Upper-class romance: homogamy at the apex of the class structure

ABSTRACT

This article focuses on social closure by way of marital homogamy within the upper class. It offers new insights into the social structuring of romantic partnerships, while drawing on research on assortative mating and contemporary elite and class analysis. The analysis is based on detailed data covering the entire Norwegian population. Our main point of focus is the upper class, whose patterns of partner choice have been little studied. By drawing on Bourdieu’s model of the social space, we move beyond conventional operationalisations of class. The analysis demonstrates that romantic partnerships within the upper class are structured along three dimensions of class: (i) vertical inter-class closure (upper-class individuals are disproportionately more likely to have partners in upper-class positions); (ii) horizontal intra-class closure (a tendency for marrying within the same upper-class fraction); (iii) closure by class trajectory (upwardly mobile newcomers are disproportionately more unlikely to have upper-class partners). We also demonstrate how class divisions intersect with gender divisions. Among the men, there are important differences between the upper-class fractions: the cultural fraction is more homogamous than the other fractions, and the economic fraction is comparatively more likely to have partners from lower down in the class structure.

KEYWORDS:

• Assortative mating
• closure
• cultural capital
• elites
• marriage

Introduction

Against the backdrop of rising inequalities, there has been increasing sociological interest in elites and the upper class (Cousin et al. 2018; Korsnes et al. 2017; Savage and Williams 2008). Dating back to the work of Weber (1978), class research and elite research have focused on various forms of social closure. The processes through which some groups enjoy privileges and opportunities that are closed off to others, and whether they form closed social and symbolic groups, have been key topics in these streams of research. Recent accounts have focused on a range of processes pertinent to social closure: for instance, how wealth is concentrated and transmitted at the top of the class structure (Hansen 2014); the perpetuation of upper-class privilege through the education system (Strømme and Hansen 2017); intra- and intergenerational mobility closure (Friedman and Laurison 2019; Toft 2019); the structure of elite networks (Larsen and Ellersgaard 2018); and, lifestyle differentiation and symbolic boundaries (Rivera 2012).

However, most studies of elites and the upper class have largely ignored group formation through marriage. This is unfortunate, because class homogamy – the tendency for individuals to marry someone in a similar class position as themselves – arguably constitutes a distinct source of group formation (Bourdieu 1976). Indeed, marriage has been considered key to upper-class formation in classical works of sociology (see e.g. Goode 1959; Mills 2000), as well as in historical elite research (see e.g. Kocka 1984). When two privileged individuals marry, they not only mutually affirm their classed ways of being (Bourdieu 1984: 243–4), they also create a highly privileged family through which resources can be combined. Joining forces consolidates privileges and advantages for the individuals involved (and their children) and it can intensify class divisions more generally.

Although they are rarely studied in contemporary elite sociology, marriage patterns have long constituted a distinct sociological research field centring on the notion of ‘assortative mating’ (see e.g. Andrade and Thomsen 2019; Blossfeld and Timm 2003; Kalmijn 1994; Mare 2016). However, although this field of research has provided much insight into the social structuring of homogamy, little is known about marriage patterns within the upper class today.

In this article, we aim to cross-fertilise insights from research on assortative mating with insights from contemporary elite and class analysis. Although both fields of research have been concerned with social closure, the analytical frameworks usual within them have been quite different. We will argue that recent developments in the wake of the ‘cultural turn’ in class analysis (Devine and Savage 2005) can provide a fruitful alternative to existing frameworks within research on assortative mating, that have predominantly relied on conventional operationalisations of class, or that have restricted analyses to single indicators such as income or education (see the discussion in Schmitz 2016). Specifically, we discuss how Bourdieu’s (1984) notion of the class structure as a multidimensional social space may direct one’s attention to horizontal intra-class divisions characterised by an opposition between cultural and economic forms of capital.

Empirically, we exploit the richness of Norwegian registry data covering the entire population. These data contain all registered romantic partnerships in Norway, including married couples and cohabitants (with children). We ask: do romantic partnerships indicate social closure between the upper class and other classes? Are there indications of closure within the upper class itself, i.e. between horizontally different class fractions? Do marriage patterns within the upper class differ according to patterns of social mobility? How do classed marriage patterns within the upper class intersect with gender differences?

Norway seems a particularly interesting case for asking such questions, partly because of the purported invisibility of concentrated privilege at the top of the class structure. In comparative research, Norway – along with the other Nordic countries – is often regarded as an outlier due to egalitarian features such as relatively higher social fluidity rates, a compressed wage distribution and extensive and universal welfare services (see e.g. Esping-Andersen 2015). However, concentrated affluence and the inheritance thereof are on a par in Norway with countries generally portrayed as more unequal. According to Hansen (2014), we are now witnessing the rise of a ‘new Nordic model’ characterised by dynastic tendencies among the very wealthy and comparatively greater equality among the population at large. A lack of attention to the upper class may obstruct an awareness of these aspects of Norwegian structures of opportunity. Indeed, Norway is distinctive in terms of the perceived equality and widely shared egalitarian and anti-elitist sentiments by the population at large (Hjellbrekke et al. 2015). These sentiments and subjective beliefs in a non-hierarchised social structure make Norway an intriguing case for analysing whether romantic partnerships at the top of the class structure perpetuate class divisions.

