Copyright Sociological Research Online, 2001


Erik Bihagen (2001) 'The Plausibility of Class Cultural Explanations: An Analysis of Social Homogeneity using Swedish Data from the Late 1990s'
Sociological Research Online, vol. 5, no. 4, <>

To cite articles published in Sociological Research Online, please reference the above information and include paragraph numbers if necessary

Received: 18/9/2000      Accepted: 14/12/2001      Published: 28/2/2001


Previous findings of large absolute mobility to service I class (the "upper" part of the "salariat") can be seen as a sign of the implausibility of class cultures. However, it is argued that these findings might be due to inappropriate divisions of class. Using Swedish data, and following a Weberian definition of class, a social class schema is derived empirically from marriage tendencies. Social homogeneity (immobility and in-marriage) is found to be relatively large in the working classes and in certain subgroups of service I. One interpretation of this, and the fact that there are few inter- marriages and a low level of mobility between the working class and these subgroups of service I, is that class structure might be bipolar such that the extremes are upholders of certain norms and cultures. The possible upper classes of service I need to be better operationalised in future research. Thus, since class cultures are plausible, and since individualistic rational action theory appears to be insufficient for explaining all possible class differentials (as earlier research has indicated), future class analysis might better rely on both rational action theory and class cultural explanations.

Class Culture; Marriage; Mobility; Rational Action; Social Homogeneity


Goldthorpe (1996: 487) argues that class theory should be reoriented away from the hitherto dominant cultural explanation based on 'distinctive class values, norms, "forms of consciousness' or other supposed aspects of class cultures or subcultures' and towards Rational Action Theory (RAT). This paper sheds light on one problematic empirical finding for cultural explanations, i.e., that classes are socially heterogeneous (Breen & Rottman 1995; Goldthorpe 1987; Jonsson & Erikson 1997. C.f. Payne 1987), and maybe even too heterogeneous to allow the maintenance of specific norms and values (Breen & Rottman 1995: 113; Featherman et al. 1988: 69; Goldthorpe 1996: 486-487). Using Swedish data on mobility and marriage patterns, I have investigated whether an empirically derived class schema might produce a generally higher level of social homogeneity and/or whether dividing the most troublesome class in terms of social homogeneity (service I) might solve the problem of heterogeneity. In conclusion, the relevance of class cultural explanations will be assessed at a more general level.

Why bother trying to solve problems using the class cultural approach; why not simply turn to individualistic explanations (e.g., RAT) of class differentials instead? The latter approach has namely been a successful turn in terms of providing plausible explanations for class differentials (Breen & Goldthorpe 1997; Breen & Rottman 1995; Erikson & Jonnson 1996; Evans 1993). However, first, it seems as though some practices, such as educational choices, cannot be explained solely in terms of rational actions (Hatcher 1998). Further examples concern class differences in consumption such as smoking habits, TV preferences and even culture consumption in general (Bihagen 1999; Bihagen & Katz-Gerro 2000), which appear to require cultural explanations, i.e. assumptions about class specific norms and values. Second, a class theory without cultural and communal aspects would seem to be a very limited project (C.f. Devine 1998; Scott 1996a), and actually recurrent news of the death of class (Clark & Lipset 1991; Nisbet 1959; Pakulski & Waters 1996) is largely related to such aspects. Thus, if we are to test this dramatic assumption, class cultures should not be abandoned a priori (Devine 1998:32).

The focus on service I (Erikson & Goldthorpe 1992a), 'the upper salariat', is based on an interesting paradox found in earlier research. This class seems to be most lacking in socialisation capacity as relatively many of its members have their origins in the other classes (Goldthorpe 1987; Jonsson & Erikson 1997: 507). At the same time, this class reveals the most distinct behaviour in terms of educational attainment, relative likelihood of immobility, political attitudes, and consumption (for Sweden see in order: Müller 1996: 176; Jonsson & Erikson 1997: 503; Svallfors 1995; 66-67; Bihagen 1999). Since service I is economically advantaged compared with other classes (Bihagen & Halleröd 2000), everything speaks in favour of a rational action explanation for its 'members'' behaviour. However, results showing social heterogeneity may be caused by a methodological artefact that will be studied in this paper.

The choice of using Swedish data is mainly pragmatically based, although Sweden is an interesting case in terms of class. Sweden has a long tradition of a relatively egalitarian income distribution (C.f. Erikson & Jonsson 1996; Esping-Andersen 1990; Goldthorpe 1996). From a Rational Action point of view (see more below), we would thus, on the one hand, expect to find comparatively tiny traces of class actions in the Swedish case. On the other hand, the level of union membership, class voting and even the awareness of classes is high in Sweden (In order: Thålin 1996: 105; Oskarsson 1994; Cigéhn 1999: 123). These could be seen as signs of strong class cultures in Sweden, although we can only speculate. Thus, the conclusions drawn here are not necessarily valid for other countries. However, classes in different Westerns countries appear to be approximately equal in terms of social heterogeneity (Breen & Rottman 1995: 113). Hence, this specific problem of cultural explanations is valid for several countries.

The Definition of Class and its Boundaries

Within class analysis, we risk committing the fallacy of reification (Frankfort- Nachmias & Nachmias 1996: 27), i.e., treating class schemas as if they completely mirrored the real phenomena of class. Although class schemas will remain abstractions of reality, careful definitions of class will improve our ability to draw more accurate class boundaries using empirical data, which in turn could increase the social homogeneity within the classes. Following John Scott's (1996b) rework of a Weberian definition of class, I will argue for a definition of class based on both economical and social aspects, which will hopefully help us to find accurate class boundaries. These aspects can also be connected with the two main explanations in class analysis: the RAT explanation and the cultural explanation.

Economic Aspects of Class and the RAT Explanation

First, class differences arguably presuppose differences in economic well-being and risk differentials on the labour market in a population, i.e., inequalities in terms of economic conditions (Devine 1997; Marshall et al. 1997; Sørensen 1991; Westergaard 1995). Second, these differences must be coherently structured by occupational characteristics, among which the most important are distinctions between the self-employed vs. employed and among the employed - distinctions related to typical educational credentials as well as the division between manual and non-manual jobs. The focus of these occupational characteristics is not very clear in Weber's writings, but is at the heart of contemporary class schemas like Goldthorpe's (Crompton 1993: 50-53).[1] These clusters of similar occupations with similar market positions (i.e., income, unemployment risk, career opportunities) can be labelled as economic classes. The underlying thought behind this definition of class is that these job characteristics generally determine the market position of the people holding them. Without an empirical validation of these occupational characteristics' links to market positions, these groups are nothing more than occupational groups (referred to as OCG below).

