AbstractThis review presents a probabilistic formulation of the decision making process, leading to a rigorous treatment of migration behaviour for projection purposes. These models are developed here at two different levels of aggregation. Aggregate‐level models may be applied to migration flows for which the objectively measured characteristics of areas of department and destination act as subjectively measured characteristics and stimuli. They generate quite different population projections than the usual Markov model: sustained cycles may appear or chaotic behaviour may even occur. Individual‐level models use event history methods of analysis to introduce a great variety of characteristics of the subject on the decision to move. They lead to projections using microsimulation models. A further step is taken in integrating macro‐ and microbehavioural models. The use of aggregate and individual characteristics simultaneously leads to more efficient and sophisticated projection models: the factors affecting behaviour at the micro‐level cannot be inferred from aggregate studies and con