Blind to culture. How classed cognition prevents cultural policy initiatives from increasing cultural participation for children

ABSTRACT

Since its first formulation in the post war years, Nordic cultural policy has had a stated aim to democratize culture, but with limited success. A growing literature argues that understanding policy failure is needed for a better functioning cultural policy. The present study uses two waves of panel data on cultural participation for parents and children in Bergen, Norway (N = 4754) to analyze the impact of three locally implemented cultural policy interventions designed in part to democratize culture for children. Using work with cultural policy failure, and tools from Bourdieu’s Sociology of Culture, this study finds that policies fail in systematic ways related to classed cognition. This leads to a ‘Matthew effect’ for cultural goods, as policies enable easier access to culture for those already invested in cultural participation, while the policies fail to even out social inequalities. There are both scientific and policy implications of the findings. First, data show that a cultural reproduction model seems a better fit than a cultural mobility model of cultural participation. Second, policy planners rarely factor in implications of the former in implementation of cultural policies designed to equalize distributions of culture. As a consequence, cultural policy initiatives to reduce inequalities in access to culture may lead to an exacerbation of cultural inequalities.

KEYWORDS:

• Cultural participation
• cultural policy
• interventions
• cultural capital
• democratization of culture

1. Introduction

The political goal to democratize access to cultural goods and to reduce cultural inequalities has a long history in Norwegian cultural policy (Bakke 2003; Danielsen 2008), in line with Nordic cultural policy in general (Duelund 2008). Despite efforts to democratize culture, the social distribution of cultural participation has remained stagnant along socioeconomic dimensions ever since the formulation of national cultural policy (Mangset 2012). This problem of socially stratified cultural participation is not unique to Norway (Donnat 1999; Mangset 2020; Warwick Commission 2015).

The failure to expand participation is one of many problems faced by cultural policy planners in an age of increasing demands on the legitimization of cultural spending (Mangset 2020). At the same time, there is a general lack of investigation into the reasons why cultural policy initiatives fail to extend participation beyond the typical participants (Jancovich and Stevenson 2021). Such an analysis depends upon understanding what causes socially unequal access to cultural goods.

Empirical research inspired by Bourdieu’s Sociology of Culture has established a formal model for the development of cultural dispositions, that is, the propensity to cultural participation and related practices. Briefly, cultural participation is theorized to be the result of cultural exposure in childhood and the duration of exposure to educational institutions, focusing on the influence of the family of origin. To Bourdieu, the social use of culture is to create difference, and to ensure that children accumulate cultural capital in its various forms is a part of the family’s social reproduction strategy. This model of cultural participation is usually labeled a cultural reproduction theory. A competing model proposed by DiMaggio (1982) understands cultural exposure in school as the core component, leading to cultural mobility. Despite cultural policy planners being familiar with Bourdieu’s sociology of culture, this research tradition has had limited impact on cultural policy planning (Dubois 2011). This is a problem because cultural policy planning is dependent on a working understanding of cultural participation as a social phenomenon. To complicate matters, there is no clear scientific consensus on whether cultural mobility or reproduction best explain differences in cultural participation, and even less so for children’s cultural participation.

To shed light on cultural policy’s failure to democratize culture, and on the phenomenon of cultural participation more broadly, this study analyzes three distinct cultural policy interventions, all implemented locally in Bergen, the second largest city in Norway. Data from these interventions, along with documents and records are paired with a panel study of parents and children’s cultural participation (N = 4754), carried out in three waves (2021, 2020, 2019). Three initiatives are analyzed, the large scale and nationally funded Den kulturelle skolesekken (DKS, literally ‘The Cultural Rucksack’), the locally administered Aktivitetskortet (the Activity Card), which provide a selection of free leisure and culture activities to children in low income families, and a random controlled trial of administering information packets on cultural offers to a randomly drawn treatment group of parents from the panel in wave 1.

Findings suggest that the policies in question enable a ‘Matthew effect’ for children’s cultural participation. To those who have, more will be given.¹ This study demonstrates how this failure to extend participation is connected to classed differences in cognition. The propensity to notice culture is socially distributed. Some notice culture, some are blind to culture. While it is strictly possible to acquire a general disposition towards culture outside the family, this study suggests the vast importance of family of origin to the development of a cultural disposition, in line with a mainly reproduction view of cultural participation. In the concluding discussion, I reflect on the challenges posed by these findings for the development of cultural policy initiatives.

2. Theoretical framework

In the theoretical framework, I briefly introduce Norwegian cultural policy in a Nordic context, and the framework for working with policy failure put forward by Jancovich and Stevenson (2023). I then present theoretical perspectives from Bourdieu’s sociology of culture, a key tool for understanding how cultural policy may fail.

2.1. Cultural policy and children’s access to culture

Duelund traces the notion of cultural policy in the Nordic welfare states back to its roots in the system of patronage and later, the European Enlightenment. Modern Norwegian cultural policy was formalized in the 1960s, making equal access to culture a part of the welfare state’s political duties. The people needed not only to be educated, but also culturally informed (Duelund 2008). In this discourse of redistribution and democratization of culture, the ‘culture’ that is contained or produced in the museums, theaters and concerts halls that receive public subsidy is understood as part of the intangible wealth of the nation and should not therefore be the preserve of any one group (Stevenson, Balling, and Kann-Rasmussen 2017, 99).

While cultural policy goals have oscillated over time, the goal of the democratization of the professional arts still stands firm as a key cultural policy objective in Norway, and the goal to ‘civilize’ is not abating but intensifying (Bjørnsen 2012). This is even more so the case for children’s cultural participation. The Norwegian Government and partners at the regional and municipal level have signed a declaration securing children’s rights to play and leisure, as well as access to participate on equal terms in culture (Norwegian Government 2022). Cultural policy has developed to include several other goals in the framing of cultural policy, such as promoting democratic values, connecting arts and culture to free speech, but central to cultural policy is still the aim that access should be relatively equal. These goals are stated quite explicitly (see e.g. Meld. St. 8 (2018–2019) 2018, 40), so also at the municipal level (Bergen kommune 2021, 10).

The democratization paradigm has been criticized for being elitist and paternalistic, for instance by advocates of cultural democracy, who aim for ‘a parity of esteem of aesthetic values (Bennett 2001, 5)’, and a more inclusive cultural policy (see e.g. Belfiore et al. 2023; Gross and Wilson 2020). In the democratization discourse, non-participation is often framed as a problem of barriers. The possibility that non-participation is an active choice made by capable adults is not considered within the cultural policy discourse (Stevenson, Balling, and Kann-Rasmussen 2017). Clearly non-users of art and culture lack information about cultural offers, with the implicit reasoning that ‘if they knew what they were missing, they would miss it (Dworkin 1986, 227).’² This criticism is perhaps less applicable to children’s non-participation. Children, being in a state of dependency, must be provided with the means to make their own choices. This argument is increasingly cited to justify state policies toward children according to Johanson (2010). As the state extend more and more rights to children, the state is made responsible for guaranteeing that children are capable of making use of these rights. The argument is reflected in the white paper on children’s culture to the Norwegian parliament (Storting), stating that ‘the content of kindergarten, school, the after-school program (SFO), the Culture School and the Cultural Rucksack (DKS), reflect this right [of equal participation, given by the Declaration of the rights of the child (UNCRC 1989)] by providing an offer to all children and youth. This is about experiencing art, learning about art and to themselves making use of artistic and cultural expressions. The training shall provide a competency for making their own choices, for further immersion and study (Meld. St. 18 (2020–2021) 2021, 28 my translation)’. As a result of this framing, the push toward greater equalization of culture appears more confident.

Cultural policy and public cultural spending clearly matter for participation. In absolute terms, cultural participation is systematically connected both to the level of cultural spending and the size of the cultural offer. Comparing OECD countries, van Hek and Kraaykamp (2013) found an important statistical relationship between the two, but could not find support for the argument that higher cultural spending in itself increases social inequality in cultural participation. The Nordic countries belong in the high spending category. Norway is estimated to spend nearly one percent of the Gross National Product on culture, second to Denmark at 1,4 percent among the Nordic countries (Berge, Hylland, and Storm 2022).

