Operational categories Analytic variables
Eight (Nine) classes Five classes Three classes
L1 Employers in large Large employers and
organisations 1.1 higher managerial
Higher managerial occupations
L2
occupations
L3 Higher professional 1.2 Higher professional
occupations occupations Managerial and Managerial and
Lower professional and 1 professional 1 professional
L4 higher technical occupations occupations
occupations Lower managerial and
L5 Lower managerial 2 professional
occupations occupations
Higher supervisory L6
occupations
L7 Intermediate 3 Intermediate 2 Intermediate
occupations occupations occupations
Employers in small Intermediate
L8
organisations
4 Small employers and 3 Small employers and
2
occupations
Own-account own-account workers own - account workers
L9
workers
L10 Lower supervisory occupations
5 Lower supervisory and 4 Lower supervisory and
L11 Lower technical technical occupations technical occupations
occupations Routine and
Semi-routine Semi-routine 3 manual
L12
occupations 6 occupations
5 Semi-routine and occupations
L13 Routine routine occupations
occupations 7 Routine occupations
L14 Never worked and 8 Never worked and Never worked and Never worked and
The National Statistics Socio-economic Classification: Origins, Development and Use Chapter 5
39 constructed a class schema that (so far as is possible within the
limitations of our LFS data discussed in the next chapter) minimises within-class and maximises between-class variation in terms of employment relations. That is, we have determined the work and market relations typical for each class. However, because they are based on social relations, class schemata are not hierarchically or linearly ordered. This is why we must collapse the NS-SEC in the manner indicated in Figure 4. For example, higher professionals and managers in large
organisations are in broadly equal positions to one another, as are intermediate employees and the selfemployed (whether small employers or own-account workers).
Of course, some classes are advantaged with respect to others; for example managers in large organisations vis-à-vis
intermediate employees and all of the working class. This follows directly from the different modes of employment regulation discussed in Chapter 4. However, we cannot wholly order a class schema such as the NS-SEC. Because the NS-SEC aims to capture qualitative differences in employment relations, the classes are not consistently ordered according to any inherent hierarchical principle. The members of different classes may be relatively advantaged or disadvantaged in different ways. Nevertheless, class schemata do not attempt to describe societies on a layer-cake model but via more subtle, relational concepts.
Finally, one of the great advantages of class schemata is that they permit us to study both the relational and the distributive aspects of social inequality, as demonstrated by the validation studies discussed in Chapter 6 (and see on all these issues Egidi and Schizzerotto 1996).
5.35A ‘health warning’ on measurement issues. Returning now to the basic measurement issues relating to the NS-SEC, the three-class model may be assumed to involve some kind of hierarchy. However, neither the five-, six-, eight- nor nine-class versions can be regarded as ordinal scales, not least (but, as we have seen, not only) because of the recognition of self-
employment as a separate class category. However, we recognise that some researchers might wish to have an ordinal scale similar to SC. This could ostensibly be achieved, for example, by combining the self-employed in NS-SEC Class 4 with the intermediate Class 3. We do not advocate this, however, not least because the self-employed are distinctive in their life chances and behaviour. On the contrary, we
recommend strongly that analysts accept the theoretical, and thus the measurement principles of the new class schema, take advantage of the conceptual base of the model for developing hypotheses linking class to outcomes of interest (see 4.15–4.22) and use analytic techniques as appropriate.
5.36(2) Population coverage. While experts were concerned with the conceptual poverty of the former classifications, users were often more worried about the practical problems they faced when they employed SC and SEG. Of most concern was the issue of population coverage. In many circumstances, occupational classifications only cover half the population. There are, of course, methods for avoiding this in data collection (for example, by using an earlier occupation or that of someone else in the household). However, this is not always possible when using administrative data sources or where classification is not seen as an objective in data collected for other purposes.
