At Duncan Green’s blog, there is a fascinating back-and-forth on the UN’s new Multidimensional Poverty Index (MPI) between its co-creator, Sabina Alkire, and the World Bank’s Martin Ravallion. This is very much a live debate in development circles. The MPI is a descendant of the earlier Human Development Index and is similar to the various Unsatisfied Basic Needs indices long used in many countries.
I agree wholeheartedly with Martin’s critique, but Sabina does offer a spirited (and highly hyperlinked) defense. Martin’s emphasizes two points: 1) what’s the point of aggregating a bunch of indicators into a single index? and 2) the choice of weights for such an index is inherently problematic.
No development economist would disagree that poverty is multidimensional. Our notion of poverty includes measurable consumption along with things like health, education, and access to infrastructure, as well as less tangible aspects like rights and opportunities. But the question is, should we try to squeeze all those measures together into one super sausage of an indicator, or should we just consider them separately?
Sabina’s response illustrates the problem with the all-in-one measure: to explain the contrasting values of the MPI for Kenyan Somalis and Kenyan Masai, she immediately goes to discussing the differences in the index’s components: child mortality, school attendance, etc. But if we want to compare the welfare of these two groups, why not start out by looking directly at the components, rather than mashing them all together and then pulling them apart again?
Sabina argues that the added value of the MPI is that it captures the overlap between its various components. But if a quarter of the children in a country are malnourished and a quarter lack access to clean water, to what extent are we talking about the same children? In practice, the correlation between such measures is likely to be high, and the best way to examine the overlap would be to consider it directly, e.g. tabulating child malnourishment vs. access to clean water.
The more powerful critique of the MPI is that there is no solid basis for the choice of index weights. When I teach poverty measurement, I give my students as a homework assignment the task of constructing their own country-level welfare indices, using whatever formula and data they like. They end up with formulas like: the square root of the child mortality rate, minus half the murder rate, plus one third the number of kilos of ice cream consumption per capita. Who am I to say that this is an adequate welfare measure? (What, you don’t like ice cream?) Unsurprisingly, country rankings vary greatly across the different measures my students concoct.
Sabina’s response to the problem that the weights are arbitrary is that she’s not claiming the final word: “we strongly encourage countries to develop national measures having richer dimensions,” and “critical scrutiny” of the weights is welcomed. The point, however, is not that some other set of weights would be better. There is simply no “right” way to come up with such an index, so the index will ultimately reflect the preferences of its designers.
I do have a bit of sympathy for one argument for the MPI: it could encourage public discussion on the fact that poverty is multidimensional. The general effect of launching a composite measure, such as the Corruption Perception Index or the Commitment to Development Index, is to focus attention on the subject matter. We’ll probably hear more people talking about multidimensional poverty in the next few months as a consequence of the MPI.
Still, my very practical worry is that the new push for multidimensional poverty indices will soak up much of the oxygen around poverty work.
Here in Kenya, the Kenya National Bureau of Statistics has given us a wealth of data to work with to measure and think about poverty. We have a number of Demographic and Health Surveys, the 2009 Population and Housing Census which will be published in the next few months, the 2005-06 Kenya Integrated Household Budget Survey (KIHBS), and a new round of the KIHBS scheduled for next year. Rather than using all this data to calculate the MPI and variations, I’d prefer to focus on, say, understanding how Kenya has achieved a stunning drop in child mortality since 2003 (more on this in a future post.)