Norms are a crucial driver of behavior, and a growing body of literature tries to understand their effects. We argue that political norms should not affect all individuals the same, but that individuals for whom deviation is more costly should be more likely to follow them. We test this expectation using observational data, leveraging a unique setting in Germany. In this country, a randomly selected sample of ballots are marked in each election, and basic sociodemographic characteristics (age, gender) of the voter are collected. This means that we can compare how voters of specific gender and age vote compared against how they declare to have voted. Using mis-declarations of one’s vote as a measure of preference falsification to follow established norms, we test which groups of voters are more likely to follow established political norms that incentivize turnout and dis-incentivize supporting the far-right. Our findings will have crucial implications to our understanding of voter’s perceptions of the social costs of different political actions, and of who is more likely to be affected by norm erosion.