We study how news coverage of immigrant criminality impacts voting in one of the most controversial referendums in recent years – the 2009 Swiss minaret ban. We combine a comprehensive crime detection dataset with detailed information on newspaper coverage. We first document a large upward distortion in media reporting of immigrant crime during the prereferendum period. Exploiting quasi-random variations in crime incidence, we find a positive first-order effect of news coverage on support for the ban. Quantification shows that, in absence of the media bias, pro-ban vote would have decreased from 57.6% to 53.5% at the national level.