Forecasting the Life Course

Frans Willekens, University of Groningen

The basic question addressed in this paper is: how to use survey data on life histories more effectively in population forecasting? The traditional cohort-component model focuses on numbers of survivors by cohort and on the intercohort variation in demographic behavior. The proposed method focuses on cohort biographies, a concept introduced by Ryder (1965). Using multistate demography and prospective or retrospective data on life histories of cohort members, it projects cohort biographies into the future. The result is a distribution of cohort survivors over the various stages of life. The method then adds intracohort variation (heterogeneity) using the theory of multistate models with covariates (multivariate survival analysis). The result is a projection of individual biographies, where an 'individual' is carrier of attributes and is defined by a set of covariates. The proposed method encompasses several existing demographic projection models, including LIPRO, PROFAMY and MUDEA, and the compartment models applied by Manton.