The Politics of the Service Class: The homology of positions and position-takings

From its inception by Karl Marx, sociological class analysis has been tied to questions of conflict and political alignments. The overall idea has been that differential positions within the production and distribution of societies’ goods and rewards produce different position-takings towards the organisation of this production and distribution. The working class would favour policies to promote the equal distribution of these goods and rewards and policies providing workers with more control over the production process, whereas the higher classes would opt for the opposite. Even though there has been considerable debate about the extent and permanency of these proclivities, they have been evident throughout the post-war period (Evans 2000; van der Waal et al. 2007).

Researchers and commentators alike were puzzled, however, when the ‘new’ middle classes arrived on the scene accompanied by forms of ‘new’ politics. What would be the political leanings of groups that found themselves at neither pole of the hierarchy? And, crucially, how is this explicable in terms of their position in the social structure? This is an important issue, since the ‘new middle class’ constitutes a sizeable segment in most Western countries, and one that is relatively privileged and influential. The puzzle was complicated by the emergence of new forms of radicalism linked to issues of warfare and environmentalism, as opposed to the bread-and-butter issues of the labour movement (Parkin 1968).

This produced a startling plurality of perceptions, fears and hopes. Several commentators expected these new middle classes to represent a new source of radicalism, as they were regarded as less tied to the mechanisms providing the power and privileges of the old middle class (Ehrenreich and Ehrenreich 1977a; Gouldner 1979). Alternatively, they were also conceived of as a ‘new working class’ and expected to align with the traditional one (Mallet 1975). These hopes were mirrored in conservative American writers’ fears of the same (see the bulk of the contributions in Bruce-Briggs 1979).

Commentators of a liberal bent emphasised that the rise of the middle classes flew in the face of Marxist expectations of a polarised class structure and corresponding political landscape. In fact, the growing influence of the values of professionalism and managerialism would create a more humane ‘post-capitalist’ social order, rendering the socialist project irrelevant. On this count, the rise of the new middle class was associated with liberal progress rather than revolution (Dahrendorf 1959; Kerr et al. 1960). Then came the 1960s, with its outbursts of radicalism among the ranks of these new middle classes (see the classic account in Parkin 1968). The emergence of the distinctively middle-class New Left lent at least some credence to the hopes and fears of the radical and conservative new-class theorists.

How could the same phenomenon be given such divergent accounts and interpretations? Probably because the social group or social groups in question do not exhibit one unitary political profile. Those who want to emphasise either the conservatism or the radicalism of the middle class will find at least some support in the evidence. The complexities involved in a differentiation of class structures and political landscapes nurtured claims about the decline or even the death of class (Clark and Lipset 1991; Pakulski and Waters 1996). In this paper, I argue that in order to understand middle-class politics, we need a more complex conception of both class and politics, as alluded to by Clark and Lipset. Taking Goldthorpe's (1982, 1995, 2007) account of the service class as a point of departure, I show that political divisions within this class are related to its internal social divisions. The data analysed is the Norwegian 2009 Electoral Survey.

Any account of the politics of the middle class hinges on how ‘the middle class’ is conceptualised. Although numerous accounts have been suggested (for example Ehrenreich and Ehrenreich 1977b; Giddens 1981; Abercrombie and Urry 1983; Wright 1985; Lash and Urry 1987), Goldthorpe's theory of the service class (1982) appeared to be the most enduring contribution, owing much to the success of the EGP class scheme more generally. The service class is seen as a class of employed executives, managers and various professionals. Goldthorpe originally conceptualised the service class as distinct because of the trust installed in them by their employers to ensure that their execution of management tasks and knowledge applications was in the firm's best interest. In later formulations this is expressed in terms of a differentiation of employment contracts (Goldthorpe 2007).

The service class, Goldthorpe argued, was in a sense similar to the old middle class in that its members were relatively privileged by the current state of affairs. They would thus seem to have little real incentive to opt for a radical recasting of the institutional order. Goldthorpe's view was accordingly that this middle class is ‘an essentially conservative element within modern societies’, and that ‘white-collar radicalism will remain minoritarian, intermittent and localised’, since it lacks a structural basis (Goldthorpe 1982: 180, 183). This position was maintained in the face of empirically based criticism (Goldthorpe 1995). It is significant to note that Goldthorpe does not deny the existence of some political heterogeneity within the service class; his point is that these variations are not related to their class position, but only to differences in situs.

