Despite the popular narrative that the United States is a "land of mobility," its internal migration rates have declined for decades, and reached a historical low. Economic and related factors were able to account for a portion of this trend, but the bulk has remained unexplained. Here, we propose a systemic, relational model of internal migration in the U.S., combining demographic, economic, political, and geographical factors with endogenous social mechanisms, with the objective of identifying factors limiting migration rates. We implement this model using valued temporal exponential-family random graph models, allowing us to calibrate it to the (valued) network of intercounty U.S. migration flows during the 2011-2015 period. Our analysis reveals a pattern of segmented immobility, where fewer people migrate between counties with dissimilar political contexts, levels of urbanization, and racial compositions. Probing our model using "knockout" experiments suggests that one would have observed approximately 3 million (17%) more intercounty migrants over the study period were the segmented immobility mechanisms inoperative. This analysis suggests that internal migration in the current era is driven not only by demographic and economic factors, but also cultural and political ones. It also reveals social and political cleavages that underlie geographical immobility in America.