Social bases of material consumption: The relationship between social groups and possession of household appliances in Denmark

Mette Hove Jacobsen, Aalborg Universitet, A.C. Meyers Vænge 15, København SV 2450, Denmark.

Email: mej@sbai.aau.dk

Practices have increasingly become the primary object of social science analysis, and the concept of practice has been applied in various empirical studies. Within the sociology of consumption, engagement with 'the practice turn' (Schatzki et al., 2001) has especially drawn attention towards the mundane activities of everyday life and the ordinary and inconspicuous aspects of consumption (for a review, see Warde, 2015).

There is no unified theoretical approach to practices, but prominent sociologists such as Pierre Bourdieu (1977 [1972]) and Anthony Giddens (1984) are often associated with practice theory. They were especially interested in reconciling individual agency and social structures (Ortner, 1984). More recently, empirical sociology has especially drawn on the work of Judith Butler (1989), as well as on ethnomethodology and science and technology studies (e.g. the work of Callon and Latour (1992), Pickering (1995) and Latour (2005)).

Within most practice theoretical accounts, practice is understood as organised bodily activity, and the bodily conduct plays a central role. Human bodies are socially constrained and at the same time the locus for agency. The ambition of reconciling agency and social structures was an important rationale for the early practice theorists to focus on bodily conduct. Arrays of human activities become intelligible practices, both for the individuals participating in the practice and for potential future practitioners. They are performed without much reflection and without the need to articulate them (for further reading on theories of practice, see, for example, Reckwitz (2002b) or Rouse (2007)).

In this process, practical understanding becomes important: knowing how to do something, and knowing what things to desire and what to avoid. This is an important theme for sociological studies drawing on theories of practice, and is variously described as practical sense (Bourdieu, 1977 [1972]), practical

consciousness (Giddens, 1984), practical understandings (Schatzki, 1996), and more recently as understandings (Warde, 2005) and competences (Shove et al., 2012). The locus of such practical understanding, however, varies across scholars. While Pierre Bourdieu examines practical understandings as embodied in humans, Bruno Latour, centred in science and technology studies and actor-network theory, studies practical understandings as embedded in material objects. In this perspective, the material objects become sites of understanding, a materialised understanding, where objects act as resources that enable and constrain the characteristic of a practice (for a more comprehensive theoretical articulation of this, see Reckwitz (2002a)). Within the last decade, the second perspective has received increasing exposure and further development within the sociology of consumption and the sociology of sustainable consumption. For example, Shove (2003) fills in a gap within studies of science and technology by studying conduct arising not just from individual devices but also from interdependent systems of materials. She demonstrates how shared understandings of good and appropriate conduct and their reproduction through practices reflect the external affordance and constraints of material arrangements. In similar ways, Shove et al. (2014) show how escalating energy demand is generally rooted in material systems. In this perspective, regularities in, and reproduction of, practices are, to a large extent, conceived and analysed as properties of the material layout over which people have limited control (see Warde (2014), for a similar discussion). The first perspective, that of Bourdieu, which emphasises the embodiment of practical understandings, has, however, received much less attention within practice theoretical accounts in general and within the sociology of sustainable consumption in particular.

According to Bourdieu (1984 [1979], 1990 [1980]), particular constellations of practical understandings are reflected by specific social trajectories and particular types of social environment. These different life experiences become embodied in the individual and constitute a stock of resources with which individuals, families or social groups act in the pursuit of establishing and reproducing social relationships. These resources provide the individual with the capability to decipher 'cultural codes' and behave accordingly. This is captured by Bourdieu's concept of habitus, a complex historical system of dispositions, which both generates and structures practices (Bourdieu, 1990 [1980]: 53). Habitus then becomes a routinised way of knowing how to act when appropriate circumstances arise, a practical understanding which is embedded in the body and acquired through experience. The experience is not solely individually stored but also collectively. The practices become identifiable to the agents participating in the specific practice as well as to observers who have been exposed to similar social trajectories and hence socialised in similar conditions. This creates homogeneity across social groups, where similar practical understanding of what constitutes competent conduct is applied in the field of practices or in its subdomains (e.g. consumption). Participation in, and performance of, practices then becomes signs of group affiliation united by individuals with similar social trajectories. Empirical research within the sociology of consumption has shown social differences in the interpretation of practices and the

use of goods. For example, Holt (1997b) demonstrates how different social trajectories, that is, different cultural frameworks, influence patterns of conduct and orientation towards specific activities and specific material objects.

