The lives of others: Class divisions, network segregation, and attachment to society in Chile.

In this paper, we examine whether social class and class divides in social networks contribute to individuals' attachment to society. We argue that network segregation restricts individuals' social worlds, thereby diminishing societal attachment. Our research site is Chile, a country with relatively low social cohesion and one of the world's highest levels of economic inequality. We use large-scale representative survey data collected in 2016 for the Chilean urban population aged 18-75 years (n = 2983) and interrelate indicators of well-established dimensions and sub-dimensions of societal attachment. Results of our regression analyses show that members of the upper middle class are more attached to society than their fellow citizens from other social strata. In addition, having more social contacts within one's own social class reduces attachment to society. In particular, network homogeneity lessens societal attachment for lower- and upper-class individuals, but not so strongly in the middle class. We conclude that social cohesion in Chilean society would be enhanced not only by more equal opportunities but also by changes to the social settings in which social class segregation is (re)produced.

INTRODUCTION

Some scholars have heralded the demise of class divisions (e.g., Clark & Lipset, 1991). However, the rise of inequality in almost every contemporary society (Piketty, 2014) has made it clear that social classes and their divides are still relevant and have powerful consequences for individuals (Savage, 2015). Class differences shape, for instance, educational opportunities, residential conditions, and well‐being (e.g., Lareau & Conley, 2008). Another aspect in which class still appears to be relevant in people's lives is the phenomenon of attachment to society.

Seminal works and contemporary research concerned with the relationship between social class and attachment to society have found that upper‐class individuals are often more strongly attached to society than their fellow citizens from the middle and lower classes, especially showing higher levels of social and institutional trust and greater civic engagement and political participation (e.g., Alesina & La Ferrara, 2002; Li et al., 2003; Meier & Bell, 1959; Srole, 1956; Verba, 2003). Although these have been valuable contributions, we note that there is a prevailing lack of work that comprehensively addresses class‐based differences in attachment to society, including a coherent conceptualization and examination of its core components simultaneously. In order to advance on this issue, the present study offers a thorough assessment of the relationship between social class and attachment to society, proposing a definition and key dimensions and sub‐dimensions relating to societal attachment.

Another outstanding—and probably more critical—issue in the current literature on the association between social class and attachment to society is the potential relevance of inter‐class social relationships. Studies on this subject have focused exclusively on class positions, while leaving aside the evaluation of whether the class composition of social networks influences societal attachment. This might be of particular interest because people have a strong tendency to interact and relate with same‐ or similar‐class others and, consequently, the social worlds of different classes do not regularly intersect (Bian et al., 2005; Lindh et al., 2021; Wright & Cho, 1992). We argue that the resulting network segregation may produce a lack of familiarity with and knowledge of the lives of other classes and diminishes the sense of belonging to wider society. Therefore, in this paper, we empirically test the hypothesis that class‐based network segregation provides an additional explanation for attachment to society beyond the individual's own social class.

Although the degree of network segregation may offer insights into variations in attachment to society, its consequences probably vary across classes. For lower‐class people, class‐based network segregation may foster feelings of being overlooked by members of higher classes and cause detachment from society. For individuals from higher classes, network segregation may not only prevent awareness of the problems and struggles of other classes but also nurture a sense of togetherness and the feeling that society generally works well. Consequently, network segregation may affect their attachment to society less negatively, or even positively. We thus additionally focus on assessing the potential interaction between network segregation and class location in shaping attachment to society.

Against this background, in this paper we examine how social class and class segregation in social networks are associated with attachment to society expressed by individuals. We also explore the novel argument that the association between network segregation and attachment to society might not hold true across all social classes. We engage in theory development and reconstruction by bringing together sociological arguments on the class divide, such as the theory of alienation and those concerning class homogeneity in social networks and closure processes. To test our ideas, we focus on Chile—an upper‐middle‐income country in Latin America. Chile is a remarkably interesting research site that combines relatively low levels of social cohesion with one of the highest levels of economic inequality in the world. This stark and long‐standing inequality has generated profound class divisions within Chilean society. We assume that in such an extremely unequal context, class divisions and their consequences will be particularly visible, and this could help scholars to understand the potential consequences of inequality in an increasingly unequal world.

Our research questions are as follows:

What is the association between social class and attachment to society?

To what extent is class‐based network segregation associated with attachment to society, above and beyond social class?

Does the association between class‐based network segregation and attachment to society vary with social class?

To answer these questions, we use large‐scale representative survey data collected in 2016 for the Chilean urban population aged 18–75 years (n = 2983). We begin by discussing the theoretical arguments concerning the links between social class, network segregation, and attachment to society in order to propose three hypotheses. We then describe the data and variables used in the study. Finally, we present our analysis, interpret our findings, and draw conclusions.

THEORETICAL BACKGROUND

Attachment to society

Attachment to society can be regarded as the individual‐level expression of social cohesion, a societal or collective‐level characteristic. The importance of social cohesion rests on its potential to contribute to the harmonious development of society, the functioning of democracy, and the well‐being of society's members (e.g., Berger‐Schmitt, 2002; Putnam, 2000). Hence, it is not surprising that the extent to which individuals are attached to society is a central issue of many sociological debates (see Friedkin, 2004; Green & Janmaat, 2011; Janmaat, 2011).

