Hence this paper will be in three parts. In the first part we shall begin with some historical background, first in relation to official socio-economic classifications (SECs) [2] in the UK and second in relation to sociological approaches. This will allow us to place the new NS-SEC in its overall context. The second part will address issues surrounding the development of the NS-SEC. Finally, in the third part we shall turn to a brief comparison of the NS-SEC and CSP, ending with a few remarks on the potential of these measures in relation to the development of an EU SEC. The paper’s aims are thus quite ambitious. As a consequence we shall have to sacrifice some depth for the sake of the paper’s breadth.

That social class should have attracted so much official, academic and, indeed, popular [3] attention in Britain is perhaps no surprise. Class has sometimes been referred to as ‘the British disease’ and the UK has often been described as a ‘classridden society’ (Halsey 1995). In part because of British sociology’s insularity (see Albrow 1989) and its obsessions with class and citizenship (Marshall 1950), Raymond Aron is reported to have said ‘The trouble is that British sociology is essentially an attempt to make intellectual sense of the problems of the Labour Party’ (Halsey 1985 : 151). Not now, perhaps ! However, it is certainly the case that the status relations of class have been more to the fore in British society than in most others. Class has perhaps been more institutionalised, visible and tangible in the UK than in other capitalist societies. So, it is unsurprising that class appeals to British sociologists as an issue of considerable intellectual importance, to government officials as a phenomenon of policy relevance, and to market researchers as an indicator of life style and taste. As a further consequence, each has produced distinct approaches to social class and its consequences, although only sociologists (and only in the last thirty years) have questioned how class arises in the first place and the nature of its dynamics. That is, in sociology there has been a turn away from a concentration on the merely distributional aspects of social stratification towards a concern for the relational ones ( cf. Goldthorpe and Bevan 1977 ; Egidi and Schizzerotto 1996). Meanwhile, UK government statisticians, unaffected by developments in sociology, maintained and enhanced their official socio-economic classifications over the course of the twentieth century. Now, however, it has been decided to abandon the RGSC and to replace it with the NS-SEC, a measure whose basis derives from recent sociological research on the relational aspects of class. Thus, two British traditions of class research, the official and the sociological, have become united in the NS-SEC. What are these traditions and how did their unification emerge ? In order to examine these questions, first we must consider each approach.

However, these were by no means the only changes made to the RGSC after 1921. At every subsequent decennial Census changes were implemented both in the method of classifying occupations and in the allocation of particular occupations to social classes. Since occupations were allocated to classes on the basis of judgements made by the Registrar General’s staff and various other experts whom they consulted, and not in accordance with any coherent body of social theory, the RGSC was rightly described by Marsh (1986a) as an intuitive or a priori scale. This is not to suggest that it made no assumptions about the structure of society and the nature of social stratification. In fact the RGSC embodied the now obsolete and discredited conceptual model of the nineteenthcentury eugenicists ; namely, that of society as a hierarchy of inherited natural abilities, these being reflected in the skill level of different occupations. Although the first published application of the class schema in 1913 was in relation to the interpretation of infant mortality statistics, the real inspiration for its construction came from the nineteenthcentury debate about differential fertility, between hereditarian eugenicists on the one hand and environmentalists on the other. T.H.C. Stevenson, an environmentalist and advocate of interventionist public health measures, developed the RGSC in order to test and disprove these eugenicist theories. To do this he had actually to measure fertility in the different occupational groups – and this is why occupation came to be used as the crucial indicator for the measurement and construction of social classes. Hence, eugenicist assumptions about society as a graded hierarchy of inherited natural abilities reflected in the skill level of occupations, remained embedded in the official, and most commonly used, measure of social class in Britain for 90 years (Szreter 1984,1993).

The earliest analyses of mortality differences were undertaken by reference to occupation and industry. However, from the beginning of the twentieth century, the development of the RGSC gave a clearer framework for identifying and understanding health differentials within the population. It was demonstrated that there was a class gradient in health – in particular in mortality rates – and despite the creation of the National Health Service in 1948, class inequalities in health and life expectancy have persisted. Overall, those in ‘partly skilled’and ‘unskilled’occupations in RGSC classes IV and V have far higher mortality rates and lower life expectancy than those in professional and managerial occupations in Classes I and II.

