1 Overall and traditional left-right class voting in eight West European countries: A comparative longitudinal study 1975-97. Oddbjørn Knutsen Department of Political Science, University of Oslo, P.O.Box 1097, Blindern N-0317 Oslo 3, Norway Tel: 47-22-854279/47-22-855181 Fax: 47-22-854411 EMAIL: oddbjorn.knutsen@stv.uio.no Paper to be presented at the workshop on Cleavage Development: Causes and Consequences at the European Consortium of Political Research s Joint Session of Workshops, Edinburgh, March 28 April 2, 2003.
1 1. Introduction Social class represents the classic structural cleavage in industrial society. In Lipset and Rokkan s work it was first and foremost a cleavage in the labour market between owners and employers on the one side and tenants, labourers and workers on the other. It sprang out of the Industrial Revolution and proved much more uniformly divisive than the other major cleavages they focused upon (Lipset & Rokkan 1967: 14, 21, 35). The rising masses of workers resented their working conditions and the insecurity of their contracts. The result was the formation of a variety of labour unions and the development of nation-wide socialist parties. The fact that the labour market cleavage was so uniformly divisive in a comparative setting implied that it tended to bring the party systems closer to each other in their basic structure. While conflicts and compromises along the other cleavages, especially the centreperiphery and the state-church cleavage lines, tended to generate national developments of the party systems in divergent directions, the owner-worker cleavages moved the party system in the opposite direction (Lipset & Rokkan 1967: 35). In this respect Rokkan and Lipset focused most on the parties of the left, neglecting to some degree to focus in detail on the parties that represented the interests of the owners and employers in a comparative context (Steed & Humphreys 1988: 400-402). The Russian Revolution, however, also brought about a more divisive party structure among parties that articulated the interests of the workers. In some countries there emerged significant communist parties which created a split among the socialist parties, while the communists became an insignificant force in other countries (Lipset & Rokkan 1967: 46-50). The cleavage in the labour market is the central class cleavage, but not the only one according to Lipset and Rokkan. The other cleavage is the conflict in the commodity market between peasants and others employed in the primary sector and those who wanted to buy the products from the primary sector, particularly the urban population. This cleavage also sprang out of the Industrial Revolution. The peasants wanted to sell their wares at the best possible prices and to buy what they needed from the industrial and urban producers at low costs, while the urban population often had opposite economic interests (Lipset & Rokkan 1967: 20-21). This is then essentially an urban-rural conflict. Such conflicts did not invariably prove partyforming. They could be dealt with within broad party fronts or could be channelled through interest organisations into more narrow arenas of functional representation and bargaining. In many countries the religious interests of the rural population were more influential than the strictly economic ones, and the economic interest articulation took place within the Christian parties. Distinct agrarian parties emerged only in some countries where strong cultural opposition had deepened and embittered the strictly economic conflicts (Lipset & Rokkan 1967: 44-46). The two class cleavages were not discussed in relation to each other in particular in Lipset and Rokkan s work, but Rokkan developed a more elaborate model based on the two economic cleavages in an important work on the Norwegian cleavage structure (Rokkan 1966: 89-105). In this paper the focus is on the relationship between social class and party choice. I use a genuine cumulative data set based on all Eurobarometers from 1970 to 1997 to examine how social class influence party choice and how this has changed from the 1970s to the late 1990s. 1 The analysis comprises eight of the nine countries that were members of the (then) 1 The integrated Eurobarometer datafile is extracted from the ZEUS database. Eurobarometers are biannual surveys carried out for the European Commission throughout the EU. From 1987 to 1998, the ZEUS Institute at the University of Mannheim integrated these data sets in a harmonised database, analysed the data and advised the Commission in matters of comparative survey research. I would like to thank Hermann Schmitt for giving me the
2 European Economic Community in the early 1970s: Belgium, Britain, Denmark, France, (West) Germany, Ireland, Italy and the Netherlands. 2 Hout, Brooks and Manza (1995: 806-809) introduced the idea of distinguishing between traditional and total class voting. By traditional class voting they mean the portion of the statistical association between class and voting behaviour that arises from the affinity of bluecollar classes for left-leaning parties and the affinity of white-collar classes for right-leaning parties. Total class voting includes, by contrast, all sources of the statistical association between social classes and party choice where all classes and parties are treated as separate categories on the two variables. Here I will advance this distinction a bit further by differentiating between total class voting when we consider class differences in voting between all the parties in the party system, and (in a somewhat more detailed way than Hout, Brooks and Manza) overall left-right class voting when the left-right voting of all social classes are examined, and traditional (left-right) class voting when we examine the left-right division of parties and only two social classes (the manual/non-manual division). This paper only analyse overall left-right and traditional class voting although I have also done analyses on overall class voting over time based on the same data material. The paper is organised as follows: First I outline the theoretical class schema that I intend to use to tap social class (section 2); then I outline the theories of changes in class voting (section 3); and I go on to outline to what extent it is possible to use this schema on the basis of the occupation variables in the Eurobarometer data sets and how this is done (section 4). In section 5 I outline the statistical measures I use to tap class voting. I then go on to the empirical analysis where I analyse overall left-right class voting (section 6) and finally traditional left-right class voting (section 7). 