions. For example, Colombia’s MPI was less than one-seventh of Lesotho?s but with the same level of inequality among the poor. Similarly, the MPI of Madagascar was more than 50 percent larger than that of Nepal, but inequality among the poor in Nepal was much higher.
- Looking at the sub-national inequality in intensity of poverty is not enough. It conceals the high level of disparity that exists across sub-national MPIs. For example, Bolivia and Zimbabwe have similar levels of between-group inequality across the poor, but sub-national disparity in Zimbabwe is more than three times worse than that in Bolivia. The illustration on India also shows a similar result but through an inter-temporal analysis. When decomposed across castes, we find that although the MPI and its components – the incidence and the intensity of poverty – went down for each subgroup, the inequality among the poor became higher for the poorest subgroup – the Schedule Tribes.
- So what is the value added of using the proposed inequality measure? First, the inequality measure adds valuable information to the adjusted headcount ratio proposed by Alkire and Foster (2011), which has been adopted by international organizations and country governments without sacrificing the dimensional breakdown property that allows understanding the contribution of each dimension to overall poverty. Second, the inequality measure does not involve any inequality-aversion parameter, whose selection may cause wide disagreement among policy makers. Thus, the additive decomposability property allows overall poverty to be decomposed into within-group and between group components. Although the contribution of within-group and between-group components to overall poverty is subject to debate (Kanbur 2006), it is not difficult to argue that understanding their change over time and across countries may provide valuable information. Finally, the inequality measure reflects the same level of inequality whether the poor are identified in an achievement/attainment space or in a deprivation space.