It’s not uncommon for sociologists to publish new social class schemes, and this week a particularly interesting contribution has been made by the research team led by Mike Savage and Fiona Devine who have analysed survey data collected through online and face-to-face interviewing facilitated by the BBC (described in http://www.bbc.co.uk/news/magazine-22000973 and in Savage et al. 2013)
In brief summary, a seven-class categorisation is advocated on the basis of applying a latent class analysis function to data collected from individuals over seven different aspects of contemporary life: Household income, household savings, house value, social contact score (based on relative advantage of the occupations of nominated friends), social contact number (the relative volume of social contacts held), ‘highbrow cultural capital’ (a measure of engagement in traditionally highbrow activities, such as visiting a museum), and ‘emerging cultural capital’ (a measure of engagement in activities sometimes thought of as ‘working class’ but now said to be more widely followed, such as playing video games, going to the pub, watching sports and eating fast food). One motivation for this approach as given by the authors is that alternative class schemes are too strongly rooted to certain aspects of employment circumstances (they cite the NS-SEC, http://www.ons.gov.uk/ons/guide-method/classifications/current-standard-classifications/soc2010/soc2010-volume-3-ns-sec–rebased-on-soc2010–user-manual/index.html , influenced in particular by John Goldthorpe’s class analysis). Though this does seem to neglect other approaches to measuring class and stratification, the new scheme proposed certainly offers some interesting alternative categories to those associated with the Goldthorpe approach.
Unlike most social class schemes in common use, allocation to the classification is not based on just one or two economic features (such as occupation or income band) but is influenced by all seven of these elements, generating what is said to be a ‘multidimensional’ categorisation of social classes. A few possible problems, however, with such a strategy include the relative practical difficulty of operationalizing the measure on other data sources, and the apparently strong relationship between category membership and, in the UK, age, gender, and region.
Like many sociologists we’ve been very interested in this new approach and the impressive analysis behind it. It is easy to imagine it has very strong empirical properties in comparison to alternative measures, and indeed we hope in our own work to soon be able to implement a version of it for existing UK surveys such as the BHPS in order to compare its placement of people against other tools of social classification. However we also have a few doubts about features of the approach. For one, like other similar classification tools designed in the social sciences (see also the paper on US data by Hennig and Liao at http://www.rss.org.uk/site/cms/contentviewarticle.asp?article=866) the scheme seems to have many similarities to the classificatory typologies which have long been used in market research to allocate people to consumer and behavioural groups (e.g. http://en.wikipedia.org/wiki/Mosaic_(geodemography) ). The class scheme, however, is much less detailed that most of these groups, and doesn’t explicitly feature age and regional criteria (its categories are clearly related to age and region, but it’s not obvious to us why these important factors shouldn’t be directly included).
An intriguing aspect of the measure to us concerns the possibility of changing social classes through the life course (or even within households). For instance, the constellation of different measures used to allocate position could lead to husbands and wives with different cultural habits being placed in different class categories, whilst it seems to us that classification hinges substantially upon measures of current earnings, household wealth and savings, to the extent that many people might change classes during their life-course on the basis of changes in these – for example by gradually paying off their mortgages whilst living otherwise stable life-courses. Traditional models of social class in sociology have been interested in allocating people to positions which serve as long-term indicators of their location within the structure of social inequality, but by contrast the recommended scheme seems to allow for people to stay in the same job continuously for 40 years whilst occasionally changing classes whilst they age. This could be either a positive or negative feature – compared to traditional models of social stratification, the confusion between life-course stage and stratification circumstances seems unhelpful, but if on the other hand we are interested in improving our estimates of the precise circumstances experienced by people in their current situations, there are many ways in which this flexibility could be attractive (it might, for example, be possible that combining this scheme with an occupation-based measure such as CAMSIS could enable analysts to compare whether occupational advantage is derived more substantially from the socio-economic position or social perception of the occupations).
A more general concern that we have with the methodology used to define the classes as it is described, however, involves the relative influence of different dimensions in the allocation to classes. From the research article describing the classes, it sounds to us as if income measures dominate the allocation in most important respects, which sounds a plausible outcome of using latent class approaches with mixed format variables, but probably isn’t consistent with the description of the class scheme as multidimensional. However to be fair to the authors of the scheme, more results are promised shortly looking at relations with measures such as of education and occupational position, which might well address this concern.
More significantly, we remain sceptical of the division of positions into the seven categories (the categorical division which has attracted much attention in the news media). The authors of the scale are trying to identify important boundaries in the social structure, and the categorisations suggested are certainly compelling and easy to identify with. Nevertheless, we suspect that both the statistical and theoretical justifications could be questioned (as is the case, we believe, with many other social class schemes which use similar numbers of categories). Statistically, the seven category latent class solution for a small datasets could be artefactual – applied to a much larger dataset the model fit statistics might well favour categorisation into a much larger number of groups. Moreover, we suspect that several of the dimensions underlying the scheme are largely gradational in nature rather than categorical, and that a metric rather than categorical functional form may well be a better statistical fit to the phenomena observed, as well as providing a better theoretical account. (Needless to say, one of the metric dimensions that we imagine would prove an important component of the structure described might be the same one that is captured by contemporary CAMSIS scales as at www.camsis.stir.ac.uk !).
In any case, it’s been impressive to see the extent of debate already published on the measure, and we certainly look forward to seeing further results using this class measure.
Paul Lambert & Dave Griffiths, 4 April 2013
Savage, M., Devine, F., Cunningham, N., Taylor, M., Li, Y., Hjellbrekke, J., Le Roux, B., Friedman, S., & Miles, A. (2013). A new model of social class: Findings from the BBC’s Great British Class Survey Experiment. Sociology, doi: 10.1177/0038038513481128, 1-32.