2. Class schema In sociological literature the concept social class is defined in many different ways. I have used a slightly modified version of the so-called Erikson/Goldthorpe (EG) class schema, which was originally developed in connection with social mobility studies (Goldthorpe 1980; Erikson, Goldthorpe & Portocarero 1979; Erikson & Goldthorpe 1992) 3 but has been used in various other studies as well. 4 It has also been used in British election studies (Heath, Jowell opportunity to use the datafile. I am also grateful to the Mannheimer Zentrum für Europeische Sozialforschung, which has hosted me as a guest professor while I was doing the research on which this work is based. 2 Luxembourg is not included in this analysis. The Eurobarometers (and the cumulative file) contain specific data for Northern Ireland, which are not included in this analysis (nor in the data for Britain). We should underscore that we use only the data for West Germany for the whole period. We refer to the country as Germany. The practice of biannual surveys started in 1974-75. Earlier surveys called European Community Studies were carried out one survey in 1970, 1971 and 1973. These surveys are considered a part of the Eurobarometer surveys and included in the integrated Eurobarometer datafile. Reliable comparative class variables were only possible to construct from 1975. I therefore use the data from 1975 to 1997 in this paper. The integrated file comprises 63.-70.000 units for each of the eight countries and 62-64.000 units from 1975. 3 It is also called the EGP class schema, owing to the contribution of Portocarero in one of the articles referred to above. 4 For an overview, see Goldthorpe and Marshall (1992).
3 & Curtice 1985: chap. 2; Heath et al. 1991: chap. 5) and comparative study of class voting in Western democracies (Nieuwbeerta 1995), and is considered the most influential conceptualisation and operationalisation of social class in European sociology (Evans 1992: 211 212). It marks a major breakthrough in class analysis and represents a major improvement in the understanding of class division (Savage 1991: 29). The principles of differentiation in the EG schema have been derived mainly from classic sources, in particular from Marx and Weber. Under the influence of various later authors, the principles have been adapted to try to meet specific requirements of analysing class mobility within the total populations of mid-20th-century industrial nations, both capitalist and state socialist (Erikson & Goldthorpe 1992: 37). The basic approach, however, remains Weberian. The aim of the class schema is to differentiate positions within labour markets and production units more specifically, to differentiate such positions in terms of the employment relations that they entail (Erikson & Goldthorpe 1992: 37). 5 The basic distinction in the schema is within the category of employees. In consequence of employer employee relations being based on quite heterogeneous principles, employees in fact occupy a range of different labourmarket and work-situations, among which meaningful distinctions can and should be made in class terms (Erikson & Goldthorpe 1992: 41). The distinction between employees involved in a service relationship with their employers and those whose employment relationships are essentially regulated by a labour contract is what underlies the way different employee classes have been delineated. A service relationship, rather than one formulated in terms of a labour contract, is found where the employees are required to exercise delegated authority or specialised knowledge and expertise in the interest of their employing organisation. Such employees must be accorded a legitimate area of autonomy and discretion, and their performance will depend on the degree of moral commitment that they feel towards the organisation rather than on the efficacy of external sanctions. The organisation must to a significant extent trust these employees to make decisions and to carry them through in ways consistent with the values and goals of that organisation (Goldthorpe 1982; Erikson & Goldthorpe 1992: 42). It is on the basis of this fundamental distinction that the class schema is drawn up. The main traits of the schema are outlined in Table 1. < Table 1 about here > Different employee classes are delineated on the basis of the theoretical distinction between employees involved in a service relationship with their employer and those whose employment relationships are essentially regulated by a labour contract. A main division is that between the predominantly salaried professional higher technical, administrative and managerial positions, and the predominantly wage-earning manual occupations. The former are positions with which a service relationship is associated, and thus constitute the basis of 5 These formulations differ from earlier discussions of the schema, which were more based on the pioneering work of David Lockwood (1989/1958), in which the divisions between social classes were based on the typical work and market situations associated with occupations (Goldthorpe 1980). The revised formulations seek to bring out more clearly that the schema is intended ultimately to apply to positions, as defined by social relations, rather than to persons (Erikson & Goldthorpe 1992: 37). Nevertheless, the allocation of individuals to locations in the schema is determined in the same way as in the earlier works, using occupational categories derived from job descriptions and employment status (e.g. self-employed, employee, management).
