Are socio-economic inequalities in diet and physical activity a matter of social distinction? A cross-sectional study

Cultural participation was measured by asking respondents how often they visit art museums, historical museums, opera or ballet, classical concerts, theatre and architecture (the most commonly used cultural activities to measure ‘highbrow’ embodied cultural capital (Kamphuis et al. 2015a). Answer categories were: 0—never, 1—once per year and 2—more than once per year.

Principal component analysis (PCA) showed that all cultural activities loaded clearly on one latent factor (Eigenvalue = 2.97) and subsequent reliability analysis showed that it was possible to construct a reliable scale from the cultural activities (Cronbach’s alpha = 0.786). Consequently, all items were summed to form a scale ranging from 0 to 14. This scale was subsequently divided into quintiles, so that one higher quintile corresponds to two additional ‘highbrow’ cultural visits.

Highest attained educational level was classified according to the International Standard Classification of Education (ISCED): 1—primary education (ISCED 0–1), 2—lower secondary education (ISCED 2), 3—upper secondary education (ISCED 3–4) and 4—tertiary education (ISCED 5–7).

Household equivalent income was measured as the level of household income per month divided by the square root of the number of people living from this income. This method is often used and takes differences in needs for households of different sizes into account (OECD 2008). Income was subsequently divided into five categories: 1—< €1000/month, 2—€1000–€1500/month, 3—€1500–€2000/month, 4—€2000–2500/month and 5— > €2500/month. Sensitivity analyses using non-equivalized income categories (i.e. €0–1200, €1200–1800, €1800–2600, €2600–4000, > €4000) showed similar results (results not shown).

Sports participation was measured using the validated Short QUestionnaire to ASsess Health-enhancing physical activity (SQUASH) (Wendel-Vos et al. 2003). Participants were asked to write down up to four sports they had participated in on a weekly basis in the last month. For each sport, they reported frequency (days per week), duration (hours and minutes per day) and intensity (light, moderate, intense). Self-reported intensity and activity-specific intensity metabolic equivalents (METs) were used to calculate the numbers of days participants were active for at least 30 min at moderate or vigorous intensity (moderate intensity = 4–6 MET for 18–55 years and 3–5 MET for 55 + years). Because of skewness, the variable was dichotomized into not doing any sports for 30 min or more at moderate intensity, versus sports participation at least once per week for 30 min or more at moderate intensity.

Walking or cycling in leisure time was also measured with the SQUASH. Participants were asked how often they walked and cycled in their leisure time. For both activities, frequency (days per week), duration (hours and minutes per day) and intensity (light, moderate and intense) were reported. Subsequently, we calculated how many days per week the participant walked or cycled for at least 30 min at moderate intensity. This cut-off was chosen because 30 min or more of walking or cycling implies that the participant made the effort to be physically active rather than just a means of local transportation (e.g. grocery shopping). Because of skewness, the variable was dichotomized into not walking or cycling in leisure time for 30 min or more at moderate intensity versus walking or cycling at least once per week for 30 min or more at moderate intensity.

Fruit and vegetable consumption was measured with a food frequency questionnaire (Bogers et al. 2004). Participants were asked how many times, on a weekly basis, they had consumed fruit and vegetables in the previous month. Subsequently they were asked to indicate how many portions (pieces of fruit and serving spoons [= 50 g] of vegetables) they ate on a typical occasion. For ease of interpretation, we calculated binary measures that indicated whether or not participants meet fruit and vegetable intake recommendations. The consumption of two pieces of fruit each day was coded as meeting recommend fruit intake. The consumption of at least 200 grams of vegetables each day was coded as meeting recommended vegetable intake.

Potential confounders included were sex (male and female), age (in 10-year age groups), living together with a partner (yes, no), country of birth (the Netherlands, other), having children living at home (yes, no), employment status (full-time employed, part-time employed, unemployed, retired, homemaker, other), social participation (at least once per week, several times per month, once per month, several times per year, once per year or less) in several associations and organizations (neighbourhood association, political organization, social club, religious organization, work-related organization [e.g. labour union], sport club or volunteer organization), mental health [five-item version of the Mental Health Inventory [MHI-5] (Berwick et al. 1991)] and self-rated health (excellent, very good, good, fair, poor).

Multivariable regression analyses were used to analyse the association between cultural participation and the health-related behaviours. First, regression models were calculated to estimate the association between cultural participation and outcomes, adjusted for potential confounders. Second, education and income were added to the model to estimate the independent associations between cultural participation, education and income and the outcomes. We used generalized linear regression models with log link function and robust variance in order to derive prevalence ratios (PR) and 95% confidence intervals (CI).