Empirical studies within the sociology of sustainable consumption often assume that practices and their social meaning are changed and reproduced through infrastructures and different constellations of material objects. When exploring the production and reproduction of conventionally shared understandings and common procedures, focus has, to a large extent, been on how infrastructures and systems of technologies co-evolve with understandings of appropriate conduct. In this sense, infrastructures and systems of materials get a background role in shaping the characteristics of a practice and in explaining the social processes that generate patterns of consumption⁶ (see e.g. Shove et al., 2015). In this context, past experience is materially stored, and systems of material objects compose a knowledge depot that is prepared to travel between individuals (this has been theoretically conceptualised by Shove et al. (2012, 2015)). To Bourdieu, meanings and practical understandings are 'objectified' in relational differences in the ways in which people act and are embodied in humans as different constellations of dispositions. Consequently, the engagement in, and performance of, practices are in this perspective interpreted as properties of social groups as opposed to material arrangements. Furthermore, regularities in practices and common procedures are conceived as a consequence of embodied systems of dispositions that vary across social groups, not a consequence of systems of technologies and material arrangements.

From the point of a theory of practice, moments of consumption occur within socially meaningful practices (Warde, 2005). Most, if not all, practices entail consumption and require goods and services in order to be performed in a satisfactory fashion. Observed patterns of consumption can be interpreted as a result of the total amount of the practices, in which a person or household engages (Warde, 2005). From a practice theoretical perspective, patterns of consumption thus provide a good indication of regularities in consumption practices.

Every individual engages in a considerable number of practices, each with its own required equipment, and that number is increasing (Warde, 2005: 141). This is echoed in the prevalence of household appliances. Figure 1 shows that the prevalence of different appliances in Danish households has increased significantly within the last 20 years, with the invention of many new technologies with new and often multiple applications. These changes have made practices increasingly subtle, advanced and complexly intertwined.

The supply and diffusion of consumer goods have made them accessible to a wide segment of the population, and in today's Western societies, the consumption of material goods plays a large role in people's daily life. Notwithstanding these trends, differences in social groups can have an impact on the type and amount of

goods one can consume. Inspired by STS, human use of machines and artefacts has become fundamental for many scholars within the sociology of sustainable consumption when exploring practices and their performance. However, in this article, patterns of possession of household appliances are seen as an indicator of a set of social practices, organised around shared understandings of appropriate conduct. It is the differentiation of these practices that this article endeavours to depict, categorise and understand.

The analysis is based on a survey on possession of household appliances in Danish households, carried out in 2014 by Statistics Denmark as part of their annual omnibus survey. This survey includes a randomly selected, national representative sample.⁷ A household is defined as one or more persons over 20 years old who are related and live together with their children.

The 20 household appliances selected from the survey reflect different practices and cover a wide range of what have been characterised by marketing writers as time-using appliances (e.g. e-book readers and Blu Ray players) and time-consuming appliances (e.g. tumble dryers). Some of the appliances are easy to group; for example, tumble dryer, dishwasher, microwave and robot vacuum cleaner are related to housework practices, telephones are related to communication and

videogame consoles to entertainment.⁸ Other appliances, like desktop computers, tablets and smartphones, reflect more than one practice. These appliances reflect practices of entertainment and information acquisition and processing as well as communication. Smartphones also reflect the practice of photography, like digital video cameras and digital cameras, and global positioning system (GPS) watches reflect navigation and sports activities. A broad range of indicators are of great importance in order to capture both the plurality of material consumption practices (not being concentrated within a particular field) and the potential mechanisms of differentiation that exist in society (not over-emphasising minority participation). Table 1 presents the 20 variables used to measure material consumption practices. The indicators are dichotomous and the households are classified as having or not having the appliances.

A substantial amount of socioeconomic and demographic information is included by combining the sample with information from the administrative registers

DAB: digital audio broadcasting; GPS: global positioning system.

