We implement a new approach for estimating the persuasive effects of political advertising. Our empirical strategy exploits FCC regulations that result in plausibly exogenous variation in the number of impressions across the borders of neighboring counties. Applying this approach to uniquely detailed data on television advertisement broadcasts and viewership patterns during the 2004 and 2008 presidential campaigns, our results indicate that total political advertising has virtually no impact on aggregate turnout. The point estimates are precise enough to rule out even moderately sized effects. By contrast, we find a positive and economically meaningful effect of advertising on candidates’ vote shares. Evidence from a regression discontinuity design with millions of observations suggests that advertising’s impact on elections is largely due to compositional changes of the electorate.