Structure and Historicity of Cultural Tastes. Uses of Multiple Correspondence Analysis and Sociological Theory on Age: The Case of Music and Movies

The ‘Omitted Axis’: Authors who, Following Bourdieu, do not Retain Age

Created in 1904 by the British psychologist Charles Spearman to measure intelligence, factorial analysis became correspondence factor analysis (CFA) in France in the 1960s under the impulse of the mathematician Jean-Paul Benzécri (Cibois, 1981). CFA was previously used in applied research (CREDOC, French Research Institute for the Study and Monitoring of Living Standards) and in the processing of economic and social data (INSEE, French National Institute of Statistics and Economic Studies), as well as in research in sociology and geography. In the early 1950s, Bourdieu and Benzécri were both students of the prestigious École normale supérieure on the rue d’Ulm in Paris and from that time onwards they never lost sight of each other (Benzécri, 2006). It is something of an ideological paradox that in the 1970s Bourdieu used multiple correspondence analysis to explain cultural practices, as he stood at the ideological antipodes of Benzécri. ² However, the paradox is easier to understand when we remember that Bourdieu (1984 [1979]: 580, footnote 6) also considered social space to be structured by ‘dimensions’ located in the factors of the analysis. The function of this analysis was to represent the levels of attraction and independence between the modalities of the variables.

On the subject of ‘Life-Styles’ in Distinction, Bourdieu argues that:

Economic and social condition, as identified by occupation, gives a specific form to all the properties of sex and age, so that it is the efficacy of the whole structure of factors associated with a position in social space which is manifested in the correlations between age or sex and practices. (Bourdieu, 1984 [1979]: 105–106)

Age is conceived as a secondary variable. For Bourdieu, age is structural, and not historical. Age refers to the succession of time and the struggle between ‘social positions’ in a cultural field, and not directly to the cultural time of history i.e. the succession of music genres, artwork and artists. Bourdieu does not retain age, but he makes it a ‘herit-age’, which on careful examination, was already being structured in the 1960s. This link between age structure and levels of education was present in Distinction through the following two multiple correspondence analyses: ‘Variants of the dominant taste’ and ‘Variants of petit-bourgeois taste’ (Bourdieu, 1984 [1979]: 262, 340). Nevertheless, it lacked further analytical and a fortiori theoretical scrutiny from Bourdieu.

In the first of the two analyses, ‘Variants of the dominant taste’, an ‘axis’ of age links the 18–30 modality, bottom left, to the ‘over 60s’ modality in the top right. An ‘axis’ of education is partly perpendicular, running from ‘postgraduate degrees’ in the top left to ‘vocational diplomas, Primary School Certificate [CEP], no qualifications’ in the bottom right. The configuration is the same as that identified earlier for tastes in music. A similar connection between age and education can be found in the ‘Variants of petit-bourgeois taste’ (Bourdieu, 1984 [1979]: 340). Bourdieu’s treatment of Axis 1 and Axis 2 does not take account of age, though there is a very clear generational effect. Age modalities indicate graphically, through their dispersion, their very strong, if not their strongest link to Axis 1 (notwithstanding the position of the individual socio-professional categories), and to Axis 2 in the case of the dominant taste. Bourdieu’s qualification of the second axis as a ‘social trajectory’ axis is problematic. Age, as correlated to the second factor in the two plane diagrams, is identified by Bourdieu, but handled in a particular way: it is removed in favor of an opposition based on social origin i.e. class seniority. Bourdieu writes:

that, as in the case of the dominant taste, the second factor expresses an opposition by age (the oldest are at the top of the second axis and towards the economic pole, the youngest at the bottom of this axis and towards the cultural pole) and, inseparable from this, an opposition by social origin: the children of big or small employers, senior executives or professionals are situated on the positive-value side, the children of manual workers, clerical workers or junior executives on the negative-value side. (Bourdieu, 1984 [1979]: 341)

It is as if, in Distinction, the old enjoy inherited status, while the young are the children of the working classes (Bourdieu, 1984 [1979]: 265–266). In summary, Bourdieu removed age from the analysis and replaced it with social origin i.e. ‘seniority in the bourgeoisie’, which describes Axis 2. In particular, he removed age as a historical and cultural generation corresponding to the frequentation of a body of work and values in relation to the cultural offer at the time.

