This paper compares class patterns of ‘conformism’ in four Western countries, taking as its point of departure arguments, which suggest that conformism/authoritarianism is more prevalent in the working class than other classes. Data from the International Social Survey Program (ISSP) are used in order to compare class attitudes towards sexual behaviour and civil rights/liberties, in Britain, Germany, Sweden and the United States. This study finds that class patterns in attitudes differ little between the countries in question, although aggregated levels vary quite substantially. It is also concluded that class differences in attitudes towards civil rights/liberties are almost exclusively a result from different levels of education, while significant class differences in attitudes towards sexual activities remain also after taking education levels into account. The implications of the results for social stratification and political articulation are discussed.
The link between positions in the stratification system, and values, commitments and attitudes is a topic which has attracted considerable research effort. These studies have mostly dealt with attitudes which relate to the (re)distribution of various goods and burdens; prominent themes have been attitudes to inequality, social justice, and the welfare state.1 Only infrequently has this research dealt with what we might call the moral order of things; that is, attitudes pertaining to such non-distributive issues as life-styles, freedom of expression, sexual orientation, and parent–child relationships.
There is nevertheless a research tradition in which such issues have been in the focus, and where the impact of social stratification on attitudes towards authority have been charted. By adopting such concepts as ‘conformity’ and ‘authoritarianism’, this research has sought to grasp the extent to which different groups espouse tolerant, democratic and open-minded attitudes.
Lipset's (1959) influential paper on ‘working-class authoritarianism’ constitutes an important starting-point for this line of research. Lipset maintains that working class support for non-democratic parties is less an unfortunate aberration than a reflection of the higher prevalence of authoritarian attitudes among workers. Distinguishing between ‘economic and non-economic liberalism’ he argues that the working class tends to score high on measures of ‘economic liberalism’ (in the American sense), but lower on measures of ‘non-economic liberalism’. In short, he observes that workers are more xenophobic, less tolerant of dissidents or minorities and less supportive of democracy than other groups. According to Lipset, such dispositions can be explained by (a) a greater economic and social insecurity among lower strata, (b) their relative isolation from ‘the activities, controversies, and organisations of democratic society’ (Lipset 1959: 490), and (c) their more restrictive and punitive child rearing practices. In sum, these factors tend to create less open-minded and tolerant citizens in the working class than in other strata.
Many of Lipset's assumptions and findings from the paper have subsequently been questioned on conceptual or empirical grounds. As will be shown, however, accumulated research findings, deriving from both survey and ethnographical research, does seem to point to a higher prevalence of conformist or authoritarian attitudes among workers than among non-manual employees. Alternative, at times competing, explanations to those offered by Lipset have been proposed, such as the liberalising effects of higher education, manual workers’ widespread experiences of powerlessness at work, and their psychological need to establish clear moral boundaries in a threatening environment.
Lipset's original formulations suggested that the degree to which the working class had been socially and politically integrated would affect authoritarian dispositions. This argument points toward a perspective, which compares class differences in authoritarianism/conformism across institutional and political contexts. The intention behind this paper is to extend existing comparative research into a comparison of attitudes in advanced capitalist societies, which differ in their institutional framework in ways which may affect class differences. Attitudes towards sexual behaviour and civil liberties/rights in different classes are compared in Sweden, Germany, Britain, and the United States, using data from the International Social Survey Program (ISSP). The importance of these attitudes in relation to issues about conformism/authoritarianism is elaborated below, where it is argued that they provide a suitable test case of some of the main issues in the debate.
The four countries chosen for this study are all long-standing capitalist democracies which share important traits in their basic economic structure; yet their political and institutional set-ups differ in significant ways. Thus, the contrast between the negligible impact of the political left in the United States and the long-standing social democratic influence in Sweden has often been noted (Bradley et al.2001); in terms of religious-cultural influence also the highly secularised Sweden contrasts strongly with the more religious US (ISSP 1998: 52, 54, 112–3). Britain and Germany fall into a middle category on both counts.
In institutional terms, Germany and in particular Sweden combine fairly regulated labour markets with extensive welfare states, whilst Britain and in particular the United States combine less regulated labour markets with more targeted, less extensive welfare states (Huber and Stephens 2001). These differences are further reflected in the lower degree of inequality of incomes in Germany and Sweden as compared to Britain and the United States (Bradley et al.2001).
It should be emphasised that the analysis here is not aimed at providing a strictly comparative analysis, where differences between countries in the class structuring of conformism/ authoritarianism are explained by cultural and institutional differences. Rather, the more modest aims are to test (a) whether such class differences remain constant across different national contexts and (b) whether variations in class differences across national contexts may be interpreted as the result of different degrees of working class integration, that is, whether a case can be made for Lipset's argument that a more integrated working class is likely to be less authoritarian.
A major problem in the literature relates to the conceptual and empirical treatment of social class. In Lipset's original article, ‘class’ is variously understood in terms of educational level, level of income, and the categories ‘workers’ and ‘the middle class’. Later research have also tended to lack conceptual clarity regarding the main independent variable, causing considerable confusion when any observed class differences need to be explained. For example, it becomes unclear whether observed class differences are to be explained mainly by different educational levels, in a way that links higher education to more liberal values, or if they should be explained with reference to the actual class experience.
The huge literature on classes in contemporary society incorporates a number of different conceptualisations of class. The one chosen here views classes as constituted by employment relations (Erikson and Goldthorpe 1992: Ch. 2; Goldthorpe 2000; Svallfors 2005). According to this perspective, classes are aggregations of positions within labour markets and production units, and can be operationalised through a combination of occupation title and employment status. This class schema has been the subject of numerous evaluations focussing on both its construct validity and its criterion validity. In general such evaluations tend to confirm the validity of the class schema on both counts. That is, classes thus defined tend to differ in terms of important social outcomes (Erikson and Goldthorpe 1992; Svallfors 1995, 1997, 2003; Marshall 1997; Marshall et al.1997; Evans 1999), as well as in terms of the employment relations they are supposed to reflect (Evans 1992, 1996; Evans and Mills 1998a, 1998b, 2000).
This paper is structured as follows: the second section discusses the concepts ‘conformism’ and ‘authoritarianism’, and it is decided that ‘conformism’ is the preferred concept for this paper. The third section summarises the various strands of relevant empirical research, and raises some problems and queries. The fourth section presents the data on which the comparative analysis is based, and discusses problems related to empirical indicators, data quality and interpretation. The fifth and sixth sections of the paper contain the empirical analysis, which focuses first on attitudes to sexual behaviour and then on attitudes to civil rights/liberties. The concluding section summarises the main findings of the paper in the light of previous research, and highlights some further implications of the findings for research on stratification and politics.