While the absolute level of participation is high in Norway, it remains a fact that the aim to democratize access to cultural goods has proven hard to realize (Mangset 2020). Despite decades of cultural policy and political efforts, the social distribution of access to cultural goods has remained to a large degree static across time and national contexts (Coulangeon 2013; Donnat 1999; Mangset 2012; Warwick Commission 2015). Despite this failure to reduce inequalities, democratization remains a central policy goal (Lindström Sol 2021). The pervasiveness of the problem of participation has led some scholars to a simple question. Why does cultural policy fail?

2.2. Working with policy failure

Jancovich and Stephenson argue that cultural policy planners and cultural institutions need a more active relationship to failure, to achieve social learning (Jancovich and Stevenson 2021, 2023). Previous research on policy evaluations argue that policies may both be successful and have failings, and often the borders between success and failure may not be clear cut. In order to reason about success and failure in a policy, it is therefore useful to identify different potential modes of success and failure. Jancovich and Stephenson propose four modes, adopted from their review of the literature on evidence-based policy. A policy may have (1) objective attainment failure, when the objectives of the policy are not met. A policy may have (2) distributional failure when some stakeholders are negatively affected by the policy. There can be (3) political or electoral failure when politicians are negatively affected, and there could be (4) implementational failure when obstacles prevent an effective implementation of the policy (Newman & Head, 2015, cited in Jancovich and Stevenson 2021, 970). The scope of the present study focuses on the first and last modes, that of objective attainment and implementational failure. Cultural policy initiatives are government sanctioned and funded interventions into citizens agency and practices. In order to understand how these initiatives may affect the practices of citizens, the analysis should contain a clear account of research attempting to explain agency and practices oriented toward cultural participation. I will focus on Bourdieu’s theoretical framework for understanding the social distribution of cultural goods.

2.3. Unequal cultural participation among children and young people

According to Bourdieu, cultural practices, i.e. the familiarity with legitimate culture that is part of a larger concept of cultural capital, is transmitted in the family (Bourdieu 1986). In this study, I take legitimate culture to be the culture that is sacralized and disseminated by the cultural field of production, that is, art and culture and educational institutions.³ The heritability of cultural dispositions in the family is central to Bourdieu’s concept of cultural capital. While economic capital, in most forms, is simply transferable without significant transaction costs, cultural capital is not. Cultural capital is embodied and must be accumulated through practice and experience. Part of cultural capital’s strength lies in its incorporation, blurring the difference between person, competence, and mode of acquiring this competence (Bourdieu 1986). In Distinction, Bourdieu laid out what can be summarized as a basic formula for the generation of a cultural disposition. A general, transposable disposition towards legitimate culture is the result of cultural transmission by the family and the cultural transmission by the school, and the efficiency of the latter depends on the accumulation in the former (Bourdieu 2010, 15). In this perspective, according to Bourdieu, cultural competence is predominantly reproduced, not acquired by merit, but with the appearance of merit.

An early adopter of Bourdieu’s ideas in the United States in the 1980s (Davies and Rizk 2018), Paul DiMaggio proposed two models explaining cultural participation’s role in attaining social status, one model based on cultural reproduction as proposed by Bourdieu, called the cultural reproduction model, and one model based on a view where the family had a smaller effect on participation, called the cultural mobility model (DiMaggio 1982). These models were originally devised to understand differences in educational attainment and social stratification (for a wider and critical discussion on mobility and reproduction in the Sociology of Culture, see Streib 2017) but have since been applied to understanding differences in the intergenerational transmission of cultural capital, especially differences in cultural participation (Nagel 2010). This is the point of interest in this study. Further reference to reproduction and mobility theories in this study applies specifically to explanations of cultural participation, which I take to be a part of accumulation of cultural capital in childhood.

In questions of accumulation of cultural capital, the cultural reproduction model assumes that the family mediation element of the equation is overwhelmingly large, while the cultural mobility model assumes family mediation is the minimal element. Simultaneously, the reproduction model assumes that the cultural capital acquired is most efficiently deployed by children who inherit large amounts of cultural capital, while the mobility model assumes that cultural newcomers are disproportionately well rewarded for their investment (DiMaggio 1982). The reproduction and mobility models are not mutually exclusive,⁴ and could theoretically and empirically coexist as explanations, but which theory is adopted as a framework has important implications for cultural policy assumptions. If the cultural reproduction thesis is correct,⁵ this means that cultural policy planners have far fewer tools at their disposal, if serious attempts should be made towards democratization of culture, taken as meaningful reduction in cultural inequality. This is true because educational settings are far more open to policy directives than the family unit. If cultural reproduction is the best framework for explaining differences in cultural participation, this also has large scale ethical implications, not only for how we understand the genesis of a cultural disposition, but for essential values such democracy, meritocracy, and the social distribution of power in society.

One necessary condition of the cultural mobility hypothesis is that children can efficiently acquire cultural capital outside the family. To extend this condition to a democratization discourse, we must be able to influence the rate of this accumulation. There have been sparse, but important research directed at this problem. In 1999, the mandatory arts course CKV1 was introduced in the Netherlands. Utilizing a panel recording cultural participation and cultural interest before, during, and 2–6 years after completing the course, Nagel, Damen and Haanstra found no evidence of heightened cultural participation or interest following the course (Nagel, Damen, and Haanstra 2010). In a similar vein, Kracman found some link between arts course attendance and heightened cultural participation, but this effect was diminutive compared to the mediating effect of family background (Kracman 1996).

The experimental work of Bowen, Kisida and Green show that it is in fact possible to influence children’s cultural capital through interventions in school settings (Bowen, Greene, and Kisida 2014; Greene, Kisida, and Bowen 2014; Kisida, Greene, and Bowen 2014). In a rigorous random controlled trial, Bowen, Green and Kisida analyzed the effect of treating school groups to a tour of a professional art institution. Even as elements of cultural reproduction were clear, such as students with higher inherited cultural capital on average showed more enthusiasm for cultural consumption, their study also shows how disadvantaged students experienced larger gains in their desire to consume culture than did advantaged students after treatment. This study could not show a change in actual consumption (as was Nagel’s chief interest), nor in the ability to deploy cultural capital (Lareau 2011; Lareau and Horvat 1999), but it did show how interventions can influence children’s valuation of cultural goods.

At the same time, this conclusion is challenged by a recent study by Coulangeon and Fougère (2022). An important question raised by these scholars is the question of within-group selection. Is this phenomenon one of cultural mobility or cultural compliance? Coulangeon and Fougére assessed the impact of a two-year learning program offered by the National Opera of Paris to middle school students in underprivileged areas of Paris, where students were introduced to the world of opera. Results show that the students who followed the program the full term had a positive and significant impact on their academic performance, while those leaving after only a year had a negative impact. Using counterfactual analysis (propensity score matching), Coulangeon and Fougére found that the contrast between the effects may be predominantly a selection effect in favor of culturally compliant students, in line with reproduction theory, not mobility theory (Coulangeon and Fougère 2022).

This aspect of the cultural mobility thesis, the possibility of interventions working through school, is in no way conclusively demonstrated or rejected.⁶ However, there is a large empirical literature working on questions of the reproduction and mobility thesis’ element of the relative importance of family transmission and educational exposure. When these approaches are contrasted in empirical research, educational exposure matters, but family transmission effects is consistently found to be dominant (Georg 2016; Kraaykamp 2003; Nagel 2010; Nagel and Ganzeboom 2002). There is also research indicating that the description of this process as an accumulation over childhood is apt (Blaabæk 2022; Jæger and Breen 2016). If the reproduction theory is dominant, this creates considerable problems for efforts to democratize culture.

3. Data, cases, and methods

3.1. Data

This study contributes to the understanding of cultural policy efficacy and aims to contribute to social learning (Jancovich and Stevenson 2021, 2023) by analyzing three different cultural policy initiatives in Norway’s second city, Bergen. In order to analyze policy efficacy, we cannot rely on ‘off the shelf’ data. Using cases which have clear geographical and temporal boundaries enables a high degree of analytical rigor, especially when matched with a unique dataset, the Families with Children survey. The Families with Children panel survey was launched in 2019, with the goal of mapping the structure of children’s access to organized leisure activities, from sports and leisure to participation in arts and culture. The panel design is equal probabilities, all biases known, meaning that every family in the city was invited to participate, and biases are known due to access to key socioeconomic variables present in the population dataset. Close to one in five families chose to participate (N = 4754). The resulting sample had a normally distributed income distribution, reaching both families with low and high income, and is skewed towards families where at least one of the parents had high levels of education. Over- and underrepresentation is handled with controls in multivariate analysis. A statistical description of the data is provided in the Appendix, Table A1.