In Distinction, the generational dimension of the age variable is clearly visible in people’s choice of favorite singer. In the ‘Variants of petit-bourgeois taste’ figure (1984 [1979]: 340), Luis Mariano (born 1914) and Georges Guétary (born 1915) are positioned close to older people (between the over-60s and the 46–60 bracket); between the 46–60s and the 30–45s we find Edith Piaf (born 1915); close to the 30–45s, Georges Brassens (born 1921), Jacques Brel (born 1929); and finally, near the 20 year olds, Françoise Hardy (born 1964). A similar structure is visible near the age modalities for ‘Variants of the dominant taste’, with Françoise Hardy bottom left, Georges Guétary on the right and Luis Mariano at the top.

It seems that a certain number of contemporary authors have renewed an approach to cultural practices by excluding age. The factorial method used by Van Eijck (2001: 1182) is based on the establishment of an ascending hierarchical classification derived from a cross-tabulation between musical genres and levels of education. The following independent variables were used in the analysis: gender, level of education and occupational status. ³ As we can see, while gender is included, age is not. Meanwhile, Warde and Modesto Gayo-Cal (2009) establish – for each musical genre – the ratio of individuals without a diploma to graduates among those who claim to like the genre, and divide the genres into three classes according to whether this ratio is between 0 and 1 (‘lowbrow’), between 1 and 2 (‘middlebrow’) or 2 and higher (‘highbrow’). Once again, only the level of education variable is used. Robette and Roueff (2017) base their sociological interpretation of the factor analysis of multiple tables of four cultural practices (television, music, movies and reading) on ‘the volume of economic and cultural capital’: ‘If Axis 1 [taste] is therefore related to the overall volume of capital, the structure of the capital is also present, in which cultural capital plays a leading role. Axis 2 is correlated primarily with age, and secondarily with gender’ (Robette and Roueff, 2017: 375). ⁴

Finally, the multiple correspondence analysis performed by Prieur et al. (2008) appears to be problematic from a sociological and epistemological point of view since it involves using the socio-demographic variables (10 in the present case) as active variables and it considers the practices as illustrative variables. ⁵ It openly flouts the principle advocated by Benzécri that ‘the model must follow the data, not the reverse’. The first axis of the analysis has been interpreted as expressing the volume of capital, and is retained as an indicator of cultural legitimacy. The musical genres are then considered as additional variables along this axis and their coordinates are read off as legitimacy scores. This way of proceeding deliberately builds a social space from a selection of sociological variables grouped into two types of capital: economic and cultural. In their article, Flemmen, Jarness and Rosenlund (2019: 161) do exactly the same by using the categories as active variables, but it seems difficult to find a ‘structure strikingly similar to the model of the social space advanced by Bourdieu’ (2019) which replaces the Guttman effect. As Desrosières writes:

A form of parable appears whenever we study a strongly hierarchical structure, which is the case with social structure, in another way. The first axis, which explains most of the variance, opposes the two ends of the hierarchy. The second, which is roughly the axis of the parable, opposes the middle categories to the extreme categories. (Desrosières, 1978: 100)

We could imagine a ‘space’ with the two variables of age and gender: there would be an axis setting the youngest and the oldest in opposition and another setting male and female in opposition. The most scientifically remarkable, and yet significant, concerns ‘the space of lifestyles’. Flemmen et al. note:

For present purposes, we focus on axes 1 and 3. Due to limited space, we only focus on the axes that correspond to the structures of the social space. The omitted axis 2 depicts a division between an established, legitimate lifestyle characterised by a taste for canonised items and a more culturally savvy and emerging lifestyle, characterised by a taste for alternative and as yet uncanonised items. This axis is strongly influenced by age differences. (Flemmen et al., 2019: 163) ⁶

So in order to find the homology thesis, Axis 2, namely the age variable, is explicitly omitted in the factor analysis.

