Intersections of class, lifestyle and politics. New observations from Norway

Although social class was once central to political sociology, it has become increasingly less so; many analysts now believe that one’s class position is less important in determining political attitudes and political party preferences. Simultaneously, more attention has been paid to what might be called the culturalization of politics, as epitomized by the US culture wars and stereotypes like the “latte-drinking liberal.” Here, political attitudes are regarded as primarily structured by people’s lifestyles and broader way of life. But do political preferences have to be explained by either relations of sociomaterial conditions (e.g. class) or cultural orientations (e.g. status and lifestyles)? In this article, we argue in favor of an approach that aims to reconcile these factors, allowing for the empirical mapping of whether and how they intersect in shaping political party preferences. We investigate this by using detailed Norwegian survey data to measure the extent to which intraclass heterogeneity in political party choices can be accounted for by the interaction between class and lifestyle. We employ a novel combination of Multiple Correspondence Analysis (MCA) and Chi-Squared Automatic Interaction Detection (CHAID). In drawing on constructed models of the social space and the space of lifestyles, we show that there are important correspondences between these spaces and that their interaction may help account for party choices. The results highlight the need for a more complex account than that suggested in recent works of cultural stratification research.


Introduction
Pivotal to the various claims about the "death of class" has been the contention that class is no longer an important source of political division and cultural identification. Several authors have argued that the bread-and-butter "politics of life chances" were giving way to a more cultural or value-based "politics of life choices," concerned with "how we should live in a world where everything that used to be natural (or traditional) now has in some sense to be chosen or decided about" (Giddens 1994, pp. 90 f.; see also Beck and Beck-Gernsheim 2002;Pakulski and Waters 1996). This contention was linked to the notion that culture and consumption were supplanting class and work as a source of self-identity and political organization (Bauman 2004;Giddens 1991) and that the link between class and culture was increasingly being disrupted (Beck 1992). In these accounts, culture was seen as trumping class as the main structuring factor in political divisions.
Though particularly influential in social theory in the 1990s, these grand claims have since been challenged. Several authors have shown that class continues to shape voting and political attitudes pertinent to the "old" politics of life chances, e.g. stances related to economic (re)distribution and state intervention (Evans 2000;Evans and Tilley 2017). Meanwhile, some analysts have claimed that the "new" politics of life choices-e.g. stances related to gender issues, environmental concerns and immigration policies-are shaped by purportedly non-class factors, such as education and status (Chan and Goldthorpe 2007;Chan et al. 2020;Houtman 2001).
It follows that, insofar as the value-based politics of life choices increasingly guide voting behavior and identities, class is becoming a less pervasive factor in contemporary political divisions. However, as we will argue here, such accounts reproduce an unfortunate bifurcation of class and culture that precludes the idea of class-cultural voting and an analysis of how political behavior may result from a systematic interplay of material and symbolic factors. Rather than a priori defining class and culture as theoretically independent factors, we argue that we need approaches that can investigate how these factors may intersect in complex ways.
In this article, we aim to develop a third approach, extending the perspective of what has been dubbed "cultural class analysis" (Atkinson 2010;Savage 2003). In this view, class divisions are understood as both sociomaterial and cultural in the sense that they involve both life chances and lifestyles. Central to this perspective is a theoretical model proposed by Bourdieu (1984), suggesting that there are homologies, or structural correspondences, between three social structures: class ("the social space"), culture ("the space of lifestyles") and politics ("the political space"). Specifically, this model suggests that the main structures of the social space-differences in terms of (i) capital volume (high versus low volumes) and (ii) capital composition (the preponderance of either cultural or economic capital)-are mirrored in the main structures of the spaces of lifestyles and politics. Although controversial, recent work has assessed this model empirically, demonstrating that both lifestyles (Flemmen et al. 2019a;Prieur et al. 2008) and political attitudes (De Keere 2018;Harrits 2013;Jarness et al. 2019) correspond to the primary dimensions of the social space.
Expanding on these accounts, we highlight a novel and crucial point: political heterogeneity within social classes (and class fractions) can be accounted for by the intersection of class and lifestyle. Specifically, despite the enduring significance of a connection between class and party choice, this connection is not absolute. A considerable proportion of people of any class do not conform to the central tendency of that class as a whole: there are manual workers who do not vote Labour and there are capitalists who do not vote Conservative. In the following analysis, we assess whether and how patterns of such class "anomalies" can be explained in terms of the lifestyles of members of specific classes.
Our case at hand is contemporary Norway. We draw on survey data from the 2011 round of the Norwegian Monitor, collected by the IPSOS MMI (N = 3980). These data are unique and well suited for our purposes here as they comprise detailed data on individuals' capital possessions, lifestyles and political party choices. To assess the intersection of positions in the social space and the space of lifestyles in accounting for party choice, we draw on previously published constructions of these two spaces (see Flemmen et al. 2018;Flemmen et al. 2019a). In the analysis we present here, we scrutinize how these spaces intersect in structuring voting by means of Chi-Squared Automatic Interaction Detection (CHAID) (Kass 1980).
We start with a general overview of the sociological research on the nexus of social class, voting and the cultural dimension (Chapter 2), then briefly sketch the Norwegian case (Chapter 3) and our data and research design (Chapter 4). In chapter 5, we present our Bourdieu-inspired analysis on the homology of social space and the space of lifestyles based on the abovementioned Norwegian data. In the main chapter 6, we then provide an in-depth discussion of the intersections of the social space and the space of lifestyles in relation to party choice, before summarizing our results in the conclusion (Chapter 7).

