While the incidence of education on the determination of personal income has been well documented by scholars, the same is not true when it comes to the effects of education on economic growth, where different data sets and models have yielded contradictory results. This paper shows that a democratic environment is favourable to the impact of education on economic growth, even if democracy itself may not have a direct effect. This can be at least part of the explanation for why education alone, when observed in a cross section of countries, has shown mixed effects on growth.
This looks like a very important piece of the puzzle, from a new paper by Armellini (Kyklos, July 2012). Alas I did not find the full paper online, and wonder how the authors define "democracy."