Regional fertility data analysis: A small area Bayesian approach

Accurate estimation of demographic variables such as mortality, fertility and migrations, by age groups and regions, is important for analyses and policy. However, traditional estimates based on within cohort counts are often inaccurate, particularly when the sub-populations considered are small. We use small area Bayesian statistics to develop a model for age-specific fertility rates. In turn, such small area estimation requires accurate descriptions of spatial and cross-section dependence. The proposed methodology uses spatial clustering methods to estimate an adjacency matrix that captures such dependence more adequately. The model is then used to estimate agespecific fertility rates and total fertility rates at the regional NUTS III area level for Portugal. The paper makes important contributions to small area Bayesian statistics in a spatial domain focusing on estimation of fertility rates. The estimates obtained are more accurate and adequately represent uncertainty in the estimates, and are therefore very useful for demographic policy in Portugal.