Class, voting and culture
Class voting has been a remarkably salient theme in the social sciences. Despite influential claims about its decline, there are marked differences in relation to social class in people's voting patterns. Whereas previous debates have focused on whether class was being dealigned or realigned with parties (see e.g. Clark and Lipset 2001), recent work seems more focused on the nature of the realignment, particularly with regard to the "new" politics of life choices, or what is sometimes called the "cultural cleavage" (Bornschier 2010). Our aim here is to contribute to the debates about class/ party realignment, especially as this pertains to what, in fairly general terms, may be thought of as the role of "culture" in shaping political preferences.
How social class is theorized and operationalized is of course of key importance to debates about the salience of social class in general and the extent of class voting in particular. This relates to the way in which the concept itself is defined and measured. It also involves broader theoretical questions about how class "works," including questions about how it is connected to other factors. Recent debates in political sociology have focused on how political cleavages involve "cultural" issues, related to what we have referred to above as the new politics of life choices. Research on class voting has struggled to deal with this in explanatory terms. Two issues are pertinent: (i) the question of whether class is itself cultural; and (ii) the related issue of how the analytical model can incorporate culture.
Are class divisions-usually thought of in terms of sociomaterial relationships-also cultural? Opinions differ in the relevant literature. On the one hand, there is Weber's (2010) classic conceptualization that insists on a strong conceptual distinction between "class" and "status." In this view, class is seen as essentially determined by market situation, and any cultural aspects of stratification should be understood in terms of status or Stand. On the other hand, there is an undercurrent of thinking that sees class as necessarily also cultural, of which the most central representative today is the tradition stemming from Bourdieu (1984).
Perspectives inspired by the work of Weber have been dominant in research on class voting. The sociology of stratification and political sociology have long relied on the (theoretically problematic) dichotomy of "manual" versus "non-manual" employees (see the critique in Parkin 1978). This was also the basis of the classic Alford index that essentially measured class voting as the extent to which manual employees voted for the Left and non-manual employees for the Right. This crude view of class differentiation was challenged by the much more sophisticated EGP class scheme (Goldthorpe 1996b), which maintains the view that class is based on production and market relations. The non-cultural nature of class has been central to Goldthorpe's approach to class, reflected for instance in his insistence on using rational choice theory to explain class inequalities in educational achievement (Goldthorpe 1996a; see also the critique in Savage 2000). As we will return to be-low, Goldthorpe has doubled down on this argument in more recent work which emphasizes even more strongly that cultural factors are independent of class (Chan and Goldthorpe 2007).
While certainly an improvement on the crude Alford index, the EGP scheme does, however, suffer from several problems of its own. One of these is related to the heterogeneity of its most privileged class categories. The service classes (I and II) have come to comprise not only a tremendous proportion of the workforce in advanced societies, they are also structurally heterogenous in ways that influence political preferences. Indeed, as argued by Heath and Savage (1995), the concept of the service class irons out the differences between fractions of the "new middle class" with different bases of privilege-e.g. professionals and managers-who are quite distinct in their voting practices and political attitudes. This critique was rejected early on-unconvincingly in our view-by Goldthorpe (1995), who argued that such results captured differences of mere situs but not class. Apart from certain "renegade" ventures (Güveli et al. 2007;Vandebroeck 2017), the principal features of the scheme remain intact, not least in its consecrated form as the official European Socio-Economic Classification (Rose and Harrison 2010).
Increased attention has also been paid to the way in which class-in both Goldthorpian and similar conceptualizations-fails to account for the new politics of life choices. Significantly, the contributions of Houtman and colleagues have pioneered the argument that while class may still explain left/right divisions in the politics of life chances, cultural capital-measured as educational level-is a better predictor of divisions in the politics of life choices, or, as they call it, "cultural voting" (Houtman 2003;van der Waal et al. 2007). A large body of work has since demonstrated that attitudes and voting are strongly related to education, especially education level (see e.g. Langsaether and Stubager 2019;Stubager 2010;Weakliem 2002). This way of thinking was followed in Chan and Goldthorpe's (2007) attempt to reinstate the class/status distinction, except that for them, it is status, not education, that is the key predictor.
However, although these studies certainly indicate that people's voting preferences and attitudes regarding the politics of life choices appear to be more strongly stratified by "cultural factors" than by economic class, they accomplish this by employing a problematic conceptual decoupling of class, education and status. Curiously, it is routinely assumed that education is independent of social class. 1 This assumption is, however, quite at odds with established theoretical accounts pointing out that educational qualifications are an absolutely central market capacity and therefore constitutive of class (see e.g. Giddens 1973;Weber 2010). Indeed, the insistence on the separateness of class and education in empirical modelling appears to be mostly conventional and commonsensical, since it is very seldom based on theoretical discussions. This is unfortunate, as it involves an a priori bifurcation of class and education at both the conceptual and methodical level, where they are typically modelled as separate independent variables. In our contribution, in contrast, we investigate how class and culture intersect in their correspondence to political voting preferences.
An alternative to this can be found in the work of Bourdieu (1984Bourdieu ( , 1986; see also the discussion in Harrits 2013). For our purposes here, a particularly important aspect of his work is how it emphasizes that some lifestyles-e.g. cultural tastes and manners-may be valorized in ways that make their mastery a valuable resource, to the extent that culture functions as a source of class privilege in its own right. This view is at least partially compatible with the Weberian position, insofar as educational credentials (or, in Bourdieusian parlance, "institutionalized cultural capital") function as a market capacity, and the practice of certain lifestyles influences the attainment of such credentials.
In Bourdieu's (1984) proposed model of social class-the social space-, cultural capital is seen as co-constitutive of multidimensional class divisions. The idea underpinning this is that social class divisions manifest themselves in a distribution of forms of capital, principally economic and cultural capital (Bourdieu 1986). The social space is a model of how these main forms of capital are distributed. Its hierarchical or vertical dimension reflects the varying levels of capital among people: the volume of capital. Its second, horizontal dimension reflects the composition of capital: the relative preponderance of either cultural or economic capital in people's capital portfolios (Bourdieu 1984). This model was developed as a means to account for variations in cultural practices and lifestyles, and a central point of Distinction is that the divisions of social space correspond to the pertinent cultural divisions of French society at the time, meaning that class and lifestyle are homologous.
While the debate about the contemporary relevance of this model is beyond our scope here, it must suffice to note that, despite the controversy, a number of contributions have found the model to be empirically generalizable to different contexts beyond Bourdieu's case of France in the 1960s and 1970s (see e.g. Prieur et al. 2008;Flemmen et al. 2018). Of more immediate relevance here is how the model of social space and its homologies have been extended to the realm of political sociology. The seeds of this can be found in the somewhat overlooked Chapter 8 of Distinction (Bourdieu 1984, pp. 397 ff.), but a number of authors have shown that a very similar homology actually pertains to the relationship between the social space and the space of political position-taking in various national contexts (see e.g. Harrits et al. 2010;De Keere 2018;Jarness et al. 2019;Lindell and Ibrahim 2021). These contributions demonstrate that a two-dimensional social space corresponds systematically to twodimensional political spaces, comprising both the "old" politics of life chances and the "new" politics of life choices.
This body of research paints a picture in which the upper-left quadrant of the social space (those rich in overall capital, with a preponderance of cultural capital) gravitates towards liberal-left position-taking, combining liberal attitudes toward the politics of life choices with leftist attitudes toward the politics of life chances. The upper-right quadrant (those rich in overall capital, with a preponderance of economic capital) gravitates towards liberal-right attitudes, combining similar liberal values with classical laissez-faire views on the politics of life chances. The lower-right quadrant (those relatively poor in overall capital, particularly in cultural capital) gravitates towards populist-right attitudes, principally defined by the politics of life choices. Finally, the lower-left quadrant (those relatively poor in overall capital, particularly in economic capital) gravitates towards the "old left" and its emphasis on social-democratic economic policies.
These patterns have clear affinities with contributions drawing on the class scheme devised by Ösch (2006). The Ösch scheme incorporates a vertical dimension, essentially derived from the EGP scheme, but also includes a cross-cutting, horizontal divide, conceptualized in terms of differing work logics-"independent," "technical," "organizational" and "interpersonal." This allows Ösch to differentiate between the fractions of the (upper-)middle class in ways that are reminiscent of the social space. A growing body of research has documented that these fractions of the middle class exhibit quite distinct voting patterns, similar to those documented by Bourdieusian scholars (see e.g. Ares 2020; Häusermann et al. 2022;Oesch and Rennwald 2018;Vestin and Oskarson 2017). This convergence suggests that these empirical patterns are quite robust, even when using different theoretical and methodological approaches (although the interpretations and explanations of these patterns differ).
In the theory of social space, then, culture is recognized as a source of privilege and power, under the condition that specific lifestyles are valorized in consequential ways in certain fields. However, not all lifestyles or tastes actually function as capital. The empirical unpacking of whether and how privileges and advantages accrue to practitioners of specific lifestyles has been a thorny issue in cultural sociology. Hence, the model of social space allows for the exploration of the way in which cultural capital is interwoven with broader conditions of existence; this facilitates a class-cultural analysis of politics. However, such an account would not exhaust the role lifestyles may play in political stances and attitudes. Therefore, our aim here is to focus on how position in social space and position in the space of lifestyles interplay in accounting for voting preferences.

