LAUSR

City elections in the U.S. are widely thought to be low-information contests decided by non-ideological factors. This consensus casts doubt on the possibility of electoral accountability in cities, and renders recent evidence of municipal responsiveness puzzling. However, our knowledge of how voters actually behave in local elections is severely limited by a lack of individual-level survey data collected from local contests. Using three such original surveys, I re-examine the role of ideology in mayoral elections, recruiting samples of local voters via geotargeted Facebook advertisements. In two large cities, I find ideology is a powerful and independent predictor of vote choice. Using a panel design, I find voters learn the relative ideological positions of candidates over the course of a campaign, and that learning causally impacts vote choice. The effect of ideology also replicates in a conjoint experiment fielded to a sample of small-city voters in another region. Electoral accountability is thus a plausible explanation for ideological responsiveness in U.S. cities, and the methodological tools introduced here can now be applied to a variety of questions about local voter behavior.