Empirically identifying the causal effect of social capital on immigrants’ economic prospects is a challenging task due to the non-random residential sorting of immigrants into locations with greater opportunities for prior or co-ethnic connections. Our study addresses this selection-bias issue by using a natural-experimental dataset of refugees and other immigrants who were exogenously allocated to their first place of residence by German authorities. This unique opportunity allows us to make an important methodological contribution to the predominantly observational knowledge about immigration and co-ethnic social capital. Although a growing body of migration studies in economics and sociology stresses the importance of social networks for migrants’ labor market integration, our results show little evidence of a causal effect of social networks themselves. Being part of a larger co-ethnic community per se does not accelerate immigrants’ labor market success except for the migrants who use the resources embedded in their social contacts when looking for a job. We conclude that further methodological advancements can be achieved by embracing recent technological developments and by combining different methods to increase both internal and external validity of findings in migration studies.