The case of Norway
Our case is Norway, a comparatively egalitarian, recognizably social-democratic society (Esping-Andersen 2015). Norway is also distinct in terms of its widely shared egalitarian and anti-elitist sentiments (Gullestad 1992;Ljunggren 2017), and Norwegians regard their country as particularly egalitarian, even compared to the other Nordic social democracies (Hjellbrekke et al. 2015). Nonetheless, Norway can still be regarded as a class society. Concentrated wealth and its inheritance are on a par with countries generally portrayed as more unequal (Hansen 2014). The classed structuring of lifestyles is also similar to patterns found in other European countries (Flemmen et al. 2019a). Significantly, there is still a comparatively high level of class voting in Norway, in terms of both occupational class and education (Knutsen 2018). The Norwegian political system is similar to that of the other Nordic countries (Berglund and Lindström 1978;Demker 2006;Lipset and Rokkan 1967): there is a variant of the "frozen" five-party model, an influential labor movement and a social democratic party (Arbeiderpartiet, the Labour Party) that has dominated postwar governments. Voter turnout is also relatively high (PewResearchCenter 2016).
K However, similar to the general Western European pattern (Kriesi 2010), the establishment of a socialist green party (Sosialistisk Venstreparti, the Socialist Left Party) and a populist right-wing party (Fremskrittspartiet, the Progress Party) in the 1960s and 1970s implies that the rise of the value-based politics of life choices has restructured the Norwegian political space. The Norwegian political system is currently dominated by three major parties, the Labour Party, the Conservatives (Høyre) and the Progress Party, and supported by a handful of smaller ones, depending on the government in power, namely the Centre Party (Senterpartiet), the Christian Democratic Party (Kristelig Folkeparti), the Socialist Left Party, the communist Red Party (Rødt) and the Green Party (Miljøpartiet De Grønne).

