Over on AidWatch, Bill Easterly and Laura Freschi describe the implications for Africa of the UNDP’s new Human Development Index. Worth reading in full, but I couldn’t agree more with one of their main conclusions:
…at the very least it’s clear that obscure choices of method make a big difference in who you celebrate – and who you make look bad.
This reminds me of the climate debate, where discussion of the discount rate (which drives almost all the results) is often relegated to an appendix and too often ignored (with some notable exceptions). Skim over the math at your peril.
I actually had a professor who used to grade using a geometric average (this will make sense after reading the post above): overall score = (homework^0.25)*(midterm^0.25)*(final^0.5). This was better than the professor who liked to give negative points if you wrote something stupid, although now that I’m on the other side of the classroom I’m much more sympathetic.
2 Responses
Interesting idea to standardize first – I suppose it depends on how much one believes that the levels (for literacy etc) have objective meanings. If so, one would probably want to endogenously give more weight to the components with higher variance. But I’m not sure if that’s the case.
I find that students rarely realize the variance angle, instead assuming that (e.g.) 25% for homework really means 25%, even if they all get close to perfect scores there.
I grade using standardized scores. Means I can’t do a geometric average as there are lots of negative numbers after multiplying.
What would it do to the HDI if one were to standardize the component scores before adding? I assume that currently the HDI gives the greatest (implicit) weight to the component that has the largest variance.