Authoritarianism and Conformism: Conceptual Distinctions
When reading Lipset's article, the lack of a clear definition of the concept ‘authoritarianism’ becomes apparent. The empirical indicators of ‘authoritarianism’ range from support for extremist rather than democratic parties, to attitudes towards civil liberties and minorities. The closest he seems to come to a definition is that authoritarianism is the opposite of ‘liberalism defined in non-economic terms – so as to support, for example, civil liberties for political dissidents, civil rights for ethnic and racial minorities, internationalist foreign policies, and liberal immigration legislation’ (Lipset 1959: 485).
Several of Lipset's critics have pointed out this lack of conceptual clarity, arguing for example that some of the elements Lipset saw as characteristic for authoritarianism are really not interrelated at all (Schegloff and Kruytbosch 1961; Hamilton 1972: ch. 11). Even if such criticisms may certainly be warranted, many of these commentators seem to miss the crucial contribution made by Lipset, namely that of identifying a fundamental attitude dimension that is virtually unrelated to attitudes towards the (re)distribution of resources. This other dimension concerns attitudes towards external authority and minorities of various kinds, and has, as Lipset points out, only a very tenuous relation to distributive matters. In contemporary electoral research, this dimension is often referred to as the ‘libertarian–authoritarian’ dimension (Evans et al.1996), or the ‘socio-cultural cleavage dimension’ (Rydgren 2002: ch 2).
From a slightly different angle, Kohn suggests that his study of conformist orientations captures essentially the same attitudinal cluster as Lipset's did. These conformist orientations are indicative of ‘intolerance of any behavior at odds with the dictates of authority’ and characterised by ‘an unwillingness to allow others to step out of narrowly confined limits of what is proper and acceptable’ (Kohn 1977: 201).
This paper adopts the concept ‘conformism’ rather than ‘authoritarianism’. The concept of ‘conformism’ bears some resemblance to that of ‘authoritarianism’, but carries stronger connotations of moral rigidity and intellectual inflexibility than the latter, which is more strongly associated with undemocratic values. The reason for choosing ‘conformism’ is two-fold. First, it is rare for members of today's Western societies to espouse outright authoritarian values. Democracy has become a non-partisan consensual value, rejected outright by only a very small group (Dalton 1999; Klingemann 1999). If we focus on the anti-democratic aspects of authoritarianism we risk missing out important dimensions of attitudes towards authority. These dimensions concern more common and mundane aspects of social attitudes, and ‘conformism’ seems better placed than ‘authoritarianism’ to capture the essentials of these attitudes.
Second, the democratic process implies more than just freedom of speech and association; it also provides groups less endowed with market resources with the means to pursue their interests more successfully than on the market (Korpi 1978). In other words, attitudes towards democracy combine conformist/authoritarian dispositions with purely re-distributive concerns, and are therefore not well suited as indicators of the issues at stake here.
The question how ‘conformism’ is best to be indicated, and what its central aspects are, is not straightforward. Following Kohn's discussion about conformist orientations it seems the concept is best captured through focusing on non-distributive issues that in different ways relate to attitudes towards authority. A very useful distinction between types of non-distributive issues is provided by Brint (1985). He argues that ‘social issues’, which he defines as ‘issues separating those who espouse relaxed restrictions on moral and cultural styles from those who defend 'traditional’ restrictive values on these matters’ (Brint 1985: 393), are of two types. The first type relates to civil liberties and civil rights; in these, ‘upper strata’ are typically more liberal than ‘lower strata’. The second set of issues relates to morality and use of military force; in these issues, workers and businessmen tend to be less liberal than professionals (Brint 1985: 393–4). As will be shown below, both these dimensions can be captured in the data on which this paper is based.
His categorisation builds mostly on how class tends to structure the issues, leaving it somewhat unclear how attitudes to civil rights/liberties and to morality/military force differ in terms of content. It could be argued that what is at stake in the case of civil rights/liberties are relations towards public authority, while in the case of morality it is the freedom to live differently from your fellow citizen. It is less clear how attitudes towards the use of military force fit into these dimensions.
Conformism and the Moral Order: Previous Findings and New Issues
The original Lipset thesis was immediately subject to criticism. As outlined above, some of them pointed to a lack of clarity in his use of the concept of authoritarianism. Others have been more concerned about the empirical findings.2 Several authors argue that education rather than class per se, is the key factor determining levels of authoritarianism or conformism (Lipzits 1965; Grabb 1979, 1980; Dekker and Ester 1987). People with higher education tend to be less authoritarian than those with lower education, and once educational differences are controlled for, there are no significant class differences. Houtman's assertion ‘that working-class authoritarianism is basically an authoritarianism of the poorly educated’ (2000: 39) summarises the prevailing wisdom in the field.
In a later paper Lipset concedes this (1981: 479–80) but maintains that the point he was making (in his original article) was that low socio-economic standing, rather than class itself, tends to be accompanied by an espousal of authoritarianism. However, this is a weak argument since his original paper mainly singles out factors in the working-class existence, and not education per se, to explain higher incidences of authoritarianism. The effect of education is never mentioned as an important key mechanism, although ‘low levels of schooling’ are suggested as one of several aspects of the working-class situation. If research would show that it is only education that matters, as much of the empirical work in the field seems to suggest, the original Lipset thesis is in need of basic revision.
This paper follows a different track in the attempt to analyse the current validity of the Lipset thesis. It could be the that the results of later analyses, along with those of Lipset himself, owe more to the indicators of class used than to any attitudinal realities. Thus the failure to find significant class differences once education has been controlled for may in part be attributable to class indicators incapable of capturing the full range of features associated with employment relations. If this is the case, some class effects may become subsumed under effects of education levels.
The research by Melvin Kohn and his associates does indeed seem to indicate that something in the work situation of different classes tends to structure levels of conformity. These work-related factors have an effect on levels of conformist orientations which is irreducible to levels of education. The key mechanism in explaining these class patterns is, they maintain, ‘occupational self-direction’, which is the extent to which one's job is characterised by ‘freedom from close supervision, substantively complex work, and a nonroutinized flow of work’ (Kohn 1977: xxxiv). From occupational self-direction flows intellectual and moral flexibility, and this flexibility is the opposite of conformism as here defined.