The assumption of economic differences (including risks and prospects) between the classes has been viewed as sufficient to make RAT explanations work (Breen & Rottman 1995; Evans 1993; Goldthorpe 1996; C.f. Devine 1998). The uneven distribution of economic resources and risks on the labour market leads to individuals being exposed to different constraints on their actions. The individual's choice of action is, thus, a result of an estimation of what can be achieved in a cost-efficient way, given these class-dependent constraints and prospects (Breen & Goldthorpe 1997; Breen & Rottman 1995; Erikson & Jonsson 1996; Evans 1993; Goldthorpe 1996). A characteristic feature of this type of explanation is that it does not assume class-specific preferences, values or norms as independent factors underlying actions and behaviours.

Since there is evidence of economic classes (for Sweden see Bihagen & Halleröd 2000), the basic requirement of RAT explanations within class analysis appears to be fulfilled. Occupational groups have consistent positions in a hierarchy from the working class with the weakest market position in terms of wage, income and unemployment risk to service I with the strongest market position, although the exact economic boundaries between the classes are difficult to specify (Bihagen & Halleröd 2000). Nonetheless, RAT explanations for class differentials focusing on individuals' economic constraints and prospects appear to have a sound basis. We now turn to the more precarious situation of social aspects of class and cultural explanations.

Social Aspects of Class and the Cultural Explanation

Class formation can be seen as the process and the result of social circulation being limited in range from one's class position (Scott 1996b).[2] There has been an emphasis on the role of a specific kind of class formation, i.e., inter-generational (parents to children) immobility, for the emergence of cultures within classes (Breen & Rottman 1995: 113; Goldthorpe 1996: 486-487; Marshall et al. 1988: 21-22). This reasoning can be reconstructed by three hypotheses:

H1, Groups with a low degree of social circulation tend to develop and maintain group-specific norms and values, i.e., cultures.

H2, People stay within their class to a large degree, i.e., they are inter-generationally immobile.

H3, Therefore, different classes tend to develop and maintain different cultures with related ways of behaving.

Thus, if we are to provide a cultural explanation, it is not enough to assume the existence of economic classes, but we also need to make assumptions about social circulation, i.e., about certain social aspects of class. The dismissal of the cultural explanation, to different degrees, appears to be due to findings inconsistent with H2 (Breen & Rottman 1995: 113; Goldthorpe 1996: 486-487), i.e., too many leave their class of origin.

However, the focus on immobility ignores the possibility of processes of secondary socialisation, i.e., it assumes a large degree of stability in values from childhood and on. For instance, the most troublesome class in terms of having a large proportion of 'outsiders', service I, has a specific feature in that many members share a life trajectory consisting of many years at university, which might unify them in terms of values and norms (C.f. Bourdieu 1984; Erickson 1996). Thus, there is a possibility of shared values and norms even in a class containing many 'outsiders', or more correctly, those with other class origins.

Whereas it might be possible to belong to a cultural group without originating from it, it is arguably unlikely that one is married outside the group, or at least not to a member of another cultural group (Merton 1941; Weber 1968: 932-938). Following Merton, it is impossible to say that there are two cultural groups when inter-marriage between them is frequent, since the norm of each cultural group is in-marriage. Thus, frequent inter-marriage between, e.g., workers and service I would be a more severe problem for the cultural explanation than would frequent working class origin within service I, even though immobility probably works in favour of the emergence and maintenance of class cultures as well. Marriage patterns are, thus, an important aspect of class formation (Scott 1996b).

In sum, the concept of social circulation, encompassing mobility, marriage and even friendship patterns, is probably an essential factor underlying the development and maintenance of norms and values within a specific group and, thus, essential for cultural explanations of behaviour to work. For the case of class, these processes of social circulation can be labelled as class formation. Finally, classes with tendencies for in-group social circulation can be called social classes (Scott 1996b).

BoundaryProblems and their Momentary Solutions

The number of classes constructed definitely affects the social homogeneity of classes. It is generally the case that the fewer groups distinguished, the higher the level of social homogeneity. To avoid constructing groups completely arbitrarily, empirical data can be used as a guide. Since the cultural explanation is at stake here, the question is how to cluster potential social classes. As a starting point, a classification of occupations (OCGs) will be used. If it is assumed that genuine groups in-marry, i.e. marry within the group, rather than inter-marry, i.e. marry outside the group (Merton 1941), marriage tendencies can be used to map out the social boundaries.[3] In the next step of the analysis, the degree of limitations in social circulation will be assessed. If it is then discovered, for example, that service I appears to be part of a larger social class containing other OCGs, social homogeneity will increase. Thus, earlier results pointing towards heterogeneity might be artefacts due to division of actual social classes.

However, the opposite has also been suggested, i.e., that more classes must be constructed than exist in ordinary class schemas (Grusky & Sørensen 1998: 1204). Focusing on service I, Goldthorpe (1987: 46) found that among high income earning service I class members, social homogeneity in terms of mobility was much higher than in the service I class as a whole. One interpretation of such findings is that the EGP (EGP: Erikson-Goldthorpe-Portocarero) schema does not define a real upper class (Breen & Rottman 1995; Crompton 1993; Savage et al. 1992; Scott 1996b), neither in economic nor in social terms. In an attempt to delineate such upper classes, the present study focused on the features of occupation, education and region, assuming that those in the upper class hold certain jobs characterised by high educational requirements and perhaps even located in certain parts of the country.

Since we do not know whether all OCGs - or even less whether sub-OCGs within service I, as indicated by education and a more fine-graded variable of occupation - are distinct economic classes, we cannot know whether they actually are social classes. Using Weberian terminology, there are probably many categories of status groups with a limited social circulation based on any one 'horizontal' feature (Bourdieu 1984; Hansen 1995; Lamont 1992; Weber 1968: 932-938). Hence, groups with a low degree of social circulation within service I may be better described as status groups, based on non-economic differences between professions or even educational qualifications, than social classes. The implication of this will be further discussed in the conclusions.

Research Questions

The following two questions will be addressed in this paper:

  1. How do different occupational groups (OCGs defined by a class schema close to EGP), aggregated according their marriage tendencies, differ in social homogeneity in terms of absolute in-/inter-marriage and im-/mobility?
  2. How do different subgroups (indicated by sub- categories of occupations, educational qualifications, and place of residence) of service I differ in social homogeneity in terms of absolute in-/inter-marriage and im-/mobility?

By addressing question (i), we can see if service I appears to be part of a larger social class with greater social homogeneity both in terms of mobility and marriage. This is made possible by using an empirically derived social class schema. By addressing question (ii), the opposite idea will be evaluated, i.e. that service I covers several socially homogeneous groups, which could be social classes (based on differences in market-positions within service I) or status groups (based on non- economic differences within service I).


