Does poor air quality affect decision-making? We study this question based on elections, in which millions of people decide on the same issue on the same day in different locations. We use county-level data from 64 federal and state elections in Germany over a nineteen- year period and exploit plausibly exogenous variation in ambient air pollution within counties across election dates. Our results show that a high concentration of particulate matter (PM10) on an election day significantly affects voting behavior. An increase in the concentration of PM10 by 10g/m3 - around two within-county standard deviations - reduces the vote share of the incumbent by 2 percentage points and increases the vote share of the established opposition by 2.8 percentage points. These are strong effects, equivalent to 4% and 7% of the respective mean vote shares. We generalize these findings by documenting similar effects with data from a weekly opinion poll and a large-scale panel survey. We provide further evidence that emotions are a likely mechanism: the survey data show that poor air quality leads to greater anxiety and unhappiness, which may reduce the support for the political status quo.