Immigrants' social networks exert considerable influence over their labor market opportunities and yet the pre-sorting of co-nationals by ability and across space, endures as a key challenge for empiricists attempting to establish causal network effects. To surmount this issue, we leverage the chaotic dispersal of Vietnamese refugees across the U.S. in 1975, which was demonstrably exogenous in both initial network size and quality, in tandem with an absence of pre-existing networks of co-nationals, to causally identify the effects of network size and network quality on refugees': occupational outcomes, skill intensity and skill upgrading. Our administrative data provide refugee's precise initial locations and pre-placement characteristics in Vietnam, which we uniquely employ as additional controls, as well as longitudinal information about their locations and occupations six years hence. We construct instruments from the initial quasi-random refugee allocations of network size and quality and leverage refugees' geo-locations to insulate our results from the Reflection Problem. Overall, network quality is a far more important determinant of refugees' labor market outcomes when compared to network size, one interpretation of which is that the type of referrals network members receive are more important than the overall number of referrals. Blue-collar networks: increase the probability of refugees' working in blue-collar jobs, draw additional workers into more manual and less complex intensive employment and serve to up-skill individuals along the manual skill dimension. Given the protracted circumstances under which the Viet Kieu entered the U.S., the composition of their networks played a pivotal role in their ultimate success.