The table displays the mean and in brackets the standard deviation.

in Denmark. The descriptive statistics for these variables are shown in Table 2. The socioeconomic variables include occupational status, education, family income, assets, building ownership and living space. Occupational status is measured by four occupational categories using the head of the household⁹ as the basis for constructing the categories. The classification of occupational status is inspired by the Erikson–Goldthorpe–Portocarero (EGP, see Erikson and Goldthorpe (1992)) social class scheme. In addition, I include a fifth category that contains the unemployed and those outside the labour market (e.g. the retired and students). Education is measured using the highest education in the household and contains three categories. Family income is measured by the logarithm of disposable household income. Assets are measured by the value of the property, bank deposits and stocks, bonds and mortgages. As socioeconomic variables, building ownership is also included, measured as a dummy variable with two categories (owner-occupied/

cooperative-occupied is coded 0, and rented housing is coded 1), and living space is measured by the logarithm of square metres.¹⁰

The demographic variables include age of the adults living in the household, measured by the average age, and ethnicity, measured by a dummy variable (coded 1 if a person in the family immigrated¹¹ to Denmark from another country and 0 otherwise).

Life-stage variables include number of people living in the household and children, measured by a dummy variable (coded 1 for having children living in the household and 0 otherwise). In addition, information on whether or not the household lives in an apartment building is included. Together, these variables measure how different patterns of consumption relate to different household characteristics.

This section presents the empirical approach used to answer the research questions. This involves finding distinct patterns of consumption and analysing the socio-economic and demographic correlates. Latent class analysis (LCA) is introduced as an approach to capturing an organising principle, latent subgroups, in rather complex response patterns. The LCA is combined with a range of logit models to estimate the relationship between the latent subgroups and explanatory variables. This empirical approach has been applied in different domains of consumption, for example, in studies of cultural consumption practices (e.g. Katz-Gerro and Jæger, 2013; López Sintas and García Álvarez, 2002).

LCA is a subset of structural equation modelling and can be categorised as the categorical counterpart to factor analysis, in the sense that LCA aims to identify a limited number of distinct latent subgroups, or classes, with similar response patterns, that is, in this case similar patterns of material consumption practices (measured by possession of household appliances). In other words, LCA is a method that constructs statistically independent and mutually exclusive subgroups, which can be characterised on the basis of the possession of household appliances and to which the households have a certain probability of belonging. This is in contrast to factor analysis that characterises the latent variable as continuous and is used to cluster practices or products together.

If we let $j$ ($j\; =\; 1,\; \backslash dots,\; J$) be the items included in the model and $r\_j$ ($r\_j\; =\; 1,\; \backslash dots,\; R\_j$) the response categories, then the latent class model with $C$ classes can be expressed as

In the model, the $\backslash gamma$ parameters are the probabilities of latent class membership and the $\backslash rho$ parameters are the conditional probabilities of having each of the

appliances given latent class membership. Consequently, the model shows the conditional probability of a particular response as a function of the probability of membership of each subgroup (Collins and Lanza, 2010).¹² Given their patterns of possession, all households are assigned to the latent subgroup that they have the highest probability of belonging to.¹³ In order to model demographic and socioeconomic differences across latent subgroups, the approach is extended with logit models of the form:

In the model, the parameter $\backslash beta$ indicates the probability of group membership controlling for socioeconomic and demographic characteristics. Consequently, the model examines the extent to which the exhibition of distinct patterns of consumption can be accounted for by relevant explanatory variables.

Denmark is a small and relatively socially and ethnically homogenous country, which is why I expect that the macro context in which I interpret the results will have less impact compared with larger countries with a more diverse national context. Denmark is a comprehensive welfare state, with the explicit aim of promoting equality both economically and educationally by means of its state-financed post-compulsory education system and redistributive policies. This makes Denmark a good case for studying social bases of consumption practices.

The previous fit statistics in Table 3 indicate that there are underlying subgroups of households with similar response patterns and hence, similar understandings of material consumption practices. Table 4 presents the conditional probability of having the different appliances.

On the basis of the patterns of possession, I have labelled the subgroups Unlimited, Outdated, Limited and Updated. The estimated sizes of the latent subgroups show that the largest group, the unlimited, represents 33% of the households. The smallest group is the updated, representing 15% of the households. The outdated represents 32% and the limited group 19% of the households. The first latent group (33%), labelled unlimited, possesses a variety of appliances. The members of this group are distinguished from the other subgroups by their relatively high probability of possessing a broad range of appliances and a higher-than-average probability of possessing all types of appliances. Derived from the conditional probabilities, the unlimited group is characterised by its members having a high probability of possessing appliances related to activities of entertainment and acquisition of information, such as e-book readers and tablets. Furthermore, the unlimited group is characterised by its relatively high probability of engaging in what can be characterised as housework activities, reflected by the possession of, for example, a tumble dryer. For this group, a normal and decent life seems to include an abundance of household appliances that reflect housework practices as well as state-of-the art equipment. Whether the appliances are more or less sustainable is not possible to comment on; however, it is well-established that overall levels of consumption are generally unsustainable and transition towards more sustainable forms of consumption is about more than just efficiency gains (Munasinghe et al., 2009).

