Geolinking UK Census 1841–1921

Putting eighty years of census returns on the map, and joining them up

The census returns for England and Wales between 1841 and 1921 record where tens of millions of people lived, worked and moved during the period in which Britain became an urban industrial society. The returns survive, and much of the text has been transcribed. What has never existed is a version in which each household is placed accurately on the map and the same person can be followed from one decade to the next.

That is what this project builds. Geocoding turns a street address written by an enumerator in 1871 into coordinates; linkage joins the same individual across successive censuses; together they turn ten static snapshots into eighty years of trajectories — of migration, occupational change, household formation and urban growth.

Why it is difficult

Both halves are hard for the same underlying reason: the identifying information is exactly the information that is least reliably recorded. Street names change, are misspelled, or refer to places that no longer exist. Names are written as the enumerator heard them. Ages are rounded, guessed or misreported. A linkage method that assumes clean identifiers will link confidently and wrongly, and the resulting errors are systematically related to literacy, mobility and class — the very things the data are used to study.

The project therefore treats geocoding and linkage as a joint statistical problem with quantified uncertainty, rather than as a data-cleaning exercise with a single "correct" output.

BDAD's contribution

The unit brings the transcription and linkage methodology developed on Danish and Swedish sources — handwritten text recognition, occupational standardisation through OccCANINE, probabilistic linkage without a personal identifier, and the inference theory needed to use machine-generated variables in downstream analysis without biasing the results.

The project is also the clearest case of the unit's methods travelling beyond Denmark. Nothing about the approach is language- or country-specific, and the UK census is a demanding test at genuine scale.

What it produces

A geocoded, linked version of the English and Welsh censuses 1841–1921, deposited with The National Archives so that it remains available beyond the life of the grant — the same principle that governs all of the unit's infrastructure work.