When does violence break out between refugees and natives in a host nation? Existing research on this topic is mostly focused on advanced industrial Western countries even though most refugees live in the non-Western developing world. We analyze this question with a new data set on native-refugee clashes between 2014 and 2018 in Turkey, which hosts the largest refugee population in the world. Our main finding is that clashes tend to spread and local authorities' response previous clashes is significantly related to the probability of clash diffusion. We present qualitative and quantitative evidence suggesting that potential rioters observe the outcomes of past clashes and adjust their behavior accordingly. When the state concedes to native rioters in one place by evicting refugees, new riots break out elsewhere aiming to provoke the state into sending more refugees away. Our findings imply that coordinating local responses to native-refugee violence may help lower the overall level of violence. More broadly, this research suggests that local authorities play an important role regarding violence between refugees and natives.