In this paper we examine the long-run relationship between religiosity and income using retrospective data on church attendance rates for a panel of countries from 1925 to 1990. We employ panel cointegration and causality techniques to control for omitted variable and endogeneity bias and test for the direction of causality. We show that there exists a negative long-run relationship between the level of religiosity, measured by church attendance, and the level of income, measured by the log of GDP per capita. The result is robust to alternative estimation methods, potential outliers, sample selection, different measures of church attendance, and alternative specifications of the income variable. Long-run causality runs in both directions, higher income leads to declining religiosity and declining religiosity leads to higher income.