There are several core advantages offered by the data set. It gathers data on children’s participation in culture, as well as parental current and historical structure of participation, so children and parents may be analyzed together. Having data on cultural practices for both children and parents is a rare quality for survey data (Octobre et al. 2011). Moreover, the geographic specificity and measurement of named cultural institutions allows highly optimal conditions for an analysis of cultural capital (DiMaggio 1982, 199). Finally, the most important quality of the data to the present study, is that it is a repeated survey. This enables controlled analysis of change over time, measures of data reliability, and experimental research designs such as randomized controlled trials, methods that I will draw upon in this analysis. All analyses are performed in R.⁷

3.2. Cases

Before reviewing the results of the study, I will briefly present the three cases. The three cases are selected for two key reasons: all three cases aim to increase cultural participation in a democratization of culture context, and ample data exist for these cases, enabling a rigorous research design. The context for the study is Bergen, Norway’s second largest city. Bergen was one of the European Capitals of Culture in 2000, and has an ambitious cultural policy, both for the quality of art and for the key role of the arts in society (Bergen kommune 2015).

3.2.1. Universal arts exposure through primary schools – DKS

Den kulturelle skolesekken (literally ‘The Cultural Rucksack’, hereafter DKS), is by far Norway’s largest cultural policy intervention. The measure was launched in 2001 and is a permanent, national scheme rooted in the tradition of seeing culture as a civilizing force (Bjørnsen 2012). The target of the policy is that all pupils in primary school should be exposed to a wide variety of professional arts and culture (Meld. St. nr. 8 2007–2008, 2007). DKS is thus a collaboration between departments of culture and education, at the state, regional and municipal levels of government. The policy is unique in a European context, as it encompasses all pupils in primary and secondary school. In a report to the governmental body in charge of DKS, the goal of the policy is summed to be to ‘arm children and young people with cultural capital … (Hauge et al. 2017, 3 my translation)’. 310 million NOK was allocated to DKS from the state in 2022, which is set aside to fund the arts in the program, while regional and local authorities fund administrative costs.

The implementation of DKS is not homogeneous. A minority of municipalities, Bergen included, receive funding from the state and manage all aspects of the policy, what is known as direct municipalities (direktekommuner). Direct municipalities evaluate and award funding to the artists who apply to provide artistic offers. As a direct municipality, Bergen enforces enhanced reporting from the artists participating in DKS. The reporting scheme includes naming which schools and which age cohorts within each school (trinn) participated in an artist’s arts offer, along with an estimated number of pupils attending.

Bergen combines mandatory and optional features in DKS. Some artistic offers are offered to citywide age cohorts, while for other age cohorts, participation in DKS is mandatory. In those cases, schools and artists find each other on Den kulturelle markedsplassen (literally ‘The Cultural Marketplace’, sic) (Hauge et al. 2017, 43). Bergen’s profile focuses on letting pupils meet artists where the art is being made, in the city’s many cultural institutions. The city therefore covers transportation costs (Hauge et al. 2017, 19).

Using artists reports, this study constructs a dataset on DKS-participation containing detailed data on school, age-cohort and participation in DKS from 2008 to 2018 (N = 8703). This allows for an estimation of publicly administered cultural exposure, which can be compared with cultural exposure provided by the family.

3.3. The activity card - targeted subsidized intervention

The second cultural policy is Aktivitetskortet (literally ‘The Activity Card’). In 2014 The Municipality of Bergen applied for funds from a state program for the alleviation of consequences of poverty in childhood. Bergen was chosen for a pilot program and a 2.5 million NOK grant. The target group was families with incomes below the EU poverty line of 60 percent of the median income. In order to practically delineate a target group, and at the same time to access a distributional mechanism, The Bergen City Government chose to distribute the card to parents who had received Sosialhjelp (emergency social aid) in the calendar year (Bergen kommune 2016a). The stated policy goal of the Activity Card was to reduce the effects of growing up in poverty by providing access to the city’s athletic- and cultural offer, and enable normal participation in social life (Bergen kommune 2016b). In 2017, the City Government briefed the City Council on the progress made by the card. The policy goal was expressed as enabling participation in cultural- and leisure activities and the reduction of social exclusion. In this briefing, the main cinema provider in the city reported 3690 free tickets cashed by the card, the main swimming hall reported 2789 free visits and the city’s aquarium reported 1150 free visits. The two largest theaters reported a combined 86 free tickets cashed (Bergen kommune 2017).

Since its adoption and expansion, the card has found its way into many of the City Government’s action plans, and the latent cultural policy aspects in the original City Government Propositions became more explicit. In a section devoted to making a case for the link between health outcomes and cultural participation, the card is presented as an important policy initiative to increase audience diversity and secure greater access to organized culture (Bergen kommune 2021). There is movement in the written record of policy goals of the card, from a very clear goal of social equalization, to including an equally clear goal of stimulating equal access to culture through combating economic barriers. I will analyze the success and failure to meet these goals in the results section.

3.4. Field experiment on the role of information campaigns

The third and final cultural policy initiative is not a formal initiative, but a field experiment utilizing the second wave of the panel, planned, and executed by the author. The field experiment models a potential cultural policy intervention, justifying inclusion in this analysis. Using respondents from wave 1, I randomly selected respondents to be in a treatment group, leaving the unselected in a control group. The treatment provided was an information packet directly mailed, postally, to respondents. The information packet advertised four cultural offers, scheduled to take place in November and December of 2019, made to conform to standard advertising done by the municipality by the department for culture itself. In this way, the field experiment mimics cultural policy initiatives already in place city wide. With experimental control of treatment and information on rates of cultural participation at time 0, this research design enables a causal estimation of the impact of information campaigns at time 1.

4. Results

I review results from empirical investigations of the three Cultural Policy Interventions in turn, starting with the DKS (The Cultural Rucksack), moving to the target cultural subsidy intervention of ‘Aktivitetskortet’ (The Activity Card), ending in an investigation of the field experiment on information distribution. Before entering into these detailed analyses, I review some basic structures of cultural participation for children in Bergen. Research on cultural participation for children in Bergen show that cultural participation is first and foremost linked to their parents having cultural capital (Roaldsnes forthcoming), an association that is mediated by parents’ access to high value social capital (Roaldsnes 2024). Table 1 demonstrates this relationship by presenting cross tables of cultural participation means by household income, and by highest educational attainment in the family. Income is only weakly related to cultural participation, while educational attainment is strongly related to cultural participation.

Table 1. Attended cultural events as an audience last four months, 2019. Averaged by household income and educational attainment, children 3–15 years old (N = 4754).

Household income before tax, divided by consumption units (NOK 1000s)	Arts Exposure
0–230	1,6
230–450	1,6
450–670	1,7
670–880	1,7
>880	1,8
Highest education level attained in the family	Arts Exposure
Secondary education or lower	1,1
University or college degree (less than four years)	1,4
University or college graduate (more than four years)	2

Table description. Cultural event categories include attending a concert, theater, dance performance, literature event, art exhibition, festival or visiting a museum or cultural heritage site. Attendance is reported to be none, once, two to three times and more than three, recoded for aggregation into means as 0, 1, 2 and 3, then averaged by child. Household income is quite unrelated to cultural participation, while highest educational attainment in the family shows large inequalities.

4.1. The Cultural Rucksack

Exposure to art and culture provided by the Cultural Rucksack (DKS) is highly stable from 2008–2018. In 2013–2014, the average pupil in primary school in Bergen was exposed to 1.37 cultural events through DKS. Table 2 displays the scope of the cultural intervention, as 10 years of DKS in Bergen has seen more than 300 000 pupils exposed to arts and culture, which is no small feat and attest to the magnitude of this cultural policy intervention.

Table 2. Yearly visits to an event organized by ‘the Cultural Rucksack’ by school year in Bergen.