The debate about middle-class politics was complicated by the emergence of ‘new politics’. Parkin's (1968) pioneering study of middle-class radicalism underlined that it was in fact quite distinct from the bread-and-butter politics of the working class. The increased salience of questions relating to the environment and climate change, (nuclear) war, rights for homosexuals, racism and xenophobia and other issues of a more ‘non-material’ nature is often seen as a redrawing of the political landscape. The coming of ‘post-materialist’ politics (Inglehart 1990) and the politics of life-style, as opposed to that of life-chances (Giddens 1994), are seen as either supplanting or supplementing the old-fashioned left-right economic divide. Many authors thus argue that political attitudes should be understood two-dimensionally, as involving a classical economic left-right axis and a ‘post-materialist’, ‘new-political’ value dimension (Flanagan 1982; Clark and Lipset 1991; Borre 1995).

The challenge for class analysis is to account for how position-takings relate to social positions. Goldthorpe's postulate of an essentially conservative service class has proved highly contestable. Two central examples will suffice. Prior to Goldthorpe's introduction of the concept, Giddens argued that the groups of professional employees should be seen as distinct from managers and executives because of their different positions in the ‘paratechnical relations’, regardless of their similar market situation (Giddens 1981: 187–8). Goldthorpe's erstwhile chief antagonist, Erik Olin Wright, argued for the necessity of distinguishing between groups based on their type of asset – property, organisational or skill assets (Wright 1985). This perspective implies that the service class is systematically divided by the type of assets members rely on. The British service class was indeed found to be politically divided by type of asset (Heath and Savage 1995). Goldthorpe was characteristically unmoved by these findings (1995: 324), maintaining it to be more fruitful to investigate political divisions between the higher and lower echelons of the service class (1995: 329, n. 10).

From outside the ranks of traditional class analysis, Bourdieu showed in Distinction that what he called the dominant class – which roughly corresponds to the service class – was internally fractured by the form of capital dominant in one's holdings. Notably, the propensity for rightist versus leftist sympathies was structured by this principle of capital composition (Bourdieu 1984: 452). In this scheme, the service class would be systematically heterogeneous: teachers and artistic producers, rich in cultural capital, adopted leftist positions, whereas those engaged in manufacturing and commerce, thus richest in economic capital, leaned to the right.

The service class is on average economically right-wing in many countries (Svallfors 2006: 52–102). However, there is much evidence of some political polarisation within this class too. Some studies point to differences between different occupational groups – notably various professionals as opposed to managers and executives (Brooks and Manza 1997; Kriesi 1998: 168–72; Güveli et al. 2007; Oesch 2008), and others point to a distinction between service-class members in the public as opposed to the private sector (Knutsen 2005). A third group of scholars has identified the type of assets or capitals as important (Heath 1991; Heath and Savage 1995; Rosenlund 2009; Harrits et al. 2010). These differing accounts have a good deal in common, proposing similar explanations. In this paper, I expand on this research by modelling a multidimensional space of political attitudes within the service class, whereas most other analyses restrict themselves to socialist versus non-socialist voting patterns. This allows one to investigate the differentiation in terms of attitudes to both ‘old’ and ‘new’ politics. The construction of an internal social space allows for a fine-grained analysis of how internal political and social differentiation are related.

The crux is then whether the internal political heterogeneity of the service class has any meaningful social structure, i.e., more than what can be attributed to the idiosyncrasies of its members. In what follows, I endeavour to show that the political divisions in the service class are homologous to its social division.

In Distinction Bourdieu argued that what he called the space of life-styles was homologous to the space of social positions, i.e., the stratification structure, meaning that a correspondence could be observed between social and symbolic divisions (Bourdieu 1984: 257–371). This claim is properly understood as pertaining to the relationship between two distinct structures – social and symbolic – and one cannot ‘reduce the homologies between systems of differences to direct, mechanical relationships between groups and properties’. The space of positions and the space of position-takings (life-styles) are claimed to be structured along the same lines. Hence, Bourdieu suggests that between these two spaces lies a third space of habitus, of the generative formulae ‘which underlie each of the classes of practices and properties’ (Bourdieu 1984: 126). The research design must therefore inspect the relationship between structures – that is, spaces.