The second latent group (32%), the outdated, differs from the unlimited, given their generally lower probability of possessing the different types of appliances and especially their relatively low probability of possessing the newest state-of-the-art equipment. Presumably, they have neither the competences nor the need to keep

LogL: Log-Likelihood; AIC: Akaike information criterion; BIC: Bayesian information criterion.

DAB: digital audio broadcasting; GPS: global positioning system.

up, but they still maintain what can be considered as an adequate amount of appliances to meet their demands and wishes, both quantitatively and qualitatively.

The third latent group (19%) is labelled limited because its members possess only a limited number of the goods from the analysed list. Households that exhibit this pattern are characterised by having a relatively low probability of possessing any of the appliances, and particularly of possessing state-of-the-art equipment and goods that reflect entertainment. Finally, the fourth latent group (15%) is labelled updated, as members of this group differ from the other groups by exhibiting a consumption pattern characterised by decidedly state-of-the-art goods and to a lesser extent possess the goods that reflect housework. They tend to have 'updated' practices compared with the outdated group and have a higher probability of possessing state-of-the-art equipment, for example, a Blu Ray player, MP3 player and smartphone, and a lower probability of possessing DVD players, CD players and digital cameras. This group seems to pursue relatively advanced and new technologies with multiple applications.

The different patterns of possession of household appliances identified and described above involve certain and different understandings of material consumption practices. In this second part of the analysis, I take a closer look at the subgroups and examine the extent to which belonging to a certain latent group can be accounted for by socioeconomic and demographic characteristics, and whether different consumption practices are socially exclusive or inclusive. My analytical strategy is to examine the relationship between group membership and a number of socioeconomic and demographic variables using logit models. The estimates presented in Table 5 show the odds ratio of belonging to each of the four groups: unlimited, outdated, limited and updated, relative to the probability of belonging to the other groups, conditioned by the selected socioeconomic and demographic characteristics.

Table 5 shows a highly significant relationship between variables that relate to the economic capital of the unlimited. Households that are likely to exhibit unlimited consumption patterns have higher incomes and live in bigger houses¹⁴ than those that are likely to exhibit the other consumption patterns. On the other hand, family income has a negative effect on the probability of belonging to the limited group, which means that members of this group have a lower income compared with those of the other subgroups. This group also tends to live in smaller houses.

Both households exhibiting outdated and households exhibiting updated consumption patterns differentiate from the other subgroups on the basis of educational level. The probability of exhibiting outdated or updated consumption patterns increases with higher levels of education, whereas the probability of exhibiting limited consumption patterns decreases with higher levels of education. The observed patterns show a significant relationship between material abundance and economic capital and a significant relationship between less-materialistic consumption patterns and cultural capital (measured by educational level). This resembles other studies of consumption (e.g. Holt, 1997b). In this sense, a key theoretical approach taken by Bourdieu in Distinction is to disaggregate culturally oriented patterns consumed by cultural elites versus economically oriented patterns consumed by economic elites. With a bigger sample size, it would be very interesting to explore how combinations of high income and high levels of cultural capital would fit into the models. It would also be interesting to explore how material consumption patterns relate to groups of people with high income and low cultural capital and the opposite, how material consumption patterns relate to groups of people with low income and high cultural capital.

Some of the household durables are technically advanced, and the competences required to use them include up-to-date knowledge. Ilmonen (2004: 32) argues that in these cases, it takes longer to learn to use them, to embody the knowledge and to transform it into routine practices. It appears from Table 5 that younger generations are more likely to exhibit updated and unlimited consumption patterns,

The table states the beta coefficient, the significance level indicated by * and standard errors in brackets *, **, ***.

while older generations are more likely to exhibit outdated and limited consumption patterns. It would be interesting to explore the relationship between consumption patterns and age in-depth. For example, it would be interesting to explore the relationship between consumption patterns and age groups with different levels of income or education. However due to the sample size, this was not possible.