Multidimensional Poverty Dynamics: Methodology and Results for 34 Countries

This paper analyses changes in multidimensional poverty over time for 34 countries and 338 sub-national regions, for which we have comparable data across at least two periods of time. The paper first describes the absolute and relative changes in the multidimensional poverty index (MPI), as well as changes in the composition of multidimensional poverty. In so doing the paper demonstrates the core statistics of dynamic multidimensional poverty analysis. Second, it presents and implements three alternative methodologies for analyzing changes across dynamic subgroups of the poor - those who entered or exited poverty - and those who stayed poor. These ‘dynamic subgroup’ decompositions cannot be implemented precisely using repeated cross-sectional data. Upper and lower estimations are constructed using empirically based assumptions and the upper and lower bounds often vary widely. Theoretically based decomposition analyses, like Shapley decomposition, are attractive in providing apparent precision. However, the paper concludes that because these are of uncertain accuracy, they are not suitable for policy monitoring. Third, moving on to more traditional decompositions by population subgroups, the paper examines changes in the MPI and its consistent partial indices over time across over 338 sub-national regions, plus a diversity of ethnic groups. In each case it identifies regions or ethnic groups where national poverty reduction is at risk of leaving the poorest subgroups behind. Finally, it describes the changes over time in Destitution, a new measure that reflects more extreme deprivations than the MPI. This extensive body of empirical evidence points to some fundamental research questions on the study of multidimensional poverty reduction.