4 the service class 6 or the salariat 7 of modern industrial society; the latter, where the labour contract usually prevails, constitute the basis of the working class. The argument for treating professional, administrative and managerial employees as holding basically similar class position is that these employees, being typically engaged in the exercise of delegated authority or in the application of specialised knowledge and expertise, operate in their work tasks and roles with a distinctive degree of autonomy and discretion; and in direct consequence of the element of trust that is thus necessarily involved in their relationship with the employing organisation, they are accorded conditions of employment which are also distinctive in both the levels and the kinds of rewards that are involved (Goldthorpe 1982: 169). I use mainly the notion of new middle class (Giddens 1973: chap. 10) instead of service class, although I also use the term service class to vary the language, but these classes are similar. The new middle class comprises administrators and managers, employed professionals, higher-grade technicians and supervisors of non-manual workers. It is divided into a higher and a lower level according to administrative responsibility and educational training. We might call them higher- and lower-level new middle class, but I prefer to use the terms higher-level and medium-level non-manual employees. The higher level has positions that typically involve the exercise of authority, within a wide range of discretion, and with considerable freedom from control by others. Typical examples are professionals who are employees, higher-grade administrators and officials in public and private enterprises (including company directors), and higher-grade administrators and officials in central and local government and in welfare institutions. Medium-level non-manual employees are lower-grade professionals (typically called semiprofessionals) and lower-grade administrators and officials. The occupational roles of the middle-level non-manual employees are located in the middle and sometimes also in the lower range of bureaucratic hierarchies of some type or another; they exercise some degree of autonomy and discretion in the performance of their work tasks, while at the same time being subject to more or less systematic, if not particularly close, control from above. Lower-level non-manual employees are largely clerical personnel, employees in administration and commerce, sales personnel and other rank-and-file employees in the service sector. Lower-level non-manual employees, or routine non-manual employees as they are called in the EG schema, do non-manual work, but they do not belong to the new middle class or the service class. They are functionally associated with, but marginal to, the service class (Goldthorpe 1980: 40). This is a class that may be regarded as intermediate in the sense that it comprises positions with employment relationships that appear to take on mixed forms. It covers the range of routine non-manual positions, usually involving clerical, sales or personalservice tasks, which exist on the fringes of professional, administrative and managerial bureaucracies. I also use the notion routine non-manual employees in addition to lower-level non-manual employees. 6 The term service class derives from the writings in the 1950s of the Austro-Marxist Karl Renner. Ralf Dahrendorf was responsible for introducing Renner s work into Englishlanguage sociology. The concept of the service class remained unpopular, however, until its use by John Goldthorpe in the course of his enquiries into social mobility (Goldthorpe 1982: 167-170; 1995: 313). 7 Salariat was introduced as a synonym of service class in order to avoid confusion between the service class and service workers and the service sector of employment (Lockwood 1995: 1; Goldthorpe 1995: 314).
5 The working-class group comprises skilled and unskilled manual wage-earners in all branches of industry, as well as supervisors of manual workers (foremen) and lower-grade technicians. I have not distinguished between skilled and unskilled workers because such a distinction does not appear relevant here. There are, then, three levels of non-manual employees. The two higher levels are the new middle class, corresponding to service class, and the lower-level non-manual employees, corresponding to the routine non-manual employees in the EG schema (see Table 1). 8 I differentiate between employers and self-employed on the one hand, and employees on the other; the four classes identified above are all employees. Employers and self-employed are divided into two categories on the basis of sector: those who work in the primary sector (primarily farmers and fishermen) as opposed to other employers and self-employed. It might have been natural to differentiate between various categories within the latter group, for example those with many employees, few employees and no employees (self-employed small proprietors, artisans, etc.). These might well have different interests, but there are also theoretical reasons for keeping them together as a joint category. The low number of cases in the empirical material makes it difficult to differentiate them. Here I deviate from the EG schema because it groups the employers with many employees (large proprietors) as well as higher-grade professionals who are self-employed or employers in the higher service class. 9 I call this category employers, not petty bourgeoisie. 3. Theories and explanations for the decline in class voting In this section I will review some general hypotheses and empirical findings about why class voting will decline in advanced industrial society. I am not able to test the various hypotheses in the empirical analysis due to lack of relevant variables in the data material. The various hypotheses will more be considered as background for the empirical analysis. Furthermore, I do not review the findings about the comparative level of class voting and the comparative decline in class voting. The reason for this is that I find it more fruitful to compare my findings with the findings from previous research as a part of the empirical analysis where I evaluate my findings on the basis of that research. 1. Social mobility. Patterns of social mobility may influence the level of class voting. Studies of social mobility and political preferences have almost invariably found that the mobile voters are likely to adopt a middle standpoint between the political preferences of their class of origin and their class of destination. Mobility processes may therefore have important consequences for levels of class voting, and they may help to explain both over-time and cross-national variations in class voting. In particular, higher absolute rates of social mobility may be expected to reduce class polarisation by increasing the proportion of mobile voters 8 Goldthorpe (1980: 40-41) refers to the service class and the routine non-manual class as the white-collar classes. 9 See the discussion and doubts about where to place the large proprietors in the EG class schema in Goldthorpe (1980: 40) and Erikson and Goldthorpe (1992: 40). Goldthorpe (1995: 314) emphasises strongly in a more recent contribution that the service class is a class of employees, and that the main boundary problems that arises is that of demarcating the service class from other employee classes. He also emphasises that there are more pragmatic reasons why large employers and all professionals (also those who are employers and self-employed) are grouped into the service class. When these categories become of sufficient size to permit separate analysis, they should certainly be excluded from the service class (Goldthorpe 1995: 314).