In Distinction, Bourdieu did not perform multiple correspondence analysis by using the socio-demographic variables as active variables; the practices are active, whereas the socio-demographic variables are illustrative (they were initially printed on a transparent page for the ‘social space’ and the two factor analyses (Bourdieu and Saint-Martin, 1976)) and it is their superimposition that supports the homological interpretation. There are good scientific reasons why sociologists do not assume a priori – even on the basis of a theory (which is also embedded in a historical epoch) – that a society is structured in such and such a way, but rather establish how a cultural practice is structured by observing the relation to a supplementary/explanatory variable. Moreover, we are not going to find any axis for age or gender, or anything else, if from the outset these variables have been excluded from the analysis, as is the case with the analyses examined here.

The ‘Descriptive Axis’: Authors who Highlight Age but do not Theorize it

Bennett and colleagues (2009) include age – as well as gender and ethnicity – in their analysis of British cultural tastes in the 2000s. They note age differences in the structure of tastes in music, and write: ‘Musical boundaries are closely associated with those of age, such that we might well talk about generations with different kinds of omnivorousness’ (Bennett et al., 2009: 82). They indicate an initial structuring principle derived from multiple correspondence analysis and based on the level of cultural participation. Subsequent lines of inquiry highlight the distinction between contemporary/commercial cultural practices and traditional practices related to age – between indoor practices and outdoor practices, and between intense practices and moderate practices. Roose et al. (2012) show that the first two strands of their analysis are very similar to those indicated by Bennett et al. (2009) in how they link to level of education and income (Axis 1), and age (Axis 2). Prieur and Savage (2013) comment that for Denmark and the UK: ‘firstly . . . the level of participation distinguishes better between survey respondents than the specific preferences or activities do . . . Secondly, age, gender and/or ethnicity may discriminate better in cultural matters than class does’ (2013: 252). They add: ‘it seems, for instance, that musical taste is strongly structured by age, and reading by gender’ (2013). Their objective is to examine the contours of a possible new cultural capital, and age does not appear to be a theoretical question in itself. This assertion is supported in the Danish studies by Prieur et al. (2008) and the Finnish studies by Purhonen et al. (2011). Ganzeboom (1990: 29) has already described how lifestyles are structured according to cultural, economic and age-related dimensions.

Several authors place particular emphasis on the role of age and generation in a process of ‘horizontal differentiation’ (Lizardo and Skiles, 2015: 10). Bellavance (2008), Berghman and Van Eijck (2009) and Purhonen et al. (2009) show that the classical distinction between ‘top’ and ‘bottom’ intermeshes with a differentiation between classical genres and contemporary genres. Age and gender appear as structuring variables in the practices of Dutch (Van Eijck and Knulst, 2005) and British people (Gayo-Cal et al., 2006). Analyses of television, music, reading, the arts and sport (Bennett, 2006; Savage, 2006; Silva, 2006; Warde, 2006; Wright, 2006) point to differences between the highbrow cultural participation of older generations and the popular participation of younger generations.

Tastes in Music

The following questions were used as active modalities: genres of music ‘listened to the most often’ (+ sign), genres of music ‘never listened to because not liked’ (- sign), and ‘having been to concerts (rock, jazz, classical or other music) during the last twelve months’. Gender, age, individual socio-professional category, socio-professional category of the respondent’s father, level of education and frequency of listening to music were used as supplementary variables. We retain the plane for the first two factors for which the first two eigenvalues (V) are greater than 0.2 (V1 = 0.38; V2 = 0.31), which means that the factors can be interpreted (Benzécri, 1973: 43). To study the first factorial plane, we selected, from among the 36 active modalities, those that made a greater than average contribution (equal to 1000/36, or 27.8) to either of the first two factors.

Figure 1 shows, on the negative side of the first factor, individuals who often listen to classical music as well as opera and jazz, who reject hip-hop and electronic music, and who have attended a classical music concert. The modalities of the supplementary variables, which were examined using the same threshold, are the highest age categories: 50–64, 65 and over, and retired. These, then, are the likes and dislikes of older people. On the positive side of the first axis we find the exact opposite: what was rejected is accepted and vice versa. At this pole is a taste for hip-hop, electronic music, international varieties and rhythm & blues, and a rejection of jazz, opera and classical music. The supplementary variables are: 15–17, 18–24 and high-school/college students. The practices of this cluster are linked to younger categories. Overall, the first factor highlights two opposing poles: the tastes and rejections of older people and younger people.

OPEN IN VIEWER

Figure 1.

