A Bayesian Coherent Framework for Forecasting Fertility by Age and Education

Felipe Sanchez , London School of Hygiene and Tropical Medicine
Arkadiusz Wisniowski, University of Manchester
José Manuel Aburto, London School of Hygiene & Tropical Medicine

Fertility forecasts play a central role in shaping public policy, influencing decisions across education, labor markets, and social protection. In Colombia, strong differences in fertility by educational attainment point to structural inequalities and shifting reproductive behavior yet most forecasting approaches overlook these dynamics. This paper investigates whether accounting for coherence across education groups improves the stability and relevance of fertility projections. It applies a Bayesian hierarchical modelling framework to evaluate the effect of coherence assumptions on forecast outcomes. The findings show that models which enforce consistency in trends across subpopulations produce more stable and interpretable projections, without masking important educational differences. Forecasts that incorporate educational gradients are also more aligned with recent demographic shifts and policy needs. These results highlight the risks of using overly simplistic or independent models in the presence of social stratification and demonstrate the added value of coherent modelling for long-term planning. By improving how inequality is represented in fertility forecasts, this research contributes to a more accurate and inclusive basis for decision making in population policy. The approach is broadly applicable to other settings where demographic behaviors vary systematically between socioeconomic groups.

See paper

 Presented in Session 2. Bayesian Demographic Modeling