These inequalities are of continuing concern. The UK Department of Health Green Paper, Our Healthier Nation, acknowledged that health inequalities in the 1990s were actually widening and that ‘the poorest in our society are hit harder than the well off by most of the major causes of death’. It also gave a firm commitment not only to improve the health of the population as a whole, but specifically ‘to improve the health of the worst off in society and to narrow the health gap’. This national pledge complements the aims of the European ‘Health For All’Strategy, to which the UK fully subscribes. This made Equity in Health its first target – specifically that ‘by the Year 2000 the differences in health status between countries and between groups within countries should be reduced by at least 25 per cent by improving the level of health of disadvantaged nations and groups’.

Quantifying the absolute and relative differences in peoples’ health within a population is thus a prerequisite for developing appropriate strategies to address them. Identifying and measuring health inequalities is essential for monitoring public health, for planning and targeting health care services and the distribution of resources, for identifying new and emerging health problems, for assisting in the discovery of causal factors, and for formulating and developing effective health service policies. In all these respects, the RGSC has played a key role.

However, dissatisfaction with the scheme has developed from a combination of theoretical, conceptual and technical grounds. This led some researchers (in epidemiology, for example) to seek other socio-economic indicators for their analyses (e.g. Osborn and Morris 1989 ; Goldblatt 1990). In sociology, where social class is such a crucial explanatory concept, alternative class and occupational scales (as discussed below) were derived on what were regarded as more satisfactory theoretical foundations.

The reasons why many researchers have sought alternatives are to be found in the now well-known limitations of the RGSC. Precisely because it had been so widely used for research, many problems with it came to light. Especially when we consider that it was created to describe an industrial society and economy, in the context of a nineteenth century debate between eugenicists and environmentalists, and in a time before serious theoretical social science had emerged in Britain, it is not surprising that the RGSC came to be considered inadequate by many academic researchers.

A plethora of articles and book chapters appeared calling attention to the RGSC’s problems. Many writers criticised it because it had no coherent theoretical basis. As Thomas (1990) conceded, even the champions of its empirical usefulness agreed on this. Others, as we have seen, have claimed that what conceptual basis it did have – a hierarchy in relation to social standing or occupational skill – in fact reflects an outmoded nineteenth-century view of social structure (Szreter 1984). Even when judged in its own terms, questions were raised regarding its validity and reliability. For example, Bland (1979) provided cogent evidence that any claim that the RGSC related to social standing could not be justified. Above all, perhaps, there was an increasing recognition that RGSC described an industrial society and economy that was fast disappearing and in which the old manual/non-manual divide was of less relevance. Inter alia, this meant that it was increasingly difficult to maintain and adapt the RGSC to new realities [4].

The problems that arose with SEG were somewhat different from those of RGSC. To begin with, there was no explanation regarding its conceptual basis. Reference to it being ‘a measure of social and economic status’ was hardly illuminating. Its seventeen groups could be collapsed to produce something like the Goldthorpe class schema on which we have based the NS-SEC (see below). However, the lack of a conceptual rationale necessarily meant that there could be no clear rules to guide researchers on how SEG might best be collapsed for analysis, hence the many, varied and often incoherent ways in which this was achieved. Nor did the SEG collapse into the RGSC classes. They were different measures. As with RGSC, SEG also relied on outmoded distinctions – those of skill and the manual/non-manual divide. Partly as a consequence of this, it reflected women’s position in the social structure very inadequately, with the heterogeneous SEGs 6 and 7 being particularly responsible for this.

The primary distinctions made in Goldthorpe’s approach are those between : (1) employers, who buy the labour of others and assume some degree of authority and control over them ; (2) self employed (or ‘own account’) workers who neither buy labour nor sell their own to an employer ; and (3) employees, who sell their labour to employers and thus place themselves under the authority of their employer. Thus any class schema based on employment relations, i.e. that defines positions in terms of social relationships at work, must include these three basic class positions. Why these basic positions exist should be obvious for any society based on the institutions of private property and a labour market. However, we can immediately note that Goldthorpe’s distinctions separately identify the self-employed, a category that was egregiously absent from RGSC.