Kohn and his collaborators are able to show that classes actually differ in the extent to which jobs are self-directed and that members of different classes do in fact exhibit different degrees of conformist orientations. Further, through longitudinal research they show that different degrees of conformity are actually an effect of being exposed to different work situations, rather than manifestations of occupational self-selection (Kohn and Schooler 1983: ch 5–7).
A quite different, yet related, strand of research is a comparative ethnography which focuses on class, morals and culture (Lamont 1992, 2000). Lamont (1992) studies upper-middle class men in France and the United States, and their socio-economic, moral and cultural ‘boundary work’ in relation to other groups in society. Lamont (2000) extends this comparison to workers in France and the United States, focusing in particular on race and immigration issues. Particularly relevant to the present argument is Lamont's observation that workers and upper middle-class respondents displayed very different levels of tolerance for ‘outgroups’ of various kinds. While the workers were inclined to establish clear moral boundaries vis-à-vis groups they considered ‘less worthy’ in various respects, the upper-middle class men were not, at least not to the same extent. Their sense of worth was largely connected to success in socio-economic terms, while the workers put morality at the centre and found their self-worth in their ability to lead disciplined and morally responsible lives. Lacking in socio-economic means, their struggle for dignity and respect often led to moralistic stances with racist overtones, entailing explicit attempts to morally distance themselves from people ‘above’ and ‘below’.
Even if Lamont points out the substantial variations between individual respondents, and between France and the United States, the thrust of her argument is nevertheless similar to that emerging from survey data: workers have a stronger tendency than middle-class people to apply a fairly rigid standard in judging moral worth.
Whilst Lamont's work is quite recent, most of the relevant survey research is now somewhat dated; the newer survey research on this topic is both scattered and scarce. It might of course be asked whether the class differences that Kohn and his associates identified still hold in a context in which levels of education have generally risen, and in which work practices and organisation have undergone considerable changes.
Similarly, it is uncertain to what extent findings about class conformism apply across national boundaries. As we saw, Lipset suggests that the prevalence of authoritarianism might be expected to be lower in contexts in which the working class is socially integrated into ‘mainstream society’; according to this reasoning, class differences in conformism ought to be less pronounced in more socially integrated political economies. Kohn, on the other hand, doubts that findings will differ between national contexts: ‘[W]e doubt that there are any sizeable industrial societies / … / in which the relationship of class to conformity is much different, or in which occupational self-direction does not play a major part in this relationship’ (Kohn 1977: 196), a claim subsequently corroborated by comparisons between the United States, Poland and Japan (Slomczynski et al.1981; Kohn et al.1990a, 1990b, 1997). Lamont's findings add further complexity, for while they show that comparatively rigid moral standards are common for both the American and the French workers, they also reveal that the application of such standards varies in line with differences in national culture and traditions.3
The present analysis is not concerned with attitudes relating to race and ethnicity as Lamont's study, nor does it use the same indicators as Kohn. It still seems relevant to ask whether the wider institutional and cultural environment gives rise to any observable national variations in the relationship between class and conformism, and specifically whether they manifest themselves in variations of attitudes towards sexuality and civil rights/liberties.
As described above, the four countries which are compared vary considerably in the extent to which national policies and institutions provide social integration and citizenship. If there is any substance to Lipset's suggestion that greater integration of the working class leads to lower levels of authoritarianism/conformism, we should expect class differences to be reduced where: (a) the level of material security is more evenly distributed in the population; (b) levels of political inclusion are higher; (c) welfare policies are more integrative and provided on a universalistic basis; (d) labour markets provide more protection and less direct exposure to market forces. As outlined in the introduction, the four countries in question could be ranked along these dimensions, from Sweden (high levels of integration/protection/participation/equality) to Germany and Britain, and with the United States ranking lowest on all counts. Accordingly, Sweden might be expected to display the smallest class differences in conformism, and the United States the greatest.
Data and Methods
The data derive from the International Social Survey Program (ISSP; <http://www.issp.org>), a comparative attitude survey inaugurated in the mid-1980s, which now involves more than 30 countries. The ISSP represents an attempt to create a truly comparative data set with which to analyse attitudes and values among the populations of industrialised countries (Davis and Jowell 1989). A wide variety of topics have been surveyed, and from 1990 previous modules have been replicated allowing comparison both between nations and over time (Svallfors 1996).
Comparative attitude research represents a promising line of social inquiry, but is at the same time fraught with difficulties which may make results and interpretations fragile (Svallfors 1996; Jowell 1998; Küchler 1998). The most important problem is probably how to establish the cross-national validity of indicators. Attitudes are by their very nature context-dependent, which is why we want to compare them across nations in the first place. This is not of course a problem in itself. A problem arises if we find that it is not values and attitudes that vary across nations, but the meaning and connotations of various concepts. There is an immanent danger of producing research design artefacts instead of comparing and explaining substantive findings.
This problem has been handled by the ISSP by making the questionnaire design a truly cross-national exercise, involving drafting groups comprised of people from several countries and requiring approval from the entire 30-country group. The ability to locate and eliminate problems of cross-national validity thus has been comparatively great within the ISSP. However, interpreting and explaining survey findings requires a thorough contextual understanding, and for this reason researchers are advised to focus their analyses on a few strategically chosen countries rather than attempting to incorporate as many countries as possible.
Translations of questionnaires pose similar problems when it comes to establishing cross-national validity. Concepts often carry nation-specific connotations and as such are difficult to translate. The ISSP has dealt with this both by carefully indicating in the source questionnaire which concept a literal expression refers to and by adopting careful translation procedures.4 It should be emphasised that the countries in question are all examples of Western capitalist democracies, and establishing cross-national validity in concepts and translations does not seem to pose an insurmountable problem.