This analytical strategy involves the independent constructions of two spaces – one of position-takings and one of positions, so that the structures of the two can be compared. This has the advantage of analysing position-takings as a space in its own right with its own logic, without reducing it to a ‘dependent variable’ or an effect of class divisions.

Optimal data for a research design such as this would include good indicators of political attitudes to economic and cultural issues, as well as good indicators of the various forms of capital. For practical purposes, one must be prioritised over the other as no suitable dataset containing both is accessible for secondary analysis. I have analysed the nationally representative Norwegian Electoral Survey of 2009,¹ which contains excellent questions about political attitudes, but fewer indicators of social position. This facilitates the construction of a fine-grained space of position-takings at the expense of a coarser space of positions.

The Electoral Survey was conducted in connection with the 2009 parliamentary election. A subset of this data is used, consisting of respondents’ classified as belonging to service class I and II. These are constructed by adapting the Trento-program syntax (Leiulfsrud et al. 2005), amounting to an operationalisation based solely on occupation, since the other indicators are unavailable or unsatisfactory. The sample analysed is hence systematically drawn from a representative sample, and is therefore considered representative of the Norwegian service class. However, the small n means that certain subdivisions to be analysed below may not readily be generalised. Table 1 shows the distribution of occupational groups – the two first digits of the ISCO-88 based STYRK code (SSB 1998) – within the service class.

The analysis proceeds by constructing two spaces: one of political position-takings – attitudes to a range of political issues – and one of social positions, hereafter called political and social space. These are constructed by Multiple Correspondence Analysis (MCA; Lebart et al. 1984: 81–108; Greenacre 2007: 137–44; Le Roux and Rouanet 2010).

MCA is a multivariate technique for the analysis of categorical data. It seeks to reveal the central latent contrasts or oppositions in the data and represent them as orthogonal axes. I use a variant called specific MCA (Le Roux and Rouanet 2010: 61–4), which allows some categories to be set as passive and thereby not influence the space. This is useful for categories such as ‘missing’ or ‘refusal [to answer]’. The variant of MCA used here produces two principal results – a cloud of individuals and a cloud of categories. In the cloud of individuals, distance reflects dissimilarity in individual response patterns: individuals that appear close have similar responses. In the cloud of categories, distance reflects the composition of the categories, so that categories appear close in the map when they often appear among the same individuals (Le Roux and Rouanet 2010).

In addition to the variables that are active in the construction of the space, one can use supplementary variables for further analysis. These variables do not affect the structure of the space, but are projected onto the space once it is constructed. Lebart et al. (1984) refer to this as using the space as a predictive map, a ‘visual regression analysis’. The space constructed by active questions can be thought of as an ‘independent variable’, and the supplementary variables as the ‘dependent’ ones.

Lennart Rosenlund (2009) developed a direct test of the homology argument, by drawing on the ‘reciprocal approach’ recommended by Lebart et al. (1984: 108). They suggest that in the analysis of survey data on the relationship between socio-demographics and attitudes, the researcher should follow a double strategy. Firstly, they recommend conducting an MCA of the relevant socio-demographic variables, and then project attitude questions onto the space as supplementary variables. This produces a ‘lattice’ of the socio-demographic variables and then the ‘weave design’ of attitude questions is placed onto it. Secondly, they recommend doing the reverse, conducting an MCA of the relevant attitude questions and then projecting the socio-demographics onto it.

Accordingly, Rosenlund constructed the space of life-styles and the space of social positions independently, using separate MCA procedures. He then projected coordinates from one space onto the other using a variable capturing coordinates from the space of positions as a supplementary variable in the space of life-styles. Hence he could show that the principal axes accounting for the variance in life-style indeed corresponded to those accounting for the variance in social positions (Rosenlund 2009). He was thus able to compare their structures. Harrits et al. (2010) applied the same logic to the issue of class and politics. A two-dimensional space of political position-takings – ‘old’ and ‘new’ politics – was shown to be homologous with a space of social positions: the volume and composition of capital are structuring principles of both. This last application serves as a key inspiration for the present study.

