Three-quarters of the global migrant population have fled fragile contexts, with 64 percent hosted by similarly vulnerable countries. These contexts account for about 75 percent of those living in extreme poverty. Despite the scale of this phenomenon, the extent to which the self-selection of South–South migrants differs from those migrating to more developed countries remains poorly understood. In this paper, I investigate the self-selection of Venezuelan migrants during the 2015–2021 crisis, which led to greater migration to less developed countries such as Colombia and Peru, compared to more developed ones like the United States and Chile. Using individual-level data representative of the Venezuelan population and similar data on Venezuelan migrants in these key destination countries, the study finds that migrants are generally positively selected in terms of education compared to those who remain in Venezuela, with migrants to developed countries being positively sorted with respect to those to developing countries. However, comparing the cumulative distribution functions of pre-migration predicted earnings reveals that migrants to developing countries are negatively selected relative to stayers, while migrants to developed countries are positively selected. This highlights the significant role of unobserved abilities in shaping South–South migration patterns. Furthermore, a discrete choice model shows that women and college graduates face lower migration costs, independent of expected earnings at the destination. Factors such as the distance to the destination country and pre-crisis networks also play a crucial role in shaping migration decisions.