In a previous post, I wrote about ongoing research to create a wealth index out of IPUMS-I data with my population center colleagues.
This summer, we learned about a hip new way to measure wealth. At Oxford, Sabina Alkire and Maria Emma Santos created the Multidimensional Poverty Index (MPI). Their index expands on the human development index as broader and more layered way to measure poverty on the individual level. In our ipums-i wealth index, we only include household variables (assets, utilities, household characteristics), the MPI includes additionally includes education and health. This multi-dimensional approach allows a researcher to zoom in a bit more on individuals, whereas our index can only tell you about households. And only for the countries who have been so kind to share their census data.
However, on a technical note (i.e. nerdwarning), the MPI has a rather simplistic methodology for calculating weights of each input variable. And it turns out choice of weights is actually the most controversial part of making a wealth index. In our index, we use first principle component analysis to determine the weights of each input variable. This captures the linear combination of all variables with the largest amount of common information (maximum variance) and then reports the scoring factor for each variable's contribution.
Lastly, our glass is clearly half full. We call it a wealth index, they call it a poverty index.
This summer, we learned about a hip new way to measure wealth. At Oxford, Sabina Alkire and Maria Emma Santos created the Multidimensional Poverty Index (MPI). Their index expands on the human development index as broader and more layered way to measure poverty on the individual level. In our ipums-i wealth index, we only include household variables (assets, utilities, household characteristics), the MPI includes additionally includes education and health. This multi-dimensional approach allows a researcher to zoom in a bit more on individuals, whereas our index can only tell you about households. And only for the countries who have been so kind to share their census data.
However, on a technical note (i.e. nerdwarning), the MPI has a rather simplistic methodology for calculating weights of each input variable. And it turns out choice of weights is actually the most controversial part of making a wealth index. In our index, we use first principle component analysis to determine the weights of each input variable. This captures the linear combination of all variables with the largest amount of common information (maximum variance) and then reports the scoring factor for each variable's contribution.
Lastly, our glass is clearly half full. We call it a wealth index, they call it a poverty index.
RSS Feed