Employees account for anything up to 90% of the active working population. Clearly, they do not all hold similar class positions. That is, employers do not treat all employees alike in respect of their relations with them as defined by the explicit and implicit terms of employment contracts. There is differentiation in employers’ relations with employees. Thus, crucial to Goldthorpe’s conception is a further level of distinction within the employment relations of employees. To observe that there are quite diverse employment relations and conditions among employees is another way of saying that they occupy different labour market situations and work situations (Lockwood 1958/1989) as expressed through employment contracts. Labour market situation equates to issues such as source of income, economic security and prospects of economic advancement. Work situation refers primarily to location in systems of authority and control at work, although degree of autonomy at work is a secondary aspect. Hence, in this conceptual construction, variation in employment contracts provides the main basis for establishing its construct validity (see Rose and O’Reilly 1998 : Appendix 10). That is, ‘membership of the classes it distinguishes, as well as having differing sources and levels of income, also have differing degrees of stability of both income and employment and differing expectations as to their economic futures that together condition both their life chances and many aspects of their attitudes and patterns of action’ (Goldthorpe 2000a : 1578-9). The Goldthorpe schema thus distinguishes broadly different positions ( not persons) as defined by social relationships in the work place – i.e. by how employees are regulated by employers through employment contracts (Goldthorpe 2000b). Three forms of employment regulation are distinguished.

First, there is the ‘service relationship’in which the employee renders ‘service’ to the employer in return for ‘compensation’ in terms of both immediate rewards (e.g. salary) and long-term or prospective benefits (e.g. assurances of security and career opportunities). This relationship ‘is likely to be found where it is required of employees that they exercise delegated authority or specialized knowledge and expertise in the interests of their employing organization’ (Erikson and Goldthorpe 1992 : 42 – emphasis in the original). Hence, within this relationship, employers must allow a certain amount of autonomy and discretion to the employee. Hence, also, employees must be encouraged to make a moral commitment to the employing organization. The service relationship is designed to create and sustain this type of commitment. The service relationship typifies higher professional, senior administrative and senior management occupations. This is where ‘the largest responsibilities in decision-making attach and which will in turn offer the fullest range of beneficial conditions associated with the service relationship’ ( ibid : 43). However, the service relationship is also found in a more restricted or attenuated form in the lower professional and managerial occupations, as well as in higher technical occupations.

In contrast with the service relationship, the ‘labour contract’entails a relatively short-term exchange of money for effort. Employees are closely supervised and give discrete amounts of labour in return for a wage (or nowadays even a ‘salary’ in the limited sense of a direct payment to a bank account). Payment is calculated on or related to the amount of work done or required or by the actual amount of time worked. The labour contract is typical of ‘working class’ occupations, but again is found in attenuated forms, for example for supervisors and ‘skilled’workers. That is, these occupations have slightly more favourable employment terms than others in the ‘working class’ where external controls can be fully effective.

Intermediate or mixed forms of employment regulation combine aspects from both the service relationship and the labour contract. These are typical for clerical occupations, as well as for some technical, sales and service occupations. They are especially prevalent in large, bureaucratic organizations.

The contrast between the service relationship and the labour contract is idealtypical. In the real world, actual employment relations may only approximate these types. Goldthorpe (2000b) discusses the reasons why these forms of employment regulation exist and are common across countries with developed market economies. Briefly, two factors are implicated in determining the form of employment regulation : (1) the degree to which work may be monitored by the employer (external controls) and (2) the specificity of human capital used by employees in their jobs. Thus, where employers have difficulty in monitoring the work of employees and employee human capital is high, a service relationship will exist. Where work is easily monitored and controlled and where human capital of employees is low, a labour contract will exist.

FIGURE 3
THE GOLDTHORPE CLASSES ( SEVEN - CATEGORY UK VERSION )

Erikson and Goldthorpe (1992 : 42) have noted that the distinction between the service relationship and the labour contract is similar to some conventional distinctions made in several European countries. France, of course, distinguishes between cadres or employ é s and ouvriers; Germany between Beamte or Angestellte and Arbeiter ; and the UK between staff and workers.

The Goldthorpe schema also separately identifies categories for the other two basic class positions : employers and the self-employed. Employers are divided between ‘large’ and ‘small’. The distinction here is between employers who delegate at least some managerial tasks (‘large’) and those who tend to undertake these tasks themselves (‘small’). The former occupations are allocated to Class I and the latter to Class IV. Similarly, because of their different market and work situations, Goldthorpe distinguishes between professional and non-professional small employers, in Classes I and IV respectively. The latter consideration also applies to the self-employed.