In order to measure both old and new politics, 13 questions about political issues are analysed. These are listed in Table 2, along with variable short names (used in the maps) and an indication of the categories of response. The choice of questions is inspired by Borre (1995: 190) but is expanded upon to produce a richer picture. Questions thus tap core ideological issues such as egalitarianism and state control over business, as well as core ‘new politics’ issues such as immigration, environmentalism and Christian values.

In the analysis, several categories are passive (Le Roux and Rouanet 2010: 61–2). This means that they are given no weight in the analysis and do not influence the structure of the space. This is the case for all missing values, as well as categories of ‘don't know’, ‘do not want to answer’ and so forth. More significant is that all ‘neither agree nor disagree’ categories are set as passive, because they destabilise the results by creating one separate dimension that distinguishes individuals with mostly ‘mixed’ opinions. This phenomenon has thus been excluded from further analysis in the present context.

The social space should capture the main divisions in the ‘capital portfolios’ of service-class members, based on the working hypothesis that the relative weight of economic and cultural capital is crucial. This requires indicators of the forms of capital and the data is less than optimal here. My selection has been inspired by Rosenlund and colleagues’ interpretation of Bourdieu (Prieur et al. 2008: 56; Rosenlund 2009: 88).

The five variables used for the social space are presented in Table 3. The most direct measure of economic capital is total household income. This has been recoded from an open question to a categorical variable. Education level is used to measure the scholastic component of cultural capital. Occupational group plays a key part: occupation indicates possession of economic capital, but also cultural capital. Occupations are coded with respect to the qualifications required (formal and technical). Typically, occupations also provide differential career prospects, including different prospects of salary increases. Hence, occupation is a good proxy measure of one's capital. As an indicator of capital composition, I use a question about whether one supervises the work of others. This is significant due to the rationale behind the theory of the service class: it comprises employees privileged because of their exercising of authority or application of specialist knowledge. People who do not supervise others’ work are included in the service class because of their application of specialist knowledge, which indicates cultural capital. Additionally, this variable can be expected to relate to the volume of capital, since supervisory responsibility typically entails more authority and a higher salary. Finally, I include a question about employment by sector – private, public or organisation/foundation. The public sector is a key component of the structural foundation of cultural capital, as in Norway large parts of the relevant labour markets are within it.

In the analysis, a number of categories are set as passive. This involves all missing categories, as well as ‘don't knows’, ‘refusals’ and so forth. Additionally, some occupational groups are set as passive due to their low frequencies, because rare categories unduly affect the analysis and destabilise results.

In all the graphical representations that follow, some axes are visually flipped in order ease interpretation and comparison across analyses. This involves no tampering with the data analysis.

The political space is three-dimensional. Its first dimension is a classical, economic left-right axis, contrasting ‘socialist’ with ‘laissez-faire’ views. Its second axis is an intensity of opinion one – this separates moderate from strong views, irrespective of the issues in question. The third axis is a ‘new politics’ dimension, separating liberals from anti-liberals. These three axes account for 92.36% of the variance in the active questions: the first accounts for 57.27%, the second for 26.86% and the third for 8.23%.² In what follows, I focus on plane 1–3, because the second axis is irrelevant for present purposes. Figure 1 shows the cloud of individuals – each respondent's position – in plane 1–3. A table of ‘explaining points’, categories with contributions above average, may be found in Appendix Table A1.

Figure 2 shows the active categories – the responses to the questions analysed – in plane 1–3. The first horizontal axis is primarily shaped by questions about economic issues, contributing 71.36% of the axis’ total variance. To the left on the map we find support for the following views: that full employment is best secured by government control over business; that societal control over business is necessary to prevent the leaders of industry and finance from becoming too powerful; that the public sector should be expanded even if this means raising taxes; and, that the government should seek to reduce economic inequalities. Similarly, we find disagreement with the views that we should accept greater economic inequalities to stimulate effort and that government control over business should be reduced. However, we also find disagreement with the view that the Norwegian Armed Forces should be strengthened and that Christian values should be more important. Significantly, the question about the environment versus the standard of living contributes most to this axis (as opposed to the third), with the more uncompromising support for environmentalism siding with socialist views. On the side of the socialist views we also find support for homosexuals to have adoption rights as well increased developmental aid – and lack of support for Christian values. The opposite is found to the right in the map.

