Arkadiusz
Wiśniowski

Even though mortality differentials by socio-economic status and educational attainment level have been widely examined, this research is often limited to developed countries and recent years. This is primarily due to the absence of consistently good-quality inherent data. Systematic studies with a broad geographical and temporal spectrum that engage with the link between educational attainment and mortality are lacking. In this paper, we propose a mortality rates reconstruction model based on multiple patchy data sources, and provide mortality rates by level of education. The proposed model is a hierarchical Bayesian model that combines the strengths of multiple sources in order to disaggregate mortality rates by time periods, age groups, sex and educational attainment. We apply the model in a case study that includes 13 countries across South-East Europe, Western Asia and North Africa, and calculate education-specific mortality rates for five-year age groups starting at age 15 for the 1980-2015 time period. Furthermore, we evaluate the model’s performance relying on standard convergence indicators and trace plots, and validate our estimates via posterior predictive checks. This study contributes to the literature by proposing a novel methodology to enhance the research on the relationship between education and adult mortality. It addresses the lack of educationspecific mortality differentials by providing a flexible method for their estimation.

The future world population growth and size will be largely determined by the pace of fertility decline in sub-Saharan Africa. Correct estimates of education-specific fertility rates are crucial for projecting the future population. Yet, consistent crosscountry, comparable estimates of education-specific fertility for sub-Saharan African countries are still lacking. We propose a flexible Bayesian hierarchical model that reconstructs education-specific fertility rates by combining the patchy Demographic and Health Surveys (DHS) data and the United Nations’ (UN) reliable estimates of total fertility rates (TFR). Our model produces estimates that match the UN TFR to different extents (in other words, estimates of varying levels of consistency with the UN). We present three model specifications: Consistent but not identical with the UN; fully-consistent (nearly identical) with the UN, and consistent with the DHS. Further, we provide a full time series of education-specific TFR estimates covering five-year periods between 1980 and 2014 for 36 sub-Saharan African countries. The results show that the DHS-consistent estimates are usually higher than the UN-fully-consistent ones. The differences between the three model estimates vary substantially in size across countries, yielding 1980–2014 fertility trends that diverge from each other—mostly in level only, but also sometimes in direction.