School year	Pupils attended DKS	N Pupils in primary school	Yearly average visits
2008–2009	23117	28713	0,81
2009–2010	26691	28824	0,93
2010–2011	26049	28720	0,91
2011–2012	30055	28653	1,05
2012–2013	34535	28478	1,21
2013–2014	39104	28474	1,37
2014–2015	32552	28666	1,14
2015–2016	38101	28836	1,32
2016–2017	36691	28952	1,27
2017–2018	30099	28986	1,04
2018–2019*	14013	29048	0,48

*Only the fall semester of the school year has been accounted for in 2018–2019, explaining the low rate.

Pupils in primary school had on average. 81 visits in the school year of 2008–2009, while the highest rate of DKS-participation was in 2013–2014, with 1.32 visits per pupil.

Does this cultural policy lift all primary school pupils equally? To shed light on this question, I aggregate and compare the private consumption to organized culture, measured in the panel data, and that gained through the cultural policy (see Appendix for methodological description). The level of cultural exposure gained from DKS and the level of cultural exposure gained privately in the family is highly correlated (.74). This correlation indicates an implementation failure. Children who have high cultural exposure through their family, also go to schools in school districts that disproportionately often sign up for arts and culture events through DKS.

The Gini-coefficient, a classic measure of inequality, for children’s private consumption of demanding cultural expressions in the city is 0.74, which represents a very unequal distribution (Roaldsnes 2024), and private cultural consumption is done largely in institutions which receive sizable public funding. At the same time, the average expenditure on a DKS-visit per pupil is 231 NOK in Bergen (Hauge et al. 2017, 39). These 231 NOKs per pupil are spent more often at culturally active school districts, leading to inequalities not only in cultural experiences, but in resource allocation. Thus, children in families with high cultural affinity are the recipients of subsidized culture on two ends.

Another key point in understanding the structure of the DKS-intervention, is analyzing the age structure in DKS-exposure to arts and the age structure in the private exposure to arts. In private consumption of the arts, the peak of exposure takes place at age five in the panel data, see Figure 1. This is true at each of the educational attainment levels in the family. In DKS, exposure is evenly distributed across age groups in school, except a peak at age 11. Using the private peak at five years old as a center and observing that private cultural exposure drops steadily after age seven, we may say that the prime period within the family for adopting arts and culture as a lifestyle happens before the age of eight. Of all recorded participation in cultural events in the panel survey, 55 percent of cultural participation takes place within the age group 3–8 years old. In contrast, 85 percent of exposure to arts through DKS happens past this age group, in children aged nine years old and older.

Figure 1. Yearly exposure to arts, private and through DKS, by highest educational attainment in family and age of children.

In families that are cultural capital rich, private exposure to culture takes place in a beneficial environment, possibly bolstering the likelihood of beneficial reception of subsequent exposure through DKS in primary school.

Furthermore, the reception in the home of arts exposure in school differs. In the survey data, families were aware of DKS at very different rates, stratified by parental educational attainment. In families where the highest educational attainment was secondary and post-secondary education, only 24 percent of parents had heard of DKS before, as opposed to 37 percent in families with at least one university or college degree, and 62 percent in families with at least one university or college graduate. From this sample, I asked whether or not what the child experienced through DKS was discussed at home. In the families with lower educational attainment who were familiar with DKS, 42 percent of the families responded that they had done so, as compared to 58 and 61 percent of the families with higher education. This means that for one out of ten children in families with less education, DKS-experiences are reinforced and valued, as opposed to nearly half of children in households with university education. The family mediation of DKS is twice as high for children in families with short college or university as the highest educational attainment, and four times higher in those with longer college or university degrees (see Appendix, Table A7).

The local implementation of DKS is thus one that favors families with larger amounts of cultural capital in multiple ways. First, there is a Matthew-effect, as those who are already well exposed to arts and culture are also more exposed in the public cultural policy initiative. Second, exposure to arts in high cultural capital families takes place at an earlier time in childhood. Socialization towards culture is a part of early life experiences, which could, in turn, enable better reception of arts exposure in school. Third, DKS is perceived and discussed in high cultural capital households, reinforcing valuation. The signal that is reinforced is that arts and culture is discernable, and worthy of time and attention. Culture is rendered visible to some children, while other children are more likely blind to culture.

4.2. The activity card and cultural participation rates

I turn now to the second cultural policy intervention. Is having access to ‘Aktivitetskortet’ systematically connected to higher cultural participation and the democratization of cultural goods? Recall the parameters set out in the introduction, that a policy intervention may have been planned to fulfill a variegated set of goals. In the case of the Activity Card, its desired outcomes are both broadly social and cultural. In the data, the prevalence of having access to the card is strongly linked to reporting low family income, along with gradually more isolated cases in families with higher incomes, see Table 3.

Table 3. Household access to the activity card by income groups, families with children panel 2020.

Household Income (1000 NOK)	N	Access to card	No Activity Card	Share of households with access to Activity Card
>880	203	0	203	0
670–880	498	3	495	0,6 %
450–670	1621	13	1608	0,8 %
230–450	1177	12	1165	1,0 %
No data	473	26	447	5,5 %
0–230	298	58	240	19,5 %

Access to the Activity Card is given if the family has received emergency social assistance (Sosialhjelp), and to refugees.

In order to analyze the possible effect of the Activity Card, I fit two regression models where the dependent variables are attendance or non-attendance for going to the movies and going to plays in the theater. Using these models, I assume that we are able to observe who takes part in the activities, and that having access to the card should be systematically related to higher likelihood of participation. In order to fit realistic models, I fit models using controls that are well attested to be systematically linked to cultural consumption, such as cultural capital, both institutionalized type (educational attainment) and embodied type, a measure of social capital, household income, children’s ages and gender, and family geographic location in the city.⁸ The dependent variable in both models is dichotomous, and numeric independent variables have been centered for ease of interpretation, making the Constant readable as the likelihood of movie-going and theatre attendance.

Access to the card increases likelihood of going to the movies and equalizes the effect of having low income (see model 1, Table 4). The same statistical association is not found for theater attendance (model 2). Families with access to the card do not have a systematically higher likelihood of attending theater performances.

Table 4. Activity card and going to the movies or theater.

	Dependent variable:
	`Movies binary`	`Theatre binary`
	OLS	OLS
	(1)	(2)
Constant	0.74*** (0.03)	0.19*** (0.04)
Access to Activity Card	0.11* (0.06)	0.05 (0.07)
Child’s age	0.02*** (0.002)	−0.03*** (0.002)
Child’s gender	0.03** (0.01)	0.09*** (0.02)
Income 0–230	−0.14*** (0.03)	−0.15*** (0.04)
Income 230–450	−0.004 (0.02)	−0.04* (0.02)
Income 670–880	0.09*** (0.02)	−0.01 (0.03)
Income >880	0.003 (0.03)	0.03 (0.04)
Income, no data	−0.07 (0.05)	−0.12* (0.07)
University or college degree	0.02 (0.02)	−0.002 (0.03)
University or college graduate	−0.01 (0.02)	0.04 (0.03)
Number of cultural institutions visited past year	0.02** (0.01)	0.11*** (0.01)
Prestige value of respondent’s social network	0.01*** (0.002)	0.01** (0.002)
Northern suburbs	0.03 (0.02)	−0.04 (0.03)
Southern suburbs	0.07*** (0.02)	−0.06** (0.03)
Western suburbs	0.05** (0.02)	0.03 (0.03)
Observations	2,461	2,461
R^2	0.07	0.16
Adjusted R^2	0.06	0.16
Residual Std. Error (df = 2445)	0.36	0.45
F Statistic (df = 15; 2445)	11.64***	31.22***

*p; **p; ***p < 0.01.

Using data on treatment and control groups from the second wave of the panel, I fit a difference-in-differences regression model, where time 0 is cultural participation as measured in our first wave of the panel, and time 1 is cultural participation as measured in the second wave of the panel. The covariate treatment denotes a dichotomous variable where those who are in the control group are 0 and those who received the treatment are 1. The treatment is shown to have had an effect if time:treated has a substantial coefficient which is statistically significant, while treated and time are controlled for.