The Data Set

The data used are from Statistics Sweden's (SCB) Survey of Living Conditions (Häll & Vogel 1997). The survey data have been collected via face to face interviews. They are suitable since they cover the 'class' of the respondent, of the spouse and of both parents.[4] The sample represents the Swedish population between the ages 16 to 84, and the average response rate is about 75 - 80 percent. To achieve a sufficient number of cases, four data sets (1994, 1995, 1996, 1997) were merged. Since the class schema is not well suited to assigning a class position to the unemployed, only those who were gainfully employed or self-employed have been included (e.g., the unemployed, retired etc. are excluded).[5] Respondents who were farmers have been omitted as well (more on that below). [6] Thus, all in all a large part of the data set is omitted, especially in analyses that require data on spouses, since respondents married to an unemployed or retired person have been excluded as well.[7]

The Variable of OccupationalGroups - OCG

The most important variable is 'class'. Actually, the variable concerns OCGs, and these are economic classes first when certain empirical outcomes are found to be related to market positions; this, however, has been largely validated in previous research (Bihagen & Halleröd 2000). In the data set, the coding used is the Swedish Socio-Economic Code (SEI) (Statistics Sweden 1995). This is not easily completely translated to the internationally well-known EGP (EGP: Erikson-Goldthorpe-Portocarero) schema for the respondent, the respondent's spouse and both parents of the respondent, although SEI is very close to the EGP schema (Erikson & Jonsson 1996:74). There are some differences, for instance concerning foremen, who are not included within the working class in the SEI code (See Diagram 1 for a broad comparison). However, the present inductive method of discovering the boundaries between social classes clearly works to our advantage, since we will learn empirically to which potential social class foremen seem to belong.[8] For the same reason, we will initially work with as many OCGs as possible, while ensuring that cell frequencies do not become too small. Consequently, the working class is divided into four groups, as a function of both general skill level and 'sector' (production and service). The intermediate class is divided into two OCGs: lower, non-manuals II, and higher, non-manuals I. The service class is divided into three groups: service II, service I-managers, and service I-other.[9] Moreover, there are two types of self-employment: self-employed without employees and with. Farmers had to be excluded from the data set since they are too few and there are very few female farmers. Compared with EGP, there are more categories of workers according to this system, and service I is divided into two groups by extracting the group of managers.

Diagram 1: A Comparison of the Present "Class" Schema (SEI) with the EGP Schema
EGP title and code SEI title and code'Typical' occupations (SEI)
Service I (I)Service I- managers (57)Managing director
Service I-other (56; 60)Physician, journalist
Service II (II)Service II (46)Nurse, police sergeant
Routine non-manual higher (IIIa)Lower non-manual I (36)Foremen, stationmaster
Routine non-manual lower (IIIb)Lower non-manual II (33)Office employees in general
Supervisors of manual workers (V)Lower non-manual I (36)Supervisors
Skilled workers (VI)Skilled workers; goods (21)Metal worker; skilled
Skilled workers; service (22)Assistant nurse, hairdresser
Semi/unskilled workers (VIIa)Non- skilled Workers; goods (11)Metal worker; non-skilled
Non-skilled Workers; service (12)Nurse's assistant, shop-assistants (generally)
Agricultural workers (VIIb)Skilled or non-skilled workers
(12 or 22)
The title reveals the kind of jobs
Small proprietors with employees (IVa)Self-employed with employees (76)The title reveals the kind of jobs
Small proprietors without employees (IVb)Self-employed no employees (78) Except independent professionals (SEI code 60)The title reveals the kind of jobs
Farmers (IVc)Excluded from these analyses.The title reveals the kind of jobs

The titles of SEI are adapted to fit the labels of the EGP schema. The similarities between EGP and SEI are not total for the equivalent groups. Among employees, requirements of education generally increase the higher up in the diagram one moves (Statistics Sweden 1995).

The problem of how to measure class in cross-class families is often referred to in class analysis (Axelsson 1992; Breen & Rottman 1995; Crompton 1993; Erikson 1984; Erikson & Goldthorpe 1992c; Leiulfsrud 1991; Scott 1996b), and is also relevant here in the analysis of class origin. The present solution is aimed at reflecting the complexity of cross-class origin - complexity that is lost using approaches such as dominant class in the household as well as class of the father (Erikson 1984; C.f. Axelsson 1992). Given a cultural approach to class analysis, a person with parents from two classes has had two cultural influences or at least a specific cross-class cultural background. Thus, class origin will be measured in terms of 'at least one parent from class x' and so on, and the frequency of cross-class origin will be indicated as well.

Additional Variables

Sex: In the analysis of in-marriage (as well as in the analysis for question ii), the variable of sex will be used, since the differing OCG distributions of men and women greatly affect the possible level of class homogeneity. As an example, the smaller proportion of women in service I make it necessary for men to marry outside this class more often. Social homogeneity in terms of class origin is not similarly sex biased by necessity.

Occupation: For question (ii) a variable of occupations (NYK 83) is used (Statistics Sweden 1995). In order to avoid disturbing the results with very diversified miscellaneous categories, all individuals who can not be included in the six large occupational categories within service I have been excluded. These categories cover 82 per cent of those with an occupation in service I and are (1) technical, chemical, biological occupations; (2) pedagogical occupations; (3) juridical occupations; (4) medical occupations; (5) administrative occupations, and (6) commercial occupations.[10] Some interesting categories that were excluded because they were too small were journalists, religious workers, social workers, and psychologists.

Education: Since the interest for question (ii) is in service I - the OCG with the highest level of education - a variable of education is used that captures a relatively extreme level of education. Longer university experience is defined as three years or more.

Region: In order to detect socially homogeneous subgroups (question ii) within service I, a variable of region is used. There are at least two reasons for including this variable. Status groups could be based on similarities in geographical residence and economic classes could be more socially homogeneous in certain geographical areas due to a regional bias in the occupational structure (Merton 1941). This variable (Statistics Sweden's variable 'H-region') covers: living in the Stockholm area, in the largest cities of the south (including Gothenburg), other large cities, other urban areas in the south, rural areas in the south, other urban areas in the north and rural areas in the north.

Finally Age is used as a control variable (ratio level) when addressing question (ii), since social homogeneity in general might be connected to both life-cycle variations and differences between generations.


The Social Homogeneity of Occupational Groups Defined by a Class Schema

In order to answer question (i), we must know which OCGs to include in the 'potential' social classes, i.e., where the social boundaries are located (Scott 1996b: 30, C.f. Goldthorpe 1987: 115). In this step, the goal is to determine which OCGs 'tend to go together'. From the discussion above, it appears as though marriage patterns are appropriate indicators of social boundaries. Thus, we can calculate distances between these groups such that if people belonging to OCG x are as likely as people in OCG y to marry people in x, then there is no social boundary between x and y. From a cross-tabulation of male and female spouses, distances are calculated for all pairs of aggregates (11 OCGs result in 55 pairs of OCGs: n*(n-1)/2, where n is the number of OCGs).