Space of tastes in music – Plane diagram of Axis 1 and Axis 2.

The second factor, being a correction of the first factor, is easier to study factorially using only active modalities with a contribution to the factor (noted as CPF) higher than or equal to the average (28) and supplementary (illustrative) modalities with a CPF higher than or equal to 13. These modalities are underlined horizontally for the first factor and vertically for the second factor.

The pole reflecting the choices and rejections of older people (bottom left) is the only one that contributes to both factors: young music is rejected and there is occasional listening to French songs. This pole also reflects infrequent listening. Conversely, in the upper right-hand corner, the highest frequency of listening (‘every day or almost every day’) also contributes to both axes, with contributions from younger age groups (aged 15–34) and associated genres. The other diagonal shows an opposition between classical and jazz concerts, which contributes to both axes (top left), and, symmetrically, the rejection of jazz in the lower right quadrant. This second diagonal is linked to educational qualifications: high in the upper left and absent in the lower right.

Connections between likes and dislikes illustrate a ‘normative value’ of music for specific groups: the first demonstrates a taste for classical, world music and French songs, and a rejection of international songs, R&B, hip-hop, rap, electro, metal and hard rock; the second demonstrates a taste for hip-hop, rap, international songs, R&B, disco and funk, and a rejection of world music, opera, classical and jazz; the third demonstrates a taste for metal and hard rock, and a dislike of French songs. Other positive tastes are not adjacent to specific music genre dislikes.

To take this further, a simplified graph (Figure 2) was produced. This represents all the supplementary socio-demographic variables, as well as the links between age and level of education (excluding the Primary School Certificate [CEP] for the oldest generation). This figure shows the orthogonality of the two dimensions of age and level of education, the first being related to Axis 1 and the second to Axis 2, but positioned on the diagonals of the graph.

OPEN IN VIEWER

Figure 2.

Space of tastes in music – Supplementary socio-demographic variables – Simplified plane diagram of Axis 1 and Axis 2.

In order to see how these two variables fit together, one solution is to create a new supplementary variable that distinguishes age categories according to level of education, since gender does not play any role in the first two factors. For each age category we constructed three new educational modalities: ‘upper’ (baccalaureate and beyond), ‘middle’ (vocational diplomas [CAP/BEP/BAC/BP/BT]) and ‘low’ (Primary School Certificate [CEP], Junior Secondary Education Certificate [BEPC], or no qualifications). To visualize the effect of this new variable we took into account an individual’s universe by drawing ellipses on the factorial space to encompass a given proportion of individuals (Figure 3). The ellipse with a median concentration of 50% of the population was chosen. ⁸

OPEN IN VIEWER

Figure 3.

Space of tastes in music – Plane diagram of Axis 1 and Axis 2 – Median ellipses.

Only the ellipses for the three levels of education of the 65-and-overs and the 18–24 year-olds were drawn (the youngest have not yet graduated). It is notable that tastes in music are not related to age or level of education, but to the intersection of the two. Both dimensions intervene: all highly educated age groups tend towards the top left. At this high level of education in a concert culture, we move from left to right (from older to younger) in the following order: classical music, jazz, rock, blues, metal and reggae. Conversely, people with lower levels of education reject the previous tastes in the case of the youngest, while the older groups replace them with infrequent or no listening. In conclusion, tastes in music depend primarily on age, but also on level of education. ⁹

Tastes in Movies

In order to understand the structuring of tastes in movies, the following questions were used as active modalities: ‘favorite movie genre’, ‘movie genre not seen because not liked’, ‘preferred movies’ (‘American movies’, ‘French movies’, ‘no preferences or other cases’), ‘particularly liked movies’ or ‘movies that you dislike’ (from a list of 14 movies). In addition, the following variable was used: ‘number of movies watched during the year’, and the following socio-demographic variables were used: gender, age, individual and father’s socio-occupational category and level of education (Figure 4).

OPEN IN VIEWER

Figure 4.

Space of tastes in movies – Plane diagram of Axis 1 and Axis 2.