The Hall-Jones scale (H-J – Hall and Jones 1950) graded occupations according to their prestige and was used by Glass (1954) in his pioneering study of social mobility. While this scale was used in some important studies, for example the Affluent Worker project (Goldthorpe et al 1969) and Townsend’s (1979) study of poverty, there were never any clear guidelines published which showed how occupations were coded to the scale by Glass ; and the degree to which different uses of the scale were truly comparable is uncertain.

The Hope-Goldthorpe scale (H-G – Goldthorpe and Hope 1974) was consciously produced to remedy the validity and reliability problems of the H-J scale and was the first step in the Oxford mobility project before Goldthorpe abandoned it in favour of his class schema. The H-G scale is derived from a survey of the social standing of occupations so that jobs are ranked in terms of their social desirability. In that sense, H-G is not a prestige scale but a cognitive judgement about the desirability of different occupations. As Goldthorpe (1981 : 9) has noted, the H-G scale can be regarded as a synthetic one which projects occupations on to the one dimension of ‘general desirability’, but with respect to a range of attributes whose selection and weighting is effectively a matter of popular opinion.

Whereas the H-G scale is an evaluation of desirability, the Cambridge Scale (CS – Stewart et al 1980 ; Prandy 1990) is an associative one. Based on the scaling of survey respondents’ occupational friendship and marriage scores, the CS is regarded by its originators as a broad measure of social stratification and social inequality. Ultimately the scale measures the market outcomes of different jobs and the lifestyle associated with them. It is not an attempt to measure the social structure and the way this creates different market capacities in different sections of the population. Indeed, the theoretical position of the authors of the CS is one that rejects class analysis on the grounds that it is a static approach to what are fundamentally problems relating to social dynamics. Nor is CS a status scale. It is a measure of lifestyle determined by social experience and, ultimately therefore, significant social processes. It is designed to unite key features of both the social and the economic ; and it raises questions about any attempt to analyse social inequality in terms of categorical measures.

In a series of articles over the last decade the authors of the Cambridge Scale have argued against the theoretical basis and empirical usefulness of the Goldthorpe class schema and its offspring (Blackburn and Prandy 1997 ; Prandy 1990,1998a, 1999 ; Prandy and Blackburn 1997). The programme of validating the Goldthorpe schema undertaken by Evans and Mills (Evans 1992,1996 ; Evans and Mills 1998, 2000) has attracted particular critical attention ( cf. Prandy and Blackburn 1997). In response, this invoked a detailed critique of the Cambridge Scale from Evans (1998). Neither has the NS-SEC escaped the attention of the Cambridge scholars, with pointed critiques to be found in Blackburn (1998) and Prandy (1998b) followed by a response from Rose (1998).

The differences between the two camps could hardly be greater. Other than their common interest in social mobility and the consequences of social stratification for individuals, there is hardly any agreement whether theoretical, conceptual, or operational. At the most abstract level, the Goldthorpe schema draws on the idea that one of the most important structuring characteristics of modern societies is given by individuals’ occupational positions within the social relationships of employment (Goldthorpe 1997,2000a and b). This approach entails the a priori definition of classes that exist independently of individuals and then the assignment of individuals, through their occupation and employment status, to these ‘empty spaces’. Because of a lack of data relating to employment relations across all occupations, initial construction of the Goldthorpe schema invariably involved expert judgements in assigning occupations to particular classes. However, these judgements have been subjected to many criterion validation studies that have sought to investigate whether or not the schema actually measures what it purports to measure (Evans 1992,1996 ; Birkelund et al 1996 ; Evans and Mills 1998,2000 ; Rose and O’Reilly 1998 : Appendix 10).