Further problems of survey research relate to sampling and response rates. All four countries in this study use random sampling methods, surveying the adult population (>18 years). Non-responses present a potentially greater threat to the representativeness of the samples. The samples achieved are reasonably representative of the adult population in the countries in question, given that response rates are reasonable and substitution of respondents is not permitted. Response rates in the surveys used here range from 56 to 58% in Britain, 53 to 63% in Germany, 63 to 73% in Sweden, and 69 to 71% in the United States. These levels are in line with the expected for these countries. As far as we know, the achieved samples are not systematically biased in terms of their age, gender and education distributions, though we still know very little about how non-responses affect the attitudinal representativeness of the samples. It seems less likely that class differences within and between countries would be much affected by differences in response rates.5
In choosing between possible indicators of ‘conformism’, there are several possibilities in the ISSP data. As argued above, attitudes towards democracy and the political system are not terribly useful indicators of, in Kohn's words, ‘intolerance of any behavior at odds with the dictates of authority’, and ‘an unwillingness to permit others to step out of narrowly defined limits of what is proper and acceptable’ (Kohn 1977: 201). Inspired by Brint's categorisation referred to above the chosen indicators are (a) attitudes towards sexual relations outside (traditional) marriage, and (b) attitudes towards civil rights/liberties. Attitudes towards sexual behaviour were mapped by the 1994 survey on ‘Family and gender’ and by the 1998 ‘Religion’ module, and attitudes towards civil rights/liberties were mapped by the ‘Role of Government’ survey, last conducted in 1996.6 At stake in both these issue areas is precisely the ‘intolerance (and perhaps fear) of non-conformity’ that Kohn argues is at the heart of conformist orientations (1977: 202).
As will be argued in the next two sections, the indicators used do not lend themselves to being summarised in indexes. Nor are they well-suited for use in standard regression models in multivariate analysis, since they do not fulfil assumptions about measurement levels and distributions. Instead the analysis proceeds by first showing class differences for individual indicators, before using ordinal regression for multivariate analysis of each indicator. Ordinal regression is a statistical technique adapted for the kind of ordinal level data present here (Winship and Mare 1984; Long 1997: ch 5; Orme and Buehler 2001).
One assumption behind ordinal regression is that categories of the dependent variable distinguish ranges of some latent variable of interest (in this case attitudes), or are qualitative distinctions which can be ordered in a meaningful way. The parameter estimates can be interpreted as logistic regression coefficients, provided that they remain the same across different thresholds of the dependent variable. The latter assumption (that estimates are more or less the same across levels of the dependent variable) can be tested by comparing the ordinal regression model with the unconstrained logistic regressions (‘test of parallel lines’).
The distribution of the manifest dependent variable (the ordered categories) may be linked to the latent variable in a number of ways. The one used here, logit, assumes fairly evenly distributed categories of the manifest variable. In cases where answers were heavily skewed between categories, other link functions were tested.7
An important drawback of ordinal (or any other non-linear) regression is that estimates are not immediately meaningful in the same way as coefficients from a linear regression. In order to increase interpretability, differences in estimated probabilities between different categories will be presented alongside estimates, as will be explained in the adjoining text.
Class – the main independent variable in the analysis – is indicated by the class schema devised by Erikson and Goldthorpe and outlined above. The schema exists in several versions; in this paper, the occupational codes used in the countries in question are recoded into five classes: unskilled workers, skilled workers, routine non-manual employees, Service class II (lower level controllers and administrators), and Service class I (higher level controllers and administrators).8 Since the debate on class and conformism is not at all clear on the status of the self-employed this category is omitted from the analyses. At times, and as indicated in the text, these five categories are collapsed into three by merging the two worker categories and the two Service class categories.
Attitudes Relating to Sexual Behaviour
The first way in which this paper approaches the study of class conformism is through an analysis of attitudes towards sexual behaviour. Both the 1994 (‘Family and Gender’) and the 1998 (‘Religion’) ISSP modules contained a battery of questions relating to this topic. These ask how respondents feel about sex outside the confines of (traditional) marriage, and more specifically, how they feel about extramarital and homosexual sexual activity. The exact wording of the questions (in English) can be found in Appendix Table A1, along with the percentage balance in the four countries studied. The percentage balance is calculated by subtracting the share answering ‘always wrong’ to a question from the share answering ‘not wrong at all’, such that the higher the value, the less conformist attitude represented.9 As the table shows, Sweden and Germany are on average more ‘liberal’ than the other two countries, whilst the United States stands out as being the least ‘liberal’ of the four. There is a clear tendency for attitudes to become less conformist between 1994 and 1998, but these differences over time should not be over-interpreted, owing to differences in questionnaire contexts. In the following analyses, the 1994 and 1998 data are merged into one data set in order to increase sample size and streamline the results.10
The three indicators are not readily summarised into indexes in all four countries.11 The average Pearson correlation is 0.38 in the United States, 0.30 in Britain, and 0.32 in Germany, but only 0.22 in Sweden. Reliability measures for summed indexes are too low to indicate well functioning indexes.12 In the following, we will therefore seek to detect class differences by analysing one indicator at a time.
Attitudes towards infidelity, by class.
Attitudes towards homosexuality, by class.
The results presented in Figure 2 reveal even clearer class differences across countries. In all four countries, the working class is the least accepting of homosexuality, and the Service class the most accepting. The considerable class differences in attitudes towards homosexuality is interesting because negative attitudes towards homosexuality have been shown to be tightly wedded to a host of other illiberal and intolerant attitudes (Herek 2000; Österman 2001).
In sum, there are quite substantial differences between classes in their views about different aspects of sexual behaviour. But is is unclear whether such differences are manifestations of class mechanisms, or whether they mainly reflect differences in educational experiences, or the age and gender composition of classes. In order to test this, a series of ordinal regression analyses were carried out, where dummy variables for class were entered alongside several other variables that were thought to be of interest. It should be noted that the main purpose here was not to explain variations in attitudes towards sexual behaviour, but rather to test whether the observed class differences can be explained by other factors than class itself.
In Table 1 below, class is entered in the models alongside variables for gender, age, and education.14 These three variables were hypothesised to correlate quite closely with class and conformity and perhaps affect the relation between them. Women are more often found in routine non-manual jobs than men are and less often in Service class positions. The young might be expected to espouse less conformist attitudes than older people, reflecting both generation and life-cycle differences. They also tend to have higher formal education but lower class positions than older people. People with higher education are more often found in Service class positions than people with lower education; they are also more likely to display non-conformist attitudes.