Building upon different schools of thought, scholars have made substantial efforts to identify the most relevant ingredients for social cohesion, including attitudes such as sense of belonging, social and institutional trust, altruistic dispositions, respect for the law, and behaviors such as civic membership, social support, and political participation (e.g., Bollen & Hoyle, 1990; Chan et al., 2006; Kearns & Forrest, 2000; Lockwood, 1999; Schmeets & Riele, 2014; Van der Meer & Tolsma, 2014). All these attempts to define and measure the concept have stressed its multidimensionality (see also Schiefer & Van der Noll, 2017).

Bearing this in mind, we regard attachment to society as a multidimensional state that indicates how well individuals relate to society. We highlight four major dimensions of societal attachment that comprise the core features with which the different approaches align: cultural, relational, political, and normative. The cultural dimension represents the emotional and ideational attachment of the individual to society, and the relational dimension the social attraction that drives people to remain in society. The political and normative dimensions reflect responsibility for the common good and the legitimacy of the public with regard to regulation and monitoring of the social order, respectively (e.g., Friedkin, 2004; Janmaat, 2011; Lockwood, 1999). These dimensions can be further divided into more specific sub‐dimensions. For instance, the cultural dimension often includes characteristics such as sense of belonging and identification (e.g., Bollen & Hoyle, 1990; Kearns & Forrest, 2000), whereas the relational dimension comprises socialities (e.g., the number of friends) and social and institutional trust (e.g., Berger‐Schmitt, 2002; Chan et al., 2006). The political dimension typically involves components such as civic engagement, political participation, and altruism (e.g., Schmeets & Riele, 2014; Van der Meer & Tolsma, 2014), while the normative dimension encompasses respect for social rules and values (e.g., Schiefer & Van der Noll, 2017).

In short, we argue that variation within these different dimensions reflects variation in the level of attachment to society. Specifically, attachment to society varies from low to high levels when an individual reports (1) a sense of belonging to and identification with society, (2) supportive social relationships and trust in societal institutions and fellow citizens, (3) participation in public/political organizations, and (4) adherence to social rules and norms. A more detailed discussion on the operationalization of attachment to society is presented in the methods section.

Social class and attachment to society

Social classes are sets of structural positions that comprise people with specific components of their objective living conditions in common, especially their positions within labor markets and production units (Goldthorpe, 2010; Wright, 1997).

The higher one's social class, the more resources one has and the greater one's access to high‐quality education and healthcare (Lareau & Conley, 2008). This implies that upper‐class people can more easily participate in all kinds of social and cultural activities, like going to a bar, theater, or sporting event, enabling them to integrate into wider society (Savage, 2015). In addition, the upper classes typically receive greater respect, have a better reputation, and are more politically visible in society (Verba et al., 1995). Therefore, upper‐class individuals usually understand the institutional “rules of the game” and manage the “appropriate” skills and cultural repertories to modify interactions to their own benefit (Lareau, 2011). This provides them with the agency to exercise control over events that affect their lives and ultimately translates into a strong sense of entitlement—a sense of empowerment through which people assume their right to pursue their own preferences (see also e.g., Sherman, 2017). In this sense, being part of the upper layers of the social hierarchy may be positively related to attachment to society in that, through the availability of socio‐economic resources, individuals achieve a sense of being a meaningful and effective part of society, a full citizen of the nation, and entitled and duty‐bound to participate in politics and to exercise a citizen's rights (Bourdieu, 1984).

Lower‐class individuals in turn not only have fewer resources but also often reside in disadvantaged or even stigmatized areas, suffer punitive responses from the criminal justice system and have poorer physical and mental health (Coburn, 2004; Wacquant, 2008). As such, they have fewer opportunities to participate fully in all aspects of society and more reasons to believe that society works against them (Bourdieu, 1999; Verba, 2003). These “misfortunes” often come with limitations to compliance with the goals and ethics of the dominant classes and institutions that make up society and can potentially result in alienation. This not only encompasses the dissociation of the individual from the “collective conscience” in terms of feelings of self‐to‐other distance—as predicted by anomie and strain theorists (e.g., McClosky & Schaar, 1965; Srole, 1956)—but also feelings of powerlessness, meaninglessness, social isolation, and self‐estrangement (Seeman, 1959, 1983). All of these are mechanisms that possibly contribute to reduced feelings of attachment to society or, alternatively, to different forms of societal engagement that are separate from middle and upper‐class individuals (see also e.g., Fischer, 1973; Neal & Groat, 1974). The excluded individual, for instance, may feel powerless because they perceive that their views or problems are not regarded as valid by state institutions, that they are not represented in the political system, and that their engagement in public affairs would be pointless.

Taken together, these arguments lead to our first hypothesis: The higher one's social class, the more attached an individual is to society (H1).

Do social networks affect attachment to society?

