Projecting proportionate age–specific fertility rates via Bayesian skewed processes

Fertility rates show dynamically–varying shapes when modeled as a function of the age at delivery. We incorporate this behavior under a novel Bayesian approach for dynamic modeling of proportionate age–specific fertility rates via skewed processes. The model assumes a skew–normal distribution for the age at the moment of childbirth, while allowing the location and the skewness parameters to evolve in time via Gaussian processes priors. Posterior inference is performed via Monte Carlo methods, leveraging results on unified skew–normal distributions. The proposed approach is illustrated on Italian age–specific fertility rates from 1991 to 2014, providing forecasts until 2030.