Understanding the mechanisms by which political advertising affects voters is crucial for evaluating the welfare effects of campaign finance and election regulation. This paper develops a method to distinguish between two alternative mechanisms for advertising influence: an “informative” channel in which voters learn about candidate attributes through advertising, and a “persuasive” channel in which voters can be directly influenced by advertising even if it makes no contribution to the quality of information they possess about the sponsoring candidate. I separately identify the impact of each mechanism on voter choice by taking advantage of variation in voters’ prior information. I first construct a dataset of all television advertisements aired in the 2002 and 2004 US Senate and Gubernatorial elections, matched with voting data at the media-market level. I then construct and estimate a structural model of vote choice that allows for both informative and persuasive effects of advertising. The results are largely consistent with the persuasive hypothesis. Using the estimated parameters, I conduct counterfactual analyses of several alternative campaign finance regimes and discuss implications for voter welfare.