Landmark paper: ‘Social Grading and the Census’
03 November 2014

I’ve already selected one Landmark paper that focused on the importance of Census data to market research, that by Baker et al (‘The utility to market research of the classification of residential neighbourhoods’) originally presented at the 1979 MRS Conference and reprinted in JMR Vol. 39 No.1 which described the then recent development of geodemographics.

The latest quarterly Landmark paper also focuses on the importance of the Census to market research, but is rather more recent than previous ones I’ve selected, published just ten years ago. But therein lies its relevance today for four reasons.

Firstly, using 2001 Census data, it is based on ground breaking work undertaken for the first time and as with the Baker et al paper it underlines the importance of the census to the market research sector (not just in the UK, but throughout the world). 

Secondly, we now know that the 2021 UK Census will continue to be based on primary data collected through the traditional household level method. However, the data will be primarily collected through the internet, supplemented by other methods to reach non-connected households, and other sources will be used to augment the data.

Thirdly, the MRS Census and Geodemographics Group (CGG) celebrates its 25th anniversary this year, with their conference 'Harnessing Open Data for Business Advantage', being held on November 5th, chaired by myself; focusing on the growing benefits of open-access data.

Fourthly, MRS has just announced that this month it has joined the Open Data Institute, providing access to training, networking events and ODI resources.The 2001 Census marked an important milestone as for the first time it contained a Social Grading classification, following the traditional format of ABC1 etc. 

In addition, the Government’s new socio-economic classification system (NS-SEC), specifically developed for that Census. As the paper describes, both methods are based on occupation, but the groupings and other contextual information were different.

At the heart of the paper is a description of the algorithm built by a CGG working party that enabled the ‘approximated’ Social Grades derived from the Census data to be turned into a form that matches as closely as possible the format normally applied in market research. Deriving a suitable algorithm required a substantial amount of work by the group, the working party for developing this having been formed as early as 1994 with the earlier phases being described in a paper presented at the 1998 MRS Conference.

The algorithm includes data from the National Readership Survey (NRS) as this is the ‘gold standard’ for social grading in the UK, and enabled the Census variables to be simulated. All the tests are described, plus the differing algorithms built to cover full-time and part-time/non workers, and how the challenges where there was uncoded data were addressed. 

When the final version was sent to ONS, this threw up further challenges due to issues with the employment data collected in the Census on those aged over 65, resulting in the recommendation that the derived grades should be confined to 16-64 age groups. 

The paper concludes with a discussion of how JICPOPS, the Joint Industry Committee that provides harmonised population statistics to the geodemographics sector and media surveys, might use the new stream of Census based social grading data, including the development of a model to provide annual updates. The rules used in building the algorithm and the validations are described in appendices.

So as a Landmark paper, it marks the start of a new, market research orientated, flow of data from the Census.

Where are we now, following on from the very successful 2011 Census? Whilst the 2011 Census did address the deficiency described above, by capturing and processing more data on the over 65s, the Approximate Social Grade outputs were still limited to the 16-64 age groups. This is simply due to need for much more data on the older age groups than can be captured in the Census to fully calibrate the model.

 The CGG social grade working party was disbanded after building the initial model. However Corrine Moy supervised the development of a new model using 2011 NRS data – which was built by Helen Lambert (GfK, NOP) and presented by Helen at the 2013 CGG conference. You can find the new model here and Helen Lambert’s presentation can be found here. 

This Landmark paper, and the recent work to update the original model, also demonstrates that despite all the changes in society and the new flows of personal level data available today, social grading for sampling, classifying participants and developing weightings remains vitally important within market and social research.

The next CGG seminar on November 5th will focus mainly on how users have been applying census and other open data, with an emphasis on 'how to' case studies and useful lessons learnt. Click here to find out more.