1. Live squirrels! Get your live squirrels here!
2. The early randomized evidence on microcredit: not so hot, notes David Roodman
(Aside: What is wrong with the aid industry that the evidence on microcredit–started in 1974– is still uncertain and early?)
3. Were we wrong to denounce the arrest warrants for Sudan’s President? Michael Kevane writes an excellent open letter
4. I sincerely hope this is a joke
There are lots of ways to f-up analysis of an RCT.
So I realized that even really basic data on countries, like GDP and population, really isn’t good for most LDCs. And I thought about it, and I also realized that the World Bank could trivially fund a statistical bureau comprised 95% of local talent with a few hired guns brought in to train them up in modern techniques.
And, of course, we don’t. Which lead me to my next conclusion — we don’t really care very much. So, um, that may extend.
It is not really that hard to find a microcredit program that did not have much impact, and I’m surprised that academics can get publishing mileage on these. Everybody knows that a lot of microfinance programs fail. The big question is: How many work, and if any, what makes them work?
Apply Bayes’ Rule. There are thousands of MFIs out there, so how far should 3 or 4 good academic studies let us generalize? Not that far.
The better lesson is that MFIs should be structured so that impartial evaluation is possible, and they can know whether they are working or not.