On the basis of own occupation, classifications most frequently exclude children, housewives, the retired, the unemployed, and the never employed, and often understate the social position of those in temporary, post-retirement or part-time work. Many of these groups are vital to policy and research interests and it is a matter of concern that they so often slip through the classificatory net. Virtually every government department made this criticism of SC and SEG, often with the comment that the new, ‘flexible’ labour market was exacerbating these problems (see 6.1). Of course, ONS provided recommendations in
Standard Occupational Classification: Vol. 3 (OPCS 1991) designed to alleviate the difficulty, and, where it was practical to follow these, then only children and adults who have never worked were excluded.
5.37(3) The disadvantaged. Academic ‘off-the-peg’ users of SECs are similarly concerned about coverage, some noting that SC and SEG did not cover the most disadvantaged and precariously placed individuals. For example, child mortality is greatest among groups such as unemployed, lone parents. Similarly the mortality rates for the unemployed and disabled were higher than those for Social Class V; including these groups would significantly affect the overall results.
5.38(4) Classifying the non-employed. In response to such criticism, and as we argued in 5.18, it is important to include within the NS-SEC as many as possible of those adults not in paid employment. To achieve this we saw that it is first necessary to create a special NS-SEC category for the ‘never worked’. For others who are non-employed, research by Marshall et al (1996) demonstrates that classifying individuals not currently in paid employment by their last main job is a satisfactory procedure, even for those who have been out of the workforce for many years.
Moreover, research reported by Arber using General Household Survey data demonstrates strong class ‘gradients’ (or, more precisely, monotonic changes) in ill-health for unemployed men and the retired (see Arber 1997b; Arber and Cooper 1998;
Chapter 5 The National Statistics Socio-economic Classification: Origins, Development and Use
40
Cooper and Arber 2003; and c.f. Arber 1996a). For the retired, class (in terms of last main occupation) continues to be associated with health throughout retirement. It is for this reason that we recommend that both the Census and
government surveys collect information on last main job for all those not in paid employment. Of course, this may seem to contradict our proposal for a class partly composed of the long-term unemployed. However, the creation of category L14 is for analysts to decide. This is why information on last main job is always required for those not in paid employment. To repeat, this strategy is one designed to create flexibility for analysts. Using these rules, only children are excluded from the NS-SEC. However, as with students, the normal practice is to classify children by reference to a parent. This raises the issue of which parent to choose and thus of how to classify households and/or families.
5.39(5) Family/household measures of class. Many analysts only perceive SECs as individual measures of labour market position. Consequently they only use them as such. In fact, however, many analyses would be improved if, when appropriate, researchers employed the household rather than the individual as the unit of class composition. We have explored some of the reasons for this elsewhere (see Rose and Pevalin 2003d). It may be asked, however, how a measure such as the NS-SEC, based as it is on employment contracts, can be anything other than an individual measure. We shall try and explain.
5.40 (6) The family/household as the unit of class composition.
Traditionally in sociology the unit of class composition (or unit of analysis) has not been the individual but the conjugal family/ household. That is, the (nuclear) family is given priority over the individual as the unit of class composition so that those living together in a family household are regarded as having the same class position. In other words, the family is the basic class structure element because of the inter-dependence and shared conditions of family members (see, for example, Goldthorpe 1983). After all, a family member’s own class position may have less relevance to his/her life chances than the position of another family member (see, for example, Vagero 2000). It is the family that is the unit of class ‘fate’ and the basic decision- making unit in terms of both consumption and labour market participation (see, for example, Erikson and Goldthorpe 1992: 232–239). Hence, lines of division run between, but not through families. This does not assume or imply that the family is egalitarian, but only that family members living in the same household share the same class fate. Therefore, we need to be able to assign a household NS-SEC value to all household members.
The simple practical solution to this problem has been to select one family or household member (usually the ‘male
breadwinner’) and take that person’s class to stand for the whole household. Recently, however, especially because of the increased participation of married women in the labour market, there has been much discussion about whether this continues to be an appropriate strategy. Some have advocated that the individual should now be the unit of class composition. Without entering the details of this controversy (see Sobel et al
2004, Rose and Pevalin 2003d, Sørensen 1994 and Erikson and Goldthorpe 1992 ibid), here we discuss different ways in which the NS-SEC can be applied to households and families.