The prospects for individuals' sense of attachment to society, as well as the ways in which they acquire such societal dispositions according to class, are not conceived in a social vacuum but depend on social relationships. Cross‐class social interactions may be particularly relevant as they create bridges between classes and allow individuals to access the worlds of others (Blau, 1977). There are many benefits of interclass relationships, including access to information for finding a (good) job and to additional opportunities and support that are unavailable within one's own social circle (Granovetter, 1973). Social network heterogeneity also provides exposure to a more eclectic range of class experiences, including values and cultural practices (Erickson, 1996). As such, people can discover alternative perspectives and reflect more carefully upon their own knowledge, gaining an improved understanding of society as a whole (Pettigrew & Tropp, 2006). Such a scenario may be adequate for the generation of societal attachment through increased feelings of national belonging, generalized trust, cooperation, and favorable dispositions towards the common good, such as civic action and support for income redistribution (e.g., Chua et al., 2020; Glanville et al., 2013; Newman, 2014).

Although cross‐class relationships and heterogeneous networks might encourage attachment to society, people's sociability patterns are generally biased in the opposite direction, that is, towards interactions with others who are similar (McPherson et al., 2001). The so‐called “principle of homophily” is a well‐established pattern in contemporary societies and implies that people end up with homogeneous networks according to class (Bian et al., 2005; Lindh et al., 2021; Otero et al., 2021; Wright & Cho, 1992). An important reason for homogeneity in networks is unequal meeting opportunities (see Mollenhorst et al., 2008). The upper, middle, and lower classes usually conduct large portions of their lives in segregated social contexts such as residential neighborhoods and educational institutions (Bottero, 2005). As a result, individuals are involved in situations that facilitate connections with others of similar class positions. These uneven experiences often imply that people from different class backgrounds are socialized in distinctive manners and morals (Lamont, 1992, 2000), cultural appreciations and worldviews (Savage, 2015), and, consequently, their lifestyles take different paths. The singular “class habitus” (Bourdieu, 1984) operates as a principle that favors mutual recognition and brings together people who have accumulated similar resources. Social distance may therefore also be driven by preferences, attraction, and habituation.

A core insight provided by the literature is that network segregation, whatever the ongoing guiding force, represents the breakdown of mutual knowledge. This means limited exposure for individuals to information from distant parts of the social structure (e.g., McPherson & Smith, 2019). The result is that the lives of others become more distant, increasingly faded from view, and unfamiliar, which potentially leads to a perception of unrelated fellow citizens as strangers (Mills, 1959). Network segregation is thus likely to hinder the coordination of activities, the development of common views, social trust, civic memberships, and a host of other joint societal dispositions and behaviors that are crucial for individuals to feel meaningfully united with society, especially because it implies that individuals have less to do with members of society of other types (Rothstein & Uslaner, 2005; Uslaner, 2012).

We suggest that there may be several other related and non‐mutually exclusive mechanisms that explain this process. One of these is the incorporation of narrow identities. When people are socially segregated from one another, they are more likely to develop a strong in‐group loyalty that works against trusting others who are different from themselves (Uslaner, 2010). Shaped by class solidarities, individuals may restrict their activities, commitments, and prosocial actions to their own social group and, as a result, probably lack the sense of connection with “strangers” that is essential to fostering societal attachment (Uslaner, 2002).

Another potential explanation is that of class prejudices. This mechanism alludes to cognitive and cultural processes whereby individuals evaluate, classify, or categorize different others through shared scripts, narratives, and repertoires (Lamont et al., 2014; Massey, 2007). People who are mostly exposed to same‐ or similar‐class contacts are more likely to deploy negative class stereotypes and feelings of hostility towards others (e.g., Edlund & Lindh, 2015). In this case, class‐based network segregation may undermine the motivational and moral basis for reaching out to those “others” in different class positions (e.g., Portes, 1998), reducing not only empathy but also sympathy—emotions that are crucial to producing civic engagement and cooperation (see also e.g., Sennett, 2012).

All these arguments lead to our second hypothesis: Higher class‐based network homogeneity is associated with lower attachment to society (H2).

The interplay between social class and network homogeneity

Although class‐based network segregation may reduce attachment to society overall, there are several reasons to suggest that this effect differs between classes.

We argue that segregation may be especially detrimental to the attachment to society of the lower classes. In general, network segregation makes lower‐class individuals less likely to have access to the resources that would help them overcome deprived living conditions (e.g., Espinoza, 1999). Here, distressing individual experiences may also be reinforced by interaction with multiple similar others who share their “misfortune”, thereby magnifying a persistent collective sense of social relegation and intensifying a prevailing atmosphere of sadness and loss among the less privileged (Bourdieu, 1999). Network segregation might additionally provide lower‐class individuals with further stereotypical information with which to construct ideas of conflicting power relations, leading to perceptions of a “class‐divided” society (e.g., Mckenzie, 2017; Wacquant, 2008). The overall system of isolated sociability may therefore cause the socially segregated to become inward‐looking, strengthen their awareness of forming a deprived group, and accentuate a sense of cohesion separate from the rest (Lockwood, 1966). In the end, feelings of distrust and political apathy among individuals from already disadvantaged lower‐class segments may be exacerbated by the segregation to which they are additionally subjected.

