The reconstruction of fertility trends with DHS birth histories. Application of Poisson regression to person-period data

Bruno D. Schoumaker, Université Catholique de Louvain

In this paper, we present a simple method to analyze and reconstruct fertility trends from birth histories. The method relies on the organisation of birth histories as person period data files, which can then be analyzed with Poisson regression. This approach proves to be very flexible in various ways. It notably makes it possible to include time-varying explanatory variables in regression models in a simple way, and to reconstruct past values of total fertility rates. We apply the method to the reconstruction of fertility trends in a variety of countries at different stages of fertility transition using DHS birth histories. We also extend the method to relax the assumption of constant fertility schedule, and to model fertility trends as spline functions of time.