Creating small-scale census zones

Census data provides crucial information about a country’s population, but once you ‘zoom in’ on a local area you hit a problem: how to keep data zones large enough to retain anonymity, and small enough to give an accurate representation. A team led by Professor David Martin at the University of Southampton solved this longstanding problem for the 2001 Census in England and Wales, by developing a geographical information system-based algorithm which divided the dataset into 175,000 small areas.

The researchers had to square the circle of retaining local anonymity while making the areas as small as possible to meet the needs of users. It had to fit exactly within existing official boundaries used by government, match as closely as possible the smallest postcode areas required by business users, and be as compact and socioeconomically homogenous as possible.

The research, originally carried out in collaboration with the Office for National Statistics (ONS) with subsequent ONS and ESRC funding, has led to an entirely new approach to the creation and management of small geographical areas for official statistics. For the first time it was possible to directly compare data between different censuses (the 2001 and 2011 UK Census).

As well as being used for other outputs such as the Neighbourhood Statistics Service and the Indices of Deprivation, the algorithm software for automated zone design has also been adopted by research groups across 10 countries.