6 with intermediate propensities to support the left and by weakening class solidarity and even pulling the non-mobile voters towards the centre. The significantly greater level of class mobility in the post-war period, resulting largely from the steady and nearly universal growth of professional and managerial employment, suggests that mobility may reduce the level of class voting (De Graaf, Nieuwbeerta & Heath 1995: 998; Manza, Hout & Brooks 1995: 143-144). However, an influential comparative study of several Western countries finds support for an acculturation hypothesis. Original class has an impact on an individual in his or her youth and early adulthood. However, the older a voter is, the more the impact of class or origin diminishes relative to that of the class of destination (De Graaf, Nieuwbeerta & Heath 1995). The impact of social mobility, then, decreases over time at the individual level, and in a sophisticated analysis of comparative levels of class voting in 14 countries, and changes of class voting in these 14 countries over time, social mobility did not influence class voting and change in class voting in the expected direction (Nieuwbeerta & Ultee 1999). 2. New social division. A second set of explanations for the decline in class voting emphasises the increased importance of cross-cutting cleavages in the politics of post-industrial democracies. These cleavages may be new or they may be old ones that are re-emerging in importance in advanced industrial society (ethnicity, linguistic differences). The working class and the new middle class are being politically fragmented due to such cross-cutting cleavages in relation to the class cleavages. For example, in Britain scholars have found support for consumption and production sector cleavages. The concept of sector means that these cleavages are vertical divisions in a society which cut across the class cleavage, such that certain common interests are shared between social classes in the same sector, while within a class, sectoral differences reflect conflicts of interest (Dunleavy 1979: 417-422). The relevant consumption sector cleavages are those forms of consumption that lead to a polarisation between individual private mode and collective public mode, together with the overtly political determination of the condition of consumption. Housing and transportation policies and voters status in these consumption locations influence voting and cut across the class cleavage. In the production sphere Dunleavy (1980a, 1980b) argues that public versus private sector is a corresponding cross-cutting sector cleavage which tends to weaken the class cleavage. On the other hand, such cross-cutting cleavages are not entirely new and can explain comparative differences in class voting in a longer perspective. For example, the important traditional religious cleavage to a large extent cuts across the class cleavage. Nieuwbeerta and Urtee (1999: 143-146) find that religious and ethnic fragmentation is a main factor for explaining the comparative strength of class voting in 14 countries, but changes in such fragmentation do not explain changes in class voting over time. 3. Cognitive mobilisation. A third set of arguments explains the decline of class in terms of the increased capacity of a better-educated citizenry to make decisions independently of the constraints of class loyalty or social cleavages. Voters are viewed as increasingly capable of rational assessment of party and candidate platforms and therefore less likely to rely on simple class-based heuristics. This approach distinguishes between voters who are shaped primarily by class and those who make electoral choices independently of class influences. The increased capacity of voters to make independent electoral choices is often linked to class dealignment (Dalton 1984; Rose & McAllister 1986; Dalton, Beck & Flanagan 1984: 18-19). 4. The embourgeoisement thesis. Growing affluence among all social classes has definitely taken place since World War II. Rising incomes, improved standards of consumption, the spread of home ownership and greater leisure in the working class lead affluent workers to identify with the middle classes and embrace its attitudes and life-styles. Theories about the
7 bourgeoisification of the working class use this as a starting-point when they focus on the lack of radical orientations in the working class during the 1970s and 1980s. Workers no longer struggle to maintain a subsistence income, and the more affluent sector of the working class has assumed a middle-class life-style (Goldthorpe et al. 1968). Feelings of class solidarity and the attachment to labour unions have also decreased in advanced industrial society (Dalton, Beck & Flanagan 1984: 15-17). 10 A major consequence is a decline in support for socialist parties. 11 More generally, rising affluence is found to be an important explanation for the decline of class voting in the comprehensive study of class voting in several countries mentioned earlier (Nieuwbeerta & Urtee 1999: 144-146). 5. New Politics and middle-class radicalism. Perhaps the most influential of the micro-level explanations for the decline of class voting emphasises the importance of value change and the rise of new attitudinal cleavages. In particular, a large literature has developed around the claim that the historical links between workers and parties of the left have weakened at the same time that a second left rooted in segments of the new middle class has grown up. It is argued that this produces a decline in left-right class voting. The increased proportions of left party support drawn from middle-class sectors have weakened the class coherence of party platforms as new issues such as concerns have arisen about environment protection, peace, civil rights for previously ignored groups such as women and gays and lesbians, and more generally quality-of-life issues. As left parties have changed to become more inclusive, their appeal to workers concerned with material issues has weakened (Lipset 1981: 503-523: Manza, Hout and Brooks 1995: 145, Parkin 1968; Weakliem 1991). The most systematic effort to account for middle-class leftism is the New Politics literature, whose proponents argue that segments of the new middle class will have post-materialist values and consequently vote for green or left-libertarian parties. According to the group polarisation hypothesis within New Politics theory, the rise of a new axis of group polarisation has been closely linked to the rise of the materialist/post-materialist dimension as a party cleavage. The social groups most likely to subscribe to post-materialist values (the new middle class, the better-educated, the post-war generations) will tend to support changeoriented parties that focus on post-materialist values (Inglehart 1984: 32-33; 1997: 248-252). 