Return is the main way refugee crises end, yet there have been few cross-national studies of when it actually occurs. A central expectation of the refugee regime is that refugees remain abroad while home is unsafe and return as security improves. Yet using a novel global dataset, I show that most returns occur during ongoing conflict, while many refugees remain abroad long after wars end. To explain these patterns, I argue that asylum countries and international organizations play major and under recognized roles in shaping return, largely overlooked by existing micro-level research focused on refugees’ concerns about security at home. Return is substantially higher when refugees face exclusion policies or conflict in asylum countries, and when donor-funded UNHCR assistance lowers the costs of return. I introduce the Global Refugee Return Dataset, covering 293 origin–asylum dyads from 1980 to 2023, which documents returns systematically undercounted by UNHCR and provides new global data on de facto refugee conditions and UNHCR financing.