Abstract:
|
It is a well-known fact that even an exceptionally well designed and implemented census of a large and complex population cannot provide an enumeration with ignorable coverage errors. To assess the extent of coverage errors we conduct a large coverage survey a few weeks after the census night. The survey data are linked to the census data to enable the estimation process. Census coverage estimation then provides the coverage error corrected population size (along with the corresponding uncertainty measures) at national and local authority levels. In this presentation we discuss the research done so far for the coverage estimation in the 2021 Census of England & Wales. We will talk about assessing the performance of methods such as a combination of dual system, ratio and synthetic estimators, logistic regression, mixed effects logistic regression and weighted logistic regression-based approaches for the coverage estimation. In addition, bias adjustment procedures will be also discussed. The quality of the the census coverage corrected population estimates are such that for most domains the population size estimates are substantially more accurate than the raw census data.
|