Abstract

Across the social sciences, mobility between social positions or physical locations is often modelled as a sequence of independent choices. In practice, however, mobility decisions are frequently interdependent: individuals respond to others’ movements, giving rise to migration chains, clustering, and other collective patterns. This insight has motivated the representation of mobility systems as networks, with locations as nodes connected by mobile individuals. Building on this perspective, a recent statistical framework combines network models and multinomial choice models to capture dependence between individuals’ mobility paths. In the first part of the talk, I show how this framework can be reformulated as a multinomial logit with dependent observations. This reformulation is useful for practitioners because it provides a transparent link between micro-level assumptions about how individuals’ mobility decisions depend on one another and the resulting model specification. By making these assumptions explicit, the approach clarifies the interpretation of model parameters and strengthens the connection between statistical modelling choices and substantive theory. In the second part, I discuss the empirical value of this modelling approach. I examine whether endogenous processes can account for structural patterns observed in real-world mobility systems and to what extent analyses that assume independent mobility decisions may lead to misleading results regarding potential drivers of mobility.

Reference

Marion Hoffman, “Modelling endogenous processes in mobility”, IAST Lunch Seminar, Toulouse: IAST, March 13, 2026, 12:45–13:45, room Auditorium 4 (First floor - TSE Building).