Network segregation may also reduce attachment to society among the upper classes. Closure and reproduction theories often underline the ability of upper‐class individuals to shape and control selective social contexts (e.g., gated communities and private schools) and thus exclude others from relationships that can transmit resources and opportunities (Bourdieu, 1984; Parkin, 1979). When upper‐class individuals “deliberately” distance themselves from others, they are more likely to develop a form of “selective belonging” in which stereotypes, prejudices, and discriminatory tendencies that work to legitimize and consolidate inequalities can prevail (e.g., Lamont, 1992; Skeggs, 2004). This may also increase self‐interest among the upper classes by means of processes of social influence, thereby reducing their sense of responsibility for the plight of invisible others (e.g., Lindh et al., 2021). While we do not deny these factors, network segregation could also provide the upper(‐middle) classes with a comfortable bubble and optimistic perspectives of overall social conditions (e.g., Thal, 2017) which in turn creates a feeling of belonging and attachment. We therefore expect network segregation to be less detrimental to their attachment to society than to that of the lower classes, and, in particular, to reduce their dispositions towards the common good.

Unlike the upper and lower classes, the middle classes are less likely to experience social closure. These groups are more flexible in terms of occupation and their frontiers are often more blurred. In this sense, the middle classes tend to represent social positions that are inescapably connected with the upper and lower classes (Wright, 1980). They have more opportunities to meet people from both the low and high ends of the spectrum by virtue of their location and, as such, often have porous networks and form a bridge between the extremes (e.g., Otero et al., 2021). This also means that, for the middle classes, there is no antagonistic class, and that both upper‐ and lower‐class traits can be relevant signals. Thus, even if certain intermediate class segments are relatively segregated in terms of their social networks, it will probably be difficult for these groups to significantly isolate themselves from the lives of others and reduce their attachment to society as a result.

In short, we expect the (negative) effect of class‐based network segregation on societal attachment to be greater for the higher classes and especially for the lower classes compared to the middle class. Thus, rather than a uniform effect of network segregation on people's attachment to society, we expect an interaction effect between class‐based network segregation and social class. In such a case, the association between network segregation and attachment to society would be moderated by class. This brings us to our final hypothesis: Class‐based network homogeneity reduces attachment to society more strongly for both upper‐ and lower‐class individuals than for the middle classes, and reduction is expected to be stronger for members of the lower classes (H3).

THE CHILEAN CONTEXT

Chile ranks among the most unequal countries around the globe, with a GINI coefficient of 0.46 (OECD, 2021). The persistence of inequality has produced high barriers to intergenerational social mobility (e.g., Núñez & Miranda, 2010), thereby producing stark class divisions. The class structure of Chilean society has been characterized by a high degree of “closure” at the top (Torche, 2005). The upper classes and established middle classes comprise about 25% of the country's population, while the rest of the intermediate classes account for around 25%–30% (Espinoza et al., 2013; Espinoza & Núñez, 2014). Although qualified workers could be classified within the middle‐class population, scholars have tended to include this group within the lower class (e.g., Barozet et al., 2021). This implies that the working class and most marginalized groups represent a large proportion of the population, ranging between 45% and 50%.

Such social divisions have been accompanied by stark segregation in the provision and quality of social protection according to social class. The rise of the Chilean neoliberal economic model, established in the 1970s during the Pinochet dictatorship, institutionalized and encouraged the privatization of basic social services provision. In time, this has allowed the upper (middle) classes to access fully private, privileged, and high‐quality healthcare and education systems, which currently serve around 10%–15% of the population and operate in parallel to the generally low‐quality public coverage (e.g., Gayo et al., 2019). By contrast, intermediate groups face significant barriers which make them highly vulnerable and impede their progress, particularly in the absence of a social welfare system to absorb unexpected shocks to people's wealth, such as health problems or job loss (Torche & Lopez‐Calva, 2013). The lower classes struggle to access adequate social security and remain trapped in highly marginalized and often stigmatized neighbourhoods, where the quality of public services is of disproportionately poorer quality than in other residential areas (Garretón, 2017).

With Chilean society fragmented in terms of social class, it is important to understand the extent to which citizens feel attached to society. In recent years, certain essential components of social cohesion have been weakened or remained relatively fragile (see MDSF, 2020 for an overview). For instance, people reported having fewer friends and less support available to deal with a given problem, and generalized trust has remained low—only about 19% of Chileans think that “most people can be trusted”. In addition, institutional trust has considerably decreased: the percentage of people reporting “strong” or “moderate” trust in key institutions (government, political parties, congress and the judiciary) is only approximately 5%. Since the beginning of the transition to democracy in 1990, a sustained decline in turnout for national elections has also been observed, falling from 84.2% in 1989 to 46.7% in 2017. This figure is one of the lowest in the OECD, where average participation in elections is approximately 66%. This decrease is accompanied by low levels of participation in social organizations (36%) and limited interest in politics. In fact, only 25.3% of the population declare themselves to be “very” or “fairly” interested in politics, which is the lowest level in Latin America.