Movies and theater are the only cultural offerings in our survey in which income is related after controls for education and embodied cultural capital. This could indicate that economic barriers are a real material concern, but if so, cost-removing initiatives like the card should also impact theater attendance. The absence of an ‘Activity Card effect’ in these results challenges the validity of the cost-hypothesis for the arts.

One alternative explanation could be differences in perceived social value. Children want to go to the movies. Using data on initiative in my panel data, where the respondent was asked to indicate whose idea it was to attend the last outing in the different categories of audience attendance, in all cases but movie-going, the respondent or the spouse (if applicable) of the respondent reported they took the initiative to the outing. In the category of movie-going, children’s initiative was reported to be 37 percent (adding friends of the children would increase the rate to nearly 50 percent). Movie-going was the sole cultural offering where children’s initiatives dominated parental initiatives.

For parents, differences in social value can be perceived in the difference of the coefficients in model 1 and 2 according to the prestige value of a sample of the respondents’ social network. Higher prestige is associated with both movie going and theater attendance, but the coefficient for theater is five times greater than for movie going relative to the constant, suggesting that prestige is more important in theater attendance. The constant of movie going (.74) and theater (.19), correspond roughly to the likelihood in percent of attending the offer, given centered independent variables and the chosen reference levels. The difference in frequency also indicates the differential potential for social distinction. ‘Everyone’ goes to the movies, while theater participation is a smaller and more capital rich audience. Having larger amounts of embodied cultural capital, measured as the number of cultural institutions visited in the past year for the respondent, is much higher for theater than for movies.⁹

The evidence from the survey data indicates that the Activity Card is partially successful. An argument can be made that it fulfills some of its policy objectives, while failing to fulfill others. Going to the movies has a distinct social value in children’s and youth’s peer groups. The Activity Card likely equalizes the distance between low-income families and median-income families, fulfilling a stated policy objective, as it reduces economic barriers to participation and allows for smaller differences between children. As for the stated cultural policy objectives, there is no evidence that the objectives are being met. Access to the card is not associated with higher cultural participation. Much like how having economic capital is the most assured way of acquiring more economic capital, having cultural capital is the surest path to acquiring more cultural capital. Some of this difference is likely linked to differences in perceived value. Movie going has a clear location in children’s value schemes, while theater attendance is valuable to parents, especially those who are richer in cultural and social capital.

4.3. Information nudge field experiment

Finally, I move on to the last cultural policy initiative, a measure to distribute information about cultural offers. Recall that in order to determine the possible effects of information campaigns on children’s cultural participation, I drew a random treatment group from the panel in Wave 1 that would receive an information packet between waves 1 and 2. After collecting Wave 2, this field experiment design would enable estimation of causal effects of certain types of information initiatives using difference in difference methods.

The Difference-in-differences model (see Table 5, model 1) shows no effect of the information treatment. This measure to communicate available cultural offers did not succeed in the goal of boosting cultural participation. A more careful analysis of what goes on in the field experiment reveals a crucial mechanism, see Table 5, model 2. The Wave 2-questionnaire contained a question on whether or not respondents had noticed the information packet. ¹⁰ The question showed a large thumbnail of the information letter, which was dispatched to the treated group, asking the respondents if they recalled received the letter.¹¹ Of the treated group of respondents, 54 percent recalled seeing the letter, while 46 percent did not. There are systematic differences in the impression and the no-impression groups, suggesting an underlying correlation with children’s cultural participation. The group on which the letter made an impression had children with 58 percent more cultural participation than the group on which it made no impression.¹² Analyzing the structure of who notices the information treatment shows that it is unlikely to be coincidental. The information packet was postally mailed to each respondent in the treatment group, using City Government marked envelopes. Assuming that most of the target group respondents opened the information packet themselves, some respondents noticed the information in the information packet well enough to recall seeing it, and some did not, the distribution of which is closely related to the respondents’ stock of cultural capital.

Table 5. Difference in difference models for effects of information campaigns to randomly drawn respondents.

	Dependent variable:
	Child’s Cultural Participation
	OLS
	(1)	(2)
Constant	0.82*** (0.02)	0.82*** (0.02)
time	0.09*** (0.03)	0.08*** (0.03)
treated	−0.04 (0.04)
time:treated	0.01 (0.06)
treated, made impression		0.10** (0.05)
treated, no impression		−0.24*** (0.05)
time:treated, made impression		0.01 (0.07)
time:treated, no impression		0.03 (0.08)
Observations	8,360	8,360
R^2	0.002	0.01
Adjusted R^2	0.001	0.01
Residual Std. Error	1.15 (df = 8356)	1.15 (df = 8354)
F Statistic	4.74*** (df = 3; 8356)	12.50*** (df = 5; 8354)

*p < 0.1; **p < 0.05; ***p < 0.01.

I fit a regression model estimating various independent variables’ link to the dependent variable, i.e. whether or not the respondent recalled seeing the information packet. The Constant is .26, corresponding to 26 percent likelihood of recalling the letter given centered independent variables and the set reference levels of categorical variables (Secondary education and lowest income bracket). Details on measurement and construction on cultural investment variables can be found in the methodological appendix.

Several variables measuring cultural capital were strongly related to who noticed the letter, see Table 6. Respondents who had university or college degrees were 60 percent more likely to notice the letter. Mothers were 40 percent more likely to notice than fathers, and younger respondents were more likely to recall the letter. In addition, being culturally active as a family and individually¹³ increases the likelihood of noticing the treatment. Standardized coefficients show that the influence of each of these variables are close to equivalent, attesting to a confluence of related forces influencing the ability to notice the treatment.

Table 6. Differences in attention to culture.

	Dependent variable:
	Likelihood of intervention making an impression
	OLS
Constant	0.26*** (0.08)
Mother	0.10*** (0.03)
Parent’s age	−0.01*** (0.002)
University or college degree	0.15*** (0.05)
University or college graduate	0.18*** (0.04)
Income 230–450	−0.01 (0.06)
Income 450–670	−0.004 (0.06)
Income 670–880	−0.08 (0.07)
Income >880	0.01 (0.09)
Family joint cultural investment	0.06*** (0.02)
Respondent’s cultural investments	0.05*** (0.02)
Observations	1,145
R^2	0.09
Adjusted R^2	0.08
Residual Std. Error	0.48 (df = 1134)
F Statistic	10.90*** (df = 10; 1134)

*p < 0.1; **p < 0.05; ***p < 0.01.

As variables are centered, the constants are interpretable as a baseline likelihood of attending theater or going to the movies. Going to the movies is a popular activity that a clear majority enjoys and is clearly related to household income. The difference between median income households (400–600) and low income families (0–230), is removed by the Card, as covariates are of similar sizes. There is no effect for theater, despite theater also being somewhat connected to household income. Theater is highly feminized and likelihood of theater going decreases with advancing ages of children, while going to the movies shows the opposite development by age. Network prestige (see Appendix) is connected both to movie going and theater, but the effect is double for theater. The same is true for differences in embodied cultural capital, but here the difference is triple.

The analysis suggests that there is a classed vision of culture that enables recognition of a cultural offer, while those who lack this classed vision are more likely to be blind to cultural offers. There is indication of abject rejection or aversion. If this was the case, and lower classed parents mobilize their identities against cultural interests, along the lines of symbolic boundaries (see e.g. Lamont, Pendergrass, and Pachucki 2015) or exasperated indifference to a difficult game (such as working class distance to culture in Bourdieu 2010), parents would more likely notice the treatment.

While this case of information treatment did not work to increase cultural participation in our case, other information treatments, including ads, could possibly work. Still, likely with important unintended consequences. If interventions came to affect behavior in other cases, it is likely that the affected individuals would already be invested in arts and culture, having larger volumes of cultural capital. Given that the structure analyzed above holds in other contexts, relying on greater outputs of information regarding the cultural offer would only increase the social stratification in cultural access, not diminish it, leaving public policy planners damned if they do and damned if they don’t. Individuals whose value schemes are less likely to recognize cultural value are unlikely to be affected by government efforts.

5. Concluding discussion

Outlook on the future of both cultural policy and the specific goal of democratization has been grim (Mangset 2020). In this paper, following Jancovich and Stephenson’s argument (2021; 2023), I not only analyze policy success or failure, but attempt an analysis of what underpins their failure. To enable such an analysis, I presented the theoretical perspectives of cultural reproduction and cultural mobility, two partly competing models of what causes cultural participation.