The distance between each pair of OCGs is calculated as a standardised odds ratio, Yules Y (Jonsson & Erikson 1997: 500):


a, b, c & d are the four cells of a 2*2 sub-table of the larger table.[11]

a = the cell that corresponds to value 1 of both variables.

d = the cell that corresponds to value 0 of both variables.

b & c = the two cells that correspond to values 1 and 0.

The variation is between - 1.00 and 1.00, where zero means no association and (+/-) 1.00 means a complete association. A positive value indicates that there is a tendency for in- marriage, whereas a negative value indicates the opposite. 'Zero' indicates that people in these two OCGs are as likely to marry outside of their group as within it. Hence, these two groups should be regarded as one group rather than as two. As a matter of fact, all values were positive in the matrix of distances analysed here. Thus, there is always a tendency towards in-marriage for the OCGs investigated, although it is occasionally very weak.

Using MDS (Multi Dimensional Scaling), these distances were, as optimally as possible, transformed into a two-dimensional map in order to indicate where the social boundaries are found between the OCGs (Diagram 2). In order to see how well the MDS solution transforms the original distances, measures of goodness of fit are often used. An alternative solution to this problem is to identify which 'case' or 'cases' disturb the map. In Diagram 2, the lines represent 'true' closeness i.e., the Yules Y measures calculated from a cross-tabulation, (Click here for Cross Tabulation), while the actual distance seen on the map is the MDS solution of adjusting all distances to a two-dimensional map. Using the software Bibexcel, lines of varying boldness (the bolder the closer) are drawn.[12] In the case of this diagram, in order to facilitate interpretation of the map, the lines describing the weakest links (the longest distances) are not drawn. Most of the lines here show that the map is successful in transforming the original distances into two dimensions, since the thick lines are generally found between OCGs that are close to each other according to the map. However, the two self-employment categories disturb the map, since they are situated near each other only due to similar distances to other OCGs and not because they are actually close to each other. If these OCGs are omitted, the map represents the distances almost perfectly with a success rate of 94 per cent; if they are included, the rate drops to 82 per cent (as indicated by a Kruskal S- stress).

Diagram 2: Tentative Map of the Swedish Social Class Structure
(MDS solution of distances calculated as Yules Y from a marriage table. See appendix)

Map of the Swedish social class structure

Comments: "Service I-other" is labelled "service I" and "service I-managers" is labelled "managers". The boldness of the lines represents the 'true' distances, i.e., as measured by Yules Y, while the actual distances on the map are the MDS solutions of adapting the 'true' distances to a two dimensional map. Only the smallest true distances are shown as lines. The size of the circles corresponds to the size of the classes.

An important conclusion from the map is that the social relations between OCGs appear to be hierarchical and that, in general, OCGs can not easily be merged into potential social classes. The hierarchical feature of the map is that the working classes (the red coloured circles) are quite close to the intermediate classes (the yellow circles, including service II), which in turn are quite close to service I (two circles in blue), whereas the working classes and service I are far apart. This hierarchy corresponds both to the hierarchy of OCGs in terms of general qualification requirements (Statistics Sweden 1995: 248), and in terms of actual market positions as indicated by measures of income and unemployment risk (Bihagen & Halleröd 2000). The OCGs easiest to merge are the two service I categories (in blue), whereas all other OCGs seem to be difficult to merge with other OCGs. The two service I categories are very close to each other, while being relatively far from other OCGs (C.f Goldthorpe 1995: 320- 321).[13] In the EGP class schema, these are already 'merged' into service I (see Diagram 1); thus this practice is empirically supported. However, the map does not support the construction of one service class, including both service I and II, although this has been done earlier (although for practical reasons; see Erikson & Goldthorpe 1992a; 1992b).

A test of significance (Chi2) between all OCG pairs indicates that only 2 of these 55 correlations are non-significant at the 95 % level, whereas the other correlations are at least significant at this level (these figures are not shown).[14] The two OCG pairs for which a significant level of in-marriage can not be established are between the two categories of working class in the production sector and between the two categories of service I. These results are coherent with the map, especially for the two service I categories, which are very close on the map. Workers in the production sector are close, although they appear to be somewhat forced from each other on the map in relation to their true closeness.

Since the intermediate classes can not be merged with service I, we can expect to reproduce the result of a considerable absolute mobility between service I and the other 'social classes', since the present schema will be similar to those used earlier (Goldthorpe 1987; Jonsson & Erikson 1997). However, in the present study, and in contrast to previous analyses, both parents' occupations count for origin, which is more appropriate (see above). Another difference in this paper is that social circulation is not restricted to mobility, but we will also look at homogeneity in terms of marriage patterns.

Given that OCGs should not generally be merged, as indicated by the previous analysis, and that the social distance between the working classes and service I is relatively large compared to all other distances, it is relevant to see whether there are many cross-class phenomena between these two extremes of the social class structure. Table 1 shows the percentages of working class origin and service I origin for each of the different occupational groups (absolute inflow percentages: Hout 1983).

Table 1: Working Class Origins, Service I Origins and Cross-Class Origins for Different Occupational Groups (percentages)
Having at least one parent from service IHaving at least one parent from the working classHaving cross-class origins: working class and service I
Non-skilled workers - goods5671
Skilled workers -goods5681
Non-skilled workers - service6621
Skilled workers -service8581
Lower non-manuals II10522
Lower non-manuals I11521
Service II13451
Service I - other25291
Service I - managers24340
- without employees7420
- with employees8391

Comments: The table consists of three cross-tabulations and it is possible to have one parent from service I and one from the working class, although this is rare, as seen in the last column. * p<0.05 **p<0.01 ***p<0.001.

First, the table indicates that very few have grown up in a family with one parent from the working class and one parent from service I.[15] Second, the table clearly reveals the hierarchical order between the OCGs. Working class origin is less common the higher the class of the respondent, whereas the opposite is true for service I origins. Somewhere around 60- 70 per cent of workers have working class origins, whereas only around 5- 10 per cent of workers have origins in service I. The working classes, seen as one unit, are clearly the most homogeneous. In service I, around 25 per cent have an origin within that OCG, whereas actually somewhat more have one working class parent, which has also been shown in previous research (Goldthorpe 1987; Jonsson & Erikson 1997: 507). At the same time, it is also evident that most of the service I class 'members' originate in non-manual classes, since around 70 per cent do not have origins in the working classes.