Despite undeniable evidence of the current fragility of social cohesion in Chile and low participation in general, social mobilization has recently increased. Since October 2019, economic inequality has been the subject of large‐scale demonstrations in Chile, with citizens calling for a new constitution and structural reform to the healthcare, education and pension systems (see Somma et al., 2021). One of the most highlighted issues in the current public debate is the disconnection and lack of awareness of the upper classes regarding economic inequality and the living conditions of others. In this context, the study of class divisions, network segregation, and their association with individuals' attachment to society is even more pertinent.

DATA AND MEASUREMENTS

Data

We use representative survey data for the Chilean urban population aged 18–75 years, collected in 2016 during the first wave of the Longitudinal Social Study of Chile (Estudio Longitudinal Social de Chile, ELSOC). The population of Chile is mostly urban (87.8%); thus, the survey is representative of a large proportion of the country's adults. The sample frame was proportionally stratified by population strata (six population scales), and 2983 respondents participated in face‐to‐face interviews following random selection of households within 1067 city blocks. The overall response rate was 62%. The ELSOC survey was designed by the Center for Social Conflict and Cohesion Studies (COES) and measures issues relating to social cohesion and conflict in Chile, including social relations, institutional trust, and political participation (see ELSOC, 2018).

To measure attachment to society, we use 32 variables associated with the four major dimensions highlighted in the theory section. The majority of these measures are often used in research on social cohesion and included in important cross‐national surveys such as the ISSP and WVS surveys (see Chan et al., 2006; Schiefer & Van der Noll, 2017). Table 1 offers a summary of our operationalization of attachment to society, which includes these four dimensions, their sub‐dimensions and their proposed indicators.

TABLE 1

Dimensions and indicators of individuals' attachment to society

Dimensions	Sub‐dimensions	Indicators
Cultural	Sense of belonging	The extent to which individuals consider themselves to be part of society
	Sense of identification	The extent to which individuals feel identification with society
Relational	Social relations	Number of friends
	Social trust	General trust in fellow citizens (including social minorities)
	Trust in major institutions	Trust in political institutions and the police
Political	Political engagement	Interest in public affairs; satisfaction with democracy
	Political participation	Voter turnout, participation in political parties, labor unions and neighborhood organizations. Signing petitions, attending demonstrations
	Pro‐social behavior	Voluntarism and donations
	Altruistic dispositions	Attitudes towards fairness
Normative	Adherence to democratic norms	Support for democracy
	Respect and compliance with the social rules	Justification of intergroup violence and crime

The cultural dimension is represented through two main sub‐dimensions: sense of belonging and identification. We use two items: whether individuals (i) “feel proud to be Chilean” and (ii) “identify with Chile”.

The relational dimension comprises three main sub‐dimensions: social relations, social trust, and institutional trust. Social relations are measured by the number of friends a person has. To measure social trust, we use the traditional item of whether “most people can be trusted” (i.e., generalized trust), along with trust exhibited towards three social minorities: (i) Mapuche people, (ii) Peruvian immigrants, and (iii) homosexuals. To measure institutional trust, we use six items: (i) the government, (ii) parliament, (iii) the courts, (iv) political parties, (v) labor unions, and (vi) the police.

To represent the political dimension, we highlight four sub‐dimensions: political engagement, political participation, pro‐social behavior, and altruistic dispositions. Political engagement is measured with two items: whether individuals (i) are interested in political affairs and (ii) are satisfied with democracy. To measure political participation, we use eight items: participation in (i) elections, (ii) political parties or social movements, (iii) labor unions, and (iv) neighborhood associations; and whether individuals (v) discuss politics with family or friends, (vi) have signed a letter supporting a cause, (vii) have attended peaceful demonstrations, and (viii) have attended a public meeting aimed at changing the Constitution. To cover pro‐social behavior, we use two items: whether people have (i) been involved in volunteering and (ii) made donations. Altruistic dispositions are explicitly addressed with two items: whether (i) “we should work to give all groups an equal chance to succeed” and (ii) “we should do what we can to equalize conditions for different groups”.

Finally, the normative dimension involves two sub‐dimensions: adherence to democratic norms and respect and compliance with social rules. For the former, we use an item on whether individuals express support for democracy. For the latter, we use three items measuring the justification of intergroup violence and one item on whether individuals have ever threatened or hit people that they know.

Table A1 in the Appendix shows details and descriptive statistics for each of the 32 measures used in the research. All of them are categorical variables that will be interrelated later in the paper using Multiple Correspondence Analysis (MCA). This analysis will show that most of the categories that theoretically represent strong attachment to society are empirically interrelated. This is important, because it means that commonly proposed features of social cohesion refer to an actual sociological reality rather than to a merely hypothetical phenomenon (Janmaat, 2011).