10 The relationship between unionism and class voting is a complex one. First, there is no general decline in union density in West European countries. The main pattern is increasing dispersion in a comparative setting. There is, however, an average decline from around 1980 to the late 1990s after a long-term increase in union density until the early 1980s (Western 1995: 180-182, Ebbinghaus & Visser 1999: 136-138: Golden, Wallerstein & Lange 1999: 198-202). Second, the relationship between class voting and union membership is complex, partly due to labour union membership of non-manual employees, who are less likely to support the socialist parties than workers. Nieuwbeerta and Ultee (1999: 140-145) find that union density is a major determinant of the cross-national level of class voting in their study of 14 advanced democracies, but union density was also strongly correlated with the decline in class voting over time in a surprising way. The decline was most pronounced in the countries with a high union density. This finding is mainly due to the high degree of union density in the Scandinavian countries, where class voting is strongly declining. The authors explain this somewhat speculatively with the successful increase in union density among non-manual workers in these countries. 11 An extensive test of the thesis on the basis of the social democratic parties in the Scandinavian countries did not support the thesis (Sainsbury 1985).
8 The rise of the materialist/post-materialist dimension can also explain the decline of class voting in post-industrial society. The new middle class and the better-educated strata are most likely to support the post-material left. And as post-materialist issues become more important, this may stimulate a materialist counter-reaction whereby some of the working class side with conservative or bourgeois parties to reaffirm the traditional materialist emphasis on economic growth, military security and law and order (Inglehart 1984: 28; 1997: 252-256). This can also be expressed by means of spatial dimensions. In advanced industrial societies the economic left-right dimension has been supplanted by a materialist/post-materialist or authoritarian-libertarian dimension (Kitschelt 1994, 1995). On this new dimension workers and people with less education tend to be located near the authoritarian pole, while the new middle class and those with higher education tend to be located near the libertarian pole. The new right parties and the left-libertarian parties have a firm location at the extremes on this new dimension, but the increasing importance of the new dimension poses strategic dilemmas for the parties of the moderate left and moderate right. The result is often a blurring of the impact of social class (Kitschelt 1994: 30-39, 149-206; 1995: 13-19). 6. The debate about the political orientation of the service class. This point examines middleclass radicalism from a somewhat different angle compared with the previous point. In the literature on the service class there is basic disagreement about the political orientations and party choice of the service class. John Goldthorpe (1982), one of the fathers of the EG class schema, has formulated an influential theory of the political orientations of the service class. According to his view, the service class will constitute an essentially conservative element within modern society (Goldthorpe 1982: 180). The service class is a class of employees who are subordinate to some form of higher agency, but the main characteristic of the serviceclass occupations is that they involve the exercise of authority and/or of specialised knowledge and expertise. Their incumbents perform their work tasks and roles with some significant degree of authority and discretion, and enjoy conditions of employment which are decidedly advantaged related to those of other grades of employees. This represents a unifying structural location that determines the basic social and political outlook of the service class, and based on these arguments, Goldthorpe s view is that the service class has a substantial stake in the status quo, and that there is little structural evidence that they will develop orientations that will result in Old Left views similar to those of the working class, or that they will generate new forms of cultural and political radicalism, as new class and New Politics theoreticians argue (Goldthorpe 1982: 168-170). Those who claim that (part of) the service class will have a radical orientation ( new class theories) are criticised for not providing an adequate account of the structural location of the political radicalism within the service class (Goldthorpe 1982: 166). Although Goldthorpe s view is that the common structural position of the service class inclines it towards voting for right-wing parties, he is well aware that substantial numbers of the service class are not rightist in their political orientations. He admits that some situs 12 divisions can occur within the service class, just as they can of course within other classes (Goldthorpe 1982: 170), but argues that these situs divisions will not prove a source of major differentiation (Goldthorpe 1982: 178). Most of the variation in political orientations within the service class can be ascribed to the fact that it has grown very fast in the post-war period. It has then been 12 By situs, Goldthorpe has clarified that he means the functional context of an occupation or group of occupations which may exert an influence on the life-styles and patterns of action of their incumbents, independent of that of class, or status, positions that they simultaneously hold (Goldthorpe 1995: 328).