Data and research design
Following Bourdieu (1984), we model the social space and the space of lifestyles by means of MCA. In the conceptualization of the space of lifestyles, the positioning of any particular taste-and any particular individual-is meaningful in relation to the other points in the space, in a manner reminiscent of Saussurean linguistics. Similarly, Bourdieu conceptualizes the social space as a system of "invisible," objective relations between the forms of capital. This means that occupying a position in the social space means being involved in objective relations of power, intelligible only relationally. This way of understanding social structures has a strong affinity with the geometric properties of MCA.
MCA can be thought of as the categorical counterpart to Principal Component Analysis. While making no assumptions about the distributions or properties of data, it aims to detect latent dimensions in the data. MCA works bottom-up, building a model from the distributions observed, as opposed to testing a predefined model against the observations (Greenacre 2007; Le Roux and Rouanet 2010). The end results are geometric spaces in which both individuals and response categories are positioned relationally. The space thus represents how (dis)similar individuals are in their responses and how (dis)similar categories are in their composition of individuals. Categories appear close in the space if they often co-appear in individuals' responses and father apart if they co-appear less frequently. Thus, the dimensions or factors of this space represent the main differences found in the data Rouanet 2004, 2010). MCA allows for the use of supplementary variables that are given no mass, meaning that they do not affect the structure of the space. This extension, called "Specific MCA," also enables the use of supplementary categories. This can be useful to prevent "junk" categories, such as missing data, from affecting the structure of the space. Supplementary categories are positioned as a reflection of how they relate to the active categories used in the construction of the space, but the positions of these categories remain unaffected.
As noted above, Bourdieu (1984) underlines that the homology between the social space and the space of lifestyles must be understood as a relationship between two distinct structures. Abiding by this places heavy demands on our operationalization. Lebart et al. (1984) have suggested that when using MCA for the analysis of relationships between sociodemographic and attitude variables, one should utilize a "reciprocal" approach, i.e. construct a space of sociodemographic variables with the attitudes as supplementary variables and then the other way round, and compare results. Drawing on Lebart et al.'s work, Rosenlund (2009) has developed a way of approaching the homology thesis in more detail. He has operationalized it by constructing each space-one social space and one space of lifestyles-with separate correspondence analysis procedures for each, and then using coordinates from the one as supplementary categories in the other to compare their structures. This has the dual advantage of providing a strict test of the homology thesis, since it deals directly with the question of whether these two structures are actually similar. Moreover, it also respects the autonomy of both spaces, treating neither as a simple "dependent" variable. We expand on this by drawing on Le Roux and Rouanet's (2004) extension of MCA methodology to include both analysis of variance and Euclidean classification, i.e. Agglomerative Hierarchical Clustering with Ward's criterion applied to the axes produced by MCA. This allows us to provide a fuller account of the relationship between the two structures and a more precise estimation of their interrelationships.
The final part of our analysis is more complicated in data-analytical terms: In what ways do the social space and the space of lifestyles intersect in "explaining" political voting patterns? Statistically speaking, this means looking for interactions between the two spaces in terms of associations regarding which party respondents would vote for. Conventional, deductive approaches would necessitate defining which interactions to test for, a process that would be cumbersome. As we will see below, this would mean testing at least 54 possible combinations of positions in the social space and the space of lifestyles. To avoid this issue, we employ a technique called "Chi-Squared Automatic Interaction Detection" (CHAID) (Kass 1980), which is a non-parametric decision-tree technique for dealing with regression problems. What it essentially does is run a series of chi-square tests on the independent and the dependent variables. The result is represented as a decision tree in which the first level represents the categories that are most strongly differentiated in terms of the dependent variable. Categories of the independent variable that are not distinct from one another are merged. The procedure then looks for interactions with the other variables so that it can split the categories on the first level by the variable with the second strongest association, but this need not be the same variable for every category on the first level. In this way, the technique can automatically reveal which interactions are important and present a visually "intuitive" representation of this as a decision tree. Thus, it is a clustering technique that clusters based on how homogenous combinations of independent variables are with respect to the dependent variable. In our case, this means creating clusters of combinations of positions in the social space with different lifestyles based on how similar they are in their voting intentions.
Our analytical strategy involves five steps. First, we construct and interpret the social space based on an MCA of indicators of economic and cultural capital. Second, we construct a space of lifestyles in a separate MCA procedure. Third, we perform a Euclidean classification of this space, producing clusters of optimally homogenous lifestyle. Fourth, we assess to what extent these two structures-the social space and the space of lifestyles-exhibit a homology, i.e. whether and how the primary structures of the spaces overlap. Finally, we assess the ways in which po-K What is the main discipline/field of your education? 6 7.7 15.9 Have you taken any education in the military/armed forces? 2 4 . 0 0 . 0

