Data collected during the administration of government procedures provides great potential to generate statistical information at low marginal cost. However, using this type of data presents challenges because the properties of the data are not tailored to statistical production, and statistics often require the integration of several data sources. Research at Southampton has examined the quality of single and linked administrative data sources, including projects on error in non-survey sources and coherence, coverage and linkage of administrative data sources.
Capture recapture methods (also called dual system estimation, dse) are effective for estimating population sizes from administrative and census data sources. Recent research includes the development of trimmed dse for the challenging problem of assessing administrative data sources prone to overcoverage, using linked administrative datasets with missing data and measurement error to provide population estimates for important population subgroups, and producing population estimates from administrative data in the absence of a person register. These approaches provide estimates of the quality of the administrative data sources, useful for understanding biases.
Southampton researchers are strongly involved in the development of population census methodology, developing a body of work that goes back to the One Number Census project in the UK in the late 1990s. Recent research includes estimation and imputation approaches in the England & Wales 2021 population census, estimation of the size of the unregistered population in the Netherlands in support of their register-based census, and work on options for an administrative data based census in Ireland.