Modeling Age Patterns of Childlessness: A Semi-Parametric Approach

Monica Alexander, University of Toronto
Benjamin-Samuel Schluter , University of Toronto

Trends and patterns in childlessness are important to measure and understand its relationship with fertility rates and fertility intentions. However, data on childlessness is often limited to survey data, which suffers from small sample sizes for key subpopulations. We propose a model estimating age-specific proportions of individuals being childless at a subnational level. The model consists of an 'expected' component -- capturing the shape of childlessness over age using a logistic function -- and deviations from the expected level, modeled with P-splines. The model estimates summary parameters which are useful to understand trends over time. The model is estimated in a Bayesian framework allowing temporal smoothing and pooling of the parameters. We apply the model to estimate the proportions of childless women by race/ethnicity in the U.S. The preliminary results are promising and the model could be expanded to model multiple parities and perform projection of childlessness.

See extended abstract

 Presented in Session 16. Modelling Kinship and Fertility Processes