From the cross-tabulation used above for tracking the boundaries between classes (Click here for Cross-tabulation), percentages of women and men married to a spouse in one of the different classes are calculated (Table 2). The table clearly indicates a hierarchical aspect of class. For the female workers, around 65-50 per cent are married to a man in the working class, whereas this is true of only around 10 per cent of women in the service I classes. A slight majority of service I women are married to a service I man, whereas very few working class women are. From the perspective of males, the largest deviation from this pattern is that it is much less common to be in- married within the service I, probably due to a shortage of women within this group. All these differences are less drastic for men, but still similar. Finally, the self-employed have an intermediate position between service I and workers with respect to both origins (Table 1) and marriage patterns (Table 2).

Table 2: Women and Men by Different Occupational Groups and Marriage to a Working Class vs. Service I Spouse (percentages)
Married to a spouse in service IMarried to a spouse in the working classMarried to a spouse in service IMarried to a spouse in the working class
Non-skilled workers - goods566266
Skilled workers -goods660359
Non-skilled workers - service855460
Skilled workers -service951550
Lower non-manuals II1439547
Lower non-manuals I2130739
Service II29221131
Service I - other55103015
Service I - managers5172712
- without employees19211032
- with employees157728

Comments: The table consists of three cross-tabulations made from the table in the appendix. * p<0.05 **p<0.01 ***p<0.001.

To sum up this part, we must interpret these results in light of question (i). People in the working classes have a relatively high level of immobility and a rather high level of in-marriage. Service I (including 'service I-other' and 'service I-managers') has a relatively low degree of immobility, and only a reasonably high level of in-marriage among women (more than 50 per cent). It is shown once again that more people within service I have an origin in the working classes as opposed to service I itself. Due to the social boundaries found, this service I should not be merged with other OCGs, which could have increased the homogeneity. However, the present results revealing a strong social relations hierarchy (mobility and marriage tend to be restricted to close OCGs in terms of qualification level and complexity of work tasks) indicate possibilities for reviving the idea of class cultures, which will be discussed at the end of the paper. First, however, the question of subgroups within the service I class will be addressed.

The Social Homogeneity of Subgroups within Service I

The subsequent analysis deals with question (ii), i.e. whether subgroups with a high level of social homogeneity can be found in service I. It is possible that this homogeneity is larger within certain service I occupation groups (C.f. Grusky & Sørensen 1998; Hansen 1995) or among those with higher levels of education (Hansen 1995:91; Müller 1996: 176). Region could also be an interesting variable (see above). In order to identify these subgroups and point out the possible internal social differentiation within service I, it would have been optimal to make a map like Diagram 2 for only those within service I, using combinations of these new variables. However, the empty cells in such a matrix are too many since the number of cases is relatively small. Instead, four logistic regressions have been made with (1) having at least one parent in service I, (2) having at least one parent in working class, (3) being married to a spouse in service I and (4) being married to a spouse in working class as four different dependent variables.[16] Logistic regressions are suitable since the dependent variables are binary, e.g. either you have a parent in service I (coded as 1) or you do not (coded as 0). The data set is restricted to respondents within service I, allowing us to see whether there are any indications of socially homogeneous subgroups within this OCG.

The independent variables of occupation, education, region (all three as dummy variables) and the two control variables of age and sex are used. It was not possible to distinguish the effects of occupation from the effect of education, since all respondents in the medical sector turned out to be physicians with longer university education. Thus, two different regressions were performed for each dependent variable (Model A with medical sector, and model B with the variable of education). From initial analyses with all independent variables and the four dependent variables, three of the independent main dummy variables created coherent patterns of social homogeneity, i.e. a higher value for the two 'service I variables' and a lower value for the two 'working class variables'. The effects of these independent variables can be seen as indicators of some socially homogeneous subgroups within service I, although this analysis is not intended to be exhaustive (see Table 3). Moreover, the control variable of sex produced some significant effects.

Table 3: Coefficients for logistic regressions with four dependent variables indicating immobility/mobility and in-marriage/inter-marriage (the data set is restricted to respondents within service I)
1.Origin in service I2. Origin in working classes3. Spouse in service I4. Spouse in working classes
Model AModel BModel AModel BModel AModel BModel AModel B
Occupational sector
(Constant: other)
Medical sector1.22***- -.61 ns-.87**- -2.13*-
Educational background
(Constant: Lt 3 years university)
At least 3 years at university-1.25***--.84***- 1.11***--.96***
(Constant: Other)
Stockholm area.77***.74***-1.02 ***-.93***.49***.38**-.70**- .58*
Sex (Constant: Male)
Female.17 ns-.05 ns-.45**-.13 ns1.26***1.04***-.92***-.61**
Constant-1.55-2.28-.37-.02-1.10- 1.77-1.53-1.23

Comments: There is a large non-response rate on the variable of social origin. For the regressions with 'spouse' non-cohabiting people are omitted. * p<0.05 **p<0.01 ***p<0.001.

From these regressions it is clear that physicians, those with long educational careers and those from Stockholm, have much higher social homogeneity than others within service I. [17] The importance of education as a determinant behind marital choice has been shown in earlier research as well (Uunk 1996: 72, 134). The effect of sex is probably due to the sex bias in occupation, whereby more men are in higher positions and, thus, women more easily find a spouse within this OCG. An estimation of proportions from these regressions indicates, for example, that among physicians in the Stockholm area, over 60 per cent have origins within the service I, around 60 to 80 percent (the higher level for women) are married within this OCG, only around 10 percent have a working class parent and even fewer have a spouse in the working class.[18] This can be compared with the considerably lower level of social homogeneity in service I in toto (Table 1 and 2). To sum up the results from these regressions, we can conclude that the social homogeneity of service I varies considerably; the present data indicate this variation, but do not describe it in depth. In other words, there are signs of socially homogeneous subgroups within service I, indicating the existence of status groups and/or social classes within service I.


An intuitively sound reason to doubt class cultural explanations, mentioned more or less in passing (Breen & Rottman 1995: 113; Featherman et al. 1988: 69; Goldthorpe 1996: 486-487), is that classes are socially heterogeneous. The class with the most distinct behaviour - service I - appears to be least socially homogeneous (see above), which would seem to make cultural explanations of that behaviour untenable. However, the findings of this paper, indicating large social differentiation within service I as well as considerable social distance to the working classes, point to another conclusion. Service I appears to encompass relatively homogeneous social cores, i.e. signs of social classes or status groups within service I, which may as well be cultural cores. Hence, the cultural explanation should not be abandoned a priori due to findings about social heterogeneity. The social heterogeneity of classes may very well be an illusion.

