Geodemographics in Five Slides
Geodemographics describes and classifies the places where people live, and the people who live in places.
The word was coined by Jonathan Robbin for an article called 'Geodemographics: The New Magic', which he wrote in 1980 for a political campaigning magazine.
He combined marketing theory (demographic segmentation) with the computational techniques he'd been learning as a PhD candidate researching geographic sociology.
But his supervisor left, so he dropped out and started his own data consultancy.
Geodemographics is arguably the first and also most accessible example of unsupervised machine learning, and therefore of AI in Geography.
It sounds like it's a subfield of Demography, but it's not really (at least yet) -- still struggling with pattern, not yet able to account for process.
Based on the Chicago School's theory of urban natural areas: social and economic forces tend to bring about an orderly and typical grouping of population and institutions.
In practice however, the default is generally arbitrary census output units.
Charles Booth's urban poverty maps of London are often named as the earliest example of geodemographics.
He mapped social conditions by the street-block. It took him seventeen years with a big team of researchers, and more than a million pounds (in modern terms) of his own money.
"Can we develop street-block geodemographics with open data?"
- Geometrically, can we define the tesselating polygons?
- Statistically, can we transform data to higher resolution without too much noise and uncertainty?
- Ontologically, does it make sense to define a neighbourhood just in terms of a street-block?
- Ontologically, would it be better to do a multi-scale analysis that treats street-blocks connected by minor streets as larger natural units? (cf. Grannis (2009), From The Ground Up).
- Ethically, what are the implications of higher resolution geodemographics?
- Ethically, what are the implications of an analytical methodology with an epistemic bias for homogeneous units?