‘Assortative mating’ and marriages within the upper class

Earlier, marriage – particularly among the upper strata – was rarely a matter of personal choice but was an overt family strategy to ensure social reproduction, partly linked to the existence of the dowry, i.e. parental property, gifts or money offered by a bride to her husband on marriage (Kocka 1984). There are, however, reasons to suspect that partner choice has become detached from such traditional arrangements: cultural norms of marriage have gradually changed from being a family matter of economic reproduction towards a more personal choice of ‘romantic love’ (Coontz 2006; Shorter 1975) and cultural shifts in ‘individualisation’ have attenuated class as a marker of social identities (Savage 2000).

A large body of contemporary research has nevertheless demonstrated a persistent connection between class and marriage (Andrade and Thomsen 2019; Blossfeld and Timm 2003; Henz and Mills 2018; Mare 2016). The literature has addressed two main questions: (i) whether people tend to marry within their class of origin; and, (ii) whether people tend to marry partners in similar class positions as themselves, i.e. within their class destination. In seeking to explain such patterns, researchers have pointed to key social determinants: the likelihood of meeting on the ‘marriage market’, geographical location, social pressure, personal autonomy and personal preferences (for an overview, see Van Leeuwen and Maas 2005).

Research has suggested that the likelihood of homogamy by class origin has decreased over time (Kalmijn 1991a), although recent trends in the US suggest a U-shaped turn toward previous levels (Mare 2016) and, in the UK, stability in more recent years (Henz and Mills 2018). Homogamy by class destination is comparatively more pronounced: people are more inclined to marry someone in the same social position, in particular someone with a similar type and level of education (Blossfeld and Timm 2003). Interestingly, research from the Nordic countries suggests a slight decrease in educational homogamy over time (Andrade and Thomsen 2019; Birkelund and Heldal 2003). Nonetheless, Norwegian studies indicate that the likelihood of marrying someone of one’s own class varies between classes: the upper class are most likely to be homogamous (Hansen 1995) and the likelihood of homogamy decreases the further one moves down the class structure (Gulbrandsen et al. 2002).

Although research on assortative mating has produced a number of key insights, there is potential for development. First, marriage patterns within the upper class in contemporary societies are largely unknown, partly because of limited data, but also possibly because most operationalisations of class do not differentiate between the upper and the middle classes. In contrast, within elite and class analysis, studies of the upper echelons of the class structure are gaining momentum (see the account in Cousin et al. 2018) and with the rise in economic inequalities (Hansen 2014; Piketty 2014), research on marriage patterns within the upper class seems apt.

Second, research on assortative mating has tended to overlook the question of horizontal intra-class divisions. This is an important omission in light of recent developments in class analysis. A number of studies have shown that upper-class fractions rich in cultural capital and upper-class fractions rich in economic capital tend to form distinct social and symbolic communities (Flemmen et al. 2019; Ljunggren and Andersen 2015; Prieur et al. 2008; Vandebroeck 2018). Because such differences pertain to aspects that are relevant when choosing a romantic partner – e.g. social mobility patterns, residential segregation, cultural tastes, body ideals and physical appearance – there is reason to suspect that intra-class divisions also structure marriages. However, the conventional operationalisations applied in the literature fall short in attempting to account for such divisions. Although some studies have differentiated between horizontal educational differences – educational homogamy is prevalent in terms of educational level and field (see e.g. Andrade and Thomsen 2019) – few studies have assessed horizontal differences in terms of class fractions divided by their possession of different forms of capital (although see Hansen 1995; Kalmijn 1994; Schmitz 2016).

Third, the dominant approach in research on assortative mating has predominantly drawn on theories of rational choice and social exchange (see the discussion in Schmitz 2016). A common assumption is that romantic relationships are driven by a maximisation of ‘utility’ (e.g. upward status mobility) and that people navigate ‘marriage markets’ strategically, calculating the costs and benefits of the various alternatives available. However, such assumptions have increasingly been questioned: research has suggested that choosing a partner is influenced by an array of cultural factors beyond the scope of rational action and exchange theories, and that social actors understand their experiences of love and marriage by drawing on diverse and often contradictory cultural repertoires (Illouz 1997; Swidler 2013).

In this article, we suggest that an approach inspired by Bourdieu (1976, 1984) can provide broader insights into class homogamy. Given widespread colloquial ideas about love as chance or romantic good fortune (Coontz 2006), we need a framework that can account for the cultural complexities of marriage patterns. One of the most significant contributions of Bourdieu’s work is the key role of classifications in the reproduction of class inequalities. Bourdieu maintains that people who have experienced similar class conditions tend to develop similar habitus or embodied dispositions. Such affinities in dispositions – e.g. material, aesthetic, moral and political inclinations – profoundly influence people’s perceptions, classifications and desires for each other, and they also stratify the likelihood of social encounters, for instance through institutional selectivity, homogenous circles of friends and geographical segregation (Lamaison and Bourdieu 1986).

Key to Bourdieu’s notion of class habitus is the model of the social space, a system of class conditions determined by a three-dimensional distribution of key forms of capital (Bourdieu 1984: 99–168). This space is shaped first by the volume of capital (high versus low levels of overall capital holdings), second by the composition of capital (the relative preponderance of either cultural or economic capital) and third by trajectories (changes in actors’ or groups’ holdings of capital over time). Because of mutual attraction and appreciation between people with similarly conditioned habitus, romantic partners tend to be situated in close proximity to each other in the social space.