The third, vertical axis separates liberal from anti-liberal views: the axis receives 73.05% of its variance from the questions tapping into ‘new politics’. Anti-liberal responses to all these questions are located on the positive side of the axis, on the lower side of Figure 2. Such responses reflect: strong agreement that immigration is a threat to national character; strong disagreement that Norwegians should aim for a more internationalist development; strong agreement that Christian values should be central; support for cutting developmental aid; strong disagreement with adoption rights for homosexuals; strong agreement that the military should be strengthened; and, an unwillingness to prioritise environmental protection at the expense of the standard of living.

Plane 1–3 can thus be read as follows: in its lower-right quadrant we find respondents with ‘laissez-faire’ views of economic issues and generally anti-liberal attitudes. In the upper-right quadrant we find respondents who are economically rightist but liberal on value issues. In the upper left quadrant we find liberal socialists, and in the lower left we find their less liberal counterparts. This interpretation is corroborated by projecting party-political preferences onto the space, using the political space to ‘predict’ party preferences. Norway's multi-party system allow for the nuances in political attitudes in both dimensions to emerge: in Figure 3, the question ‘which party best represents your views?’ is projected as a supplementary variable onto plane 1–3. Along the horizontal left–right axis, the parties are ranked according to conventional wisdom, from socialist left to laissez-faire right. Along the third, (anti-)liberal dimension, we can see that the Socialist Left Party (a typical new-left party founded in the 1960s in opposition to Norwegian NATO membership), the social-liberal Liberal Party and the Conservatives are preferred by the service class’ liberal elements, whereas the Centre Party, the Christian Democratic Party and the new rightist Progress Party are preferred at the opposite end. The Labour Party occupies the middle ground on axis 3 and is only slightly to the left on axis 1.

The social space is two-dimensional, the first representing the composition and the second the volume of capital. These account for 80.87% of the variance – 55.36 and 25.05%, respectively.³ The cloud of individuals and the cloud of categories are shown in Figure 4, and explaining points are shown in Appendix Table A2.

The horizontal axis 1 is shaped particularly by occupation (36.96%), sector (27.28%) and supervision (22.01%), and less so by educational level. Four occupational groups contribute above average: corporate managers, public service administration professionals, life science and health professionals and teaching professionals. This constitutes a clear pattern of differentiation according to the relative weight of cultural capital.

Axis 2 (volume of capital) is shaped by occupation (35.68%), educational level (27.56%), household income (19.56%) and supervision (10.78%). Four occupational groups contribute above average to axis 2 as well: corporate managers, teaching professionals, engineering science associate professionals and executive officers in administration, business services, social work and entertainment. This constitutes a clear hierarchical difference, a division by the volume of capital.

The space of positions encompasses the public/private sector divide, so central to recent debates on voter behaviour. Nevertheless, it is clear that axis 1 is not reducible to the sector divide, given the large part played by both occupation and supervision. The constructed space is theoretically meaningful and provides a good summary of central measures of social position, but as a social space of capitals, it is a suboptimal construction, owing to the limited range of indicators available.

I described homology as meaning that position-takings and positions are structured along similar lines. The two spaces constructed should then exhibit clear similarity of structure. In what follows, I will compare the two spaces. The social space is sliced into nine fractions, based on coordinates in the cloud of individuals. Specifically, this is performed by cutting each of the axes of interest into three equally sized categories. All three-by-three combinations are coded into nine class fractions. This splitting of the clouds of individuals is shown in Figure 5. This merely logical division serves to compare the structures of the spaces.

The next step is to project the class fractions onto the political space. This is performed by using the class fractions as supplementary variables in the political space. The homology is thus examined by inspecting the mean points of the class fraction in the political space, and vice versa. As a rule of thumb, distances less than 0.5 should be considered small, above 0.5 as notable and above 1 as large (Le Roux and Rouanet 2010: 46).

