Forecasters could easily be overwhelmed by the plethora of methods for predicting cohort and period fertility. To shed light on which method to choose when predicting fertility we validate their forecast accuracy based on data of the Human Fertility Database and the UN World Population Prospects. We found for cohort fertility completion that forecast accuracy does not necessarily improve with method complexity. That is, the baseline Freeze rates (that holds latest fertility rates constant) belongs to the top methods in all world regions, and it is only outperformed by two extrapolation methods and two Bayesian approaches. As a follow-up we investigate if this finding for cohort fertility completion also holds true for period fertility forecasting, adopting the same study design. We introduce the forecast performance spectrum as a visualization tool that supports forecasters in their decision for the most appropriate method(s) in various fertility settings by joining all evaluation results.
Presented in Session 3. Population, Development, & the Environment; Data & Methods; Applied Demography