There are several advantages to employing a Bourdieusian framework to study class homogamy. First, the notion of habitus can help us sidestep problematic assumptions about people’s strategic decision-making, and theorise how class may lubricate class homogamy without rational calculations to maximise utility. Specifically, the notion of habitus suggests that homogamy is facilitated by people’s practical sense of whom they belong with. While not rejecting the idea that rational calculations are involved in the process of choosing a partner – indeed, a ‘utilitarian’ mode of classification can itself be incorporated in the habitus (Schmitz 2016) – the notion of class habitus offers an encompassing theory of action that: (i) lends due space to types of preferences and choices that are apprehended subjectively as a ‘sense of the miraculous’ and not just ones that are aimed at maximising utility (Bourdieu 1984: 241–4; see also the discussions in Bouchet-Valat 2014; Bozon and Héran 1989); (ii) links different modes of preferences with different types of class conditions and class cultures. The model thus suggests that people in similar class positions are disproportionately likely to marry one another.

Second, it offers a multidimensional model of the class structure, encompassing both vertical and horizontal class divisions. A particular advantage over conventional operationalisations of class is that the social space model allows one to assess whether there are systematic intra-class divisions within the upper class according to the ‘capital-composition’ principle (Bourdieu 1984: 114–25). Thus, the model implies fraction-specific homogamy: people situated within a given upper-class fraction are more likely to find a partner within that class fraction than within other upper-class fractions.¹ Recent Norwegian research has suggested that class fractions rich in cultural capital are comparatively more homogenous in their lifestyles, attitudes and outlooks than their counterparts rich in economic capital (see e.g. Flemmen et al. 2019). This leads us to expect that the cultural upper class will be comparatively more likely than the other upper-class fractions to have a partner in the same class fraction.

Moreover, the model can shed light on the connection between class origin and class destination. The development of classed dispositions is a continuous process through which class conditions shape individuals throughout their life course: the more similar the class trajectories, the more similar the dispositions (Bourdieu 1984: 114–24, 69–207). The model thus suggests that established upper-class individuals originating from upper-class positions will be more likely to have a partner in the upper class than upwardly mobile newcomers to that class.

Third, the Bourdieusian framework can also shed light on the way in which class intersects with gender to produce different dispositions. Gender properties, Bourdieu claims, ‘are as inseparable from class properties as the yellowness of the lemon is from its acidity’, meaning that there are as many ways of realising femininity and masculinity ‘as there are classes and class fractions’ (Bourdieu 1984: 107–8). This means that cultural processes – linked for instance to different modes of socialisation and gender roles and expectations (Ridgeway 2011; Skeggs 1997) – produce gendered preferences that manifest themselves differently across the class structure.

Studies suggest that the Norwegian welfare state offers arrangements that help people combine work and family; that the gendered division of domestic labour is more equal in Norway compared to other countries; that Norwegian women’s participation in the labour market is comparatively high; and, that women with higher education are more likely than women with lower education to stay on in the labour market after marriage and childbirth and exhibit more positive attitudes towards gender equality in the domestic sphere (see e.g. Ellingsæter and Leira 2006, Knudsen & Wærness 2001). However, studies of economically rewarding elite careers in Norway suggest that within this segment of the labour market, women are comparatively more likely than men to reduce their commitment to work or drop out of the labour market entirely after childbirth (Halrynjo and Lyng 2010). Moreover, ethnographic research indicates that traditional gender norms are practised in very wealthy families, partly due to collective strategies to preserve dynastic lineage (Aarseth 2016; Glucksberg 2018). Wives in such families tend to opt out of their own careers to take part in the emotional and social sides of domestic duties, ensuring social reproduction as ‘housewives’, while freeing their husbands to invest their time in continuing to accumulate wealth, ensuring economic reproduction. This indicates that there are constraints related to the labour market and gendered work orientations that affect the intersection between class and gender in different ways in different upper-class fractions. Given the studies mentioned above, we would expect that within the upper class, men in the economic fraction are the least likely to have a partner in an upper-class position.

In sum, the theoretical framework leads to the following expectations: (i) because of the similarity in the class conditioning of dispositions (i.e. similar modes of attraction and appreciation), those in upper-class positions are more likely than those in other class positions to have an upper-class partner; (ii) because the cultural upper class is comparatively more homogenous in terms of other manifestations of dispositions (e.g. lifestyles, attitudes and outlooks), the cultural upper class is more likely than the other upper-class fractions to have a partner in a general upper-class position and to have a partner in the same upper-class fraction; (iii) because the similarity in class trajectories means similarity in the class conditioning of dispositions over the life course, ‘established’ upper-class individuals originating from upper-class positions are more likely than upwardly mobile ‘newcomers’ to the upper class to have a partner in that class; (iv) because constraints related to the labour market, gendered work orientations and dynastic wealth management within families tend to affect women’s careers negatively in terms of class destination, men in the economic upper class are the least likely to have a partner in an upper-class position.

Data and methods

We exploit the richness of administrative registry data covering the entire Norwegian population. The data were anonymised by Statistics Norway and consist of various registers such as employment, marriage and tax records. In addition to married couples, the data also offer information about cohabiting couples with children. The registers are demographically linked, allowing for class origins to be identified based on parents’ occupations.

