Investigating address accuracy in a key administrative data source

Due to rising cost, respondent burden and decreasing response rates associated with a traditional population census and social surveys, many national statistical institutes (NSI), including those in the UK, US, Canada and Australia, are increasingly using administrative data to inform official statistics on population and society. For example, the NSI of the UK constituent countries will draw on administrative data to support various elements of the 2021 Census operation and subsequent statistical outputs. It is therefore imperative that the demographic and geographic information held in these sources is reliable, both at individual- and household-level. This research is based on record linkage between the 2011 Northern Ireland (NI) Census and the NI Health Card Registration System (HCRS); taking the former as a reference facilitates an assessment of the accuracy of address information in the HCRS, a key administrative data source providing a high degree of population coverage. The analysis is based on a comparison of the anonymised version of the Unique Property Reference Number (UPRN) assigned to the census and HCRS address of records linked across both data sources. Specific research objectives are to: (a) investigate if households can be accurately recreated in the HCRS given the value and importance of reliable household-level statistics; and (b) quantify the extent of and identify the factors associated with record-level address inaccuracy in the HCRS for the reference time point of April 2011. The research complements ongoing work by the UK’s NSI on the feasibility of developing an alternative approach to the traditional population census that is underpinned by administrative data; accurate census statistics have a key role in society with regard to policy making, funding allocations to local areas and planning service delivery.