The plane of the first two factors forms the best possible summary of the data. The first two eigenvalues (V1 = 0.31; V2 = 0.24) are greater than 0.2, meaning that the relationship can be interpreted. The above-average contributions to the first factor on the negative side (1,000/62 active modalities i.e. 16.1 per thousand) are: ‘prefer French movies’, ‘dislike horror movies’ and ‘dislike sci-fi movies’, but ‘like romantic (love story) movies’, ‘like documentaries’ or ‘like no genre in particular’, ‘appreciate Titanic’, ‘do not want to watch Star Wars or Lord of the Rings’ and finally, the supplementary modality ‘have not been to the cinema over the last twelve months’. The supplementary socio-demographic modalities (with a threshold reduced to 8 per thousand) are: ‘retired’, ‘65 and over with lower or middle educational level’, ¹⁰ ‘50 to 64 with lower educational level’, ‘woman’, ‘housewife’ or ‘father owns a farm’. The negative side of the first factor is thus related to the highest ages and to the lower or middle levels of education; however, age, education and gender can interact. For age and level of education, as in the previous analysis, a crossover operation has been performed to make the factorial plane diagram easier to interpret.

For the positive side of the first factor, the higher than average contributions are: ‘prefers American movies’, ‘does not like musicals’, ‘does not like romantic movies’ or ‘does not like art-house movies’, ‘likes ‘sci-fi’, ‘likes horror’ or ‘likes action’ movies’; favorite movies were Star Wars, Matrix, Lord of the Rings, Pirates and Shrek. The supplementary modalities (threshold of 8 per thousand) are: ‘high-school/college students’, ‘18–24 years, medium education level’, ‘15–17 years, lower education level’ and ‘male’. The first factor therefore represents an opposition between young and old for the middle and lower levels of education.

The second factor, on the negative side (at the bottom) for the most popular active modalities, corresponds to: ‘prefers comedies’, ‘particularly liked Les Bronzés, Camping, Les Visiteurs, Asterix’, and ‘likes Lord of the Rings’. There are no supplementary contributory modalities above the threshold. On the positive side are the following active modalities (same contribution threshold): ‘prefers art-house movies and drama’, ‘especially likes The Lives of Others, Brokeback Mountain’, and ‘does not want to see Camping, Les Bronzés’. The associated supplementary variables are: ‘goes to the cinema more than once a month’, ‘postgraduate level and higher’, ‘first (bachelor’s) degree’, ‘aged 34–49’ and ‘aged 50–64 with higher educational level’, and the socio-professional categories ‘management and liberal professions’, and ‘father in management and liberal professions’.

In summary, if we look at the modalities that make the largest contribution, we find a triangular opposition: the most educated at the top prefer Auteur movies and dramatic comedies; they liked The Lives of Others and Brokeback Mountain, and did not like Les Bronzés, Camping or The Brice Man. The younger people on the right prefer action movies, sci-fi movies, and movies in their original language: they liked Star Wars, Matrix and The Lord of the Rings; they do not like musicals. These two poles often go to the cinema (5 to 12 times per year). The third pole is represented by older people with a low level of education who prefer French movies and comedies, and who do not like horror movies or science fiction. They loved Love Story, Les Bronzés and Camping, and did not like Star Wars.

The only slight difference between men and women is that women watch fewer movies than men (an average of 5.6 movies per year compared to 6.1 for men). Conversely, both men and women from all age groups with a high level of education are frequent watchers (women aged 65 and over watch more than men).

Figure 5 shows a triangular structure with a visible distinction between the age opposition for the first factor and the ‘cultured’ movie pole at the top. This result becomes clearer on examination of the median ellipses. By taking the age limits available for the three levels of education, we see that the three ellipses for each age category are on the left for the oldest respondents and on the right for the youngest (first factor age opposition), but that they overlap more closely at the ‘cultured’ pole of the second factor, as the level of education increases.

OPEN IN VIEWER

Figure 5.

Tastes in movies – Plane diagram of Axis 1 and Axis 2 – Median ellipses.

Tastes in music and movies both reflect the age structure, which appears first but cannot be examined in isolation – only in conjunction with levels of education. Glevarec and Pinet (2013) have shown that level of education is the most contributory socio-demographic variable to Axis 1 of the space of French cultural practices, and age in the case of Axis 2. For tastes in music and movies, analysis shows that the priority of these variables is reversed, with age coming before level of education.