William Easterly recently had a post on the "mystery of the benevolent autocrat" that illustrated the fact that non-democratic countries seem to have a higher variance in growth rates than democratic countries. Since I've been playing with visualization software, I thought I'd try to replicate his analysis and produce some pretty graphs that I could use in my dictatorships class. The results are here (make sure to click - the full-screen versions are much prettier than the embedded versions below, and the software has some nice features that make it easy to explore the data).
The first view replicates Easterly's scatterplot, but using data from the Penn World Table from 1950-2007. (At least, I think it does. My econometric skills are obviously much worse than those of a former World Bank economist, so take that with a grain of salt). It adds, however, a color for the Polity2 score - greener is more democratic here - and the size of each dot is proportional to the number of years of data available for the comparison (some countries have only a few years of data available, others have more than 50). I use the same transformation of the Polity2 score that Easterly uses - Polity2/(11-Polity2). As you can see, countries that have been on average autocratic have a higher growth variance - some have had very large growth rates over the period in question, others have had very low rates. Democracies, however, seem to be clustered towards the average growth rate of the world economy - around 2.5% per year or so or so.
second view shows the same scatterplot, except using the average of the raw Polity2 score rather than Easterly's transformation. The greatest amount of variance seems to be found in the countries that have an average Polity2 score between 1 and -5 or so - the more autocratic "anocracies" (many of them "hybrid regimes") rather than the full "autocracies" of Polity's classification (though of course a scatterplot is not a statistical analysis).
last view shows the data on a map, where the size of the dot corresponds to the average (geometric mean) of the growth rate over the period in question, and the color corresponds to the average Polity2 score - green is more democratic (at least in the Polity2 scheme).
Why do non-democracies seem to have greater growth variance than democracies? I suspect the incentives of leaders in democracies prevent large policy changes (both good and bad), whereas more autocratic systems may have greater variance in the quality of leadership and in the incentives they provide to these leaders. (Maybe I will try a more fine-grained visualization using data on types of political regime to see what comes out - later).