Mon compte |
Mon panier d'achats |
Ma bibliographie 
Contacts |
Aide 
Sociétés contemporaines
Presses de Sc. Po.

I.S.B.N.2747530272
252 pages

p. 187 à 189
doi: en cours

Veille sur la revue
Veille sur l'auteur
S'abonner à la revue
Vous consultez

no 45-46 2002/1-2


Discipline > Revue > Sommaire > Article
Police normale Agrandir la police Imprimer l’article. Nouvelle fenêtre


2002 SOCIÉTÉS CONTEMPORAINES

On official social classifications in france and in britain

John Goldthorpe Nuffield College New Road OXFORD OX1 1NF
In Britain for most of the XXth century social classifications were developed on essentially the same lines as the CSP (and PCS) in France, although, to be sure, in a far less systematic and institutionalised way : i.e. they involved consultations and negotiations with employers, unions, professional organisations etc. And, in turn, the Registrar General’s Social Classes were defended on essentially the same lines – though with less elegance and sophistication – as are the CSP in the report by Faucheux and Neyret : “everybody knows what they mean” (“un ‘langage’ naturel et familier à tous”) ; “they just correlate well with everything” (“... élément de repérage primordial des disparités de comportement, d’opinions, de revenus, d’inégalité de chances”) etc.
However, what became increasingly clear to many sociologists, myself included, and then to statisticians and others within the British ONS, was that this approach had many inherent weaknesses and deficiencies, and that, as social scientists, we had a responsibility to try to do something better. (After all, biological scientists have in their field managed to go a good way beyond “formes primitives de classification” even while paying them all due intellectual respect). In this regard, the following would seem the most important points. Of further interest would be the two volumes that resulted from the ESRC-ONS (Economic and Social Research Council-Office for National Statistics) review of British social classifications : D. Rose and K. O’Reilly, Constructing Classes, 1997 ; and The ESRC Review of Government Social Classifications, 1998.
  1. Any social classification resulting from a constructivist approach must be `synthetic’ in character : i.e. it will be based on a number of criteria of, it may be supposed, differing degrees of interest and importance to the various social actors involved in the `construction’. Thus, the CSP reflect employment status, legal status, sector, level of qualification, skill etc. However, the question then arises of how, in the final product, these different criteria are to be combined and weighted, and especially if they are not simply to be “croisés de façon mécanique”. In the case of the CSP I do not find any clear answer to this question ( cf Faucheux and Neyret, p. 69) ; and it seems to me that, with any synthetic classification, the problem can only be dealt with either by further socio-political negotiation (what one might call “constructivisme jusqu’au bout”) rather than on any rational basis (despite claims of “une construction raisonnée”) or by social scientists in the end imposing their own theoretical ideas, explicitly or – in the circumstance, more likely – implicitly and thus without transparency.
  2. In reaction to this difficulty with synthetic classifications, my colleagues and I have in effect taken up the maxim of “one concept, one criterion, one measure”. The concept ( e.g. class) and what is taken to be the relevant criterion ( e.g. employment relations) are indeed specified quite openly on theoretical grounds, and the classification then constructed accordingly. However, there is no need here to equate “theory” with “ideology” as Faucheux and Neyret seem to do (or for that matter to suggest that my own class schema is Marxist, which is quite mistaken). Further, and more importantly, the theoretically informed classification is then open to critical assessment through tests of its criterion validity : i.e. it can be investigated empirically whether the classification, through the proxies used in making it operational (e.g. occupation and employment status), does indeed capture what it is conceptually supposed to capture (e.g. differences in employment relations). I might add that the decision by ONS to take my class schema as the basis for its new classification, to replace the Registrar General’s Social Classes, was essentially the result of the schema performing rather well in the variety of tests for criterion validity to which it was subjected. In contrast, in the case of synthetic classifications, tests of criterion validity are scarcely possible because of the multiplicity of criteria and uncertainties over weighting (and I now see that my request regarding the results of such tests for the CSP was beside the point). Rather, only the issue of construct validity can be raised : i.e. does the classification give rise to associations and correlations with other variables of a theoretically expected kind. In this regard, it is true that synthetic scales usually – though scarcely surprisingly – perform well, especially when allied with inductivist, data-dredging techniques such as correspondence analysis.
  3. However, this success comes at a heavy price. To begin with, there is always the problem of variables being confounded. For example, if qualifications are already one element in the construction of the CSP, it is difficult to see how the relationship between CSP and educational level can then be empirically investigated without bias. Further, even if this problem does not arise, it is difficult to derive from a connection between a synthetic classification and some `dependent’ variable of interest any indication of the nature of the causal processes that might be involved. Because of the diversity of the criteria involved, very many different causal narratives would be possible. It is notable that social epidemiologists now seem very enthusiastic about our new classification because it gives them the possibility of examining, through appropriate multivariate techniques, the links between class and aspects of morbidity or mortality while also bringing other variables apart from class into the analysis : education, income, housing standards, diet etc. Assessing how far an association with class does or does not remain when these other variables are controlled (or vice versa) can then give valuable clues to aetiology. In short, while a synthetic classification may well appear attractive as a basis of description, its value for analytic purposes is likely to be very limited.
  4. Finally, it is difficult to see how a constructivist synthetic classification could be developed for cross-national comparative purposes – as Faucheux and Neyret seem to some extent aware. If such a classification is the “fruit d’une longue histoire, et de très nombreuses consultations et concertations”‘, it is difficult to see how it could “travel” ; it is too much the product of national specificities. With the kind of classification that my colleagues and I have tried to develop, the question of its applicability in other societies does at all events remain open, and could be decided by empirical enquiry : i.e. by carrying out elsewhere the same kind of tests of its criterion validity as were made in Britain. This is in fact already being done in some cases with, I understand, quite encouraging results.
© Cairn 2007 Vie privée | Conditions d’utilisation | Conditions générales de vente
À propos | Éditeurs | Bibliothèques | Aide à la navigation | Plan du site | Raccourcis