The Plausibility of Small Area Estimates of Demographic Quantities from Census and Survey Data in Sub-Saharan Africa

Dennis Feehan, Princeton University

In many parts of the world, reliable direct estimates of demographic quantities like births and deaths are unavailable for small geographical areas, yet this information is essential to good policy and research. Model-based strategies for combining different data sources may help in this situation, provided the assumptions they rely upon hold. I investigate the plausibility of some of these methods using the example of district-level fertility in Uganda. I fit a Bayesian hierarchical model to birth history data from a national survey, and then use the distribution of women by age and district to estimate fertility rates at the district level. I evaluate the model's performance using district-level fertility rates computed from the census. I also explore and compare more complex specifications.