What is the effect of political violence on electoral support for anti-system parties? We argue that the effect of violence is asymmetric and predominantly benefits nationalist, radical-right parties. Posing as defenders of the nation, nationalist parties benefit from violence by their political opponents and violence targeted against perceived threats to the nation. To test this conjecture, we collect novel actor-based and geospatial data of political violence in Weimar Germany. Using a staggered difference-in-differences estimator, we model the effect of violence on vote shares of anti-system parties at the municipality-level in all nine parliamentary elections between 1919 and 1933. Our results indicate increasing electoral support for the extreme nationalist parties in municipalities that experienced violence committed by either the far right or left. In contrast, the radical left Communist party loses electoral support if violence occurs. We investigate heterogeneous effects with respect to the type of violence, across elections, and the socio-economic characteristics of municipalities. We conclude by discussing the relevance of our finding for increasing violence in liberal democracies today.