This paper uses a historical quasi-natural experiment to provide causal evidence on how networks and diversity affect migrants' economic outcomes in a rural setting. Between 1932 and 1941, Italy's Fascist government relocated about 4,000 families to the Pontine Marshes, a formerly malarial, sparsely inhabited area undergoing massive reclamation. Settlers from different provinces were quasi-randomly assigned to plots, providing plausibly exogenous variation in neighborhood composition. Using data on the universe of settlers, we find that larger common-origin networks significantly increase the likelihood of long-term settlement and land purchase. Greater neighborhood diversity worsens economic performance on average, largely by reducing network externalities, but benefits settlers who lack common-origin ties. Network effects operate, at least partly, through higher productivity, with no evidence of knowledge transmission. Instead, our results point to cooperation and mutual support as the main channel. Effects are concentrated among families under the greatest strain and those from origins historically more reliant on bonding social capital. Network effects are also stronger when neighbors arrive at the same time and live nearby.