Missing data were handled via multiple imputations. Twenty imputed datasets were created. Information from all variables described in the methods and several other variables is used in the imputation model. Respondents with missing data on an outcome variable were left out of the analyses concerning that outcome variable (9.7% missing on sports participation, 3.8% missing on walking and cycling in leisure time, 2.0% missing on recommended fruit intake and 2.2% missing on recommended vegetable intake). The analyses were weighted by respondent-level sample weights to account for the sampling strategy within the GLOBE study. All analyses were performed using Stata, version 15 (StataCorp LP, College Station, Texas).

In the present sample of Dutch adults, those who had a higher level of cultural participation were more likely to participate in sports, to walk and cycle in leisure time and to meet fruit and vegetable intake recommendations. These associations showed a clear gradient, even when adjusted for education and income and a range of potential confounders.

The strong association between cultural participation and the health-related behaviours presumably does not suggest that engaging in cultural activities will directly motivate individuals to engage in more healthy behaviour. Rather, it suggests that socialization in high-status environments results in tastes, preferences and attitudes that induces a tendency to consume more highbrow cultural activities as well as a tendency to eat more healthily and to be more physically active during leisure time. This argument is perhaps best exemplified by the results pertaining to walking and cycling in leisure time. Our study found that walking and cycling were independently associated with cultural participation, but not with education and income. Since these activities do not require many financial resources, or specific knowledge or skills, there is no reason for a strong causal relationship with income or education. Going for a walk or a bicycle ride in leisure time does, however, require a certain appreciation for spending leisure time carrying out these activities. This appreciation is likely developed by the same socialization processes in which enjoyment for cultural activities is attained. Walking and cycling in leisure time can thus be viewed as an expressions of taste which is also associated with being culturally active and which allows higher-status groups to distinguish themselves from others.

The association between cultural participation and sports participation similarly suggests that sports are an important part of lifestyles and act as a marker of distinction. Indeed, research has shown that different social groups participate in different types of sport (Bourdieu 1984; Kraaykamp et al. 2013; Pampel et al. 2010; Stempel 2005; Wilson 2002). For the purposes of our study, we did not specifically examine the distribution of different types of sports, but rather we examined whether the socio-economic gradient in sports involvement reflects cultural differences between socio-economic groups. Similarly, Christensen and Carpiano (2014) also found that those with higher levels of cultural capital were more likely to be involved in sports. However, in line with studies that have observed social class differentiation in types of sports (Bourdieu 1984; Scheerder et al. 2002; Stempel 2005; Wilson 2002), restricting our analysis to ‘high-status’ sports (e.g. golf, tennis, sailing) also resulted in a strong association between cultural participation and sports involvement (Online Resource Table S8). Moreover, associations with cultural participation were not observed when we restricted the analysis to sports that were not defined as ‘high-status’ sports (Online Resource Table S9). In addition to a mechanism of distinction, the association between educational and income level and sports participation suggests that other determinants related to education or income, such as financial resources or social capital (e.g. social support), are also important for sports participation. Again, this is also in line with existing research, since economic capital (i.e. income-related resources) has been found to be associated with a higher probability of being involved in sports, independent of other capital measures (Christensen and Carpiano 2014; Wilson 2002).

The strong association between cultural participation and meeting fruit and vegetable intake recommendations again likely represents the tastes and attitudes of different social groups (Fismen et al. 2012). Eating patterns and food preferences are subject to socialization processes and are internalized and converted into dispositions. Several studies have shown that high cultural capital is associated with healthier food consumption (Fismen et al. 2012; Kamphuis et al. 2015a; Kraaykamp 2002). For example, Van Otterloo and Ogtrop (1989) found that food choices among Dutch mothers from lower social classes could be characterized by a regime of ‘much, fat and sweet’, whereas mothers from higher social classes paid more attention to the healthiness of their nutritional patterns. Kraaykamp (2002) also observed taste differences between social groups in Dutch society and showed that cultural capital provides higher-status groups with resources to adopt a more exclusive consumption pattern. Kamphuis et al. (Kamphuis et al. 2015b) have recently showed that higher educational groups in the Netherlands were more concerned with healthiness than lower education groups. These results are in line with observations by Bourdieu, who suggested that lower-status groups are more inclined to eat traditional, heavier and cheaper foods, whereas higher-status group has a preference for more delicate, refined and healthy foods (Bourdieu 1984). Although differences in food consumption between social groups are much more encompassing, we restricted this study to fruit and vegetable consumption because differences in fruit and vegetable consumption are the most profound contributors to socio-economic inequalities in healthy dietary intake (Giskes et al. 2010). To test the hypothesis related to food preferences more directly we have previously examined the association between cultural participation and so-called superfoods (i.e. spelt, quinoa, goji berry, chia seeds and wheatgrass), the consumption of which can, in our view, be seen as a contemporary expression of social distinction. These results also showed a strong association between cultural participation and the consumption of superfoods, net of education and income (Oude Groeniger et al. 2017), supporting our interpretation that differences in healthy food consumption can be seen as an expression of distinction.