We study the birth cohorts of 1955–1965 with upper-class destinations in 2012 (i.e. ages 47–57). To map class position, we use the Oslo Register Data Class scheme (ORDC), developed by Hansen et al. (2009) and used for a number of purposes in class analysis, for instance mapping class mobility (Flemmen et al. 2017), educational inequalities (Andersen and Hansen 2012) and residential segregation (Ljunggren and Andersen 2015).

This class scheme is inspired by Bourdieu’s (1984) model of the social space and distinguishes between classes and class fractions along two dimensions. It has a primary hierarchical dimension of the total amount of capital that differentiates between four main classes: the upper, the upper-middle, the lower-middle and the working class (see Figure 1). A second dimension of capital composition crosscuts these classes: moving from left to right in the figure, the preponderance of economic capital increases, while the preponderance of cultural capital decreases.

Figure 1. The ORDC scheme with examples of the most dominant occupations.

ORDC relies on occupational classification, with the supplementary use of information about income (earnings, capital income and income from self-employment). Our analysis focuses on the upper class – defined as those possessing the highest volumes of capital – and the three class fractions therein. The economic upper class is defined as dominant positions in business, such as chief executives, financial brokers and owners of large businesses or rentiers. Information about individuals’ income allows one to categorise the self-employed, proprietors and rentiers by singling out individuals with no registered information about occupation but who nonetheless have substantial capital income or self-employed income. This is a major advantage, since class schemes relying solely on occupational classifications tend to neglect such propertied classes. ORDC relies on information about income to establish a capital volume principle within the business sector, and only those with income in the top ten per cent are assigned an upper-class position.

The cultural upper class is defined as those possessing the most cultural capital and those who possess considerable symbolic power over national cultural production and representation. Specifically, this class fraction is defined as consisting of top academic positions (e.g. professors), the field of cultural production (e.g. top artists) and cultural institutions (e.g. museum directors).

Finally, the balanced upper class is defined as those who possess large volumes of capital, with a fairly balanced composition of both cultural and economic capital. This fraction comprises top positions within the state bureaucracy (e.g. top-level civil servants and ministers) and the elite professions (e.g. surgeons and judges).

We classify individuals’ class positions, their class origins and their partner’s class positions according to the ORDC class scheme. Given the aforementioned likelihood of the women of upper-class households being ‘housewives’, we classify partners with no occupational information and no class position as ‘outside of the labour market’ (including welfare recipients), as we do not wish to exclude these partnerships from our analysis. To record class origin, we have used data about parents’ occupations; these data were retrieved from censuses in 1970 and 1980. To classify them according to the ORDC scheme, we have opted for a ‘dominance approach’, using the highest registered vertical class position of either parent in either year. If both parents were on the same vertical level, we have prioritised economic capital as the dominant position.

Information about partnerships was gathered from the register on civil status. This register started in 1992 and includes information about marriages, registered cohabitees with children and the various forms of marriage dissolution (whether widowed, separated, divorced, etc.). We study civil status in 2012. We have included information about the previous partner if the status of ‘unmarried’ was recorded in 2012, irrespective of whether this was due to bereavement or some other form of dissolution.

There are, however, some limitations to our study. First, because of the way the official registers are organised, we lack information about cohabiting couples without children. Second, it is likely that homosexual households are underestimated (approximately 0.48 per cent of the upper-class population is registered in same-sex marriages, and such marriages have been registered only since 2009). Third, we record only the most recent marriage observed, and we do not consider whether previous relationships were homogamous. If, for instance, there are systematic differences in the durability of homogamous marriages, we may be underestimating the occurrence of heterogamous marriages (see the discussion in Schwartz and Mare 2012). Related to this, there might be systematic differences in homogamy by marriage order. For instance, if later marriages are less homogamous than first marriages, analysing current marriages may result in lower homogamy estimates compared to estimates based only on first marriages. Finally, we do not have information about class positions before marriage. This may lead to an underestimation of homogamous relationships. If, for instance, a partner with an initial upper-class position has opted to drop out of the labour market after marrying an upper-class individual, the partner will be classified as ‘outside the labour market’ in our analysis.

We analyse the relative tendency among people in upper-class positions to have partners in that class. In other words, we do not study whether the upper class is more or less homogamous compared to other classes but ask whether the upper class is comparatively more likely to have an upper-class partner.

First, we assess whether there are differences among people in different class positions and their likelihood of having a partner in the upper class. To this end, we analyse vertical class differences and disregard horizontal differences between the class fractions, i.e. the dependent variable only distinguishes between upper-class partners (regardless of fraction) and partners who are not in an upper-class position. For the independent variable, we distinguish between four classes: the upper class, the upper-middle class, the lower-middle class and the working class.² The working-class category is the baseline.

Second, we estimate whether there are differences in the likelihood of having an upper-class partner according to one’s class trajectory, i.e. among people who are currently in the upper class but who have different class backgrounds. Here, we distinguish between three types of class trajectory: a stable upper-class trajectory (those currently in and originating from the upper class), the short-range mobile (upper-class individuals with upper-middle class origins), the mid-range mobile (upper-class individuals with lower-middle class origins), and the long-range mobile (upper-class individuals with working-class origins). The last of these is the baseline category. Thus, by estimating the likelihood of being in a homogamous relationship according to one’s own class trajectory, we can assess differences between ‘heirs’ and different types of ‘newcomers’ to the upper class.