9 recruited widely from other social classes, with the consequence that only about one-third of its members are the offspring of parents who held similar class position; in turn, this means that the political views of its upwardly mobile members may well be linked to their class or origin and hence less likely to be conservative (Goldthorpe 1982: 174-177). According to Goldthorpe, it is more personal characteristics of service-class members, like social mobility, educational differences and processes of occupational self-selection, rather than any structural feature, that contribute to political differentiation within the service class (Goldthorpe 1995: 323-324). In the British debate, Lash and Urry (1987: 209-231) have argued that the service class will be an innovative and disruptive force in capitalism. These authors couple the service class with the rise of new social movements and the weakening of the old patterns of political alignments. Perhaps more fruitful is the argument that it is structural forces within the service class that tend to fragment the political alignments of the middle classes. Savage (1991) has questioned Goldthorpe s argument about a basic conservative orientation of the service class in the British context, and found support for the argument of considerable internal divisions within the service class. With regard to party choice, he finds that divisions within the service class are considerable greater than within any of the other social classes. He questions the unitary political orientation of managers, administrators and professionals by showing that the professional group is considerably less likely to support the Conservative Party than the other groups, and finds that sector employment is an important structural division within the service class. Heath et al. (1991: chap. 6) made a parallel finding, demonstrating that the so-called creative and welfare professions appear to be politically distinct within the service class because of their relatively low support for the Conservative Party. Heath and Savage (1995) also find considerable variations within the service class s party identification by examining 39 different middle-class occupation groups. They find, however, that structural characteristics like sector employment and employment status have only limited power for explaining these differences. Similar findings about divisions within the service class with regard to participation in new social movements (Kriesi 1989) and party choice are found for several Continental European countries (Müller 1999; Kriesi 1998: 168-172). 7. Party strategies, class appeal and changes in class structure. Another way of hypothesising that class voting will decline is related to the party programmes and appeals that parties make in relation to their voters. Parties no longer appeal to class issues and values that are related to the economic left-right dimension to the same degree as previously. One line of argument is that the main socialist parties, the social democrats, cannot successfully appeal to all groups and that they therefore face an electoral trade-off appealing to different groups of voters. The best-known perspective in this respect is formulated by Przeworski (1980) and Przeworski and Sprague (1986). Left parties whose core constituencies are manual workers face an electoral dilemma: Since the working class (defined narrowly by Przeworski and Sprague by not including service workers) is an electoral minority, left parties must appeal to middle-class voters as well if they are to win elections. Such strategic manoeuvres have, however, important class-demobilising effects because these parties presumably would lose support from workers when they appeal to the middle classes. In the long run socialist parties cannot extend their electoral support beyond the working class and simultaneously mobilise the workers vote to the same degree as previously. The electoral dilemma thus undermines class voting in two ways. It forces parties originally based in the working class to appeal to members of other social classes. To the extent they
10 succeed, the class-based electoral cleavage declines. The dilemma also discourages classbased appeals in politics generally in favour of supra-class themes. When social democrats extend their class appeal to other social classes, they cannot appeal to interests and issues specific to workers as a collectivity those that constitute the public good for workers as a class but only those which workers share as individuals with members of other social classes (Przeworski 1980: 43). Workers are no longer mobilised as workers but as consumers, taxpayers, parents, economic prosperity, the poor, the people, etc. (Manza, Hout & Brooks 1995: 146; Sainsbury 1990: 31). These arguments that left parties are unable to appeal successfully both to large segments of the working class and to the middle classes have been challenged seriously. Sainsbury (1990) shows, for example, that there is little if any evidence that the Scandinavian leftist parties lose working-class votes when they adopt strategies to gain middle-class votes, and Kitschelt (1993: 320-337) formulates some important theoretical arguments against the theory. He also offers some alternative dilemmas that are related to his own two-dimensional conflict structure, which is briefly outlined above. In a competitive space comprising the economic left-right dimension and the new libertarian-authoritarian dimension, social democratic parties in advanced industrial society have to move beyond their appeal to the working class by emphasising leftist economic orientations. They can appeal strategically to more middle-class segments by moving to a left-libertarian position, or they can move towards the centre on the left-right dimension to compete with the non-socialist centrist parties. The alternative strategic appeals are unlikely to involve electoral