Figure 6 shows the class fractions inserted into the political space. Along the horizontal axis 1, the class fractions are placed according to their leanings in terms of economic issues, whereas they are placed along the vertical axis 3 based on their attitudes to issues of culture and values. To the left of the map, indicating leftist attitudes to the economy, we find all three cultural capital fractions, with the lower volume one closer to the centre. To the right we find all the economic capital fractions. Fractions with more balanced capital portfolios are closer to the centre. At the top, indicating generally liberal views, we find all higher class fractions, characterised by high volume. Moving down the axis towards less liberal attitudes, we also find fractions with lower volumes of capital.

Overall, there are clear signs of homology. The divisions by the predominant form of capital – economic versus cultural – manifest themselves in the economic left versus right division within the service class. The volume of capital manifests itself in marked differences in opinion on ‘value’ issues. These differences are much more pronounced for the right-wingers at the economic capital pole.

A t-test of the class fractions on the two dimensions of the political space⁴ indicates that five groups exhibit distinct profiles. Upper and middle cultural fractions are markedly more leftist than the mean. All the economic fractions are also distinct: The upper and middle are distinct on both dimensions, while the lower economic is only significantly distinct in their opposition to liberal values. While not every category is found statistically significantly different from the mean, we find significant differences along both dimensions. The homology is thus likely to apply to the service class and not simply the sample analysed, even if the small n warrants cautionary conclusions.

The analyses demonstrate a clear and systematic relationship between social and political divisions within the service class. But how strong is it? Norway's historical Red-Green coalition government (Labour, the Socialist Left and the Centre party) came to power in 2005 and was re-elected in 2009. According to the survey, almost twice as many in the cultural capital fractions than in the economic one reported voting for that government (63.4 versus 33.1%).⁵ Hence, the fractions of the service class not only express different opinions, but seemingly also act on them in the polling booth (Table 4).

This paper has shown that the social divisions in the service class appear homologous to its political divisions in ways that resonate well with earlier studies. This is similar to what has been found in the broader population in Norway (Haakestad 2012) and the Danish city of Aalborg (Harrits et al. 2010). There is a systematic heterogeneity in the political leanings of the service class that follow the divisions constituted by the possession of forms of capital.

The Norwegian service class is thus similar to the American, British, German, Dutch and Danish ones in that it is systematically split on economic issues, and not ‘essentially conservative’ as Goldthorpe claimed. In 1982 he could reasonably argue that heterogeneity in the service class could be chalked up to its newness: because of its expansion during the preceding decades, many service-class members were necessarily from lower-class origins, and hence the class had not yet been properly formed (Goldthorpe 1982: 174–5).⁶ Thirty years later this seems less convincing. But if this were to explain the leftism of the cultural capital fractions – the deviant case in Goldthorpe's view – one would have to postulate that these groups were relatively more often of lower-class origin, and hence less prone to the laissez-faire attitudes deemed natural for the higher classes. Evidence about class mobility in Norway does not support this: the rates of self-recruitment in the cultural and economic fractions are quite similar (Hansen and Wiborg 2010: 204). Class theory will need to account for the sources of structured heterogeneity within classes, which have been in evidence for decades.

This study has expanded upon earlier research into the middle classes by analysing political attitudes as multidimensional and revealing how ‘new’ politics followed the internal differentiation by volume of capital. The hierarchical differentiation of attitudes to ‘value’ issues is yesterday's news (Lipset 1959; Kohn 1969). It is, however, remarkable that this is so pronounced within the generally liberal service class (Svallfors 2006: 130–9). Goldthorpe, dismissive of any ‘horizontal’ divisions, suggested it would be more fruitful to investigate the difference between the higher and lower echelons of the service class (1995: 329, n. 10). Indeed, a division along these lines does show up in the analyses above. Two types of explanations are generally offered for the hierarchical side of the value issues. Those following Kohn point to aspects of the work situation: workers are less liberal because of their lack of self-direction. Others follow Lipset in explaining liberal attitudes with level of education, claiming that high education produces liberal attitudes and low or no education the opposite. (Houtman 2003: 24–65; Svallfors 2006: 121–43). In the case of the Norwegian service class, explanations in terms of one's work situation seem less pertinent: anti-liberal attitudes are most clearly related to low education, and the most markedly anti-liberal occupational group is general managers of small firms.⁷ Whatever may be said of such general managers of small firms, they are arguably not marked by the lack of occupational self-direction pointed to by Kohn.