Whether or not each of the cases of cultural policy intervention was a failure or success, or something in between, was an open question. While this analysis has pointed to both objective attainment failure and implementation failure in each case, there are obvious positives as well. The Cultural Rucksack (DKS) creates an unprecedented exposure to arts and culture for children in absolute terms. The Activity Card likely equalizes income differences for popular culture and leisure activities for children and young people.

At the same time, these policies did not bridge the social gap in the use of arts and culture. To some degree, they exacerbated social differences. The policies had a ‘Matthew effect’, i.e. ‘to those who have, more will be given’. Pupils who are already exposed to arts and culture through their family had twice the exposure to arts and culture through DKS. Having the Activity Card was not associated with higher participation in legitimate culture, and those who already were familiar with the city’s cultural offer simply got a reminder through the field experiment, while those uninvested in culture was unlikely to even recognize cultural offer ads. In short, the cultural policy initiatives did not live up to the task of reducing inequalities in children’s access to arts and culture. To the contrary, these policies arguably contributed to more inequality.

My findings suggest that policy planners are damned if they do fund cultural policy interventions to reduce inequalities, and damned if they don’t. In the former case, inequalities could easily become more entrenched by inadvertently funding the culturally invested. In the latter, authorities failing to act to reduce cultural inequalities could lead to a drop in cultural participation, recalling that less funding could be followed by reduction in cultural participation in general, deducing from the findings of van Hek and Kraaykamp (2013). The way out of this dilemma is perhaps to rethink the typical instruments at disposal for cultural policy planners.

To be sure, it is indeed possible to work within what appears to be the current paradigm of ‘barrier to participation,’ and simply tweak policies. In the case of the Activity Card, digital implementation of such a policy would allow policy planners to implement gamification elements (for an overview of gamification in education, see Majuri, Koivisto, and Hamari 2018), attempting to stimulate greater cultural participation in families with children. In the Cultural Rucksack, policy planners may take into account that children are far more receptive to culture at lower ages, and, since part of the mission of the Cultural Rucksack is to socialize children to culture, a socialization they are not guaranteed to receive from the family, this exposure and socialization should come at an age when children are most receptive. While these policies may be viable, they raise ethical questions, and moreover, they do not address the root of the problem of failure to reduce cultural inequalities.

What are the common attributes of failures in these three cases, and what kinds of lessons may be drawn from them? In all three cases, instances of ‘misrecognition’ appear to lie at the root of the practical problem of facilitating increased cultural participation for the target group. While the term ‘misrecognition,’ in the Bourdieusian tradition, usually denotes the process in which higher or lower status is imputed on the basis of some arbitrary attribute, falsely recognized as legitimate (Bourdieu 2010; Bourdieu and Passeron 1990; see; Jæger, Rasmussen, and Holm 2023 for a succinct presentation), the term can also mean blindness, or ‘subjective non-recognition (Bourdieu 2010, 559 fn46)’. Families are unlikely to participate in culture unless cultural participation is seen as valuable. This valuation (and devaluation or non-valuation) can take many forms, but a classed cultural disposition is at the core for all cases analyzed in this study.

Educated parents who are culturally invested, even more so those with privileged social networks, are more likely to value theater attendance. Cultural capital rich parents notice and reinforce children’s cultural experiences in school. Cultural capital rich parents recognize cultural offers. Sources of value apart from classed dispositions direct children otherwise. Kids want to go to the movies, and this influences family practices. Cultural practices are highly feminized (Donnat 2005; Octobre 2005), impacting boys’ valuation schemes. Children turn away from cultural practices with advancing age (Nagel 2002; Roaldsnes forthcoming), which in turn affects valuation beyond class and intersects with class.

This study has focused to some degree on the theoretical difference between a cultural reproduction and mobility model of children’s cultural participation. While this might at first glance appear to be of little consequence to practical cultural policy, I argue that such perspectives are of vital importance. This study suggests that planners should assume cultural reproduction to be the core condition for cultural policy. Taking seriously ‘blindness to culture’ restricts some of the classic policy tools and forces authorities aiming to reduce cultural inequalities to think again. Cultural policy planning has to some degree assumed that ‘non-users’ are empty vessels that would take to culture as soon as they encounter it. This study argues that ‘non-users’ might not even notice the world of cultural offers. Planners attempting to influence the role of art and culture in children and young people’s lives, regardless of background, would perhaps do well to reconsider these conditions.

Acknowledgments

I wish to thank the two anonymous reviewers at International Journal of Cultural Policy for their insightful comments. I am indebted to Jan Fredrik Hovden and Torgeir Uberg Nærland for providing critical comments at various stages of the construction of my argument. Finally, I would like to thank the City of Bergen for creating and funding the Families with Children panel, and especially Harm-Christian Tolden and William Hazell for their belief in, and support of, the project.

Disclosure statement

The author is on leave from his employer, the City of Bergen, while the Ph.d-project is ongoing.

Additional information

Funding

This work was supported by the Norges Forskningsråd [312039].

Notes on contributors

Andreas Roaldsnes

Andreas Roaldsnes is a Ph.d candidate at the department of Information Science and Media Studies at University of Bergen. His main research interests are social inequality in childhood, especially classed differences in access to arts and culture, leisure and extra-curricular activities.

Notes

1. ‘For to everyone who has, more will be given, and he will have abundance; but from him who does not have, even what he has will be taken away (Matthew 13:12)’

2. Which is equivalent to saying that if they did want it, they would want it, which Dworkin labels ‘true, but unhelpful’. The soundness of Dworkin’s tentative affirmative to his own question: can a liberal state support art, is up to the reader to validate, but as Dworkin notes, even if we agree that a liberal state can legitimately support art, his argument provides no answer as to by what scale of funding, and how to distinguish worthy recipients from unworthy recipients.

3. Bourdieu presents somewhat different images of legitimate culture in the course of his work. In Reproduction (Bourdieu and Passeron 1990), legitimate culture is very clearly associated with the symbolic goods protected, produced and disseminated by institutions and agents in who are dominant in the field of cultural production, while in Distinction (Bourdieu 2010), legitimate culture is more oriented towards symbolic goods valued at the top strata of social space. In much recent empirical research, legitimacy has taken on this last interpretation (Jæger, Rasmussen, and Holm 2023; see e.g. discussions on legitimacy in; Robette and Roueff 2014).

4. They are not mutually exclusive, but they do invert each other at key points.

5. I do not mean to claim that one is true and the other false, but correct in the Bachelardian (Bachelard 2002) sense of our least wrong way of understanding this process.

6. The veracity of the second element of cultural mobility theory, that children of disadvantaged backgrounds gain more from cultural capital than children of advantageous backgrounds is also unclear, but there is empirical evidence that support both mobility and reproduction (see e.g. Jæger and Karlson 2018; Jæger and Møllegaard 2017).

7. To perform the Multiple Correspondence Analyses, the package GDAtools was utilized (Robette 2023). The stargazer package was utilized for well formatted regression tables (Hlavac 2022).

8. Embodied cultural capital, or familiarity with socially valuable lifestyles (Bourdieu 1986), is measured as the parent’s number of visits to named cultural institutions in Bergen that receive public funding. Social capital is measured through the use of a position generator (see Son 2020 for an introduction). In the position generator, each respondent was asked to indicate whether they knew someone holding certain occupations from a list of 17 occupations, then each observed social tie was given a prestige score (Ganzeboom, De Graaf, and Treiman 1992). The prestige variable in the models is each respondent’s total prestige score derived from the observed network.

9. The variable for embodied cultural capital effectively removes any link between institutionalized cultural capital and participation for both genres, all the while institutional cultural capital is strongly connected to the distribution of embodied cultural capital (Roaldsnes forthcoming). None of these independent variables exhibit signs of multicollinearity, see Appendix. Standardized coefficients for numeric and pseudo-numeric variables show that the embodied cultural capital, along with children’s ages are the two (almost equally) important variables to theater attendance, while prestige in social capital and gender play a subsidiary role.

10. Information packets that failed delivery was returned and respondents struck from treatment groups. This applied to few cases, and is related to the fact that a minority of respondents moved between wave 1 and wave 2.

