Skip to menu | Skip to content |

giCentre - Department of Information Science

Exploring the geodemographics of Britain

Latest news: this work has been accepted at InfoVis 2011 and will appear in Transactions in Visualization and Computer Graphics 17 (6) in October 2011. A video demonstrating these techniques can be found here.

We use a hierarchical rectangular cartogram to explore the demographics of Britain, showing how typical each of the >200,000 output areas is to the seven super-groups in the Output Area Classification (OAC). OAC is a geodemographic classification scheme used widely in local and national government for population profiling. We use freely available data about OAC that are difficult to collate and present in a digestible form.

Each of the small coloured squares represents an output area and these are organised into their postcode hierarchy. Each is sized by its population and is arranged geographically. Highly populated postcode areas (e.g. N, W and SW) are towards the bottom right (SE). Scottish postcodes are smaller because of their comparably low population density and are at the top (N). (See an example of what happens if you don't use a population cartogram).

The Output Area Classification (OAC) classifies Output Areas according to the socioeconomic characteristics of the population that live there. OAC defines seven super-groups: Prospering Suburbs, Blue Collar, Typical Traits, Countryside, Constrained by Circumstances, City Living and Multicultural. These are convenient labels to attach to areas that indicate characteristics of population and is a fascinating insight into the British population. This approach to population profiling is known as geodemographics.

Populations classified in this way (also using alternative classifiers and different numbers of classes) are often used in statistics and reporting in both the public and private sectors. However since populations contain a huge amount of variation, allocating to one of the seven super-groups represents a huge amount of simplification. Although this is obvious, the details of this are not widely known or appreciated. Variation within reported classes are rarely shown. These in themselves are fascinating, but are also important for those using and interpreting geodemographics.

Showing this variation in a digestable form in challenging. Our poster attempts to do this with hue to show the most similar geodemographic class and lightness to show how typical the local population is to that class, where dark is more typical. We do this for each of the seven classes as miniature maps. Note that some classes are either very typical of local population or not at all ('Multicultural' and 'Countryside'). 'Typical Traits' is at the opposite extreme, and shows little variation in population typicality. See our paper for how we chose the colours. You can also see cartograms for each postcode area individually.

But exactly how do these populations vary?

OAC Explorer helps us find this out. It is an interactive interface that provide instant access to the original census variables that were used to construct OAC and the values in the typical cases for each group as parallel plots. This allows spatial variation to be explored, e.g. where the classification provides a good summary of the local population and where it does not; important knowledge for geodemographic practitioners. Use of OAC Explorer highlights, for example:

OAC Explorer was presented at the GISRUK conference in 2010, where it won Best Paper. There is a short paper in the proceedings. It also demonstrates how that interactive visualisation can be used to help understand more about the characteristics of classification schemes.

OAC Explorer is interactive. Below are some static screenshots and some commentary.

This is a rectangular hierarchical cartogram of Britain (England, Scotland and Wales), where areas are sized by their population and are geographically arranged. The letters refer to UK postcode areas. The colours refer to the OAC group to which the local area is classified.

OAC is hierarchical, being classified into 7 super-groups, 21 groups and 52 sub-groups. These can be seen on the right, again sized by population. Britain's large areas of low population density makes conventional maps at a national scale problematic.

The core-periphery structure can be seen in many of the cities. For example, much of the core of Birmingham (B) is classified as "Multicultural" with other OAC groups on the periphery. Note that larger areas refer to postcode area, not city boundaries.

Here, colour lightness is used to indicate how good a summary of the local population that assigned OAC category is. Where the colour is strongest, the local population has very similar characteristics to the typical case for that OAC category. Where the colour has faded to white, the local population shares the characteristics of multiple OAC categories. Geodemographic practitioners often use the assigned demodemographic category without appreciating that this might not reflect the characteristics of the local area very well.

Interactively zooming an panning reveals detail allowing individual output areas to be resolved and identified through their postcode hierarchy.

Here, the barchart (top right) shows the similarity to each OAC super-group. The mouse cursor (not visible in these slides) is over a pale area classified as 'Multicultural' (pink). The graph shows that it is not similar to any of the categories and that it is almost as similar to 'Proposing Suburbs' (red). Instant access is offered to this information for any autput area.

As above, but the mouse cursor is over a dark coloured 'Multicultural' output area.

Here, we are colouring by similarity to 'Multicultural'.

Here, we are colouring by similarity to 'Countryside'. This unsurprisingly broadly gives the inverse of the above.

All output areas should a strong similarity to 'Typical Traits'. This group does not appear to be very discriminatory.

Here, we have selected some postcode areas from around the country.

Here, we have excluded the other postcode areas to allow their comparison.

Here's a parallel plot of the 41 census variables. The thick pink and green lines represent the typical case for 'Multicultural' and 'Countryside', respectively. The thin red line corresponds to the population of the area over which the mouse is positioned (not shown in this screenshot) showin similarities with 'Multicultural'.

The axes in the parallel plot can be sorted on the values of the variables for the area over which the mouse pointer is positioned. This help in the identification of key or outlying variables.

Here, the baseline for the axes has been set to 'Multicultural' and all values should are deviations from these. Variables near the centrelines are similar to those of 'Multicultural'; those at the extremities are very different from the values for 'Multicultural'.