We provide the first large-scale evidence on self-selection of refugees and irregular migrants who arrived in Europe in 2015 or 2016. Our analysis uses unique datasets from the International Organization for Migration and Gallup World Polls. We find that refugees are positively self-selected with respect to human capital, as are female irregular migrants. Male irregular migrants are negatively self-selected. These patterns hold whether analyzing individually stated main reason to emigrate, country-level conflict intensity, or sub-regional conflict intensity. Several additional analyses show that our results are unlikely to be driven by omitted variable bias or liquidity constraints. We offer a theoretical framework to explain these patterns, by extending the Roy-Borjas model to include risks related to staying in an unsafe country of origin, risks related to migration, and gender specific returns to human capital.