This deductive method stands in contrast to the inductive nature of the Cambridge scale. The authors of the CS go to some lengths to distance themselves from any theoretical or a priori assumptions and any theoretical basis for the scale appears to rest on ‘the reasoning that incumbents of occupations that are socially similar would tend to interact more than incumbents of those that are dissimilar’ (Prandy 1990 : 630). From a variety of data sources (see Prandy 1990 ; Evans 1998) the occupations of friends and spouses are scaled and found to be arranged along a single dimension. The authors assert that the scale actually measures ‘stratification arrangements’ or ‘generalised advantage’ (Prandy 1990 : 635) and more recently it has been compared with Bourdieu’s idea of the volume of global capital (Prandy 1999).

Criticisms of the CS tend to be based on the lack of theoretical justification for friendship choices as a primary method of structuring society and why this should be so (Evans 1998) and also to the lack of any criterion validation work due to the inductive nature of the CS’s construction. Prandy (1998b) and Blackburn (1998) view the separation of the economic and social dimensions of social stratification embedded in the Goldthorpe schemas as fundamentally flawed. Along with others, they have produced studies that provide evidence of stronger associations between the outcomes of interest and the CS over that of other classifications. This has simply led to further disagreements over measurement issues.

Thus, since the late 1990s a virtual cottage industry has been established surrounding the assessment of the competing claims of various social classifications (and other measures of socio-economic advantage such as car ownership and housing tenure) and how they relate to the outcomes of interest in each case (e.g. Bartley et al, 1999 ; Chandola, 1998,2000 – see Rose and Pevalin, 2000 for a reply – Prandy, 1999 ; Sacker et al, 2000). Breen and Goldthorpe (1999 : 7) have characterised this type of assessment, using independent variables with different metrics, as a ‘variable race’. They have noted that ‘assessing the relative importance of independent variables, whether in a regression context or otherwise, is a much more complex and difficult matter than has often been supposed’, especially when the comparisons are between categorical and continuous variables. Whatever the relative strengths of the associations and the care with which they are determined, the fundamental point remains that explaining variance in outcomes sheds little, if any, light on the validity of any schema or scale in terms of what it claims to measure. This can only be achieved through theoretical reasoning and criterion validity exercises (see also Evans, 1998 ; Rose, 1998).

Naturally, there have been many other conceptual and methodological disputes between sociologists in the UK surrounding and arising from the issues discussed in this section. In particular the continuing relevance of class analysis has been challenged. Since these are not exclusively British debates, we shall not address them here, but for UK perspectives readers are referred to the work of Pahl (1989 and 1993), Goldthorpe and Marshall (1992), Savage et al (1992), Butler and Savage (1995), Lee and Turner (1996), Scott (1996), Marshall (1997 : Ch. 1), Halpin (1997 and 1999), Prandy (1998b), Blackburn (1998), Crompton (1998), Rose (1998) and Crompton et al (2000). Similarly, there is not space in this paper to address the recent debates on ‘meritocracy’ and social mobility, initially stimulated by the work of Saunders (1996). This has led to vigorous responses from Marshall et al (1997), Breen and Goldthorpe (1999) and Savage and Egerton (1997).

We have discussed the two main traditions of socio-economic classification in the UK. We can now proceed to a discussion of the NS-SEC, the new UK government SEC that replaced both RGSC and SEG in April 2001. The intellectual origin of the NS-SEC is the Goldthorpe schema. The review that led to its creation was established by ONS as a result of recognition of the shortcomings of RGSC and SEG as already described above.

Figure 4 offers a diagrammatic representation of the way in which the NS-SEC is derived. As with the Goldthorpe schema, the primary distinction made by the NS-SEC is between employers, employees and the self-employed. To these we added a fourth basic position for those who are involuntarily excluded from employment relations altogether. However, such a classification is not exhaustive, as figure 4 shows.

The operational categories (indicated in bold in figure 5) represent a variety of labour market positions and employment statuses which can be collapsed into socio-economic classes as defined by an employment relations approach (see below ; and see also Goldthorpe 1997). L14 is an optional category. L15, L16 and L17 are the residual categories that are excluded when the classification is collapsed into classes. All the sub-categories are component codes required for bridging and continuity to RGSC and SEG rather than necessary sub-categories in terms of the conceptual base of the NS-SEC. For example, L3 is sub-divided between positions which were recognised by both SEG and RGSC as professional – ‘traditional professionals’ – and those (for example, computer analysts) which now appear to be professional positions on the basis of research conducted to produce the NS-SEC – ‘new professionals’.

