Flexible Modeling of Demographic Transition Processes with a Bayesian Hierarchical Penalized B-splines Model

Several demographic and health indicators, including the total fertility rate (TFR) and modern contraceptive use rate (mCPR), exhibit similar patterns in their evolution over time, characterized by a transition between stable states. Existing statistical methods for estimation or projection are based on using parametric functions to model the transition. We introduce a more flexible model for transition processes based on B-splines called the B-spline Transition Model. We customize the model to estimate mCPR in 174 countries from 1970-2030 using data from the United Nations Population Division and validate performance with a set of out-of-sample model checks. While estimates and projections are comparable between the two approaches, we find the the B-spline approach improves out-of-sample predictions. Illustrative results for a model for TFR are also presented and show larger differences in projections when relaxing parametric assumptions.