This study is not without limitations. First, we have interpreted the strong association between cultural participation and health-related behaviour as evidence for a social distinction mechanism. We feel confident about this interpretation since ‘highbrow’ cultural participation is a well-known indicator of embodied cultural capital and a clear marker of distinction. Moreover, cultural participation is not likely associated with health behaviour in a causal way (i.e. the mere act of cultural participation, such as visiting museums or classical concerts, is not likely to cause a strong increase in fruit consumption or leisure-time walking). However, it could be argued that other determinants may be responsible for the association between cultural participation and health-related behaviours. One is the amount of leisure time available, as more leisure time may clearly increase one’s available time to spend on both cultural participation and health-related behaviours, like walking, cycling and sports participation. We had no information on respondents’ amount of leisure time, but we tried to adjust for this factor by controlling all analyses for two important time restriction variables: employment status and whether the respondent had children living at home (Kraaykamp et al. 2008). Cognitive ability may also have confounded our results. It has been argued that cultural participation is higher among people with greater information processing capacities, because they seek more complex cultural activities to fulfil their needs (Notten et al. 2015), and cognitive ability may also result in a more healthy lifestyle. However, by conditioning on educational level we have likely accounted for a substantial part of this confounding effect. Further, cultural participation may reflect an active lifestyle rather than high cultural tastes, and this active lifestyle may also result in more sports participation and more leisure-time physical activity. Although we adjusted for social participation in several associations and organizations, there may be other determinants that affect both cultural participation and health-related behaviours that were not controlled for in our analyses.

Second, we only used a limited number of cultural activities to represent cultural participation, and cultural participation only measures one facet of the sociocultural environment of the participants. For our purposes, trying to reflect a social distinction mechanism, ‘highbrow’ cultural participation is an important and strong indicator (Bourdieu 1984; Bourdieu et al. 1991; Nagel and Ganzeboom 2015; Notten et al. 2015). Moreover, we have used the most commonly applied ‘highbrow’ cultural activities in our study (Kamphuis et al. 2015a). Notwithstanding, there are likely more relevant cultural activities and social distinction instruments (e.g. possession of arts at home, music preferences) that we have not captured in our analyses.

Third, we used self-reported information about the health behaviours. If higher socio-economic groups indeed value a healthy lifestyle more than people from lower socio-economic groups, as we hypothesized, they may be more inclined to overestimate their healthy behaviours or underreport their weight. However, the use of the SQUASH for physical activity has been shown to be reliable and valid (Wendel-Vos et al. 2003) and the use of a food frequency questionnaire was considered suitable for ranking individuals and establishing relative intake (Bogers et al. 2004).

Fourth, the use of cross-sectional data only offers a snapshot of the social and cultural processes through which dietary habits and physical activity are formed and limits the ability to make any causal interpretations.

The proposed mechanism of social distinction reflects a long-lasting process of socialization that constrains and guides behaviours throughout the life course. If behaviours are not consistent with the habits and preferences of one’s sociocultural environment, it is unlikely that they will be materialized or maintained. This may contribute to the poor effectiveness of many interventions aimed at improving health behaviours, especially among lower socio-economic groups (Beauchamp et al. 2014). To improve their effectiveness, interventions should be better aligned to what is considered meaningful and practicable in different sociocultural environments (Delormier et al. 2009). In addition, because socialization plays a strong role in the adoption of health-related behaviours, interventions focusing on young children and their parents may be most successful, as incorporating healthy habits then becomes part of the socialization process.

More research, however, is needed for more in-depth examinations of these mechanisms and the potential to intervene on them. The hypothesis that social distinction induces a healthy lifestyle in higher socio-economic groups should also be tested in a more direct manner. Since the process of distinction can be implicit, this will require more sophisticated (experimental) methods combined with more qualitative studies. One method that may be able to measure a tendency for social distinction more directly—even if this tendency is unconscious—is the use of Implicit Association Tests. Implicit Association Tests are specifically designed to include unconscious attitudes of subjects (Greenwald et al. 1998), which could be very useful when assessing a tendency for social distinction. Techniques that are used in research on social comparison may also be helpful in this line of inquiry. For instance, prototype perceptions (i.e. social images of typical persons engaging in certain behaviours) have been used to study health risk behaviours (Gibbons and Gerrard 1995). Similar techniques may also be effective in assessing social distinction tendencies. In addition, more research is needed to disentangle how these mechanisms may differ across gender and age groups.