FIGURE 6
NATIONAL STATISTICS SOCIO - ECONOMIC CLASSIFICATION

Finally, we should note that the NS-SEC has been subject to a full range of criterion and construct validation analyses (Rose and O’Reilly 1998 ; Rose and Pevalin 2003). It has been shown both to be a good measure of employment relations and a sound predictor of life chances.

Of the nine EU member states that have a SEC, Grais (1999) notes two contrasting approaches to the derivation of national classifications, what he calls the ‘theoretical’and the ‘intuitive/empirical’. The latter are more common. Only the UK (for the NS-SEC), the Netherlands and Sweden have adopted a ‘theoretical’approach, although we should note that the French CSP could be said to be ‘theoretical’, although of a more inductive type. Nevertheless, the real difference between the French approach and those of the UK, Netherlands and Sweden [all of which are based on or closely related to the Erikson-Goldthorpe-Portocarero (EGP) schema] is perhaps best summed up as a difference between emic and etic approaches to social science [6].

We would agree with Brauns (1999), however, when she argues that the basic principles of CSP and Goldthorpe’s approach are relatively similar. It was for this reason that she and her colleagues were able to construct the Goldthorpe schema based on French data. Therefore, we should be careful not to over-emphasise the differences between national approaches. What should be obvious are the implied similarities of the classifications themselves in terms of categories and meanings. Very similar variables enter into the various SECs – occupation, activity status, status in employment, enterprise size, agricultural and non-agricultural sectors – and the categories of the theoretically based SECs are generally interpreted to have meaning as ‘social’units. All the SECs, except for the Spanish, are what Grais refers to as ‘multi-dimensional’; and also they are ‘partially ordered’classifications. Nearly all have the flexibility previously referred to, with more and less aggregated versions. Thus, if we examine Grais’ analysis, it is clear that :

Thus, there are many common basic features to national SECs. To be sure, details differ on issues such as size cut-off for enterprises, definition of household reference persons and the precise treatment of the inactive. However, the principles, whether implicit or explicit, appear to be similar. Naturally, there needs to be a common language to describe the concepts embodied in these similar principles. Nevertheless, in this respect it is clear, for example, that both CSP and SECs related to Goldthorpe’s schema are similar in the following important ways :

Nevertheless, it would be interesting to undertake analyses that compare CSP and NS-SEC in various ways [7]. This may be possible as work on a European SEC develops further (see Rose et al 2001 : Sections 9 and 10).

It has often been remarked that almost any sensibly derived intuitive SEC will have the capacity to display variation. The Social Grade scheme used in UK market research (MRS 1991) is an example. However, it will not have analytic transparency. That is, without a clear conceptual rationale, we cannot understand the causal pathways which lead to the regular patterns revealed by its use in research ; that is, the processes that generate empirical regularities (Breen and Rottman 1995 ; Bartley et al 1999). In addition, if we cannot get a handle on causal pathways, then it is not apparent how recommendations can be provided on relevant policy actions that might address these persistent variations. Examples include the difficulties encountered in setting targets for reducing health variations between states that can be linked to achievable policies and, more generally, in developing policies to target deprived groups. Needless to say, any SEC must also be used and interpreted correctly by analysts, if the benefits of analytic transparency are to be realised.

Obviously, as we have already observed, a clear conceptual rationale does not thereby remove all barriers to explanation. There are many bases to social stratification, not all of which will be measured by a particular SEC. Nevertheless, we would argue that a conceptually clear, properly constructed and well-validated SEC facilitates a focus on other variables when searching for explanations of remaining differences.

Finally, the lack of a conceptual rationale renders the task of validating a classification impossible and of maintaining a classification over time much harder. Validation involves both demonstrating that a measure does indeed measure what it purports to measure (criterion validity) and that it usefully discriminates in theoretically predicted ways (construct validity). In addition, once (criterion) validated, a measure may be re-validated to assist with maintenance over time.

So far as the development of an EU SEC is concerned, much work remains to be achieved. The task of applying an outline theoretical model to the different member states requires new empirical research. However, we would not wish to claim that an EU SEC would ever be superior to a national classification when the requirement is solely to analyse national data. The role of an EU SEC would be a comparative one. If it were created, there can be little doubt that it would prove a useful tool for both policy and academic purposes.