We develop a theory of political support for religious parties. We predict that the probability that a voter supports a religious party (i) increases in income for the poorest voters, but decreases in income for the richest; and (ii) is greater for voters in richer districts. We test these predictions on original micro-level data in a nationally representative sample of 600 individuals in 30 districts in Tunisia. Our empirical results confirm both predictions and suggest that belonging to the middle class and living in a richer district together affect voting decisions more than being a religious voter. We document similar patterns in other key elections in the Muslim world.