We estimate the causal e ect of a large-scale cash assistance program for refugee re- turnees on con ict in Afghanistan. The program led to a signi cant increase in repa- triation. Leveraging historical returnee settlement patterns and previously unreleased combat records, we nd that policy-induced refugee return had cross-cutting e ects, reducing insurgent violence, but increasing social con ict. The program’s cash bene ts were substantial and may have raised reservation wages in communities where returnees repatriated. Consistent with this hypothesis, policy-induced return had heterogeneous e ects on insurgent violence, decreasing use of labor-intensive combat, increasing the lethality of capital-intensive insurgent attacks, and reducing the e ectiveness of coun- terinsurgent bomb neutralization missions. Additionally, social capital and the quality of local institutions signi cantly o set the risks of refugee return for communal vio- lence. Our study provides the rst causal evidence demonstrating the link between aid-induced refugee return and political and social con ict. These results are economi- cally signi cant, highlighting unintended consequences of repatriation aid and clarifying the conditions under which refugee return a ects con ict. Supporting social capital and legitimate, local institutions are key antecedents for safe refugee repatriation.