trade-offs between workers and non-workers, but between different segments of the working class as well as different segments of the service class (Kitschelt 1993: 300-313; 1995: chap. 1; 1999). Are there not arguments indicating that class voting should persist, not decline? Manza, Hout and Brooks (1995) review the literature on the reasons for the decline in class voting by focusing on arguments very similar to those presented above. They then use counterarguments or question some of main theoretical arguments and the empirical findings related to some of the main work in the decline of class literature. They refer to some of the central sceptical scholars who deny or question the decline of class thesis, and also point to other literature where some of the central assumptions and findings are questioned (Manza, Hout & Brooks 1995: 148-150). They basically question that the degree of social and political change has the consequences for class voting that supporters of the decline of class thesis advance. They also refer to empirical findings from Britain and the USA which show very little decline in class voting over time. I will briefly review this debate here. The debate about decline or persistence of class voting. What they refer to is indeed the now-classic debate on the topic of the decline of persistence of class voting. While it was once considered conventional wisdom that class voting was declining in all advanced democracies, this wisdom has been seriously challenged by American and British scholars in particular. The conventional view has been advanced by leading scholars like Seymour Martin Lipset (1981: 503-507) and Ronald Inglehart (1984: 29-31), who have documented a decline in class voting in several countries by mainly relying on published material from other scholars and by using a simple measure of social class (the socalled Alford index; see below). In the United States it is first and foremost the research by Clem Brooks, Michael Hout and Jeff Manza that has challenged the conventional view. By using a more detailed class schema than the so-called Alford index, as well as more sophisticated statistical techniques which measure relative class voting instead of absolute class voting, they argue in several central works that social class has a quite permanent overall impact on party choice in the USA
11 (Brooks & Manza 1997a; 1997b; Manza & Brooks 1999: chap. 3; Hout, Brooks & Manza 1995). In Britain the conventional view was expressed by leading electoral researchers in the 1970s and 1980s (Franklin 1984; Rose & McAllister 1986: chap. 3), including those who conducted the British Election Studies in the 1970s, Bo Särlvik and Ivor Crewe (Crewe, Särlvik & Alt 1977: 168-181). However, those who became responsible for the British Election Studies in the 1980s and 1990s the so-called Nuffield team disputed the claims about class dealignments for some of the same reasons as those put forward by their American colleagues. Using newer statistical techniques and a more elaborate class schema (the EG class schema), they differentiated between absolute and relative class voting and found more trendless fluctuation than decline in class voting (Heath, Jowell & Curtice 1985: chap. 3; Heath et al. 1991: chap. 5; Evans, Heath & Payne 1991). These findings based on other statistical methods have resulted in an interesting debate about the persistent impact of social class or its declining impact between Ivor Crewe (1986), one of those who conducted the British Election Studies in the 1970s, and the Nuffield group (Heath, Jowell & Curtice 1987). It is, however, interesting that at least some of the members of the Nuffield group admit that there is a secular decline in class voting when the more recent results from elections in the 1990s are analysed from a long-term perspective, in addition to considerable trendless fluctuation. There is in particular a very clear decline during the 1980s and 1990s, according to their analysis (Evans, Heath and Payne 1999). In more recent publications Anthony Heath has also shown and admitted that there is considerable long-term decline in class voting in both Britain and the USA, as well as in France, somewhat contrary to his previous position (Wiekliem & Heath 1999a; 1999b). Others within the Nuffield group, however, have defended the view that class voting in Britain is constant (Goldthorpe 1999) and argued that the comparative studies of the decline in class voting are disputable (Evans 1999a; 1999b). 13 4. Operationalisations of social class: Occupation variables and construction of class variables in the Eurobarometer data set Is it possible to construct a class variable based on the categories on which the EG class schema is based, from the information about occupation in the Eurobarometers? The answer is only partly yes. The categories on the occupation variable have changed twice, and for the first (long) period for which occupation is available, it is not possible to differentiate between lower-level non-manual and medium-level non-manuals. Let us look at the occupation categories in more detail. There were some occupation variables available in the first European Community Studies, but these were country-specific, and it is impossible to construct a common class variable on the basis of them. From EB 3 (conducted in 1975) the response categories in the first version in Table 2A were used. These categories were used until EB 30 (conducted in autumn 1988), where a set of more detailed categories was used. In EB 37 (conducted in 1991) the categories 13 The debate about class voting is also reflected in other discussion between the main contributors within the field. See, for example, the debate in International Sociology between Clark and Lipset (1991; Clark, Lipset & Rempel 1993) and Hout, Brooks and Manza (1993), and between some other main scholars within the field (Mair, Lipset, Hout & Goldthorpe 1999).
