January 30th, 2016
(written by lawrence krubner, however indented passages are often quotes). You can contact lawrence at: email@example.com
Half the variation in income per capita in 2002 is associated with variation in technology in 1500AD. It is worth stopping here to say something about what CEG are saying empirically. This is not a “policy experiment” paper, and I don’t think it is appropriate to evaluate it as such. This is a paper about forecasting, basically. What their result says is that if you tell me the level of technology in 1500AD, I can predict with a good amount of accuracy your income per capita in 2002.
That’s different than saying that if one could increase technology in 1500AD by “1 unit”, you could raise income per capita today by a factor of 26. First, as is obvious, there is no way to actually do that. Second, this isn’t the interesting research question here. CEG are trying to establish that old technology levels have predictive power for current income per capita. They are not looking for explanatory power, or forecasting power, and not a strict causal effect.
Should they be looking for a strict causal impact? That’s a fair question, but one without an obvious answer. Ultimately, you might want to argue that we want strict causal explanations for why some countries are rich in 2002. But an explanatory paper like this is valuable to that search. Knowing that anything in 1500AD has strong predictive power for incomes today is informative. It tells us that we have to look back to 1500AD for at least some of those causal forces. Even if it turns out that technology in 1500AD is just a proxy for institutions, or geography, or culture, or whatever, this is a time period we should be looking at.
All this means that CEG have evidence of cross-sectional persistence. Gaps in technology in 1500AD appear to persist for centuries. This doesn’t mean the gaps are perfectly persistent (i.e. will last forever), just that they take a loooooong time to dissipate. Is this surprising? Perhaps not, if you think that there is a lot of time-series persistence in technology.