Retrospective questions from recent surveys let us estimate rates of church attendance among children and their parents in ten Western democracies throughout most of the 20th century. We combine these time series with standard sources to test competing theories of religious change. Although our attendance estimates affirm the prevalence of religious decline, our statistical tests offer no support for traditional theories of secularization (which link decline to changes in income, education, industrialization, urbanization, and family life). Nor can we attribute much of the observed decline to growth in the welfare state. But increased school spending by governments does reduce church attendance, and this effect is not the result of greater educational attainment. In shaping the content of schooling, governments may strongly influence long-run religious trends.