The findings interpreted with the theoretical framework offered point to the possibility of a bipolar class structure, where the class schema (the OCGs) used does not succeed in extracting economic upper class(es) from the occupational group of service I. Earlier research indicates that the classes are hierarchically ordered in terms of market positions, with the working classes as the least advantaged and the upper classes as the most advantaged economic classes, hidden under the umbrella of service I (Bihagen & Halleröd 2000). The findings of this paper indicate a corresponding social hierarchy of these economic classes, and even the possibility of several social classes within service I. The economic extremes are represented by at least two social classes with a high degree of social homogeneity. The intermediate classes are found between these two poles (including some occupations within service I). These classes might serve as a kind of socio-cultural buffer zone into which both inter- marriage and mobility from the extremes are quite frequent. Given the assumption that in-group cultures develop by social isolation from other such groups, this would make class cultures possible.

However, it might be the case that it is more appropriate to understand social stratification in terms of a continuum of economic class positions paralleled by a social continuum, instead of social and economic classes with clear boundaries to each other. Social circulation would then be common between close economic class positions and uncommon between distant economic positions, which the findings of the low level of social circulation between the working class and service I indicate. In other words, a gradational disaggregated model may be preferable to a categorical disaggregated class schema (Grusky & Sørensen 1998: 1190). Based on evidence of mobility derived using the Cambridge Scale of Occupations, Prandy (1998) has argued for a gradational disaggregated schema, while abandoning the concept of class. Since social circulation between the poles of the social class hierarchy is very limited, such a gradational schema would still make cultural explanations plausible.

Another scenario is that the present findings indicate the existence of status groups, or combinations of status groups and social classes within the occupational structure, and especially within service I. Thus, certain parts of service I, such as physicians, those with higher levels of education and those living in Stockholm, would be characterised as one or several status groups based on mutual esteem (Weber 1968: 932), rather than as relatively economically advantaged economic classes that constitute social classes. However, even such status groups could be upholders of group-specific norms and values, explaining why those within service I behave in a specific way. Therefore, with the terminology offered here, there might be in previous research on class differentials a conflation between class action and status group action.

An enduring result from the analyses is that social homogeneity is always higher when looking at in- marriage as opposed to immobility. This supports the plausibility of class cultures since frequent inter-marriage would be an anomaly, given the assumption that cultural groups do not generally inter-marry (Merton 1941). The relatively high level of in- marriage, compared with mobility, in the subgroups of service I might indicate that those with working class origins who have moved to service I have generally assimilated the values of these subgroups through processes of secondary socialisation.

I have argued this far that the cultural explanation should not be abandoned based on previous findings of social heterogeneity within higher classes. However, Goldthorpe (1996) put more emphasis on abandoning cultural explanations, since they are not consistent with the finding of relatively constant class differentials in educational attainment over time (the explanandum). According to Goldthorpe (1996: 490), cultural explanations instead predict increased class differentials with the expansion of the educational system, since only service class children can benefit from this (for a similar view, see Erikson & Jonsson 1996: 31). A number of counter-arguments are possible. First, the finding of constant class differentials could be a methodological artefact of the fallacy of reification. It could be the case that the OCGs used are continuously less valid measures of class over time, since the actual job characteristics and market positions of each OCG would increasingly become more varied within each group. Thus, with a valid indicator of class over time, there could be growing class differentials over time. Second, the prevalence of class norms does not necessarily mean that everyone in every class follows the rules (Devine 1998: 35; Merton 1968: 215-248; Scott 1996a: 511). The growth of the educational system can be assumed to have made the transition to higher education easier and, thus, non-conformist working class children will benefit from this, just as do children from other classes. Thus, educational expansion might not necessarily result in substantially growing class differentials as Goldthorpe assumes; this is strengthened by the additional assumption that the upper classes (located within service I) will not benefit much from educational expansion, since almost all of their children already continue to higher education. Third, and finally, it may be the case that class cultures are experiencing some erosion over time, which explains the status quo in class differentials.

Sweden is actually described as a deviant case in terms of this explanandum (Goldthorpe 1996), i.e., until the 1970's, class differentials have diminished in Sweden (Erikson & Jonnson 1996), although some other countries may have had similar experiences (Müller 1996; C.f. Breen & Goldthorpe 1997).[19] The non-culturalist explanation for this is assumed to be an equalisation of conditions (e.g., income) in Sweden over time, due to a Social Democratic regime (Goldthorpe 1996. C.f. Esping-Andersen 1990). However, if we adopt cultural explanations, we could instead stress factors such as a possible decline of the family as a socialisation agent, and investigate why this hypothetical decline may have gone further in Sweden than elsewhere.

A completely different but still interesting explanandum, at least concerning Sweden, is that the OCG of service I appears to deviate considerably from other classes in a wide range of areas, i.e., educational attainment, relative immobility, political attitudes, consumption (In order: Müller 1996: 176; Jonsson & Eriksson 1997: 503; Svallfors 1995: 66-67; Bihagen 1999). Given that the socially homogeneous cores within service I revealed here are characterised by certain values and norms, future research may benefit from more detailed study of the variation in behaviour within service I. However, there also signs of varying economic conditions within service I - variation that is increasing (Bihagen & Halleröd 2000; le Grand 1994;). Thus, if our aim is to compare the explanatory power of RAT and cultural approaches, a finding of varied behaviour within service I can not directly be taken as evidence in favour of one of these explanations. A starting point in dismantling such social and economic aspects of behaviour would be to divide service I into several occupational groups (C.f. Devine 1998: 32; Grusky & Sørensen 1998).

Future research on class differences might benefit from taking class cultural explanations into account and not only relying on RAT theory, which appears to be Goldthorpe's (1996) advice (Cf. Devine 1998). Qualitative research has, for instance, indicated that rational choices only play a partial role in educational decision- making and, thus, RAT would seem to be too narrow for explaining these choices (Hatcher 1998). In order to test if educational attainment, for example, is best explained by rational actions or in terms of culture, the cultural explanation should not be abandoned a priori (Devine 1998: 31). Another similar example is class-related consumption patterns that do not appear to be completely reducible to RAT explanations, e.g. the stronger tendency for service I to visit the theatre even at equal income levels as other 'classes' (Bihagen 1999; Bihagen & Katz-Gerro 2000). Hence, based on the findings of this paper the conclusion is that class cultural explanations are plausible, and based on findings of earlier research; class cultural explanations even appear to be necessary. Finally, I wish to mention again, that I have restricted this paper to the question of social homogeneity, which is arguably a precondition for class cultural explanations, and I do not offer explanations for variations in social homogeneity. A further restriction on my conclusions stems from the use of Swedish data; patterns of social homogeneity might be different elsewhere. In spite of this limitation, however, this paper has shown that cultural explanations are still plausible and, hence, that a complete turn to individualistic explanations in class analysis needs to be reconsidered.


1I deviate from the Goldtorpe-Erikson (Erikson & Goldthorpe 1992a) theoretical definition of class in some respects, but accept their operational definition of class. This may appear to be an odd decision, but the Goldthorpe definition of class has also changed considerably over time, while the operational definition has been maintained (Evans 1992). The theoretical definition of economic class in this paper partly resembles an earlier definition of class by Goldthorpe (1987) and Lockwood (1989: 15).

