Oliver mapped the ONS Output Area Classification (OAC), which is a free geodemographic 2 dataset. I don’t make much use of the OAC at work, but I do use Experian’s ‘MOSAIC’ product instead which is very similiar 3. Experian produce the data at postcode level and household level.
A quick overview of MOSAIC
MOSAIC is a dataset that classifies every household or postcode area into one of fifteen lifestyle groups, or one of sixty-nine lifestyle types.The types have names like ‘affluent singles in new build areas’ or ‘older people reliant on friends and family’ and a statistical profile of characteristics behind those names. The profiles allow you to make some broad assumptions about the nature, characteristics and interests of those people. It can be a useful research tool. For example by using the MOSAIC data and comparing it to Child Poverty, we’ve identified that some groups are more likely 4 to have children in poverty than others. This means we can then focus resource on these groups first, rather than blanketing everyone in Blackpool 5.
A challenge to present
When trying to provide a summary of MOSAIC for Blackpool at group level though, I have always found it challenging to present the information in a meaningful way to complete geophobes. At household level it becomes a blur of colour 6, while at postcode level it is difficult to relate the postcode districts back to individual areas and the postcode sectors can make some groups appear more important than others.
I have tried various approaches, from dot plots to thematically shaded postcode polygons or larger areas colour coded to the most prevalent MOSAIC type. The most useful to date ended up being a very blocky postcode sector image a snippet of which is reproduced below – each colour corresponds to a MOSAIC group. Messy, impractical to identify and a bit ugly (but functional) 7
When I saw Oliver’s approach I thought it would be an interesting experiment to learn how to map data in that way, and to use the same approach to mapping MOSAIC household data particularly.
Learning and Doing, Breaking and fixing
The process is actually more straightforward than I realised, if you can get to grips with software like Quantum GIS, an open source geographic information system. Essentially for me it went like this (skip this if you have no interest in doing GIS work):
- I created a new map layer using MOSAIC household data which was basically a single point for every household in Blackpool.
- I added in map layers from the Ordnance Survery OpenData – specifically the ‘Vectormap District’ data which is a broadly detailed dataset containing building outlines, streets, landscapes, water features.
- I tried to spatial join the building outlines to the MOSAIC household data (essentially ‘linking’ data in one dataset to the other). This means I tried to give every building a MOSAIC attribute which I could use to colour code the map. This didn’t work well! The problem being that the first ‘point’ encountered is used to join on – so a street with 9 Group A households and 1 Group B would sometimes be categorised as Group B. Terribly misleading.
- Plan B then! – I added building references to every point – I then exported the data, and with a bit of industrial light and magic, identified the dominant mosaic type for each building. In the example above this would result in Group A being selected.
- I then merged this new ‘dominant’ data set to the building outlines using Quantum’s join tool.
- The resulting map then is a map of the dominant MOSAIC group for each block of buildings.
Results below. Image on the left is the basic point map. Image on the right is the new Booth style map after joining data together. Which works for you? 8.
I think I prefer this to the raw point data and to the original postcode sector map. I feel I can connect better with the places and streets. If I have a criticism it’s that some detail is lost in the process. In the example above a lot of the ‘green’ households are lost – you could even suspect there was none. Again feel free to ask if you’d prefer to see full maps.
There is More to maps than colour
There’s a wider issue there. The point data is often falsely assumed to be really accurate – people often point out that their household MOSAIC type is wrong. The reasons for this are a) it’s not directly tied to actual people, but to idealistic profiles so it’s a mistake to assume everything in the profiles applies to every person in the group. b) it’s often a ‘best fit’ of data – which MOSAIC group is the closest match based on data available? c)in the worst case it may just be an infilled point where no data was available and so it’s been classed as the same as the neighbouring household. d) previous residents have written to Experian to demand they are given a more affluent group so they can get that new sofa on credit 9.
For me then, I prefer the blurring of the detail to create an impression of an area rather letting people reach false conclusions about individual households. Does aggregating point data make MOSAIC data appear less accurate than it is? Interestingly the debate went slightly the other way round on the spatial analysis blog – does applying OAC data to buildings make OAC appear more accurate than it is?
I think that one is a challenge for educators and presenters – asking people to think about the data behind the picture helps.
- See the wikipedia article on Charles Booth ↩
- it classifies people by where they live and their social characteristics essentially ↩
- but also very expensive ↩
- relatively ↩
- In theory at least though it isn’t always that straightforward ↩
- Pointillism at it’s best ↩
- Due to my own ignorance about copyright, I haven’t made full maps available but if anyone is interested feel free to ask. ↩
- I’m not an expert in Quantum GIS – it would have been nice to get the street names to scale better and similiar with road lines ↩
- Completely innacurrate and untrue. It was for the TV. ↩