Oxford researchers have developed a new Multidimensional Poverty Index (MPI) for the UN. The Economist provides a general guide for the newbie, and Duncan Green gives more analysis. The MPI brings together indicators on health, education, and living standards to calculate an aggregate number for the level of poverty affecting by an individual, group, or nation. Green sees it as a small step forward on the previous HDI, but he wants to see a more radical rethink that includes a more expansive set of indicators.
Green also hosted a pair of dueling guest posts on the MPI. The criticism comes from the World Bank’s Martin Ravallion. He points to the obvious complications of using arbitrary weights to compare different outcomes on one scale (i.e. what’s the relative value of one year of a child’s education vs. having regular electricity in the home?). More importantly, he argues, there isn’t much to be gained from a multidimensional index: “Being multidimensional about poverty is not about adding up fundamentally different things in arbitrary ways. Rather it is about explicitly recognizing that there are important aspects of welfare that cannot be captured in a single index.”
The rebuttal comes from Sabina Alkire, co-creator of the MPI. She defends the utility of the MPI. Unlike income poverty measures, the MPI allows one to drill-down into the constituent factors. She further says that using arbitrary weights is a strength, not a weakness, as long as the choices are explicit and open to debate.
I think Alkire’s point about drilling-down actually supports Ravallion’s criticisms: the interesting information comes when the data is looked at separately, not when it is aggregated. Ravallion is right that any single measure for poverty has few uses. When thinking about policies or programs, you need to deal with individual problems, not an aggregation.
It seems like these two are arguing past each other.* The real issue is whether it makes any sense to even attempt cross-country (or cross-group) comparisons. Context and historical moment are so important in development, yet we’re constantly tempted into deducing universal solutions from the average experiences of countries. Gleaning policy lessons from growth regressions seems to have fallen out of favor. A new index may give new life to the practice. Has Alkire provided a measure that is enough better than HDI and GDP to make such comparisons useful? Or does encouraging the practice just waste energy and intellect that could be more productively spent on other endeavors?
I’m leaning toward Ravallion on this, but I’m still sitting on the fence. So here’s a bit of homework: For the international development crowd, read the posts linked above and let me know what you think. For the non-development readers, think about how this debate relates to measurement issues in other fields such as education (GPA/standardized test scores vs. concepts of multiple intelligences) or corporate finance (quarterly earnings vs. a range of financial ratios). Report back, 250 words or less.
P.S. Matt at Aid Thoughts also recaps the debate, and comes down on Ravallion’s side, concluding that such indices are no more than intellectual exercises.
* This one is for the math nerds: I once heard Ray Fisman use a great analogy in the context of another development debate. He said it was like the debaters are two-dimensional creatures, but they only share one dimension with their opponent. So they each think they’re hurling huge blocks at the other, but they see only slivers coming at them, which they easily dodge.