The paper contrasts Lipset’s modernization hypothesis and Przeworski- Limongi hypothesis that entries into democracy are random with respect to income. We use data on income and democracy going back to 1820, multiple definitions of democracy, and non-parametric testing focusing on the distribution of entrants’ incomes. We find that income matters for entry into higher levels of democracy; but if we control for the previously achieved level of democracy, the income effect vanishes. This means that countries that enter into higher levels of democracy are not a random draw from the universe of all country incomes but are a random draw from the joint distribution of previous level of democracy and income. These results are compatible with the presence of a subgroup of (low) income and (low) democracy countries from which recruitment into democracy is seldom made. But for other countries, accession to higher levels of democracy is income-random. Income seems therefore both to matter (probably explaining why poor countries cannot improve their democracy levels) and not matter (explaining why for other countries improvements in democracy are income-random).