General indicator of capital
What is your occupation? 10 13.4 6.1 sitions in the social space and the space of lifestyles are combined in the structuring of political party preferences. We do this by means of standard CHAID, looking for interaction effects between the social space and the space of lifestyles in accounting for which party respondents would vote for. The dependent variable has seven categories: Labour, Conservative, Progress Party, Christian Democrat, Socialist, Centre Party and the Liberal Party. 2 This procedure requires rich data, with ample indicators of both the forms of capital and a wide range of lifestyle variables. We use the 2011 round of Norsk Monitor, carried out by the Ipsos MMI. Respondents were selected by simple, random sampling from telephone directories and recruited through telephone calls. An interviewer asked introductory questions and most of the survey was conducted through self-completed questionnaires. The response rate for the telephone interviews was approximately 25%, and 10% filled out and returned the questionnaires, amounting  to 3980 respondents. As could be expected from this relatively low response rate, the resulting sample was sociodemographically skewed. For our analysis, we correct for this by adding weights by gender, age, geographical region, education level and the number of rooms in the home. Thus, we can make our sample nationally representative in these respects. 3 For the construction of the social space, we use thirteen questions as indicators of economic and cultural capital (see Tab. 1 and a more elaborate account of the construction of the space in Flemmen et al. 2019a). To construct the space of lifestyle, we use a wide range of indicators to move beyond the undue narrowing of scope of much previous research and its exclusive focus on cultural consumption. We thus follow Bourdieu's (1984) broad focus on "culture" in the "anthropological sense", while heeding Scott's (2002) warning that many analysts of cultural taste have an unduly narrow understanding of Weber's notion of Lebensstil. The variables are listed thematically in Tab. 2 (for further details on the construction of the space, see Flemmen et al. 2019a).

The social space
With two dimensions, there is a cumulated rate of Benzécri's modified eigenvalues of 53.0%. The positions of all categories of our active variables-the cloud of categories-are shown in Fig. 1, with the first axis shown vertically and the second horizontally. Both axes are shaped by economic and cultural capital; thus, both reflect differences in the possession of both forms of capital. The structure of the space is strikingly similar to the model of the social space proposed by Bourdieu (1984). The first axis, shown vertically, depicts differences in the volume of capital. Moving from the bottom and upwards in the space, the total volume of both cultural 3 However, it could be the case that, despite our weighting, there is some form of unobserved selection. We have measured this by checking how our data compare with high-quality, nationally representative surveys collected by Statistics Norway. Specifically, we have compared the distributions of responses with two survey questions relevant to our topic: which political party one would vote for and how often one reads news online. The responses to both of these questions are practically identical in our survey and our external yardstick surveys. This indicates that the weighting works, and we thus have a dataset that is nationally representative in terms of the issues we are looking at as well.     Fig. 3 The space of lifestyles, explicative points for Axis 2, factorial plane 1-2

The space of lifestyles
In this section, we briefly outline the main structures of the space of lifestyles. A more detailed interpretation of the space can be found in Flemmen et al. (2019a). Axis 1 (see Fig. 2) depicts a division between a taste for "legitimate" items and a taste for items that are considerably less so. Axis 2 (Fig. 3) displays a division between a cosmopolitan lifestyle and a lifestyle that is considerably less so. Axis 3 (Fig. 4) is shaped by a division between an ascetic, intellectually oriented lifestyle on the one hand and an excitement-seeking, bodily oriented and expensive lifestyle on   Fig. 4 The space of lifestyles, explicative points for Axis 3, factorial plane 1-3 the other. Finally, Axis 4 (Fig. 5) is a more specific media axis, depicting a division between a rational-intellectual and a more alternative-emotional lifestyle.

Clustering lifestyles
Next, we conduct a Euclidean classification of the space of lifestyles based on the four dimensions explained above. Six clusters are optimal to account for the variance of the space of lifestyles, and these reflect the main oppositions of the space. The interpretation and labelling of the clusters are based on their most distinct lifestyle categories and their position in the space.
Cluster 1, "the emotional-traditional," is characterized by enjoying entertainments, love/romance books, diet books, reading about celebrities in newspapers, family movies, following fashion, reading self-development books and listening to pop music. The strongest dislikes include stereotypically "masculine" goods like

Fig. 5
The space of lifestyles, explicative points for Axis 4, factorial plane 1-4 K hunting and fishing books, war movies, reading about cars in newspapers and debate books. Cluster 2, "the legitimate and savvy," is characterized by tastes for legitimate items, both in the traditional sense (i.e. "high culture") and in a more recent sense (i.e. the "hip" and "savvy"). The most characteristic preferences include art and culture books, in-depth articles and items about culture in newspapers, Moroccan cookery, philosophy books, classical music, raw food and French and Japanese cookery. Although not strongly marked by dislikes, some typically denigrated and mocked items are steered clear of: country and western and "dansband" music, microwave cookery and, interestingly, Interrailing and backpacking holidays.
Cluster 3, "the action and excitement seekers," is characterized by stereotypically "masculine" goods and activities, e.g. reading about cars in newspapers, hunting and fishing books, war movies, TV sports, westerns and ball games. Dislikes include legitimate items, e.g. foreign contemporary literature, drama/"serious quality" movies, classical music and books about culture, art and poetry.
Cluster 4, "the emerging and commercial-technological," is characterized by sci-fi movies, owning a smartphone, fantasy movies, American cookery, action movies, hip hop and rap music, reading about personal computers and technology in newspapers, animated movies and owning a home cinema. Dislikes include regional television programs, hiking in the forests, foreign contemporary literature, TV programs about art, and diet books.
Cluster 5, "the traditional and established," is characterized by dislikes for a range of "cosmopolitan" and more "popular" goods and activities, e.g. action movies, pop music, lottery, home cinema and microwave cookery. Interests in established and legitimate items are also prominent: classical music, spending days off outdoors, drama/"serious quality" movies, poetry books, op-eds in newspapers and television programs about art.
Cluster 6, "the illegitimates," is characterized by disliking a wide range of items, including themed holidays, foreign contemporary literature, culture in newspapers, diet and self-development books and in-depth articles in newspapers. Only a few, select items are liked, exclusively encompassing goods and activities frequently mocked and stereotyped as being in "bad taste": "dansband" music, lottery, caravans, traditional dance music, westerns, microwave cookery and motor homes.