TABLE A1

Descriptive statistics for categorical variables representing attachment to society

Dimensions and variables	%	MCA category labels	Category included in the count variable
Cultural
Sense of identification
I feel proud to be Chilean
Disagree	9.9	belong‐prouds(−)
Agree	55.6	belong‐proud(=)
Strongly agree	34.5	belong‐proud(+)
I identify with Chile
Disagree	9.7	belong‐ident(−)
Agree	58.7	belong‐ident(=)
Strongly agree	31.6	belong‐ident(+)
Relational
Social relations
Number of friends
None	15.1	Friends(−−)
Between 1 and 2	27.4	Friends(−)
Between 3 and 5	33.9	Friends(=)
Between 6 and 10	15.4	Friends(+)	Yes
11 or more	8.2	Friends(++)	Yes
Social trust
Generalized trust
Most people can be trusted	10.3	soc‐trust(+)	Yes
Need to be very careful	8.3	soc‐trust(−)
Depends	81.4	soc‐trust(=)	Yes
Trust towards Mapuche people
No	58.8	tr‐map(−)
Yes	41.2	tr‐map(+)	Yes
Trust towards Peruvian immigrants
No	76.5	tr‐peru(−)
Yes	23.5	tr‐peru(+)	Yes
Trust towards homosexuals
No	68.3	tr‐hom(−)
Yes	31.7	tr‐hom(+)	Yes
Trust in major institutions
Trust in the government
Do not trust at all	46.2	intr‐gob(−)
Trust somewhat	48.0	intr‐gob(=)
Trust very much	5.8	intr‐gob(+)	Yes
Trust in parliament
Do not trust at all	54.6	intr‐par(−)
Trust somewhat	41.7	intr‐par(=)	Yes
Trust very much	3.7	intr‐par(+)	Yes
Trust in the courts
Do not trust at all	44.4	intr‐jud(−)
Trust somewhat	47.8	intr‐jud(=)	Yes
Trust very much	7.8	intr‐jud(+)	Yes
Trust in political parties
Do not trust at all	69.5	intr‐pp(−)
Trust somewhat/very much	30.5	intr‐pp(+)	Yes
Trust in labor unions
Do not trust at all	35.1	intr‐uni(−)
Trust somewhat	48.2	intr‐uni(=)
Trust very much	16.7	intr‐uni(+)	Yes
Trust in the police
Do not trust at all	13.0	intr‐pol(−)
Trust somewhat	46.0	intr‐pol(=)
Trust very much	41.0	intr‐pol(+)	Yes
Political
Political engagement
Are you interested in political affairs?
Not very interested	56.8	int‐pol(−)
Somewhat interested	31.5	int‐pol(=)
Very interested	11.7	int‐pol(+)	Yes
Satisfaction with democracy
Completely dissatisfied	43.1	sat‐dem(‐‐)
Rather dissatisfied	25.1	sat‐dem(−)
Rather satisfied	22.0	sat‐dem(+)
Completely satisfied	9.9	sat‐dem(++)	Yes
Political participation
Conventional political participation
Did you vote in the last presidential elections?
No	34.1	Vote(−)
Yes	65.9	Vote(+)	Yes
Participation in political parties or social movements
No	78.9	MovParty(−)
Yes	21.1	MovParty(+)	Yes
Participation in labor unions
No	83.5	PartUnion(−)
Yes	16.5	PartUnion(+)	Yes
Membership of neighborhood associations
No	72.6	PartNeigh(−)
Yes	27.4	PartNeigh(+)	Yes
Unconventional political participation
How often do you discuss politics with family or friends?
Never	53.2	PoltTalk(−)
Rarely/sometimes	28.3	PoltTalk(=)
Very often/always	18.5	PoltTalk(+)	Yes
Over the past 12 months, did you sign a letter supporting a cause?
No	75.1	PoltLett(−)
Yes	24.9	PoltLett(+)	Yes
Did you attend peaceful demonstrations?
No	88.4	PoltMarch(−)
Yes	11.6	PoltMarch(+)	Yes
Did you attend a public meeting aimed at changing the constitution?
No	89.4	PoltConst(−)
Yes	10.6	PoltConst(+)	Yes
Pro‐social behavior
Over the past 12 months, did you do any voluntary work?
No	86.2	ProVolunt(−)
Yes	13.8	ProVolunt(+)	Yes
Over the past 12 months, did you donate money to a social project or charity?
No	54.9	ProDonat(−)
Yes	45.1	ProDonat(+)	Yes
Altruistic dispositions
We should work to give all groups an equal chance to succeed
Disagree	9.3	same‐opp(−)
Neither agree nor disagree	74.6	same‐opp(=)
Agree	16.1	same‐opp(+)	Yes
We should do what we can to equalize conditions for different groups
Disagree	8.5	equal‐cond(−)
Neither agree nor disagree	71.6	equal‐cond(=)
Agree	19.9	equal‐cond(+)	Yes
Normative
Support for democracy
Democracy is always preferable	48.2	demo(+)	Yes
Authoritarian rule can be preferable	12.3	demo(−)
It does not matter	39.5	demo(=)
Justification of violence
Pursuing and beating a person that has just committed a mugging
Never justified	24.1	jviol‐beat(−)	Yes
Sometimes justified	75.9	jviol‐beat(+)
Tying to a pole and undressing a person that has just committed a mugging
Never justified	49.5	jviol‐tiepole(−)	Yes
Sometimes justified	50.5	jviol‐tiepole(+)
That police use force to repress a peaceful demonstration
Never justified	65.5	jviol‐repress(−)	Yes
Sometimes justified	34.5	jviol‐repress(+)
Aggressive and violent behavior
I have threatened or hit people that I know
No	71.8	Violence(−)	Yes
Yes	28.2	Violence(+)

Note: ELSOC Survey, n = 2770. Weighted figures.

