Abstract

This paper investigates how networks formed during World War II Navy service shaped U.S. internal migration patterns. Using a novel dataset of over 1.3 million Navy sailors and exploiting quasi-random assignment to ships, I identify the causal impact of these newly formed geographic networks on post-war migration. Results show that a one standard deviation increase in out-of-state and out-of-region ties raises out-migration by 4-5% by 1950. Ship networks prove even more influential in directed migration, increasing the likelihood of moving to specific Census divisions by 2-15% and to specific states by up to 30%. Employing a gravity model with embedded networks, I demonstrate that Navy networks served as substitutes for existing migration networks, with long-distance ties approximately three times as valuable as ties to neighboring states. Moreover, by utilizing these random network shocks, I estimate the lifetime income returns to this network-facilitated migration to be 95%.