Multistate Transition Models in Demography

Multistate models deal with the transitions that individuals living in social or geographical groups experience during their lifetimes, as they move from one state of existence to another. Increment-decrement life tables constitute a first approach: they generalize the basic life table to more complex state spaces. These linear models applied to homogeneous groups use aggregate level data and are elaborated within the framework of Markov chains. They are linked to multistate projection models, which lead to stable or quasistable asymptotic populations. A generalization is, however, possible to nonlinear models. A second approach uses event history analysis in multistate models. These models using individual data, permit the handling interactions between different processes and the introduction of population heterogeneity. This heterogeneity may be time dependent and may differ according to origin or destination state. These models are linked to microsimulation for population projection. Aggregated over individuals the process is no more Markovian. A third approach includes both individual and aggregate measures in the same analysis. Contextual analysis is a simple extension of usual modeling methods, introducing aggregate characteristics. Multilevel analysis introduces also a within-group dependence, adding random effects to the previous model. These models consider small groups without structure or larger ones for which a structure cannot be introduced realistically. Finally, the generalization of these models to structured groups is discussed