This paper develops a dynamic spatial equilibrium model of a multi-region, multi-sector
open economy where heterogeneous agents optimally choose their job, making forwardlooking decisions under aggregate uncertainty. We propose a solution of the system of individual dynamic optimal-control problems under rational expectations as a Mean Field Game, in discrete time and state space, preserving the full nonlinear structure of the problem. Using a calibration for France, we demonstrate that welfare outcomes and the spatial distribution of labor differ systematically from those under perfect foresight. Accounting for decisions made under aggregate uncertainty significantly reshapes the patterns of labor reallocation, both during transitions and in the long run.