Third, we examine whether the patterns we have revealed thus far are different if we include horizontal differences between the three upper-class fractions. For both the individual and the partner, we distinguish between the cultural, the economic and the balanced fractions. The dependent variable differentiates between the three upper-class fractions and one non-upper class category. The independent variable differentiates between the three fractions, and we pool the remaining population as a baseline due to the rarity of working-class individuals with partners in each upper-class fraction.³

Fourth, we investigate the likelihood of having a partner in the same upper-class fraction as oneself according to one’s class trajectory. The independent variable distinguishes between four class trajectories: a stable fraction-specific trajectory (those currently in and originating from the same upper-class fraction); a stable upper-class trajectory, but originating from a different upper-class fraction; the short-range mobile (upper-class individuals with upper-middle class origins), the mid-range mobile (upper-class individuals with lower-middle class origins), and the long-range mobile (upper-class individuals with working-class origins). The last of these is the baseline category.

In all the analyses, each estimate is calculated separately for men and women. We also include dummies to account for structural mobility between eleven birth cohorts.

We employ logistic regression to assess the odds ratio for having a partner in the upper class and multinomial logistic regression to assess the relative risk ratio for having a partner in each fraction of the upper class.⁴ As the logistic regressions provide an account of the relative tendencies, we also provide an initial account of the absolute proportions of homogamous upper-class men and women.

Table 1 shows the percentage shares and frequencies for the upper-class population, amounting to 5.2 per cent of the relevant age groups. The table indicates in percentage points how these individuals are different from their age group with respect to their civil status, gender composition, class origin, as well as the class position of their partners. Compared to the remaining birth cohorts, those in upper-class destinations disproportionately originate from upper-class families. They are also disproportionately more likely to be married rather than unmarried, divorced or widowed. Their partners are also overrepresented in upper-class positions. Moreover, upper-class recruitment is profoundly gender skewed, and this echoes previous Norwegian studies (Toft and Flemmen 2018). As a staggering 72.3 per cent of the upper class are men, this means that ‘perfect’ upper-class homogamy is logically impossible: many upper-class (heterosexual) men ‘must’ necessarily have partners outside the upper class.

Table 1. Descriptive statistics for upper-class subpopulation.

	Frequency	Per cent	Cum.	Percentage point difference difference
Class position
Cultural fraction (1.12 % of age group)	9,702	21.47	21.47
Balanced fraction (2.30 % of age group)	20,008	44.29	65.76
Economic fraction (1.78 % of age group)	15,469	34.24	100.00
Gender
Men	32,658	72.29	72.29	21.13
Women	12,521	27.71	100.00
Class origin
Upper class: culture	1,700	3.76	3.76	2.84
Upper class: balanced	5,630	12.46	16.22	9.22
Upper class: economic	2,040	4.52	20.74	3.36
Upper-middle class: culture	2,409	5.33	26.07	3.42
Upper-middle class: balanced	4,269	9.45	35.52	3.63
Upper-middle class: economic	5,079	11.24	46.76	5.89
Lower-middle class: culture	1,378	3.05	49.81	1.25
Lower-middle class: balanced	3,487	7.72	57.53	−0.39
Lower-middle class: economic	2,067	4.58	62.11	0.85
Skilled working class	4,863	10.76	72.87	−4.22
Unskilled working class	5,656	12.52	85.39	−12.58
Farming/forestry/fisheries	1,622	3.59	88.98	−1.47
Missing^a	4,979	11.02	100.00	−11.80
Civil status (2012)
Spouse or partner	31,186	69.03	69.03	17.95
Divorced, widowed, etc.	7,463	16.52	85.55	−3.92
Unmarried	6,846	14.36	99.91	−5.51
Missing	42	0.09	100.00	−8.52
Partner’s class position
Upper class: culture	1,967	4.35	4.35	3.71
Upper class: balanced	3,064	6.78	11.14	5.29
Upper class: economic	1,613	3.57	14.71	2.23
Upper-middle class: culture	3,050	6.75	21.46	3.98
Upper-middle class: balanced	6,377	14.11	35.57	7.14
Upper-middle class: economic	3,106	6.87	42.45	2.58
Lower-middle class: culture	982	2.17	44.62	1.04
Lower-middle class: balanced	4,064	9.00	53.62	3.67
Lower-middle class: economic	2,509	5.55	59.17	1.21
Skilled working class	2,275	5.04	64.20	−4.57
Unskilled working class	1,833	4.06	68.26	−5.35
Farming/forestry/fisheries	127	0.28	68.54	−0.67
Outside of the labour market	4,429	9.80	78.35	−1.81
Missing	9,783	21.65	100.00	−18.47
Total N:	45,179	100.00
Total partner's N	38,649	100.00

^aAmong the upper-class subpopulation for which parental data are lacking, 91 per cent are first-generation immigrants from European countries, many of which are culturally similar to Norway (e.g. the remaining Nordic countries, the UK, and Germany). Only 75 individuals are descendants of two foreign-born parents. The ethnic composition of the Norwegian upper class is in other words overwhelmingly homogeneous. Thus, it is highly unlikely that ethnic divisions stratify the marriage patterns observed.

Results

Partnerships within the upper class: absolute rates

Figure 2 shows the distribution of the vertical class positions of the partners of the upper class by gender. The majority of both men and women have partners outside their own class. Partly reflecting the vast gender skewedness of upper-class recruitment, there are also substantial gender differences: upper-class women tend be more homogamous than upper-class men. 36 per cent of the women are homogamous, compared to 13 per cent of the men.

