This project is building the first fully geocoded, longitudinally linked census microdata for England and Wales covering the period 1841–1921. We are developing novel computational and statistical methods to link individuals across multiple census enumerations, incorporating machine learning, handwritten text recognition, geocoding, and probabilistic record linkage.

The project combines expertise in historical demography, economic history, machine learning, and record linkage, delivered in partnership with The National Archives and major census data providers including FindMyPast, Ancestry, and FreeBMD. We also work closely with the Ministry of Justice to develop and enhance FSEM (Fellegi–Sunter Expectation Maximisation) methods for probabilistic linkage.

Our aims are to produce high-precision, low-bias census links; improve linkage outcomes for women by incorporating marriage and death registration records at scale; and geocode individuals to street and building level using historical gazetteers and digitised Ordnance Survey maps. This linked and geocoded dataset will enable unprecedented research on internal migration, urbanisation, women’s work, and demographic outcomes across two centuries of British history.

The resulting datasets, documentation, and technical outputs will be released via this site as the project progresses. All data and code will be made available as open access resources to support ongoing research and collaboration.