I study the effects of direct democracy on economic policy in a novel setting. In Spain, national law determines that municipalities follow either direct or representative democracy, depending on their population size. Using a fixed-effect regression discontinuity design, I find that direct democracy leads to a smaller government, reducing public spending by around 8 percent. Revenues decrease by a similar amount and, therefore, there is no effect on budget deficits. These findings can be explained by a model in which direct democracy allows voters to enforce lower special-interest spending. I provide several additional results and discuss alternative mechanisms.