A response to the weaknesses of the Multidimensional Poverty Index (MPI): the correlation Sensitive Poverty Index (CSPI)

In 2010, the Human Development Report introduced the Multidimensional Poverty Index (MPI). The MPI complements traditional income-based poverty indices by measuring the multiple deprivations that households face at the same time. Besides other appealing properties, in particular its decomposability, the MPI has the advantage that it is very easy to calculate. At the same time, the simplicity of its approach causes a number of methodological weaknesses. The calculation of the MPI is based on ten vital items that are weighted differently according to their importance. The MPI is a counting index, as it simply counts the number of weighted items that households lack. All households for which this number is at least 30% are considered poor. All other households are considered non-poor and therefore excluded from the calculations. The appealing simplicity of the MPI, however, comes at a cost. The MPI has four main methodological weaknesses: Since the MPI simply counts the number of items lacked by households, it assumes that no correlation exists between them. This assumption is not realistic. It is rather safe to say that, for instance, proper sanitation and safe drinking water are related to health as well as education indicators. The MPI is unable to capture inequality. In other words, transferring items from a poor to a less poor household does not change the poverty index as long as both households remain poor according to the MPI. The cut-off level of 30% is an arbitrary choice; changing it would affect poverty rates and even country rankings. The specific structure of the MPI implies problematic distortions. It leads to an inflation in poverty rates that increases the poorer a country and thus the severer its budget constraints. This results in less attention paid to the neediest of the needy. Rippin (2010) introduced the Correlation Sensitive Poverty Index, a new index that shares the appealing properties of the MPI but none of its weaknesses. The CSPI is a counting index like the MPI and therefore shares its decomposability as well as its simplicity. However, the CSPI does not require a cut-off. Instead of excluding households from the calculations, it weights each household according to the number of weighted items that it lacks. This unique structure leads to the following advantages compared to the MPI: The CSPI is able to capture the correlation between the poverty indicators. The CSPI captures inequality among the poor; it increases whenever items are transferred from a poor to a less poor household. The CSPI does not require the arbitrary cut-off but instead provides policy makers with the opportunity to deliberately choose the level of importance they want to attribute to inequality among the poor. Finally, the new index avoids the inflation of poverty rates for poorer countries and puts a greater emphasis on the neediest of the needy in those countries than the MPI.