The correspondence between the space of lifestyles and social space
We now briefly outline the homology, or the structural similarity, between the social space and the space of lifestyle. Again, we refer the reader to Flemmen et al. (2019a) for a more detailed explanation of this relationship.
In Tab. 3, we present a summary of this relationship. The results depict marked divisions in lifestyles according to both dimensions of the social space: the volume and composition of capital. Specifically, the dimension of capital volume can be interpreted as largely mirroring a division between legitimate cultural tastes and less legitimate ones, while the dimension of capital composition largely mirrors the division between different levels of cultural legitimacy, as well as between intellectual activities associated with the faculties of the mind and sporty activities associated Cramer's V: 0.150 with the body. As can be seen from the table, the social space is divided into nine (roughly) equally sized regions: both the capital volume and capital composition dimensions are sliced into three class categories before being crossed with each other. The class categories are then cross-tabulated with the lifestyle clusters. The results indicate that the overall association in the table is moderate. 5 However, the associations are unevenly distributed: some regions of the social space exhibit very distinct lifestyles, while others do not. This is shown by the adjusted standardized residuals demonstrating which specific associations are significant. In Tab. 4, we show how the class categories relate to political parties. The overall association is weaker than with lifestyles. However, the associations are unevenly distributed in ways that closely resemble those of lifestyles. In both tables, we can see the strongest associations on the fringes of the social space: the most distinct preferences in both lifestyle and politics are found among those with high and low volumes of capital and those with either a preponderance of cultural or economic capital. Thus, the general impression is that the divisions of the social space are reflected in both lifestyles and politics: diametrically opposed lifestyles and political party preferences are clearly differentiated in their social positioning.

The intersection of social space and lifestyles in party choice
We use the CHAID procedure in SPSS with Bonferroni adjustment. The dependent variable is political party choice (which party the respondent would vote for if there were a parliamentary election held next Monday). The model is estimated with four key predictor variables: social-space categories, the clusters of lifestyles, and the volume and composition of capital as continuous variables. The latter are included to allow for the possibility that the class categories might be too big or heterogeneous with respect to political party choice. In this way, the CHAID procedure can split the classes into more finely grained groups, by volume and/or composition of capital, if this would account more accurately for party preferences. This reflects the fact that the social space is a space of differences in which boundaries are only contingent. The use of the class categories is only a heuristic tool to ease interpretation and analysis. As we will see, there are instances where these categories are not always distinct from each other. There are also instances where some of the categories are shown to be too heterogeneous.
In the analysis, we forced the first variable. Without it, CHAID simply starts with the predictor variable that has the strongest association with the dependent variable. In this case, this would have been the lifestyle variable. However, in this analysis we want to start from the position in the social space and then investigate how lifestyle divisions "moderate" its politics. Using the aforementioned feature, we therefore forced the analysis to start with the social space.
The full CHAID tree can be seen in Figure s1. Unfortunately, it is difficult to reproduce the figure here, but it can be found in the Online Appendix for a more thorough inspection. The figure starts by identifying which of the class categories are significantly different from the overall distribution in the sample in terms of party choice. As we can see, the figure shows that almost all classes have a distinct political profile. There are, however, two exceptions. First, at this general level, the cultural fraction of the upper class and the cultural fraction of the middle class are not significantly different. Second, and somewhat more surprisingly, the balanced fraction of the lower class and the economic fraction of the middle class are quite similar. The CHAID procedure thus merges these categories on this first level. In the following, we present the categories in terms of their position in the social space and not their position in the CHAID figure.

Upper and middle class, preponderance of cultural capital
Beginning at the top left, we can see that the cultural fractions of the upper and middle class have been merged (see Fig. 6, Node 7 in Fig. s1) due to the similarity in their political profile (see Fig. 6). First, the most marked difference from the sample as a whole is their dislike of the populist right: they are about 0.3 as likely to vote for the Progress Party. Second, they are half as likely to vote for the Centre Party. Third, they are close to 1.9 times more likely to vote Socialist. They are 1.4 times more likely to vote Liberal and 1.3 times more likely to vote for the Christian Democrats.
Notably, the politics of those rich in cultural capital are clearly divided by their lifestyles. We have seen in Tab. 3 that these class categories are quite strongly associated with variants of legitimate cultural taste (i.e. the "traditional and established" and "legitimate and savvy"). The CHAID analysis reveals that an affiliation with these lifestyles forms significant political dividing lines in this region of the social space. Those who exhibit an "established and traditional" taste (Node 18) within these class categories differ from the class as a whole by being almost half as likely to vote Conservative and almost twice as likely to vote Christian Democrat. They are also more than 1.4 times as likely to vote Socialist.
Those who exhibit "the legitimate and savvy" taste (Node 17) within these class categories, meanwhile, stand out from the rest in their distaste for the populist right: none of them would vote for the Progress Party. They are also much less likely to vote for the Christian Democrats (0.4) or the Centre Party (0.5). They are, however, more likely to vote Socialist (1.4) or Liberal (1.3). Those associated with any of the other less legitimate taste clusters (Node 16) within these class categories are differentiated from the others by a much lower propensity to vote Socialist (0.4) and a much more favorable attitude toward the Progress Party (1.9). They are also more likely to vote Conservative or for the Centre Party (1.3 for each).