Figure 2. Distribution of upper-class partner’s vertical class position, by gender.

Next, we assess the horizontal differences between class fractions. Figure 3 shows the tendency for having a partner in the same upper-class fraction, denoted by the dark grey bars. There are marked gender differences, particularly in the economic fraction: 24 per cent of the women in the economic upper class have partners in the same class fraction, compared to 3 per cent of the men. There are also large differences among upper-class men: men in the economic fraction are particularly likely to have a partner in the lower-middle or the working class, whereas men in the cultural fraction are comparatively more likely to have a partner in the upper-middle or the upper class. Among upper-class men, the tendency for fraction-specific homogamy is strongest in the cultural fraction and weakest in the economic fraction. Among upper-class women, this tendency is strongest in the economic fraction and weakest in the balanced fraction, though the differences are smaller. As anticipated, we can also see that men in the economic fraction of the upper class are the most likely to have a partner outside the labour market.

Figure 3. Distribution of upper-class partner’s horizontal class position, by gender and upper-class fraction. Fraction-specific homogamy marked in dark grey.

Upper-class homogamy

Differences in percentages mask the different ‘marriage markets’ facing men and women, since they do not account for the sizes of the class fractions. Because of the substantial gender skew in each fraction, the pool of potential upper-class partners varies extensively for (heterosexual) men and women. Men constitute 72 per cent of the upper class as a whole, and male dominance is particularly striking in the economic fraction (86 per cent). We thus turn to the relative estimates of having an upper-class partner (for the complete regression tables, see Appendix A). To ease interpretation, we present our estimates in odds ratios (for the average marginal effects, see Appendix B).⁵

Figure 4 shows that, compared to the working class, all the other classes have greater odds of having an upper-class partner. The biggest class differences are found among men: the odds of having a partner in an upper-class position are 14.7 times higher for upper-class men than for working-class men (12.2 times higher among women).

Figure 4. Odds ratios for having a partner in the upper class, by gender and vertical class position. 95% confidence intervals.

Is the tendency for having a partner in the upper class conditioned on one’s own class trajectory? Figure 5 shows that the tendency for upper-class homogamy is significantly higher for stable upper-class men and women compared to the upwardly mobile with working-class origins. This is also the case for those originating from the middle classes, although the coefficients are smaller.

Figure 5. Odds ratios for having a partner in the upper class, by gender and class trajectory of upper-class subpopulation. 95% confidence intervals.

Fraction-specific homogamy

When the upper class marry homogamously, is the partner located in the same upper-class fraction, and does the likelihood of such fraction-specific homogamy vary between the fractions? Figure 6 depicts how individuals in the three upper-class fractions compare to the remaining population in their tendency for having upper-class partners.

Figure 6. Relative risk ratios for having a partner in the upper class (first row) and in the same upper-class fraction (second row), by gender and upper-class fraction. 95% confidence intervals.

The top section of the figure shows, separately for men and women, the differences between the upper-class fractions in their relative odds for having a partner in the upper class, regardless of the fraction of the partner. Men in the cultural fraction have the highest relative odds for having an upper-class partner as opposed to a partner outside of that class, whereas men in the economic fraction have the lowest. Among the women, the differences between the class fractions are smaller and the coefficients are not significantly different from each other.

The bottom section of Figure 6 illustrates the relative risk ratio of having a partner within each of the upper-class fractions compared to having a partner outside the upper class. For each upper-class fraction, there is a tendency for fraction-specific matching: the relative risk ratio for having a partner in the cultural fraction is largest for men and women in the cultural fraction, and a similar pattern is found in both the balanced and the economic fractions. Models using each upper-class fraction as a baseline confirm that such fraction-specific homogamy is statistically significant (see Appendix C).

Figure 6 also shows that among the upper-class fractions, the relative tendency for marrying within one’s own fraction is clearly strongest in the cultural fraction. The relative risk ratio is about 24 times higher than in the population as a whole. Within the upper class, we also find that men and women in the cultural fraction are the least likely to have partners in the economic fraction. Conversely, men and women in the economic fraction are, compared to men and women in the other upper-class fractions, the least likely to have partners in the cultural fraction. The fraction-specific patterns in these associations are similar for both men and women.

Are there also differences between the upper-class fractions in terms of how their class origins intersect with the tendency for fraction-specific homogamy? In Figure 7, we estimate the same relationships as in Figure 5, but we now introduce intra-class differences between the fractions. The top coefficient in each section of the figure denotes the relative estimate for having a partner in the same upper-class fraction among those with a stable fraction-specific trajectory (i.e. those with both origin and destination in the same upper-class fraction). In comparison to the upwardly mobile, a stable fraction-specific trajectory does indeed strengthen the tendency for fraction-specific homogamy. This is true of all three upper-class fractions and both sexes.

Figure 7. Relative risk ratios for having a partner in the same upper-class fraction, by gender and fraction-specific class trajectory. 95% confidence intervals.

However, only in the balanced upper-class fraction do we find that the fraction-specific measure of class origin yields higher estimates compared to a measure of general upper-class origin, but these coefficients are not significantly different from each other. The confidence intervals highlight that each of the upper-class fractions amounts to less than two per cent of the population. Thus, we do not gain statistical power by singling out fraction-specific trajectories when studying marriage within the upper class. Nonetheless, the estimates are less vulnerable to sampling errors, as they encompass the whole population and they signify sociologically meaningful patterns.

