Abstract

I revisit Aidt and Frank's study on revolutionary threats and democratization. The authors posit that the reform-friendly Whigs obtained a majority of seats in the House of Commons in the 1831 election due to the violence of the Swing riots. This conclusion hinges on two-stage least squares (2SLS) estimates that are about two to five times larger than their corresponding uninstrumented ordinary least squares (OLS) ones. My empirical examination reveals that the OLS/instrumental variable (IV) gap is an artifact of ignoring non-linearities, as well as heterogeneity, in treatment effects. An appropriate interpretation of the 2SLS estimates reveals that neither the claim that the riots were instrumental in passing the Reform Bill nor the notion that voters' support for the Whigs was driven by "revolutionary fears," are supported by the historical evidence. These findings have important implications for the hypothesis that revolutions, riots, and other violent protests can instigate democratic change. They also underscore the risks of making causal inferences from observational data.