Upper class, balanced composition of capital
Moving on to the balanced fraction of the upper class (see Fig. 7, Node 2 in Fig. s1), these respondents differ from the sample as a whole in their strong disinclination to vote Christian Democrat (0.3) or for the Progress Party (0.5) and their strong disposition to vote Liberal (2.2).
Within this class category, respondents are divided in their lifestyles. The respondents in the "action and excitement seeking" and "commercial technological" lifestyle clusters (Node 11) stand apart from the others in the class category with a strong over-representation of Progress Party voters (3.4 times). Indeed, all the Progress Party voting respondents in this class category exhibit these lifestyles. Conversely, they are much less likely to vote for the Centre Party (0.3), Socialist (0.6) or Labour (0.7). Correspondingly, the rest of the respondents in the class cat- Fig. 7 Upper class, balanced composition of capital K egory (all other lifestyles; Node 10) are rather similar to the category as a whole, except for the fact that none of the respondents would vote for the Progress Party.

Upper class, preponderance of economic capital
In the economic fraction of the upper class (see Fig. 8), political attitudes are not divided by lifestyle. However, the CHAID procedure indicates that this class category is too broad and Node 4 is split in two: those with somewhat lower volumes of capital (Node 12) and those with higher volumes (Node 13). We will comment on these directly, since we are not dealing with subdivisions in lifestyle but with what is more likely to be a more accurate representation of the political divides in the social space.
The category with a preponderance of economic capital and the highest volume of capital (Node 13) clearly exhibits a conservative tendency, and it is the most homogenous group identified here: 57.9% of the category report that they would vote Conservative. All other party choices are under-represented, and the respondents are distinct from the sample in their disinclination to vote Christian Democrat (0.36), Socialist (0.4) or for the Centre Party (0.5) or the Progress Party (0.56).
The category below-what we may think of as the economic fraction of the uppermiddle class (Node 12)-has a less distinct political profile. They are evenly split between the Conservatives and Labour: 29.4% for each. They are also distinct from the sample in terms of their strong preference for the Christian Democrats (2.0), the Progress Party (1.7) and the Centre Party (1.4), although it should be noted that we are dealing with rather small numbers of respondents here.

Middle class, balanced composition of capital
The balanced fraction of the middle class (see Fig. 9, Node 6 in Fig. s1) is positioned in the middle of the social space, a region where one would expect to find a more "average" political profile. Moreover, this is the only class category that exhibits no significant association with any of the lifestyle clusters. Indeed, it is rather similar to the overall sample. Politically, it is most distinct in terms of a slight distaste for the Progress Party (0.7) and an inclination toward the Centre Party and the Christian Democrats (about 1.4). There is also some disinclination to vote Socialist (0.8).
This class category exhibits a link between lifestyle divisions and political divisions, although it is difficult to interpret. The CHAID procedure suggests that, within this class category, those who express polar-opposite preferences for variants of legitimate tastes and least legitimate tastes (Node 15) have a fairly diverse political profile: aside from nearly one-third favoring Labour, most other options are present in roughly equal proportions. This class category is distinct from the class as a whole in its inclination to vote Socialist (1.8) and a dislike of the Conservatives (0.6). There are only three respondents who report that they would vote for the Progress Party.
The rest of the balanced fraction of the middle class-those with all other lifestyles (Node 14)-do not differ as markedly from the class as a whole. Basically, they are split evenly between the Conservatives and Labour: the vast majority would vote for one of these parties. They are about as likely to vote Labour as the class as a whole but are clearly more in favor of the Conservatives (1.4). Somewhat surprisingly, the CHAID procedure indicates that these two fractions (see Fig. 10) are politically indistinguishable (Node 1). Tab. 3 above indicates that that they are somewhat different in lifestyle: the economic fraction of the middle class is drawn towards "action and excitement" and "illegitimate" tastes, whereas the balanced fraction of the lower class is drawn towards both the "emotional and traditional" and the "illegitimate" cluster, thus caught in the middle of the lifestyle oppositions salient in the lower regions of the social space. Compared to the sample, these class categories are much less likely to vote Socialist (0.6) or Liberal (0.4) and markedly more likely to vote for the Centre (1.6) and Progress parties (1.5).
Lifestyle divisions within this class category set those who enjoy the two variants of "legitimate" taste (Node 9) apart from the rest (Node 8). The situation presents itself as if legitimate taste pushes or pulls away from the right wing: none of the respondents would vote for the Progress Party. Correspondingly, the node is distinctive in being highly likely to vote Socialist (3.8), likely to vote for the Centre Party (1.4) and less likely to vote Conservative (0.6). Those who would vote for other parties are too few in number to comment on. The rest of the class category (Node 8) is not very distinct from the class category as a whole, except that there are few respondents who would vote Socialist (0.4).

Lower class, preponderance of economic capital
The economic fraction of the lower class (See Fig. 11. Node 5 in Fig. s1) stands out from the sample, with only three respondents planning to vote Liberal, making this the most under-represented option. This class category also stands out in being likely  to vote for the Progress Party (2.1) and Labour (1.4) and less likely to vote Socialist, Christian Democrat or Conservative (all 0.6). The CHAID procedure indicates no further differentiation in political party preferences by lifestyle within this category.

