Modernization and Discrete Measures of Democracy

We reassess the empirical evidence for a positive relationship between income and democracy, commonly known as the ?modernization hypothesis,? using discrete democracy measures. While discrete measures have been advocated in the literature, they pose estimation problems under fixed effects due to incidental parameter issues. We use two methods to address these issues, the bias-correction method of Fernandez-Val, which directly computes the marginal effects, and the parameterized Wooldridge method. Estimation under the Fernandez-Val method consistently indicates a statistically and economically important role for income in democracy, while under the Wooldridge method we obtain much smaller and not always statistically significant coefficients. A likelihood ratio test rejects the pooled full sample used under the Wooldridge estimation method against the smaller fixed effects sample that only admits observations with changing democracy measures. Our analysis therefore favors a positive role for income in promoting democracy, but does not preclude a role for institutions in determining democratic status as the omitted countries under Fernandez Val-fixed effect method appear to differ systematically by institutional quality measures which have a positive impact on democratization.