Culture which evolves over time and is affected by many factors, including geographical adaptation, is the MAIN issue that everyone is afraid to look at when talking about the state of Africa today. Democracy doesn’t work because it is culturally irrelevant in Africa.

Your loaded question(“And are you really in agreement with Ms Moyo”) indicates that you’re missing a key part of the debate by (I assume) not having read Dambisa’s book.

On p.44 she gives an example of a local African net maker employing 10 staff being put out of business by a movie star arranging for 100,000 $10 nets to be dropped in as aid.

She calls this the micro-macro paradox. The well-intentioned aid has put the local firm out of business creating more need for aid and knocking out local capabilities. If the $1 million had been invested in the local firm think of the possibilities. Another example of “what is unseen” in the aid debate.

@Matt: I think I see… if the effects of environment were not restricted to mortality of settlers only, they might as well be correlated with income, leaving environment an invalid instrument. Thanks again for pointing that out.

No need to stop

Your posts are very much appreciated!

You’re right, we want to know the total effect of institutions, but my point is that a crappy instrument could easily lead to an over-estimation of the effect of institutions. AJR use settler as a proxy an “unhealthy settler environment.”

The crucial assumption though, is that that “unhealthy settler environment” only applies to settlers, and so determines their decision to set up beneficial institutions, or something closer to an extractive state. But, due to the validity restriction, AJR assume that the “unhealthy environment” doesn’t apply to the local population, at least not enough to affect the long-run growth path. They “prove” this by including measures of health, malaria incidence, etc in their regressions.

If AJR are wrong about their restriction (as I think many suspect) then their instrument isn’t valid, which, combined with the fact that it is a weak instrument to begin with, can lead to an overestimation of institutions.

I’m not, by any means, defending Sach’s view. I think that Bill is right in that institutions matter more than any other factor in development, and a lot of other factors actually flow through institutions. But you can’t reject all other factors based on a few cross-country studies – it’s not much more defensible than claiming aid is good for growth!

I’ll stop now

Isn’t the total effect of institutions what we want to know?

Say geography matters, but it did so only through institutions. Then would this not favor (or actually be) Bill’s view? As opposed to if geography mattered in any direct or indirect way other than through institutions (which would more be Sachs’ view).

I thought that we’d like to regress income on geography and institutions where institutions may be endogenous. And that this is where Acemoglu et al use “healthy environment” as an instrument for institutions. What exactly do you criticize about that?

I’m not the greatest econometrician so please correct me if I’m mistaken somewhere.

You are totally right: AJR bring to their work an econometric rigour that have yet to see in the Collier work. They also have a general theory at their back when they dive into their regression work, so it’s unlikely that they are engaged in any data mining.

However, tiny mistakes in even more rigorous studies can still bring us to the wrong conclusions (we can even be worse off, i.e. with a weak/invalid instrument problem).

My original point was: nearly all econometric studies are flawed – *especially* cross country ones. We need to be cautious when completely rejecting factors like geography based on just a few of these studies.

You are right there are some question marks about Acemoglu, Johnson, Robinson. But there is no comparison with Collier, they are in different universes. AJR are serious about causality and Collier is

not. AJR are not data mining and Collier is. AJR have a serious theory guiding empirical work, and Collier does not.

Regards, Bill Easterly