| Infidelity | Homosexuality | |||||
|---|---|---|---|---|---|---|
| Variables in the model: | I: Class | II: (I)+gender, age | III: (II)+ education | I: Class | II: (I)+gender, age | III: II+education |
| Britain | ||||||
| Worker | −0.60*** | −0.57*** | −0.45** | −0.65*** | −0.58*** | −0.37** |
| Routine non-manual | −0.30* | −0.20 | −0.14 | −0.21 | −0.41** | −0.28* |
| Nagelkerke Pseudo-R2 | 0.021 | 0.080 | 0.086 | 0.027 | 0.186 | 0.198 |
| Germany | ||||||
| Worker | −0.60*** | −0.58*** | −0.44*** | −0.69*** | −0.68*** | −0.45*** |
| Routine non-manual | −0.37*** | −0.31*** | −0.20* | −0.16* | −0.26** | −0.06 |
| Nagelkerke Pseudo-R2 | 0.021 | 0.059 | 0.067 | 0.031 | 0.127 | 0.150 |
| Sweden | ||||||
| Worker | −0.97*** | −0.88*** | −0.54*** | −0.61*** | −0.67*** | −0.23* |
| Routine non-manual | −0.76*** | −0.63*** | −0.40** | −0.26* | −0.45*** | −0.15 |
| Nagelkerke Pseudo-R2 | 0.055 | 0.087 | 0.101 | 0.022 | 0.091 | 0.111 |
| USA | ||||||
| Worker | −0.39** | −0.47*** | −0.19 | −0.80*** | −0.78*** | −0.45*** |
| Routine non-manual | −0.49*** | −0.33*** | −0.07 | −0.48*** | −0.54*** | −0.26* |
| Nagelkerke Pseudo-R2 | 0.012 | 0.021 | 0.053 | 0.036 | 0.091 | 0.115 |
| Infidelity | Homosexuality | |||||
|---|---|---|---|---|---|---|
| Variables in the model: | I: Class | II: (I)+gender, age | III: (II)+ education | I: Class | II: (I)+gender, age | III: II+education |
| Britain | ||||||
| Worker | −0.60*** | −0.57*** | −0.45** | −0.65*** | −0.58*** | −0.37** |
| Routine non-manual | −0.30* | −0.20 | −0.14 | −0.21 | −0.41** | −0.28* |
| Nagelkerke Pseudo-R2 | 0.021 | 0.080 | 0.086 | 0.027 | 0.186 | 0.198 |
| Germany | ||||||
| Worker | −0.60*** | −0.58*** | −0.44*** | −0.69*** | −0.68*** | −0.45*** |
| Routine non-manual | −0.37*** | −0.31*** | −0.20* | −0.16* | −0.26** | −0.06 |
| Nagelkerke Pseudo-R2 | 0.021 | 0.059 | 0.067 | 0.031 | 0.127 | 0.150 |
| Sweden | ||||||
| Worker | −0.97*** | −0.88*** | −0.54*** | −0.61*** | −0.67*** | −0.23* |
| Routine non-manual | −0.76*** | −0.63*** | −0.40** | −0.26* | −0.45*** | −0.15 |
| Nagelkerke Pseudo-R2 | 0.055 | 0.087 | 0.101 | 0.022 | 0.091 | 0.111 |
| USA | ||||||
| Worker | −0.39** | −0.47*** | −0.19 | −0.80*** | −0.78*** | −0.45*** |
| Routine non-manual | −0.49*** | −0.33*** | −0.07 | −0.48*** | −0.54*** | −0.26* |
| Nagelkerke Pseudo-R2 | 0.012 | 0.021 | 0.053 | 0.036 | 0.091 | 0.115 |
*** = Wald statistic significant at 0.001-level; **=0.01-level; *=0.05-level.
Models estimated with SPSS 10 (‘PLUM’). Link function: logit. Reference category: Service Class.
In Table 1 coefficients for class dummy variables are therefore displayed, first with no other variables in the models, then including dummy variables for gender and age, and lastly including in addition dummy variables for education. Only the coefficients for class are displayed (information on model specifications and output is available from the author). Preliminary tests (along with inspection of figures) showed that differences in attitudes between Service class I and II, and between skilled and unskilled workers, were mostly rather small. In order to reduce the amount of displayed output, and to simplify the models, the two Service class groups were therefore collapsed into one category, as were the two worker groups.
The estimates presented in Table 1 reveal some interesting relationships. Regarding attitudes towards infidelity we find clear and consistent differences between class categories. Workers are significantly more likely than the Service class to take a conformist position. These differences are reduced by taking education into account, but they only become insignificant in the United States.15
At the right-hand side of the table, showing results for attitudes towards homosexuality, estimates for class categories are clearly significant for all four countries before education is introduced in the models. Once we control for education, however, these class effects become much weaker (though they are still significant). This shows that much of the relatively large class differences in attitudes towards homosexuality identified in Figure 1c is due to educational differences rather than class per se.
In order to provide a more meaningful interpretation than the one given by the estimates, expected probabilities can be estimated for different combinations of categories and expressed as ‘expected percentage balances’. The expected percentage balance is calculated by subtracting the expected probability for having a ‘conformist’ attitude from the expected probability for having a ‘non-conformist’ attitude.
Attitudes towards infidelity.
Attitudes towards homosexuality.
Although age and gender have not been included in the models, the Net Class Difference (as expected) mainly reproduces the patterns found in Table 1. Class and education differences are clear in the European countries for attitudes to infidelity, while in the United States, education differences are clear but class differences negligible (Figure 3). The Net Class Difference for attitudes to homosexuality (Figure 4) is on average larger than that for attitudes to infidelity, Sweden being the only country demonstrating small class differences on this issue. Differences between education categories are also much larger in attitudes to homosexuality than for infidelity, and dwarf the class differences.
Attitudes Towards Civil Rights/Liberties
In order to attitudes towards civil rights/liberties, two items from the 1996 ISSP survey ‘Role of Government’ were used. One asked whether ‘people should obey the law without exception, or whether there are exceptional occasions on which people should follow their consciences even if it means breaking the law’? The other asked whether it is worse ‘to convict an innocent person’ or ‘to let a guilty person go free’? The answers were recoded so that respondents choosing the ‘liberal’ answer were juxtaposed to those choosing the ‘illiberal’ answer. Those responding ‘Can't choose’ were placed in a middle category. The percentage balance was then calculated by subtracting the share choosing the ‘illiberal’ answer from the share choosing the ‘liberal’ answer. As shown in appendix Table A2, some country differences are evident, Sweden and Germany appearing less conformist than Britain and the United States.
Attitudes towards following the law without exception, by class.
Attitudes towards miscarriages of justice, by class.
Figure 5 illustrates the substantial class differences observed in all the countries except Sweden, where no significant differences were found. In the other three countries workers were more inclined to choose the ‘follow the law without exception’ option than members of the Service class, with lower level non-manuals occupying a position between these two.