5.41 (7) Assigning household class: (1) ‘highest income householder’. Because of the overt sexism involved in the male breadwinner approach to the definition of the household reference person (HRP), a new method has been developed by official statisticians (see Martin 1995 and 1998, Martin and Barton 1996). ONS has decided that, in the final instance, the HRP should be the ‘Highest Income Householder’ (HIH), thus removing sex as a criterion for determining head of household. Here the householder is regarded as the person responsible for owning or renting or who is otherwise responsible for the accommodation. Where this definition yields joint householders, the person with the highest income takes precedence and becomes the HRP. Where incomes are equal, the older is taken as the HRP. This procedure increases the likelihood both that a female will be the HRP and that the HRP better characterises the household’s social position. Analysts will generally use this procedure for determining household class. When using most government datasets they will have no choice but to do this. However, it should be noted that any definition based on income is likely to reduce the number of HRPs classified as self- employed, since they tend to have (or to declare) low incomes.
5.42(8) Assigning household class: (2) the ‘dominance’ approach. There is an alternative approach that regards the household reference person as the one who is dominant in the labour market, the so-called ‘dominance’ approach (see Erikson 1984). From a social scientific perspective, this procedure is preferable to one that relies on income in the determination of household class, but it does require that NS- SEC values must first be established for all household members. The dominance approach advocates two class concepts. In the first, work-related concept, individuals are the unit of classification because work is uniquely related to individuals. Hence it does not matter whether the individuals are male or female; each can be assigned a work position. In the second, market-related concept, families are seen as the classification unit. This is called the class position. Everyone has a class position, whether or not they are in the labour market.
The National Statistics Socio-economic Classification: Origins, Development and Use Chapter 5
41 The problem is thus how to determine a class position for the
family and then assign it to men and women alike. Erikson argues that class position may be derived as a function of individual family members’ work positions based on an order of dominance. At first sight, this may appear to contradict our earlier statement that the NS-SEC classes cannot be ordered. However, it should be recalled that the employment relation or contract only applies to individuals, that is, to the work position. Employment relations do not exist within families and so do not play a part in determining which family member’s work position best represents the family’s class position.
5.43(9) Determining family class position. If only one household member is in paid employment, that person’s work position becomes the family’s (household’s) class position. Similarly, if two generations are present in the household and each has a representative in employment, the person of the senior or primary generation takes precedence. However, where each of two or more members of this primary
generation has work positions and these positions are different (i.e. place them in different individual-level class categories), we need another dominance rule to determine (household or family) class position.
As with any other method for determining the HRP, ultimately we need an ordinal variable to make the final selection. If the work positions are the same (as they often will be) then this becomes the family class position. Otherwise, we need to decide for each possible pairing of different work positions, which is likely to have the ‘the greatest impact upon ideology, attitudes, behaviour and consumption patterns of the family members’ (Erikson 1984:504). Note this ordering is not based on work position as determined by employment relations but on the basis of the life chances known to be associated with work positions. It is in this sense that Erikson assumes there are dominance relations on various dimensions in which work positions may differ. Thus, higher qualifications dominate lower ones; non-manual work dominates over manual work; self- employment dominates over being employed; employers dominate over own-account workers; and managerial work dominates all other forms of work. Finally, the active are dominant over the inactive. All these assumptions flow from long-established results of research on the relationships between class position and life chances. Erikson then tested these assumptions using data from the Swedish Level of Living Survey (ibid:507–511).
5.44 (10) Dominance method for the NS-SEC. On the basis of Erikson’s research, where the NS-SEC work positions (that is, the individual class assignments) differ, the rules of precedence
we suggest are as follows. First, individual work positions derived from full-time work are dominant over those from part-time work. Second, if each is in full-time work, or each is in part-time work something like the order of precedence in Figure 5 should prevail from highest to lowest. Note, however, that this order has not yet been validated, but could be by following similar procedures to Erikson’s.