Joanne Ellison , University of Southampton
Ann M. Berrington, University of Southampton
Jakub Bijak, University of Southampton
Erengul Dodd, University of Southampton
Fertility projections are a key determinant of population projections; they are also vital to anticipate demand for maternity and childcare services. While there is a large and diverse literature concerning proposed fertility forecasting models, these tend to use aggregate population-level data indexed by age, and period or cohort, alone. In this way they neglect to include information about birth order (parity), despite parity-specific data being collected by many countries and the evidence supporting greatly differing determinants of childbearing by parity. This omission risks ignoring a crucial mechanism of fertility dynamics, and producing biased predictions as a result. To this end, in this paper we develop fertility forecasting methods that incorporate parity information within a Bayesian framework. Preliminary work has focused on the use of Bayesian Generalised Additive Models (GAMs) to model parity-specific rate estimates for England and Wales from the Office for National Statistics (ONS). To reflect our prior knowledge that completed cohort fertility changes slowly over time, we have investigated the use of a random walk to constrain this aggregate measure, finding that it can lead to increased forecast precision and accuracy. We develop this model further by integrating rate estimates from alternative data sources including the ONS Longitudinal Study and the United Kingdom Household Longitudinal Study. Then, we compare our forecasts with the current best-performing models to quantify the impact of including the parity dimension on predictive accuracy. Our findings have the potential to lead to more reliable fertility projections, aiding government policymakers and planners in their decision-making.
Presented in Session 78. Improvements in forecasting methods