Lower class, preponderance of cultural capital
The cultural fraction of the lower class (see Fig. 12) is most clearly distinguished from the sample by their strong inclination to vote Christian Democrat (1.8) or  4). They are also less likely to vote Conservative (0.6) or Liberal (0.7). The CHAID procedure does not reveal any significant lifestyle differences with respect to political party choice within this category.

Conclusion
In this article, we have used a combination of MCA and CHAID to show that there are important interactions between positions in the social space and in the space of lifestyles and that this can help account for class voting. An inspection of the interactions between the two spaces reveals some important patterns. Previous research has established that class fractions rich in cultural capital have a markedly leftist orientation; this is corroborated here in their pronounced tendency to vote Socialist (i.e. Socialist Left Party or Red). The analysis reveals that this overlaps strongly with their orientations towards two variants of "legitimate" cultural taste. In earlier work, we have argued that "cultural capital leftism" might be interpreted as a negation of the power of economic capital, much like scholastic-intellectual tastes are (Jarness et al. 2019). Favoring stances such as stronger workers' rights, market regulation and increased taxes to fund social spending represents a challenge to the power of economic capital, similar to the way in which support for stronger public funding of cultural institutions is favorable to holders of cultural capital, since this shields this group from market competition. This interpretation seems to be bolstered by the observation that, within the class category richest in cultural capital, leftist voting is strongest among those whose lifestyle also negates economic capital.
The nature of the association between social space, the space of lifestyles and party choice is more complex than recent accounts of "omnivorousness" in cultural stratification research suggest (see e.g. Chan 2019). In this view, "cultural omnivores"-cultural consumers who exhibit eclectic, broad tastes straddling hierarchical divisions between legitimate and less legitimate cultural goods and activities-are essentially tolerant and open-minded, by virtue of their liberal political stances. Our analysis demonstrates that these kinds of (left) liberal stances are associated with cultural capital, especially among respondents who also display variants of legitimate taste. In our earlier work, we have presented reasons to doubt the interpretation of "omnivorous" taste patterns as indicative of an open or tolerant cultural orientation (Flemmen et al. 2018;Jarness 2015). Recently, we have also argued that it is implausible to support this view by referring to cultural omnivores' endorsement of immigration, environmental protection and the EU (Flemmen et al. 2019b). This is further corroborated in the present analysis: voting for parties representing such political stances is most strongly associated with upper-and middle-class fractions rich in cultural capital, especially those who participate in legitimate culture and who shun less legitimate cultural forms.
This interpretation is further strengthened by the observation that an affiliation with legitimate lifestyles also has a similar "effect" in other regions of the social space. For instance, an affiliation with "legitimate" cultural taste accounts for almost all of the Socialist votes among the balanced fraction of the lower class and the economic fraction of the middle class. This is also striking in the case of the politically rather neutral middle region of the social space, i.e. the balanced fraction of the middle class. In this region, an affiliation with legitimate culture is associated with a pronounced tendency to vote for leftist parties. Interestingly and perhaps somewhat complicating the picture, this political tendency is associated to a roughly equal degree with affiliations with its polar opposite: "illegitimate" culture.
Conversely, an affiliation with lifestyles more frequently associated with economic capital (and distanced from the culturally legitimate and savvy) seems to "pull" towards the political Right. This is evident in the case of the balanced fraction of the upper class. In terms of the theory of social space, this class fraction would be caught between the two poles of economic and cultural capital and their symbolic expressions. However, the analysis reveals that, within the context of this tension, there is a symbolic affiliation with their counterparts to their "right" in the social space, as well as a clear orientation towards the political Right.
Finally, it should be underscored that we are not proposing a causal model, in the sense that class position in social space somehow determines political party choice, and that this causal relationship is in some sense mediated by lifestyle. Rather, we regard political taste as analytically similar to lifestyle, a scholarly position that is also implied by Weber (2010). What our analysis does suggest, however, is that there are systematic configurations in the affinity between cultural taste and political taste (as expressed through voting). We remain agnostic about the "mechanisms" producing these associations. On the one hand, we consider it likely that political taste may be shaped by concerns of self-interest, for instance when those rich in economic capital oppose progressive taxation, or when those rich in cultural capital favor generous public spending on cultural production. This interpretation can, however, be defended without postulating explanations in terms of rational action theory: it does not seem necessary to have some cost-benefit calculation in mind to experience a desire to keep more of one's pay or to want more funding for one's sector of employment. However, as we have implied above, we consider it likely that these political tastes can also be understood more in terms of "symbolizations" of economic and cultural capital, that is, not as conscious attempts to defend their respective value but as a practical sense or an emotional attachment that valorizes these forms of capital and their modes of accumulation, such as Bildung versus market competition.
We think our analysis supports the kind of symbolic interpretation alluded to here, while not debunking or weakening explanations relating to self-interest. The case of the balanced fraction of the upper class would seem to confirm this. With the same amount and composition of capital, their self-interest would predispose them to go either way: favoring public spending supporting "culture" and/or the public sector, while also preferring to pay less in taxes. These respondents more often support the Right if they are symbolically aligned with economic capital, if we may be excused this slightly circular-sounding expression.

Supplementary Information
The online version of this article (https://doi.org/10.1007/s11609-022-00472-y) contains supplementary material, which is available to authorized users.
Funding Open access funding provided by University of Oslo (incl Oslo University Hospital).

K
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4. 0/.