12 used in the former version were more detailed and some new ones were added. 14 This version has been used in all later Eurobarometers. We have, then, three different categories used in three different time periods, and to do analyses over time we have to code the occupation categories into some common class categories. There were two questions about occupation with the same classification schema for all surveys, one about the respondent s own occupation and one about the occupation of the head of the household (if the respondent was not defined as head of the household). The coding schemata for these two variables change at the same time. < Table 2 about here > The first occupation codes that were used contained only six categories. There is one general category for White collar office workers and one category for the higher-level salariat: Executive, top management, directors. The first category in many class schemata would normally be for non-manual workers or lower-level non-manual workers, but the EG schema is more sophisticated, using three categories for the employed non-manual workers, and it is simply not possible to differentiate between the routine non-manual group and the lower level of the salariat on the basis of the categories used. Another problem is that it in contrast to the later schemata, there is no differentiation between employed and self-employed/employer professionals, as there is in the later categories. I have decided to group this category in the employer group (together with category 3 in Table 2A), not in the higher-level non-manual group, since the later surveys show that a majority of the professionals were selfemployed/employers. 15 The categories that were introduced in 1988 (from EB 30 to EB 36) are much more detailed, and regarding the classification of the categories into the EG classes, they were much improved over those used previously. The changes and improvements can be summed up as follows. 1. Employed and self-employed professionals are differentiated (categories 5 and 3, respectively). 2. There are now four categories for the non-manual employees (categories 6-9), and it is possible to differentiate between the service class and the routine non-manual workers. This is outlined below. 14 The respondents are not shown a card for this question. They are asked about their present occupation, and then the interviewer classifies the answers into the categories in the questionnaire. 15 In Belgium, France, Germany, Italy and the Netherlands, 60-80% of the professionals are self-employed/employers in the surveys from 1988, when there is differentiation between self-employed and employed professionals. In Denmark the self-employed and employee professional groups are fairly equal, while only in Britain and Ireland are a majority of the professionals employed; only about 30% are self-employed. The other category, which comprises the group of employers outside the primary industries in the first occupation schema (category 3 in Table 2), is a considerably larger category than the professionals (category 2) in all countries. The latter constitute 12-22% of the employer category in six countries, and about 30% in two countries (Britain and Germany). Thus, only a fairly small portion of the employer category might actually be professional employees in the relevant period (1975-88).
13 3. There are several categories for workers (categories 10-12), and skilled and unskilled workers are differentiated. However, since I use only one category for the working class, these categories are collapsed. Table 2B shows the categories that are placed in the various classes. The various categories are numbered, referring to the category numbers in Table 2A, and the table also describes in detail which categories are grouped together in the various classes. Here we will discuss only some major points and some borderline problems. In order to have an equivalent class variable for the whole period, we have to keep the distinction between higher-level non-manuals and other non-manuals in the main variable, but I have decided to create a new class variable based on the categories used since 1988, because it is now possible to use the class schema that was outlined in the theoretical discussion above. We can differentiate between the lower level of the service class and the routine nonmanual employees, and this is of some importance. I will call the two class variables the main class variable (since most of the analysis will be based on it) or the five-class variable, which covers the whole period, 1975-97; and the six-class variable, which will cover only the period 1988-97. We will mainly use the five-class variable for analyses over time, while the six-class variable will mainly be used for analyses of the whole period for which the variable is available, to show the impact of social class based on a class schema that is in accordance with our theoretical discussion. The time span for which this variable is available is too short to base any trend analyses on. Categories 7-9 used in EB 30 to EB 36 are grouped into the broad category "other nonmanual" employees, since all these categories cover the same group as category 5 in the first occupation code. As to the new six-category variable, it is natural to group the middlemanagement category as the medium-level non-manual category, while there is some problem with the next category, Other official employees. Should this category be placed among the lower level of the service class or among the routine non-manual group? Possibly the borderline goes somewhere within this category. Central for grouping this category in the routine non-manual category is that those who have some administrative responsibility are probably placed in the middle-management category, so that those who are placed in this category are the remaining personnel who work in offices, that is, secretaries and other rank-and-file personnel. The names of categories 8 and 9 are also very similar to those in the coding schema; there is one category for those rank-and-file-personnel who work in offices and one for those who work other places. The service class is admittedly defined somewhat narrowly by this decision. It comprises employed professionals, general and medium-level management. The newest version from 1991 (EB 37) is even more detailed. The differences between the categories in this third version and the previous one can be summed up as follows: 1. Employed and self-employed/employers are even more explicitly differentiated. 2. Two groups of owners, small and large, are differentiated (categories 4 and 5). 3. The description of the categories general and middle management is much more precise. In EB 30 to EB 36 only the titles general management and middle management are used. 4. Three categories of lower-level non-manual employees are differentiated (previously, two), and the descriptions of these categories are more detailed. The questionable classification involves again the borderline between the lower level of the service class and the routine non-manual group in the six-class variable, and in accordance with the decision above, the three groups in the categories beginning with employed