Oil and democracy: More than a cross-country correlation?

Abstract

This article revisits the empirical relationship between oil and democracy. Existing studies establish a negative cross-country correlation between oil and democracy in a Pooled OLS framework. This insight has recently been challenged by claiming that it suffers from omitted variables, and that including country fixed effects removes the statistical association between oil and democracy. This article argues that because of the high persistence of democracy and oil, the findings of no effect of oil on democracy suffers from weak instrument problems and is not informative. Indeed, this article shows that levels of oil systematically predict both levels and changes in democracy in a sample of up to 156 countries between 1972 and 2002. By considering a different identification assumption for oil and by using additional and more informative moment conditions to instrument the regressors, this article demonstrates that the relationship between oil and democracy is negative also when country fixed effects are taken into account. This result holds for alternative measures of democracy and alternative measures of oil abundance; it is robust when including additional covariates and when removing major oil producers.