Multiple regression equations are developed for U. S. graduate student migration by state, using 1958 state of origin and destination data, for estimating the number of out-migrants, the destination of out-migrants from one state, and for predicting in-migrants to states. Factors found to be associated with graduate student migration include the number of faculty in the state, institutional density, urbanization, college educated adult males in the population, number of fellowships and assistantships, and the Zipf hypothesis. This application of regression is evaluated in relation to H. ter Heide's review of migration models.