Small area estimation strategy for the 2011 Census in England and Wales

Abstract

The use of model-based small area estimation for adjusting census results in the UK was first introduced in the 2001 Census. The aim was to obtain local level population estimates by age-sex groups, adjusted for the level of undercount that combined results from the Census and the Census Coverage Survey. A similar approach was adopted for the 2011 Census but with new features and this paper describes the work carried out to arrive at the chosen small area strategy. Simulation studies are used to investigate three proposed small area estimation methods: a local fixed effects model (the 2001 Census approach), a direct estimator and a synthetic estimator. The results indicate that both the synthetic and the local fixed effect models constitute good options to produce accurate and reliable local authority population estimates. A proposal is made to implement a small area estimation procedure that accommodates both the synthetic and local fixed models, as in some selected areas with differing local authority under-coverage rates a local fixed effects model may perform best. We examine this strategy under real census conditions based on the final results from the 2011 census.