The Multidimensional Poverty Index: Achievements, Conceptual and Empirical Issues

The MPI has been an interesting and important effort to provide a household-level multidimensional poverty measure that can compete in depth and coverage with the widely used (and problematic) $1.25 a day income poverty indicator. We strongly suggest that Human Development Report Office (HDRO) continues to use an MPI-type indicator in its future Human Development Reports. At the same time, there are many open questions and issues regarding the conceptual underpinning and alternative formulations of the MPI. We suggest that these issues are carefully considered and possibly a revised MPI be produced that reflects different choices. Among the issues we would flag particularly are the use of the union (instead of the dual cut-off) method for identification and considering inequality in deprivations across people in the MPI (at least in some version of the MPI). We also believe that the headcount would be simpler and more transparent as the headline indicator, with a second measure reflecting intensity (and possibly inequality). Regarding the empirical implementation of the MPI, we propose a number of changes, including dropping the WHS as one of the data sources, dropping the BMI as a nutrition indicator, and changing the age ranges and cut-offs for the education and mortality indicators. In line with the request to simplify the living standards measure, we also recommend to focus on only three living standard indicators (water, floor, and assets). We illustrate the empirical relevance of these changes using the Demographic and Health Surveys (DHS) for Armenia, Ethiopia, and India. We believe that these changes would represent improvements over the current formulation but want to emphasize that one would need to investigate these proposed changes in more detail to come to more definitive conclusions about this. In a final section, we briefly present and comment on the way the HDRO has revised the MPI in the 2014 Human Development Report which has been partly based on the recommendations made in this paper.