Concluding discussion

Our analysis contributes several key insights about class inequalities today. First, while most upper-class individuals marry across classes, our analysis suggests tendencies for upper-class closure: upper-class individuals are disproportionately more likely to have partners in upper-class positions. Second, our analysis indicates horizontal intra-class divisions: there is a systematic tendency within the three fractions of the upper class to have a partner in the same fraction. This aspect of class homogamy has mostly been neglected in previous research. This is striking, since a number of studies have demonstrated connections between horizontal class divisions and a range of aspects of people’s lives, such as cultural consumption, physical activities, bodily ideals, political stances and spatial whereabouts (Ljunggren and Andersen 2015; Prieur et al. 2008; Vandebroeck 2017). Insofar as class homogamy is, at least in part, produced and reproduced through dispositional affinities in ‘schemes of perception and appreciation’ (Bourdieu 1984: 241–4), there is reason to suspect that horizontal class divisions are at work beyond our particular empirical case.

In line with our expectations, the cultural fraction of the upper class stands out from the economic and balanced fractions in terms of fraction-specific homogamy. This resonates with the few studies that have operationalised horizontally differentiated class fractions (Hansen 1995; Kalmijn 1994). This may be partly attributed to the fact that the lifestyle of the cultural fraction of the Norwegian upper class is comparatively more distinct than the lifestyle of the economic fraction (Flemmen et al. 2019). It may also be the case that people in the cultural upper class are comparatively less desirable outside their own circles. Ljunggren (2017) has found that the Norwegian cultural elite has expressed frustration over a general lack of status and recognition, a finding that resonates well with Krogstad’s (2019) study of the portrayal of the cultural elite in the media as ‘politically correct’, ‘high-cultured’ and ‘arrogant’.

There may also be differences linked to the social characteristics of where and when people form romantic relationships. Given the key role of the education system as a site for finding a partner (Blossfeld and Timm 2003), it seems understandable why the cultural fraction stands out. Distinctive in terms of their possession of higher-education credentials, it is plausible that the educational settings of the cultural fraction (and, later, their work settings) are more socially homogenous than those of the economic fraction: as demonstrated by Flemmen (2012), parts of the Norwegian economic upper class have never attended institutions of higher education.

Yet we can also observe a notable relative inclination for fraction-specific homogamy within the economic upper class. We would thus highlight that cultural matching – i.e. a mutual attraction arising from classed affinities in habitus – might be at work across the class structure (Bourdieu 1984: 169–225). According to some influential researchers, however, the process of cultural matching is regarded as linked exclusively to cultural capital (see e.g. DiMaggio and Mohr 1985; Kalmijn 1994). Kalmijn (1994) suggests that matching linked to similarities in outlooks, behaviour, opinions, tastes, styles of speech and so forth, applies only to occupations in the cultural domain, whereas occupations in the economic domain are governed by a ‘competition regime’ that entails more rational calculation in marrying ‘upwards’ in the quest for economic resources and status. However, as we understand the notion of cultural matching, it suggests that individuals in economic class fractions are characterised by their own distinct lifestyles and ways of being, an understanding that is supported by a range of empirical studies (see e.g. Prieur et al. 2008; Rivera 2012; Vandebroeck 2017; Jarness et al. 2019).

Third, our analysis indicates closure by class trajectory. Those with a stable upper-class trajectory are disproportionally more likely to have a partner in the upper class. Moving down the vertical dimension of class origin, the tendency for having a partner in the upper class decreases systematically. A probable reason for this, we would argue, is that individuals with similar class trajectories are endowed with similar dispositions, which facilitate homogenising experiences, mutual attraction and sympathies. If the upper-class trajectory has been stable, individuals will have shared a persistent ‘distance from necessity’ (Bourdieu 1984: 53–6), a type of class condition newcomers have only experienced for parts of their lives.

Additionally, the classed conditioning of habitus is arguably an important factor in structuring the pool of potential partners over the life course, through processes of differential association, selective educational trajectories and spatial segregation. In Norway’s capital, Oslo, adolescents from upper-class families tend to live in class-segregated areas over the life course, which means that they are persistently surrounded by affluent neighbours (Toft 2018). Educational choices are also stratified by class origin, and narrow recruitment to elite education persists over time (Strømme and Hansen 2017). The geographical and institutional sites for meeting a potential partner thus seem to be contingent upon class trajectory: the whereabouts of individuals originating from the upper class are distinctly different over time from the whereabouts of newcomers. Thus, a similar class origin may not only predispose upper-class individuals to appreciate each other; it may also increase the likelihood of having met and intermingled at specific geographical sites and within institutional milieux.

Crucially, this way of understanding the connection between class origin and class destination is distinct from understanding the two in terms of ‘ascribed’ and ‘achieved’ dimensions of assortative mating (see e.g. Blau and Duncan 1967; Kalmijn 1991b). According to these authors, the relative strength of the two dimensions is indicative of the pervasiveness of meritocracy in society: if homogamy by class origin is high (and homogamy by class destination is correspondingly low), this would indicate that people are under the influence of largely constraining social structures and kinship traditions. The opposite scenario would indicate that people have become less bound by constraining structures and traditions, and thus choose partners according to their own ‘achievement’ (Kalmijn 1991b: 497).

