In this thesis, new methodology for the estimation of small-area indicators of population health has been developed, making use of Bayesian random-effects specifications that can pool strength and make use of correlations between the various dimensions of the required mortality and morbidity data; for example, between sexes, age groups, and contiguous areas.

This methodology constitutes a new class of small-area estimations that forms a welcome addition to the existing small-area estimation literature due to its excellent accommodation of sparse data problems while being solely based on the observed small-area data, i.e. without reliance on area-level covariates or assumptions about the transferability of correlations between aggregate-level data to the small-area level.