Easterly’s blog recently had a terrific post explaining some of the downsides of the new HDI methodology – which now employs a geometric average (taking the cubic root of three separate 0-1 indices), rather than their mean.
It has received this criticism from others as well – a race of highly-educated immortals living in a utopia in which money doesn’t exist would score 0 (rather than .67 in the last methodology). Admittedly, it’s always a challenge to keep the index somewhat comprehensible while still getting it to do what you want. Essentially, the index now penalizes countries for not keeping the 3 composite indices (education, GDP per capita, and life expectancy) in relative step with one another. In a way, that’s too bad because (as Easterly points out), countries making progress in education and life expectancy but still poor don’t get as much recognition for their progress. This is a problem because that’s normally how development works: among the poorest countries, small increases in GDP lead to large increases in “what matters” like education and life expectancy. Rejiggering it in effect re-linearizes that relationship. It’s also too bad in that every time UNDP switch methodologies, history is rewritten – again.
As for the “top movers” deviation problem – the idea that the 1970-2010 HDI change should be reported as a percentage of the average change in your 1970 HDI bracket – it’s just stupid to try to build into the index a control for one of the very things you want to look at… namely, whether worse-off countries are making progress vis-à-vis better-off countries. If you control for how badly off you were in the first place, it no longer gives you a look at the convergence hypothesis. I agree with one of the comment posters that indices are not meant to replace multivariate analysis.