However, this conceptual distinction may risk glossing over the importance of class trajectory. For instance, preferences, lifestyles and ways of being are at least partly conditioned by class conditions early in life: as highlighted by the notion of habitus clivé, newcomers to the upper class tend to feel ‘out of place’ when navigating unfamiliar cultural and social terrains (Friedman 2016). Moreover, attendance at and one’s level of success within institutions, such as the education system and the labour market, cannot reasonably be seen as solely attributed to individuals’ idiosyncratic ‘achievement’, since both attendance and success are widely known to be stratified by class origin (Friedman and Laurison 2019; Rivera 2012). We would thus suggest that the connection between homogamy by class origin and homogamy by class destination is better understood as an indicator of the interplay between early and later class conditioning.

Fourth, our analysis demonstrates interesting intersections between class and gender. In particular, the odds ratios indicate that upper-class men are relatively more homogamous than their female counterparts. Moreover, among upper-class men, there are important intra-class differences: men in the cultural fraction are the most likely to have an upper-class partner, while men in the economic fraction are comparatively more likely to have partners in other regions of the class structure.⁶ This latter tendency may partly reflect gendered work orientations within elite business circles, making women more likely to opt out of their careers, or reduce their commitment to work, after marrying economically privileged upper-class men (or after childbirth) (see e.g. Halrynjo and Lyng 2010; Glucksberg 2018; Aarseth 2016). Thus, the higher level of ‘class mismatch’ observed among men in the economic fraction might be partly attributed to the gendered division of labour within the household regarded as necessary for dynastic wealth management.

In sum, our analysis lends credence to Bourdieu’s (1976, 1984) framework for studying class homogamy. Specifically, it has demonstrated the pervasiveness of the model of the social space in contemporary patterns of romantic relationships. Although our data do not allow for the investigation of specific modes of partner preferences, nor the contents of subjective classifications, our analysis can serve as a working hypothesis for further investigation of the idea that the embodiment of distinct classificatory schemes of attraction and aversion vary according to intersections between capital volume, capital composition, class trajectory and gender.

Acknowledgments

The article is part of the research project 'Paradoxes of Wealth and Class: Historical Conditions and Contemporary Figurations' (HISTCLASS) at the Department of Sociology and Human Geography, University of Oslo. We would like thank Statistics Norway (SSB) for providing the administrative registry data. We would also would like to thank Sam Friedman, Magne P. Flemmen, Marianne N. Hansen, Jørn Ljunggren, Thea B. Strømme, Håvard Helland, Patrick L. Andersen, organisers and participants at the Studying Elites workshop at Copenhagen Business School in April 2019, as well as organisers and participants at the Research Seminar at Centre for the Study of Professions at Oslo Metropolitan University in September 2019 for insightful comments on earlier draft of this article.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes on contributors

Maren Toft is a Postdoctoral Research Fellow in sociology at the Department of Sociology and Human Geography at the University of Oslo. Her research interests lie in stratification research and cultural sociology, with a focus on processes of class formation. Her previous work places emphasis on biographical divisions within the upper class, while her more recent work concerns historical dimensions of upper–class formation. Recent publications have appeared in Sociology, British Journal of Sociology, Socio–Economic Review and The Sociological Review.

Vegard Jarness is a Researcher at Centre for the Study of Professions at Oslo Metropolitan University. His research interests include class, status, and cultural and political divisions. Recent publications have appeared in Sociology, British Journal of Sociology and Poetics.

Notes

1 The expectation of fraction-specific homogamy runs counter to what might be inferred from theories of assortative mating that rely on rational action and exchange theories. According to Becker (1981: Chapter 4), ‘superior’ people with complementary traits will tend to marry each other. For instance, if one is well-educated and knowledgeable within the domain of legitimate culture, the utility of this competence increases if the spouse is economically wealthy: this enables investment in pieces of art and the chance to entertain and shine in high-status social circles (see the discussion in Hansen 1995: 208-9). Although Becker never considered complementary exchange in terms of both vertical and horizontal class divisions, the theory arguably entails class homogamy, but not fraction-specific homogamy. Assuming that all upper-class individuals can be classified as ‘superior’, cross-fraction marriages within the upper class are, according to the theory, rational, since a match between partners situated in the cultural and the economic upper class is mutually beneficial in terms of complementary traits. Thus, it might be inferred from the theory that cross-fraction relationships within the upper class will be more salient than fraction-specific homogamy – i.e. the opposite of Bourdieu’s theory.

2 Here, we collapse skilled workers, unskilled workers and welfare recipients.

3 The results are also robust with a working-class baseline. However, changing the baseline circumvents having to map very large relative risk ratios. For instance, the relative risk ratio for having a partner in the cultural fraction of the upper class is 62 when comparing men in the cultural upper class to men in the working class.

4 See Appendix B for average marginal effects (AME).

5 The results presented are corroborated by the AME estimates, although women’s relatively greater likelihood of homogamy is more pronounced than the estimates using odds ratios show.

6 The differences we observe relate only to the class destination of the partner. However, homogamy might also be facilitated by class origin. For instance, as demonstrated by Wagner et al. (2020) in the case of Denmark, offspring of parents at the top of the wealth distribution tend to marry among themselves. Thus, it may be the case that by measuring the class destination of the partner, we are underestimating the level of homogamy, particularly within the economic fraction of the upper class. We have, however, conducted additional analyses using the partner’s class origin as the dependent variable and the results corroborate our findings: men in the economic fraction have a relatively greater inclination to have partners outside the upper class than men in the cultural fraction.