The most significant finding, however, is the correspondence between the composition of capital and the socialist versus laissez-faire dimension. This supports Bourdieu's emphasis on the multidimensionality of class – which is overlooked even in the most comprehensive follow-up of his work on culture and class (Bennett et al. 2009). This calls for further work on whether and how this applies for populations at large, relating these findings to the upper and lower class(es).

Given the connection between capital composition and the public/private sector divide, this is in a sense in line with recent politological work (e.g., Knutsen 2001, 2005). However, the leftism associated with cultural capital cannot be wholly attributed to sector affiliation. Employees in the public sector are also diverse: there are significant political differences between, say, bureaucrats and teaching professionals. The relative leftism of the latter compared to the former is intelligible in terms of capital composition, but not sector.

The leftism of teachers and other health and cultural professionals constitutes a challenge to the power of economic capital. Redistribution, various efforts by the state to counter inequalities and the expansion of various state services, especially in the intersection between welfare services and the labour market, lessen the potency of economic divisions. In extension of this, such efforts attack the power of economic capital, specifically in terms of transactions on the labour market. Hence, cultural capital leftism may be interpreted as a strategy in the struggles in the field of power (see the brief discussion in Lamont 1987).⁸

The claims of the death of class were fuelled in part by the difficulties of accounting for contemporary politics with the old-fashioned class schemes. I have shown how this can be more satisfactorily dealt with by drawing on Bourdieu's conception of social space and forms of capital. However, this raises a fundamental question: what is the relation between these concepts and class theory?

The service class has distinct market situations and employment relationships in common, but this manifestly does not ‘translate’ into a unified class along the lines suggested by Goldthorpe. While the groups that make up the service class are similar in these respects, they are situated in different fields – (higher) education fields, the bureaucratic field, the economic field and so forth – and thus they rely on different forms of capital. Because of this, the service class becomes systematically fractured by the principle of capital composition. Certain fractions of the service class are privileged primarily through their possession of cultural capital, whereas others base their position on the possession of, and/or control over, economic capital. Goldthorpe's differentiation between the higher and lower echelons of the service class, while certainly important, is insufficient to account for its systematic heterogeneity.

In this paper, I have shown that Bourdieu's theoretical and methodical toolbox can be used to account for the fractioning of the service class. However, sociologists who take their main lead from Bourdieu pay little attention to the types of issues that animated the theorisation of the service class. The force of Goldthorpe's theory was his account of the genesis of the particular positions in the social structure, through an interpretation of the development of Western capitalism in the twentieth century – particularly the rise of impersonal ownership, increased bureaucratisation (including the internal division of labour) and an increased demand for specialised expertise. Such an account cannot be replaced by describing the middle class in terms of their capital portfolios. In my view, this points to a need to combine core features of ‘traditional’ class analysis – of which the theory of the service class is an excellent example – with the ‘Bourdieusian’ emphasis on fields, capitals and the multi-dimensionality of class (Flemmen 2013). One way of conceptualising this is to see it as the intersection of class structuration (Giddens 1981: 105–12) with the dynamics in different fields. This, however, is an issue that calls for more work.

I would like to thank Marianne Nordli Hansen and Vegard Jarness for important comments on the paper, and Johs Hjellbrekke for crucial critical feedback on the analysis. I am grateful to the participants at the SCUD workshop on Field analysis, boundary drawing and socio-cultural inequality in York 23–25 May 2011 for comments on an earlier version. Gitte Sommer Harrits and Tor Bjørklund gave generous advice on literature. I would also like to acknowledge the crucial inspiration from Harrits, Annick Prieur, Lennart Rosenlund and Jakob Skjøtt-Larsen.

Magne Flemmen (b. 1980) is Researcher at the Department of Sociology, University of Bergen. His research interests are in class theory, social stratification and sociological theory.