The main cross-sectional and time-series properties of state-level gross and net worker flows, wages, and rents are characterized using micro-data from the U.S. Census. A dynamic general equilibrium model of worker migration is introduced to explain the stylized facts. In the model, a location may experience simultaneous inflows and outflows of workers. Recent migrants choose to migrate more often than incumbent workers. Thus, locations that attract high numbers of migrants also tend to experience high outflow rates. This pattern is a robust feature of the data and cannot be explained by models of net flows only. (This abstract was borrowed from another version of this item.)