Figure 6 showing class attitudes relating to criminal justice, reveals significant class differences in all four countries, albeit the differences are smaller in Germany than in the other three countries. Workers tend to be less troubled by the thought of convicting an innocent person than members of the Service class, with the routine non-manuals mostly occupying a mid-positition between workers and the Service class. The magnitude of class differences is roughly similar for all the countries, with the exception of Germany.
As in the previous section, it is pertinent here to ask to what extent these class differences can be explained by the different age, gender and educational composition of classes. To test this, the same set of variables was entered in the same set of models as in the previous section. The results are displayed in Table 2.
| Follow the law without exception | Miscarriage of justice | |||||
|---|---|---|---|---|---|---|
| Variables in the model: | I: Class | II: (I)+gender, age | III: (II)+ education | I: Class | II: (I)+gender, age | III: II+education |
| Britain | ||||||
| Worker | −0.47** | −0.42* | −0.02 | −0.62*** | −0.59*** | −0.18 |
| Routine non-manual | −0.23 | −0.26 | 0.07 | −0.54** | −0.40* | −0.08 |
| Nagelkerke Pseudo-R2 | 0.012 | 0.053 | 0.074 | 0.023 | 0.038 | 0.059 |
| Germany | ||||||
| Worker | −0.59*** | −0.57*** | −0.34** | −0.26** | −0.25** | −0.19 |
| Routine non-manual | −0.28** | −0.26* | −0.08 | −0.09 | −0.08 | −0.03 |
| Nagelkerke Pseudo-R2 | 0.018 | 0.034 | 0.041 | 0.004 | 0.006 | 0.007 |
| Sweden | ||||||
| Worker | 0.04 | −0.02 | 0.18 | −0.72*** | −0.72*** | −0.27 |
| Routine non-manual | 0.19 | 0.24 | 0.45* | −0.68** | −0.69** | −0.30 |
| Nagelkerke Pseudo-R2 | 0.001 | 0.082 | 0.083 | 0.027 | 0.032 | 0.043 |
| USA | ||||||
| Worker | −0.57*** | −0.70*** | −0.21 | −0.49*** | −0.55*** | −0.24 |
| Routine non-manual | −0.50** | −0.42** | −0.03 | −0.42** | −0.30 | −0.07 |
| Nagelkerke Pseudo-R2 | 0.022 | 0.062 | 0.100 | 0.015 | 0.029 | 0.046 |
| Follow the law without exception | Miscarriage of justice | |||||
|---|---|---|---|---|---|---|
| Variables in the model: | I: Class | II: (I)+gender, age | III: (II)+ education | I: Class | II: (I)+gender, age | III: II+education |
| Britain | ||||||
| Worker | −0.47** | −0.42* | −0.02 | −0.62*** | −0.59*** | −0.18 |
| Routine non-manual | −0.23 | −0.26 | 0.07 | −0.54** | −0.40* | −0.08 |
| Nagelkerke Pseudo-R2 | 0.012 | 0.053 | 0.074 | 0.023 | 0.038 | 0.059 |
| Germany | ||||||
| Worker | −0.59*** | −0.57*** | −0.34** | −0.26** | −0.25** | −0.19 |
| Routine non-manual | −0.28** | −0.26* | −0.08 | −0.09 | −0.08 | −0.03 |
| Nagelkerke Pseudo-R2 | 0.018 | 0.034 | 0.041 | 0.004 | 0.006 | 0.007 |
| Sweden | ||||||
| Worker | 0.04 | −0.02 | 0.18 | −0.72*** | −0.72*** | −0.27 |
| Routine non-manual | 0.19 | 0.24 | 0.45* | −0.68** | −0.69** | −0.30 |
| Nagelkerke Pseudo-R2 | 0.001 | 0.082 | 0.083 | 0.027 | 0.032 | 0.043 |
| USA | ||||||
| Worker | −0.57*** | −0.70*** | −0.21 | −0.49*** | −0.55*** | −0.24 |
| Routine non-manual | −0.50** | −0.42** | −0.03 | −0.42** | −0.30 | −0.07 |
| Nagelkerke Pseudo-R2 | 0.022 | 0.062 | 0.100 | 0.015 | 0.029 | 0.046 |
*** = Wald statistic significant at 0.001-level; **=0.01-level; *=0.05-level.
Models estimated with SPSS 10 (‘PLUM’). Link function: logit. Reference category: Service Class.
Results for the two indicators in Table 2 tell similar stories. Initially (model I) we find clear class differences in all cases except for one of the indicators in Sweden. Introducing gender and age into the model does not substantially change this pattern (model II), but once education is introduced (model III) virtually all class differences disappear, with the exception of those relating to one of the indicators in Germany. 18 The differences between the Service class and workers observed in Figures 5 and 6 are almost entirely attributable to different educational levels that typify different classes.
Attitudes towards following the law without exception.
Attitudes towards miscarriages of justice.
Concluding Discussion
Is it still the case, as Lipset argued more than four decades ago, that the working class is less tolerant and open-minded than other groups, and more conformist in its orientations? And do any such class differences vary across national contexts?
The analyses presented here indicate that the answer to the second question is clearly ‘no’. It is hard to detect any systematic country differences in the class structure of conformism. While there are certainly large country differences in the general level of conformism, with the United States standing out from the European countries in the study, the relationship between class and conformism differs little between countries. Thus the attitudes of the Swedish working class are no closer to those of the Swedish Service class than the attitudes of the US workers are to those of the American Service class. It is true that Swedish workers are less conformist in absolute terms than their American counterparts, but the relative distance to other classes is more or less the same. The initial assumption that class differences in conformism might be affected by the extent to which workers are economically and socially integrated thus seems ill-founded. Instead, Melvin Kohn's assumption that class differences in this respect would remain more or less constant across national contexts finds greater support.
This is not to imply that national contexts are without effects for attitudes towards these issues. National contexts do seem to impact significantly on aggregate levels, and in ways that clearly sets the United States apart from the European countries. National institutions and cultural frameworks would obviously feature strongly in any attempt to explain aggregate levels of conformism.19
The first question requires a more equivocal answer. Class differences certainly occur in the expected direction, with workers being more conformist than members of the Service class. The empirical analysis showed that class differences in attitudes in the sexual realm become less evident once we take education into account, but that significant class differences nevertheless remain. In contrast, attitudes in questions of civil rights/liberties do not differ between classes once we control for different levels of education. It seems there are some aspects of working-class conformism that are not reducible to differences in educational levels; these differences may have gone unnoticed in other recent survey research owing to the class variables used or the indicators chosen for conformism/authoritarianism.