2Goldthorpe (1987: 330-331) distinguished between class formation in a demographic sense and as a socio-political process. In this paper, class formation is used in the first sense and this also resembles Giddens's (1981: 107-110) use of the concept class structuration.

3These tendencies are equivalent to the concept of relative mobility, i.e. the statistical association between class of origin and class of destination, whereas the concept of social homogeneity used here is equivalent to absolute mobility, i.e. marginal- distribution-sensitive mobility (Breen & Rottman 1995; Hout 1983).

4This could be described consistently in terms of OCG instead of class. However, these aggregates are commonly described in terms of class and class schemas, and therefore I use these terms in this section. Throughout the paper I will also use the combined labels with class for the different OCGs as shown in Diagram 1 (See also the section below).

5Determining the class positions of the non-employed is problematic within class analysis (Breen & Rottman 1995: 92).

6The original data set contains 23,692 responses. When farmers (n=189, see more below), unemployed and other non-working categories are excluded (i.e. self- employed and employees are included), the data set contains 13,622. A further exclusion, which is used in the marriage analysis, excludes all but cohabiting/married couples where both work, and the data set is then based on 7,971 responses. In some analyses the n is smaller due to non-responses.

7In Sweden, it is common for unmarried couples to live together as if they were married. Thus, cohabiting people are treated as married in the present analyses.

8It could be argued that the EGP schema was not intended to locate social classes (Erikson & Goldthorpe 1992a), and thus, these analyses might not be able to indicate the position of foremen (C.f. Scott 1996a, 1996b). Moreover, 'non-manuals I' is not only comprised of foremen.

9The idea was to have independent professionals as a separate group as well, but there were too few women in this group for the analyses. Initial analyses indicated that the independent professionals were more similar to service I than to managers. Both independent professionals and managers are part of service I according to EGP.

10These groups correspond to the following NYK (Nordisk Yrkes Klassificering) codes: (1): codes 00001-02999); (2) codes 03000-03999); (3) codes 05000-05999; (4) codes 10000-10999); (5) codes 20000-29999; (6) codes 30000-39999.

11As an example, the distance between Lower non-manuals I and Managers is calculated from the following sub-table made from the complete marriage table (Click here for complete Cross-tabulation)

Men's class
Low. Non-m. 1Managers
Women's classLow. Non-m. 1(a) 108(b) 48
Managers(c) 5(d) 12

12Olle Persson constructed this soft-ware (, which covers considerably more than the mapping techniques used here. For more information see <>.

13I separated these slightly in order to enhance readability; this is the only manipulation of this kind made on the map. Thus, these classes are even closer than the map reveals.

14Some of these cross-tabulations do have an expected frequency below 5 in some cells, which makes this test troublesome in certain respects. However, this concerns OCGs that are far apart on the map, and especially between managers and some distant OCGs. Thus, this problem does not appear to be of great importance.

15Actually, the number of respondents from cross-class origins is probably slightly underestimated, due to the handling of non-responses. Only respondents for whom no occupation is reported for either of the parents are excluded from the analysis and, thus, respondents who actually have cross-class origins but missing information for one parent are not included in the columns for cross-class origins. From the marriage table (Click here for Cross Tabulation) we see that around 4 percentages of all marriages have this cross-class combination. However, since inter-marriage is increasing over time (Bihagen & Halleröd 2000) and since many adults have grown up in families where the mother was a housewife (Axelsson 1992), we expect to find a smaller degree of this cross-class experience when asking about class origin than when asking about the respondent's own marriage combination.

16It would be more convincing to show the degree of immobility and in-marriage within the subgroups. However, since the subgroups are very tentative - i.e. I do not know if the right social barriers have been constructed - it is enough to show that they are more frequently immobile and in-marry within the OCG of service I.

17The only significant interaction effect was between having a long university education and being a woman in the regression with having at least one parent in service I class as dependent variable. The implication of this interaction effect is that there are no gender differences among those with a higher education.

18For the regressions in Table 3 the log-odds (L) were first calculated as: L =a0 (constant) + b1 'Medical sector (alternatively 'At least 3 years at university')' + b2 'Stockholm area' + b3 'Female'. Then the proportions were estimated as: Proportion =exp(L)/(1+exp(L)).

19The general trend of stability, with the exception of Sweden, appears to be the case for relative patterns of inter-generational mobility as well (Erikson & Goldthorpe 1992a).


I would like to thank Björn Halleröd, Tally Katz-Gerro, Kenneth Prandy, and Stefan Svallfors for their helpful comments. I greatly appreciate the thorough comments of the anonymous reviewers. I would also like to thank Rickard Danell and Olle Persson for advice and assistance with the software Bibexcel.


AXELSSON, C. 1992. Hemmafrun som försvann: Övergången till lönearbete bland gifta kvinnor i Sverige 1968-1981. Institutets för social forskning avhandlingsserie 21. Stockholm: Institute for Social Research.

BIHAGEN, E. 1999. 'How Do Classes Make Use of their Incomes?' Social Indicators Research, 47(2) 119-151.

BIHAGEN, E. and KATZ-GERRO, B. 2000. 'Culture Consumption in Sweden'. Poetics, 27 327-349.

BIHAGEN, E. and HALLEROD, B. 2000. 'The Crucial Aspects of Class'. Work, Employment & Society, 14(2) 307- 330.

BOURDIEU, P. 1984. Distinction: A Social Critique on the Judgement of Taste. Cambridge: Harvard University Press.

BREEN, R. and GOLDTHORPE, J. H. 1997. 'Explaining educational differentials: Towards a formal rational action theory'. Rationality & Society, 9(3) 275-306.

BREEN, R. & ROTTMAN, D. B. 1995. Class Stratification: A comparative Perspective. Hemel Hempstead: Harvester Wheatsheaf.

CIGÉHN, G. 1999. 'Klassidentitet vid seklets slut'. Sociologisk Forskning (1): 121-139.

CLARK, Terry N., & LIPSET, Seymour M. 1991. 'Are Social Classes Dying?' International Sociology. 6 (4), 397-410.

CROMPTON, R. 1993. Class and stratification: an introduction to current debates. Cambridge: Polity Press.

DEVINE, F. 1997. 'Towards a classless society?' In: Jones, H. (ed). Towards a classless society? Pp 201-224. London: Routledge.

DEVINE, F. 1998. 'Class Analysis and the Stability of Class Relations'. Sociology, 32 (1) 23-42.

ERICKSON, B. H. 1996. 'Culture, Class, and Connections'. American Journal of Sociology, 102 (1) 217-251.