On this basis, we decided to create an integrated measure of attachment to society for use in the subsequent regression analyses as the main dependent variable. This measure counts the most relevant categories identified in the MCA that define a strong attachment to society. We used clustering analyses to concretely formalize the interrelated categories that characterize the pattern of strong societal attachment. Here, we noted that categories involving “strong” feelings of national identification or national pride are not associated with the common core of relationships representing attachment to society. Therefore, these were not included in the count variable. For details of the categories counted in the construction of our integrated measure of attachment to society, see the last column in Table A1 in the Appendix.

The typology of class structure

In order to measure social class, we constructed an indicator that combines three main variables: educational level, occupational class, and household income. Education is measured using five levels of attainment ranging from no formal education to university education. Occupational class is measured using the 7‐class version of the scheme devised by the UK National Statistics Socio‐Economic Classification (NS‐SEC), developed from the Goldthorpe class scheme (Goldthorpe, 2010). We also incorporated the unemployed and the retired. Household income was converted into five categories to facilitate their analysis alongside the categorical variables.

We applied the Hierarchical Clustering on Principal Components technique (HCPC), which allows the identification of classes that are homogeneous with respect to each other but heterogeneous with respect to others. This approach combines three different methods: principal component analysis and two clustering algorithms (see Husson et al., 2011). It is important to note that this method has previously been used to define classes but also to identify spatial and social mobilities (e.g., Méndez & Gayo, 2019; Otero et al., 2021). Here, we have identified six social classes that broadly constitute the stratification system in Chile. Table 2 summarizes these classes, while Table A2 in the Appendix provides more detailed information regarding the classification. As can be seen, there are noticeable differences among the social classes in terms of both resources and group size.

TABLE 2

Summary of social classes

	Percent	Description
Upper middle class	9.3	Very high economic capital. Higher/lower managerial, administrative, and professional occupations (e.g., engineering professionals, psychologists). University education.
Established middle class	14.5	High economic capital, and mostly university education. Lower managerial, administrative, and professional occupations (e.g., business and related associate professionals).
Emerging middle class	19.5	Moderate economic capital. Vocational or secondary education. Intermediate occupations (e.g., sales administrators); lower managerial, administrative, and professional occupations; or lower supervisory/technical occupations.
Working class	29.8	Low economic capital. Primary or secondary education. Routine or semi‐routine occupations (e.g., cleaners, domestic staff, carpenters, joiners).
Precariat	20.1	Moderately poor economic capital. Very low educational level. Small employers and self‐employed workers, unemployed, or retired.
Poor	6.8	Poor economic capital. No formal education. Retired.

Note: ELSOC Survey, n = 2983. Weighted figures.

TABLE A2

Educational level, occupational class and income by social class (percentages)

	Upper middle class	Established middle class	Emerging middle class	Working class	Precariat	Poor	Total
Educational level
No formal education	0.0	0.6	0.0	0.1	19.6	60.8	8.5
Primary	0.0	1.6	1.0	37.8	44.4	27.3	22.0
Secondary	3.0	25.2	51.8	54.1	30.8	9.7	36.8
Vocational	7.2	15.8	47.2	8.0	5.1	0.0	16.0
University	89.8	56.8	0.0	0.0	0.0	2.3	16.7
Occupational class
Higher managers/professionals	47.6	5.0	0.0	0.0	0.0	0.0	5.5
Lower managers/professionals	35.9	42.1	22.0	0.3	1.9	0.0	14.3
Intermediate occupations	10.8	12.9	35.6	9.2	1.7	0.0	13.4
Small employers/self‐employed	3.0	5.3	8.0	22.0	27.4	8.6	14.9
Lower supervisory/technical occupations	0.9	5.3	20.0	7.9	5.0	2.3	8.4
Semi‐routine occupations	0.0	5.3	6.6	25.7	8.8	5.7	11.7
Routine occupations	0.0	6.0	7.8	30.9	20.8	4.6	15.7
Never worked or unemployed	0.4	4.0	0.0	2.2	12.6	17.2	4.9
Retired	1.3	14.2	0.0	1.8	21.7	61.5	11.2
Household monthly income
USD 0–400	0.9	2.4	0.0	1.2	44.2	87.9	15.9
USD 400–800	0.0	10.0	8.9	44.5	40.6	9.8	24.5
USD 800–1500	4.4	32.4	65.7	53.7	12.3	2.3	36.7
USD 1500–2500	28.1	25.2	25.3	0.6	2.9	0.0	12.2
USD 2500+	66.7	30.0	0.0	0.0	0.0	0.0	10.7

Note: ELSOC Survey, n = 2983. Bolded numbers are statistically higher than the percentage of category in set (<0.01).