If national institutions cannot help elucidate the actual mechanisms through which class differences in conformism are produced, we must look elsewhere in order to explain (a) why such differences appear at all and (b) why they become less evident, and sometimes disappear altogether, once education is controlled for.
It is clear that any such explanation should acknowledge the liberating effects of higher education,20 and thus extend beyond the immediate class situation. The significance of education should be understood here in two different ways. On the one hand, we might expect the content of higher education to engender greater intellectual and moral flexibility in those exposed to it. These effects are likely to be stronger in some fields of higher education than in others, though lack of data precludes us from testing this assumption. On the other hand, a likely degree of self-selection into higher education suggests that open-minded and curious individuals will be over-represented among the higher educated. The content effects and selection effects are likely to work in tandem, with educational content and peer influence re-enforcing selection effects.
Some class differences, however, remain after taking education into account, and here it seems that additional mechanisms to the liberating/selecting effects of education are at play. Two possible mechanisms have already been suggested: one which reflects workers’ needs to maintain moral boundaries vis-à-vis other groups in society; and one through which experiences from work can affect degrees of intellectual and moral flexibility. The available data do not permit us to pronounce on the validity of these explanations. Better survey data could probably go some way in establishing the extent to which these are competing or complementary explanatory mechanisms, by studying what kind of variables intervene on the path from class to conformism. But detailed ethnographic work, taking as its point of departure the findings of survey research, could also provide valuable insight into the operation of these and other mechanisms (Goldthorpe 2000a: 86–7).
Finally, what are the wider implications of all this, and why should anyone not specifically interested in studying conformism care about the results? There are two answers, one comments on the implications of the findings for understanding the nature of social stratification, and the other highlights the political implications of the study.
Firstly, the results presented indicate that the effects of stratification extend beyond issues directly connected to class. In the case of attitudes towards (re)distribution, it is easy to see the self-interest involved on the part of different classes. Their livelihood and material security are directly affected by issues such as income distribution, welfare state intervention, and work conditions, and it is therefore not surprising that class differences in attitudes occur. However, the research tradition of which this paper is a part has demonstrated that class differences also occur on issues far removed from questions of material well-being and security.
Even if some, or even most, of these differences are the result of different exposure to higher education, it is still the case that as a group workers tend to be more morally rigid than higher level non-manuals. Such differences in conformism are then a class effect of the second order, whereby differences in material resources, success and security are translated into differing moral stances. It might even be ventured that in an increasingly cosmopolitan and flexible international economy, such moral rigidity may act as a further impediment to educational and occupational success.21
A second implication of the findings is that they illustrate the complexity of the relation between class and party politics. Lipset (1959: 499–500) observed that leaders of social democratic parties tended to be ‘liberal’ (in the American sense) in both economic and non-economic terms. Their core voters, the workers, tended however to be economically liberal but morally conservative, opening a space for other parties to appeal to workers on grounds other than issues of (re)distribution.
This is the case also in the democracies of the early twenty-first century. While non-democratic leftist parties, which were Lipset's main concern, have more or less perished, the rise of the populist right in many Western democracies is a related phenomenon. Their pro-redistribution stances combined with the tendency to draw sharp moral boundaries in relation to various kinds of ‘other’ sometimes strikes a chord among workers and those with low levels of education (Rydgren 2002). In contrast, the democratic left in the Western world tends to cultivate multi-cultural and libertarian values in issues relating to life-style and minorities. The results of this paper and other research suggest that workers’ support for these latter parties tends to be not because of, but in spite of, these values. As long as the democratic left can persuade individuals with few marketable resources that their material security and livelihood are best secured by choosing social democracy, their moral illiberalism may remain politically inert. But when such attempts fail, other political entrepreneurs may succeed in convincing them that their allegiance should lie elsewhere.
Acknowledgements
The research has been financed by the Swedish Council for Working Life and Social Research (FAS) and the Bank of Sweden Tercentenary Foundation. I thank Christina Axelsson, Rickard Danell, Jonas Edlund, Robert Erikson, Nils Eriksson, David Grusky, Björn Halleröd, Mike Hout, Lena Karlsson, Jonathan Kelley, Staffan Kumlin, Mikael Nordenmark, Bo Rothstein, Jens Rydgren, Al Simkus, and, in particular, Ingrid Schild for useful advice and comments.
Footnotes
For example, many white American workers were occupied with what they perceived as the low work ethic and lax morals of blacks, while black workers were concerned with the lack of companionship and caring among whites. Immigrants, on the other hand, were not considered morally suspicious. French white workers were less obsessed with race, but were worried about immigrants from North Africa, to whom they attributed many of the same negative traits that white Americans ascribed to blacks.
The procedures for translating the questionnaire are very similar in Germany and Sweden, the two countries whose translations of the source questionnaire are relevant here. Independent translations by research team members are compared, discussed and decided on in a wider forum. Specific translation problems are discussed with the ISSP secretariat. No reverse translation is carried out.
No weighting to correct for different response rates has been applied, since such weighting (a) builds on the unfounded assumption that respondents and non-respondents have the same response pattern, and (b) makes little difference to results. Weighting has only been used to correct for different sampling probabilities in Germany (where former East Germany was oversampled) and in Sweden (where a subsample of non-respondents was selected for telephone interview). Differences between weighted and unweighted samples are negligible in Sweden, but make some difference in Germany. (Detailed information about weighting is available from the author.)
These attitudes were also surveyed in 1990, but at that time Sweden was not yet a member, and the occupational classification was not well suited to the re-classification conducted here.
The class variable is based on the occupation of the individual. Individuals who are not currently working are classified according to their most recent occupation (if any). Individuals who have never worked but whose spouse works or has worked are classified according their spouses’ occupations. For Germany, the US and the Swedish 1998 sample recodings were made from the International Standard Classification of Occupations (ISCO). The logic and procedures for reclassifying ISCO into Goldthorpe classes are adapted from Ganzeboom and Treiman (1996; 2003). Minor deviations between their recodings and the ones used here occur due to national deviations from the standard ISCO classification. In Britain, recodings were based on the British Standard Occupational Classification (SOC, 3-digit level) and Employment Status Variables, following a scheme provided by John Goldthorpe and Anthony Heath (1992). The data for this reclassification was supplied by the National Centre for Social Research, since the ISSP archive data file only included SOC at the two-digit level. The Swedish 1994 and 1996 data uses the Socio-economic Classification (SEI) and the Nordic Occupational Classification (NYK). The former occupation schema is very similar to the Goldthorpe schema, while the latter is a national adaptation of ISCO68. The reclassification from the Swedish occupational classifications was adapted from Robert Erikson, Janne Jonsson and Michael Tåhlin at the Swedish Institute for Social Research. The classification used here is not as detailed as theirs, since some of the relevant information is missing from the Swedish ISSP data. Files of recodings may be obtained from the author.
