The hypothesis that political violence deters tourism is mainly based on case study evidence and a few quantitative studies confined to a small sample of countries. This is the first comprehensive, general quantitative test of the impact of various forms of political violence on tourist arrivals. We employ two estimation techniques: a fixed-effects panel estimator with contemporaneous effects only and a dynamic generalized method of moments estimator, which allows for lagged effects of political violence on tourism. In both model specifications, we find strong evidence that human rights violations, conflict and other politically motivated violent events negatively impact upon tourist arrivals. In a dynamic model, autocratic regimes, even if they do not resort to violence, have lower numbers of tourist arrivals than more democratic regimes. We also find evidence for intra-regional negative spill-over and cross-regional substitution effects.