A colleague the other day mentioned his annoyance with the Hans Rosling TED talks on global poverty. His annoyance generally stems from treating developing nations’ GDP estimates as anything more than numerical hocus-pocus.

A few things I know basically nothing about:

* I really have no idea how hocus-pocusy these GDP estimates are.
* I also have no idea how hocus-pocusy the U.S.’s own GDP estimates are.
* Another thing I have no idea about is whether every year’s GDP estimates from a given country are mangled in the same way, so that (estimated GDP in year 2) minus (estimated GDP in year 1) is actually a reasonably accurate measure of year-over-year change in GDP.
* Per-capita GDP estimates might introduce another source of uncertainty, namely uncertainty in the population estimates. I likewise have no idea how accurate most nations’ population estimates are. And I have no idea whether per-capita-GDP estimates come from sampling individual people on their incomes, or estimating the country’s aggregate GDP and dividing by an estimate of the population.

I guess what I’d like, then, is a good introduction to the problems of measurement in countries with not-very-well-established economic-measurement systems — and for that matter, an introduction to how the U.S. statistical-measurement agencies do their work. Paging Chris Blattman