This remaining answer categories are ‘almost always wrong’ and ‘wrong only sometimes’. They were merged with the ‘can't choose’ category into a middle category.
Separate analyses of the 1994 and 1998 data are available from author. None of the conclusions drawn in this paper changes is affected by any differences between the two years.
The low scalability of the indicators in this section and the following is a cause of concern, since it implies that the indicators do not measure exactly the same thing, in spite of their face validity. A couple of observations need to be made in order to defend the procedures applied here:
(a) As will be shown, class patterns in attitudes are essentially the same across different indicators for each dimension. This also means that the use of summarised indexes – although it means violating the assumptions behind the construction of such indexes–yields essentially similar results to the ones presented here (SPSS runs available from author).
(b) Changing analytical techniques from the ones applied here to techniques using latent variables (such as LISREL) would not be appropriate, since the low reliability of the manifest indexes would lead to doubts about the content of latent variables. Instead, it was decided to stay with manifest indicators and display results item-by-item.
Cronbach's α runs from 0.30 (Sweden 1998) to 0.50 (Germany 1994).
All identified class differences are statistically significant (0.05-level).
Education is coded into three categories: (a) those with the socially expected minimum in each country, (b) those with secondary or tertiary education without university degree, (c) those who have a university degree. This is the most detailed coding that can be made on a comparative basis. Age is coded into 6 age groups: 18-25, 26-35, 36-45, 46-55, 56-65, 66-. This grouping is preferred to applying age as a continuous variable, in order to take into account possible curve-linear relationships and to make computations of probabilities easier and goodness-of-fit measures more reliable. Men, low education, and 66- are used as reference categories in the models.
It is mainly age that accounts for different attitudes to infidelity. Education also matters for attitudes towards infidelity: the highly educated are more tolerant than those with less education. The same pattern is found in attitudes to homosexuality but here we also find substantial gender differences, with women being more accepting than men (SPSS runs available from author).
The Net Class Difference is achieved by the following steps:
1. Estimate the expected probabilities for conformist and non-conformist attitudes, in three classes at three different education levels (for education levels, cf. note 14).
2. Calculate the expected percentage balance, for each class at each education level, by subtracting the expected probability for having a ‘conformist’ attitude from the expected probability for having a ‘non-conformist’ attitude.
3. Calculate the difference between the expected percentage balances for the Service class and workers, for each of the three education levels.
4. Sum the three differences achieved in step 3 and divide the sum by three to get the Net Class Difference.
The Net Education Difference is achieved by replacing step 3 with ‘Calculate the difference between the expected percentage balances for university graduates and those with low education, for each of the three classes.’
Pearson's R varies from 0.06 in Germany to 0.12 in Sweden and the US. Cronbach's α varies from 0.11 in Germany to 0.21 in Sweden and the US.
Attitudes to following the law without exception are heavily structured by age, with education playing some role. Attitudes towards miscarriages of justice differ less by age category; being mostly affected by education (except in Germany where absolutely no differences are found between any of the categories).
As an example of an important institutional difference in these matters one could point to the fact that homosexual relations are still formally illegal in a large number of American states.
A parallel might here be drawn to Paul Willis’ (1977) classic ethnographic study of working-class boys at school, which illustrates how the racist and sexist attitudes of ‘the lads’ tend to hasten their exit from the education system and their entrance into working class jobs.
| Balance* | Britain | Germany | Sweden | USA | ||||
|---|---|---|---|---|---|---|---|---|
| 1994 | 1998 | 1994 | 1998 | 1994 | 1998 | 1994 | 1998 | |
| Do you think it is wrong or not wrong if a married person have sexual relations with someone else than his or her husband or wife? | −63 | −54 | −46 | −31 | −65 | −55 | −77 | −77 |
| And what about sexual relations between two adults of the same sex? | −27 | −8 | −8 | +6 | −19 | +13 | −46 | −34 |
| (n) (min) | 978 | 794 | 3410 | 1993 | 1464 | 1182 | 1358 | 1262 |
| Balance* | Britain | Germany | Sweden | USA | ||||
|---|---|---|---|---|---|---|---|---|
| 1994 | 1998 | 1994 | 1998 | 1994 | 1998 | 1994 | 1998 | |
| Do you think it is wrong or not wrong if a married person have sexual relations with someone else than his or her husband or wife? | −63 | −54 | −46 | −31 | −65 | −55 | −77 | −77 |
| And what about sexual relations between two adults of the same sex? | −27 | −8 | −8 | +6 | −19 | +13 | −46 | −34 |
| (n) (min) | 978 | 794 | 3410 | 1993 | 1464 | 1182 | 1358 | 1262 |
*=Share stating ‘Not wrong at all’ minus share stating ‘always wrong’.
| Balance* | Britain 1996 | Germany 1996 | Sweden 1996 | USA 1996 |
|---|---|---|---|---|
| In general, would you say that people should obey the law without exception, or are there exceptional occasions on which people should follow their consciencies even if it means breaking the law? | +22 | +39 | +33 | +14 |
| All systems of justice make mistakes, but which do you think is worse … to convict an innocent person?/to let a guilty person go free? | +30 | +53 | +53 | +39 |
| (n) (min) | 975 | 3330 | 1340 | 1303 |
| Balance* | Britain 1996 | Germany 1996 | Sweden 1996 | USA 1996 |
|---|---|---|---|---|
| In general, would you say that people should obey the law without exception, or are there exceptional occasions on which people should follow their consciencies even if it means breaking the law? | +22 | +39 | +33 | +14 |
| All systems of justice make mistakes, but which do you think is worse … to convict an innocent person?/to let a guilty person go free? | +30 | +53 | +53 | +39 |
| (n) (min) | 975 | 3330 | 1340 | 1303 |
*=Share choosing ‘liberal’ answer minus share choosing ‘illiberal’ answer.
References
Stefan Svallfors is Professor in Sociology at Umeå University, Sweden. His research field is the comparative study of attitudes and values and their links to social structure and institutions. He is involved in two major comparative attitude research programmes: the International Social Survey Program and the European Social Survey. His current research mainly concerns class differences in attitudes in Western countries.