Wednesday, April 1, 2009

Root causes of African underdevelopment

What is the main cause of underdevelopment? This is a question that has gotten a lot of attention in the development literature.
  • Acemoglu et al. and Rodrik et al. argue the most important determinant is institutions, instrumented by settle mortality.
  • Nunn, who I've noted before, finds evidence that in Africa, the slave trade had a huge impact on the legacy of development.
  • Sachs believes its disease, and specifically malaria, that is holding Africa back.
Which of these theories is correct? First, lets ignore the problem that the popular settler mortality data is in fact seriously flawed. How do we choose which answer is right?

The most common method so far is to just dump all of these variables into a regression and see what comes out as significant. That's what Carstensen and Gundlach have done for institutions and malaria and Bhattacharyya has done for institutions, malaria and the slave trade in Africa.

The results strongly suggest that malaria does a better job of predicting current development than any of the other arguments.

Just running regressions and looking for statistical fit though, as I have mentioned before, is a bad way of approaching the question. A better way is to ask which of the models themselves makes more sense. This question is then best answered by the Bayesian Information Criterion (BIC).

Luckily, Bhattacharyya has his data online, and even uses limited information maximum likelihood (LIML) estimation, which, for those Bayesian fans out there, has the same properties as the posterior distribution if the Jeffrey's prior is used. This is not a bad assumption to use.

So, to test model fitness becomes really quite easy. I ran the BIC test on Bhattacharyya's data, and the results back up his findings very nicely. The model comparison suggests that the malaria only model is the best, followed by dropping settler mortality, followed by dropping distance, followed by the full model.

While this all sounds very technical, the take away is that, even though I hate to admit it, Sachs is correct, malaria rules, not institutions, and not the slave trade.

No comments: