Optimising record linkage techniques for local geographies and local population estimations

In response to growing evidence that official population statistics based on the decennial census are inaccurate and not fit-for-purpose at the Local Authority level - the fundamental administrative unit of the UK, the author has established an innovative method of using locally available administrative data sets for estimating populations. The method combines different data sources with different population coverage according to a defined set of rules. The result is timelier and geographically flexible data which is more cost-effective to undertake than a survey-based census. The ONS in the UK are now committed to exploring the use of administrative data as a replacement for the traditional survey-based census, and their findings are to be given in 2014. Key to this is the linkage of disparate datasets that are captured for other purposes to determine when the same entity (person) is present. This paper presents the author’s experience of linking administrative datasets and investigates how geography is not only the common link, but also an inherent bias that affects matching rates in terms of neighbourhood effects such as population turnover and ethnic diversity. An optimised algorithm for linking local administrative datasets is put forward, based on the results of sensitivity analyses.