11. I am indebted to Erik Knutsen for this suggestion.

12. (Impression coefficient .10 + intercept .82/no-impression coefficient −.24 - intercept .82)

13. Family and respondent’s cultural investments are the first dimension in two Principal Components Analyses done on the general use and non-use of a variety of cultural offers and cultural institutions measured in wave 1. I display the construction of these PCA’s in the methodological appendix.

Appendix

Table A1. Descriptive statistics.

Categorical variables			N	Share
Respondent educational attainment
Primary education only			104	2,5 %
Secondary education			723	17,6 %
University or college, less than three years			772	18,8 %
University or college, four or more			1960	47,6 %
Research level (Ph.d)			222	5,4 %
Missing data			335	8,1 %
Second adult in household educational attainment
Primary education only			121	2,9 %
Secondary education			845	20,5 %
University or college, less than three years			611	14,8 %
University or college, four or more			1373	33,4 %
Research level (Ph.d)			188	4,6 %
Missing data			978	23,8 %
Household income
0–227 (EU-50 approximation)			306	7,6 %
227–400			752	18,7 %
400–600			1602	39,8 %
600–780			760	18,9 %
780–1600			496	12,3 %
Missing data			113	2,8 %
Respondent age
Female			3141	76,3 %
Numeric variables	Mean	Min	Max	SD
Respondent age	42	20	73	6,8
Children’s age	9	3	16	3,8

Table A1 provides a brief descriptive summary of the sample. Table A1 describes the second wave of the panel, which has suffered some attrition since wave 1, see Table A2. Household income is distributed in a reasonably representative manner, but educational attainment is skewed towards higher education, as 47 percent of respondents have educational attainment of four or more years of university or college. In the specific case of Bergen, however, it is important to note that the majority of respondents would have completed higher education according to national statistics on educational attainment of the population’s dominant age group (Statistics Norway 2023b). The levels of secondary education and shorter university or college completion are representative. It is difficult to recruit a large number of respondents with only primary school education (Savage 2015; Stedman et al. 2019). In the analyses, primary and secondary education have been combined and contrasted with university or college educational attainment. This makes sense in these analyses, as there is little difference in cultural orientation between these levels of educational attainment (Roaldsnes forthcoming). In the population dataset, only married fathers were linked to the pool of children from which I drew focal children, as parents only report activities for a single target child. This led to mothers becoming the majority responder, as only married men were invited to respond. There is no systematic difference in the data due to this artefact, which has been validated using regression analysis. The gender of the respondent did not influence recall, for instance, as one could hypothesise that mothers have better insight into children’s social calendars. These hypotheses were unfounded. The skew towards higher education is lower in the second adult in the household, because men to a lesser degree complete higher education in Norway. The number of missing in this category is due to the (representative) level of single parent households, which is 17 percent of the sample, contrasted with 15,8 percent nationally in 2019 (Statistics Norway 2023a).

Table A2. Overview of panel attrition.

Educational attainment	Wave 1	Wave 2	Wave 3
Secondary or lower	100	73 %	47 %
Uni/College short	100	81 %	55 %
Uni/College long	100	82 %	61 %
Mean retention		79 %	54 %

Table A2 describes panel attrition in the Family with Children Panel. Attrition follows expectations and projections made by the social research institute, Ideas2evidence, which carried out the survey on behalf of the City of Bergen. Respondents with higher educational attainment are more likely to keep responding. The difference is slightly more pronounced in wave 3 than wave 2.

Table A3. Model comparison.

	Dependent variable:
	`Theatre binary`
	OLS
	(1)	(2)	(3)
Constant	0.16*** (0.03)	0.19*** (0.04)	0.19*** (0.04)
Access to Activity Card	0.03 (0.05)	0.05 (0.07)	0.04 (0.07)
Child’s age	−0.03*** (0.002)	−0.03*** (0.002)	−0.03*** (0.002)
Child’s gender	0.08*** (0.01)	0.09*** (0.02)	0.09*** (0.02)
Income 0–230	−0.15*** (0.03)	−0.15*** (0.04)	−0.15*** (0.04)
Income 230–450	−0.03* (0.02)	−0.04* (0.02)	−0.04* (0.02)
Income 670–880	0.02 (0.02)	−0.01 (0.03)	−0.001 (0.03)
Income >880	0.07** (0.03)	0.03 (0.04)	0.04 (0.04)
Income, no data	−0.06 (0.05)	−0.12* (0.07)	−0.12* (0.07)
University or college degree	0.02 (0.02)	−0.002 (0.03)	0.001 (0.03)
University or college graduate	0.05** (0.02)	0.04 (0.03)	0.05* (0.03)
Number of cultural institutions visited past year	0.11*** (0.01)	0.11*** (0.01)	0.11*** (0.01)
Prestige value of respondent’s social network		0.01** (0.002)
Northern suburbs	−0.04* (0.02)	−0.04 (0.03)	−0.04 (0.03)
Southern suburbs	−0.05*** (0.02)	−0.06** (0.03)	−0.06** (0.03)
Western suburbs	0.03 (0.02)	0.03 (0.03)	0.03 (0.03)
Observations	3,860	2,461	2,461
R^2	0.16	0.16	0.16
Adjusted R^2	0.16	0.16	0.15
Residual Std. Error	0.44 (df = 3845)	0.45 (df = 2445)	0.45 (df = 2446)
F Statistic	53.10*** (df = 14; 3845)	31.22*** (df = 15; 2445)	33.04*** (df = 14; 2446)

Note: *p; **p; ***p< 0.01.

Table A3 displays a comparison between theatre participation models (Table 4 in section 4.2) using wave 2 and wave 3 of the panel. In order to run regressions where prestige of social networks is included in the model fit, wave 3 has been utilized, which results in dropping almost 1400 observations. Table A3 is included in the appendix in order to show that results are not sensitive to this drop in N. Findings are robust across specifications and subsets of respondents. Model 1 fits a model without prestige scores of networks, increasing the N from 2461 to 3860. There are no substantive changes in model interpretation following this. Model 2 is the model presented in section 4.2 and Model 3 is the same model, without network prestige scores, but where the model data is filtered for non-missings in the network prestige scores variable, allowing comparison that highlights the role of this independent variable. That network prestige influences theatre more than cinema, is a part of the argument of the article, which justifies the model with the smaller N, especially because overall results do not substantially change. That social networks have effect independent of income, education and cultural orientation is moreover justified in previous research (Roaldsnes 2024).

Methodological appendix

Case 1 - Comparing mean cultural consumption in the private realm and that provided by the Cultural Rucksack (DKS).

To conduct a comparison between private (in the family) and public (DKS) exposure to arts and culture for children, I aggregate two data sets to comparable unit levels, which were suburbs and school districts. In the first operation, I aggregate the two first waves of the Families with Children panel to these two unit levels, using weighted observations (weights were constructed based on educational levels), then subtracting each observation on the aggregate level (suburbs or school district administrative zone) from the citywide mean. This provides an overuse and underuse scale for each observation in the aggregate level. In the next step, I do the same operation for DKS, except that there are no weights in the DKS-data. The full population of completed artistic offers provided by DKS, along with which schools attended and which age cohorts within the school, are reported. As with the previous dataset, I construct an overuse and underuse scale against the citywide mean. In the comparison of these sets of deviances from the mean (each year of DKS-reports and each wave of the panel), the correlation is large at .74.

There are limitations to these data. There is no overlap between the observed time series of cultural exposure. DKS-data spans 2008-2018, while the panel data spans 2019-2021. In order to justify comparison, I next show that each of these datasets are stable over time and not likely to have been drastically changed, allowing this comparison (with the limitation in mind). Across the DKS-set, the Pearsons product-moment correlation in deviance from the school year mean by school is .67 (p.value of .17e-6). For private consumption patterns the Pearson product-moment correlation between private consumption in school districts across wave 1 and wave 2 is .86 (with a p.value of .29e-7). Ideally, we would be able to also add children’s cultural exposure in wave 3, but this set is not included as it was collected in COVID-19 conditions. Despite these limitations. I make an assumption that the patterns are stable enough to merit comparison.

