An extensive body of work has found that countries with relatively educated populations are more likely to be democratically governed. Further, a large body of work argues that education is associated with a host of individual-level factors, such as political participation and democratic values, which provides a micro-level mechanism to explain the link between education and democracy. The  central claim is that education universally engenders democratic values, which in turn, drives individuals to make claim for democratic governance.  I build on this prior research in three respects. First, in Paper 1, using a sample of identical twins I show that the impact of education on political knowledge is highly confounded by family background.  Education has a positive impact on knowledge for those individuals that were not exposed to political discussion in the home during upbringing.  But for those that discussed politics with family, education has no impact on political knowledge.  Second, I challenge the claim that education has a universally positive effect by examining the role of political context.  In Papers 2 and 3 I leverage education reforms as quasi-experiments to study how the effect of education on political attitudes varies in authoritarian and democratic countries.  In Paper 2 we find that education at the primary and secondary level has no impact on support for democracy in principle, but that education in an authoritarian context leads to less satisfaction with democracy after a country transitions, whereas education in a democratic context leads to greater satisfaction with democracy.  In Paper 3 I find that higher education in an authoritarian context weakens support for authoritarian rule, but that this effect is mitigated by a strong economy. Finally, in Paper 4 we focus on the validity of survey measures of regime support in authoritarian states.  Through a series of list experiments implemented in a novel web-based survey in China we find that respondents self-censor their true level of regime support to a large degree.  Further, the level of self-censorship varies greatly by income, age, residence status, and education.