ERIKSON, R. 1984. 'Social Class of Men, Women and Families'. Sociology, 18 (4) 501-514.

ERIKSON, R. and GOLDTHORPE, J.H. 1992a. The Constant Flux: A Study of Class Mobility in Industrial Societies. Oxford: Clarendon Press.

ERIKSON, R. and GOLDTHORPE, J.H. 1992b. 'The CASMIN project and the American dream'. European Sociological Review, 8(3) 283-305.

ERIKSON, R. and GOLDTHORPE, J.H. 1992c. 'Individual or Family? Results from two approaches to class assignment'. Acta Sociologica, 35(2) 95-105.

ERIKSON, R. and JONSSON, J. (eds) 1996. Can Education Be Equalized? The Swedish Case in Comparative Perspective. Oxford: Westview Press.

ESPING-ANDERSEN, G. 1990. The Three Worlds of Welfare Capitalism. New Jersey: Princeton University Press.

EVANS, G. 1992. 'Testing the validity of the Goldthorpe class schema'. European Sociological Review, 8(3). 211-232.

EVANS, G. 1993. 'Class, Prospects and the Life-cycle'. Acta Sociologica, 36(3) 263-276.

FEATHERMAN, D,L, SPENNER, K,I & TSUNEMATSU, N 1988. 'Class and the socialization of children'. In: Hetherington E. M, Lerner R.M & Perlmutter M (eds) Child Development in Life-Span Perspective. Pp 67-90. Hillsdale: Lawrence Erlbaum Associates, Inc.

FRANKFORT-NACHMIAS, C. and NACHMIAS, D. 1996. Research Methods in the Social Sciences (5th ed.). London: Arnold.

GIDDENS, A 1981. The Class structure of the Advanced Societies (2nd ed.). London: Hutchinson.

GOLDTHORPE, J.H. 1987. Social Mobility and Class Structure in Modern Britain (2nd ed.). Oxford: Clarendon Press.

GOLDTHORPE, J.H. 1995. 'The Service class revisited'. In: Butler and Savage (eds) Social Change and the Middle Classes. Pp 313-329. London: UCL Press.

GOLDTHORPE, J.H. 1996. 'Class Analysis and the Reorientation of Class Theory: The Case of Persisting Differentials in Educational Attainment'. British Journal of Sociology, 47(3), 481- 505.

GRUSKY, D. B. and SORENSEN, J. B. 1998. 'Can Class Analysis be Salvaged?' American Journal of Sociology, 103(5) 1187-1234.

HANSEN, M. N. 1995. Class and Inequality in Norway. Oslo: Institute for Social Research.

HATCHER, R. 1998. 'Class Differentiation in Education: Rational Choices?' British Journal of Sociology of Education, 19(1) 5-20.

HOUT, M. 1983. Mobility Tables. Series: Quantitative Applications in the Social Sciences. Newbury Park: SAGE.

HÄLL, L. and VOGEL, J. 1997. Välfärd och ojämlikhet i 20-års perspektiv 1975-1995. Stockholm: Statistics Sweden.

JONSSON and J. ERIKSON, R. 1997. 'Klasstruktur och social rörlighet i Sverige under 1900-talet'. In: Häll, L. and Vogel, J. (eds). Välfärd och ojämlikhet i 20-årsperspektiv 1975-1995. Pp 491- 512. Stockholm: Statistics Sweden.

LAMONT, M. 1992. Money, Morals, and Manners. Chicago: University of Chicago Press.

LE GRAND, C. 1994. 'Löneskillnaderna i Sverige'. In: Fritzell, J. & Lundberg, O. (eds.) Vardagens villkor. Stockholm: Brombergs Bokförlag.

LEIULFSRUD, H. 1991. Det familjära klassamhället. Lund: Arkiv.

LOCKWOOD, D. 1989. The Blackcoated Worker (2nd ed.). London: Allen & Unwin Ltd.

MARSHALL G., Rose D., Newby H. & Vogler C. 1988. Social Class in Modern Britain. London: Routledge.

MARSHALL, G, SWIFT, A. & ROBERTS, S. 1997. Against the Odds: Social Class and Social Justice in Industrial Countries. Oxford: Clarendon Press.

MERTON, R. K. 1941. 'Intermarriage and the social structure'. Psychiatry, 4, 361-374.

MERTON, R. K. 1968. Social Theory and Social Structure (5th ed.). New York: The Free Press.

MÜLLER, W. 1996. 'Class inequalities in educational outcomes' in Erikson and Jonsson (eds) Can Education be Equalized? Oxford: Westview Press.

NISBET, R. A. 1959. 'The decline and fall of social class'. Pacific Sociological review, 2(1): 11-28.

OSKARSSON, M. 1994. Klassröstning i Sverige. Nerenius & Santerus Förlag AB.

PAKULSKI, J. & WATERS, M. 1996. The Death of Class. London: Sage.

PAYNE, G. 1987. Mobility and Change in Modern Society. London: The Mac Millan Press LTD.

PRANDY, K. 1998. 'Class and Continuity in Social Reproduction: an Empirical Investigation'. The Sociological Review 46 (2) 340-364.

SAVAGE, M., BARLOW, J., DICKENS, P. and FIELDING, T. 1992. Property, Bureaucracy and Culture. London: Routlege.

SCOTT, J. 1996a. 'Comment on Goldthorpe'. British Journal of Sociology 47(3) 507-512.

SCOTT, J.1996b. Stratification and Power; Structures of Class, Status and Command. Cambridge: Polity Press.

SØRENSEN, A. 1991. 'On the Usefulness of Class Analysis in Research on Social Mobility and Socioeconomic Inequality'. Acta Sociologica. 34(2), 71-87.

STATISTICS SWEDEN 1995 Yrkesklassificeringar i FoB 85 enligt Nordisk yrkesklassificering (NYK) och Socioekonomisk indelning (SEI), Stockholm: Statistics Sweden.

SVALLFORS, S. 1995. 'The End of Class Politics? - Structural Cleavages and Attitudes to Swedish Welfare Policies'. Acta Sociologica, 38, 53-74.

THALIN, M. 1993. 'Class Inequality and Post-industrial Employment in Sweden'. In: Esping-Andersen, G. (ed.) Changing Classes. Pp 80-108. London: Sage.

UUNK, W. 1996. Who Marries Whom: The role of Social origin, Education and high culture in Mate selection of industrial societies during the twentieth century? Nijmegen University, Dissertation.

WEBER, M. 1968. Economy and Society - An Outline of Interpretative Sociology (G. Roth & C. Wittich (ed)), volume 2, New York: Bedminster Press.

WESTERGAARD, John 1995. Who Gets What : The Hardening of Class Inequality in the Late Twentieth Century. Cambridge: Polity.

Copyright Sociological Research Online, 2001