Case III – Attention to culture

There are two independent variables in case III (Table 6) that measure the cultural orientation of the respondent. First, there is a measure of the respondent’s relative (to other respondents) use of cultural institutions in Bergen measured in wave 1, that is, before treatment. The second is the relative cultural orientation measured by the target child’s (and therefore the family’s) cultural practices in wave 1. Both of these variables are constructed by conducting an MCA. In the first case, dichotomized participation in named cultural institutions, and in the second case, dichotomized participation in cultural participation categories. An MCA conducted on only dichotomous variables is mathematically equivalent to a PCA, and in both cases I find a very strong first axis, which is then retained for regression analysis. In the following, I describe and interpret these axes.

Respondent’s cultural investment

Active variables in the MCA: ‘Akvariet’; “Bergen Internasjonale Kultursenter; ‘Bergen kino’; ‘Bergen Nasjonale Opera’; ‘Borealis’; ‘Bymuseene i Bergen – Gamle Bergen Museum’; ‘Cornerteateret’; ‘Kode – Kunstmuseer i Bergen’; ‘Kulturhuset USF’.

Table A4. Coordinates and contributions of active variables in Respondent's cultural investments variable.

Active variable	Coordinate	Contribution	Relative Contribution
Borealis_Yes	3,3	4,2	0,09
Bergen Internasjonale Kultursenter_Yes	2,3	9,9	0,21
Bergen Nasjonale Opera_Yes	1,7	2,5	0,05
Kulturhuset USF_Yes	1,7	12,1	0,27
Museum Vest - Norges Fiskerimuseum_Yes	1,5	12,2	0,28
Cornerteateret_Yes	1,4	10,5	0,24
Kode - Kunstmuseer i Bergen_Yes	1,2	15,3	0,40
Bymuseene i Bergen - Gamle Bergen Museum_Yes	1,1	11,3	0,29
Akvariet_Yes	0,3	3,1	0,15
Bergen kino_Yes	0,2	0,9	0,07
Borealis_No	0,0	0,0	0,09
Bergen Nasjonale Opera_No	0,0	0,0	0,05
Bergen Internasjonale Kultursenter_No	−0,1	0,4	0,21
Kulturhuset USF_No	−0,2	1,2	0,27
Cornerteateret_No	−0,2	1,3	0,24
Museum Vest - Norges Fiskerimuseum_No	−0,2	1,6	0,28
Bymuseene i Bergen - Gamle Bergen Museum_No	−0,3	2,8	0,29
Kode - Kunstmuseer i Bergen_No	−0,3	4,1	0,40
Akvariet_No	−0,4	4,0	0,15
Bergen kino_No	−0,5	2,7	0,07

Table A4 describes the construction of the regression variable in Table 6. The first axis explains 97 percent of the variance of importance, according to Benzecri’s modified eigenvalues (Le Roux and Rouanet 2004). The axis clearly separates the users from the non-users of the measured set of cultural institutions in Bergen. In addition to organizing respondent’s profiles from none, few, to many, there is additional information in the axis. Users of Borealis – a festival of experimental music – are the rightmost respondents of this cultural orientation axis. However, the cultural institution with the highest contribution to the axis is Kode, Bergen’s premiere art museum (which is also the institution that receives most public funding of these institutions). The relative contribution of Kode versus the Opera or Borealis suggests that these are rather eclectic. Respondents who use mostly the Aquarium or the Cinema are not non-users, but do not have the same high-brow orientation. At the left side of the axis, the no-responses have been gathered. The axis thus reflect levels of participation, but also legitimate culture orientation.

Table A5. Summary statistics, respondent's cultural investment variable.

	Mean	Min	Max	SD
Axis 1	0	−0,48	2,83	0,45
Scaled Axis 1	0	−1,05	6,24	1

Table A5 describes summary statistics of the axis described in Table A5, transformed by a scaling function for use in regression analysis. Scaling is performed for ease of interpretation.

Family cultural investments

I turn now to describing the second variable. The set of active variables are categories of outings where the target child did or did not attend in wave 1 of the panel.

Active variables in the MCA: ‘Concert’; “Theater; ‘Dance event’; ‘Literature event’; ‘Movies’; ‘Art exhibition’; ‘Museum’; ‘Public holiday event’; ‘Festival’; ‘Sports event’; ‘Arts and crafts workshop’; ‘Open gymnasium’; ‘Library, borrowed any medium’; ‘Library, event’

Table A6. Coordinates and contributions for active variables in Family cultural investment variable.

Active variables, audience category	Coordinate	Contribution	Relative Contribution
Literature event_Yes	2,3	11,0	0,28
Art exhibition_Yes	1,7	10,7	0,28
Arts and crafts workshop_Yes	1,6	6,0	0,15
Library, event_Yes	1,6	10,2	0,27
Festival_Yes	1,5	5,5	0,14
Dance event_Yes	1,1	3,5	0,09
Museum_Yes	1,1	9,7	0,30
Concert_Yes	0,9	6,6	0,20
Theater_Yes	0,7	6,2	0,22
Public holiday event_Yes	0,7	4,9	0,16
Open gymnasium_Yes	0,5	3,1	0,11
Library, borrowed any medium_Yes	0,5	4,0	0,19
Movies_Yes	0,1	0,3	0,02
Sports event_Yes	0,1	0,1	0,00
Sports event_No	0,0	0,0	0,00
Dance event_No	−0,1	0,3	0,09
Arts and crafts workshop_No	−0,1	0,4	0,15
Festival_No	−0,1	0,4	0,14
Literature event_No	−0,1	0,6	0,28
Art exhibition_No	−0,2	1,1	0,28
Library, event_No	−0,2	1,1	0,27
Movies_No	−0,2	0,5	0,02
Concert_No	−0,2	1,5	0,20
Open gymnasium_No	−0,2	1,4	0,11
Public holiday event_No	−0,2	1,9	0,16
Museum_No	−0,3	2,6	0,30
Theater_No	−0,3	2,9	0,22
Library, borrowed any medium_No	−0,4	3,7	0,19

Table A6 describes the coordinates, contributions, and relative contributions for all active variables in the MCA. The MCA is as unidimensional as the previous MCA, with Axis 1 explaining 96 percent of the variance of importance. The axis sorts non-participation from participation, as all no-responses are located at the left side (negative) of the axis, and all participation is located at the right (positive) side. As with Table A4, there is much more information than simply use and non-use. The highest contributing active variable is attending literature events, and together with art exhibitions, these are the variables that claim the most positive positions. Art and literature are the more intellectually demanding audience categories in the set. Going to the movies, or going to sports events have no influence on the axis, just as with the institutional MCA in Table A4. Non-attendance in museums, theatre and not borrowing any medium at the library marks the furthest distance from cultural orientations at the negative side of the axis.

Table A7. Summary statistics, family cultural investments variable.

	Mean	Min	Max	SD
Axis 1	0	−0,46	2,42	0,41
Scaled Axis 1	0	−1,1	5,83	1

Table A7 describes summary statistics of the axis described in Table A6, transformed by a scaling function for use in regression analysis. Scaling is performed for ease of interpretation.

Variance inflation tests

Using independent variables that are correlated risks inflating estimates. In the model displayed in Table 6, I deploy two different measures of cultural orientation, the family’s and the respondent’s. The reported model in Table 6 suggests that each variable explains some of the likelihood of noticing the intervention. It is quite possible for a respondent to not use cultural institutions themselves, and still be a part of a culturally oriented household. Variance inflation tests, run with the car package in R, shows all variables to have less than 1.3 test scores, indicating no problems of variance inflation due to multicollinearity.

Table A8. Differential awareness of DKS by household education.

Highest education level attained in the family	Aware of DKS	Discussed DKS	Rate of family mediation
Secondary education or lower	24 %	42 %	10 %
University or college degree (less than four years)	37 %	58 %	21 %
University or college graduate (more than four years)	62 %	61 %	39 %

Table A8 calculates the differential rates of cultural mediation of DKS in the family. To find the rate of family mediation, I multiply the percent cognizant of DKS with the percent who discussed it at home (only respondents cognizant of DKS were asked if DKS was discussed in the home), which for the lowest education attainment level is (.24 × .42) = .10. The rates of family mediation in Table A8 have a near perfect correlation (.98) with private arts exposure, see Table 6.