We examine crop choice as a dynamic optimization problem over an infinite time horizon, taking into account the effects over time that corn-soybean rotations have on soil quality, which manifest in yield and therefore profit impacts. We show how the efficient decision rule depends on model parameters and how it compares to those characteristic of static models of supply. The model is parameterized for a representative acre of Iowa cropland and used to predict actual crop choices in a panel of over 6500 Iowa plots during 1979–1997 surprisingly well.