Class homogeneity in social networks

We used the position generator to measure the kind of social contacts that form people's networks (see Lin, 2001). This survey instrument works by asking respondents: “How many people do you know who are X?”, where X indicates a variety of occupations of differing status.

In our case, respondents were shown a list with the following 13 occupations: (1) manager of a large firm, (2) street vendor, (3) secretary, (4) car mechanic, (5) shop assistant, (6) lawyer, (7) office cleaner, (8) doctor, (9) preschool teacher, (10) taxi driver, (11) waiter, (12) accountant, and (13) university professor. Seven options were offered as answers: none, one, 2–4, 5–7, 8–10, 11–15, and 16 or more. To perform the analysis, we assigned midpoints to the original intervals (e.g., range 2–4 is coded as “3”), and we also assigned the value of 17 contacts to the “16 or more” category.

In order to measure class homogeneity in social networks, we began by classifying the occupations included in the position generator into three status positions, closely resembling Goldthorpe's class scheme: higher (ISEI range: 65–88), medium (ISEI range: 40–64), and lower (ISEI range: 16–39) positions (see Table A3 in the Appendix). We then counted the total number of people known in these three status positions and assigned each respondent the proportion of same‐class contacts, that is, the proportion of low‐status contacts in the social network for those respondents from the poor, the precariat and the working class; the proportion of medium‐status contacts for those from the emerging and established middle classes; and the proportion of high‐status contacts for those from the upper middle class. In essence, the proportion of same‐class contacts indicates the class homogeneity of an individual's social network and can be regarded as a proxy measure of network segregation (e.g., Otero et al., 2021).

TABLE A3

The average number of people known with certain occupations

	ISEI	Min	Max	Mean	SD
Higher status
Doctor	88	0	17	2.1	3.3
Lawyer	85	0	17	1.5	2.4
University professor	77	0	17	1.7	3.3
Manager of a large firm	70	0	17	1.7	3.2
Medium status
Accountant	60	0	17	1.7	2.7
Secretary	53	0	17	2.6	3.6
Shop assistant	43	0	17	2.7	3.4
Preschool teacher	43	0	17	2.0	2.9
Lower status
Waiter	34	0	17	1.0	2.3
Car mechanic	34	0	17	2.1	2.7
Taxi driver	30	0	17	2.3	3.3
Street vendor	29	0	17	2.3	3.8
Office cleaner	16	0	17	1.7	3.0

Note: ELSOC Survey, n = 2831. Weighted figures.

Analytical strategy

The first stage of our analysis is conducted by exploring the structure of relationships among the different dimensions of attachment to society. MCA is an appropriate method to accomplish this goal when categorical data is available, such as in our case, while graphically showing the relative association between properties in a multidimensional space (Le Roux & Rouanet, 2010). We performed MCA to explore the interrelation between 32 variables across the four major dimensions of attachment to society. This method not only allows us to develop a comprehensive exploratory analysis, but also to justify the creation of an integrated measure of attachment to society.

The second stage of our analysis consists of estimation of ordinary least squares (OLS) regression models. These models are especially used to examine the joint and interaction effects of social class and network homogeneity on the level of attachment to society. Our integrated measure of attachment to society is used as the main dependent variable. All models include control variables for gender, age, being married or cohabitating, and political orientation. The latter variable is used because it strongly reduces the association between class and attachment to society, thus allowing us to provide more reliable results. We also considered geographical strata because it constitutes the key stratification factor for sample selection.

Table 3 shows the descriptive statistics of the variables used in the final models of the study. We note that the average level of attachment to society is about 9 points, with a maximum score of 26 points. This means that no person reached the potential score of 30 points. As noted, the working class is the largest group, while the poor is the smallest one. With respect to class homogeneity in social networks, we report that Chileans have, on average, 43% similar‐class contacts.

TABLE 3

Descriptive statistics of variables included in the regression models

	Min	Max	Mean/proportion	SD
Attachment to society	0	26	8.92	3.98
Social class
Poor	0	1	0.07
Precariat	0	1	0.20
Working class	0	1	0.30
Emerging middle class	0	1	0.20
Established middle class	0	1	0.15
Upper middle class	0	1	0.09
Class‐based network homogeneity	0	1	0.43	0.24
Control variables
Women	0	1	0.50
Age	18	88	44.42	15.40
Partner	0	1	0.55
Political orientation
Left‐wing	0	1	0.22
Center	0	1	0.24
Right‐wing	0	1	0.16
Independent	0	1	0.06
None	0	1	0.32
Geographical strata
Santiago	0	1	0.42
Valparaíso	0	1	0.10
Concepción	0	1	0.12
Large cities	0	1	0.28
Intermediate cities	0	1	0.04
Small cities	0	1	0.04

Note: ELSOC Survey, n = 2770. Weighted figures.

RESULTS

Descriptive analyses

We start with an MCA that interrelates 32 variables concerning individuals' attachment to society. In Figure 1, categories with a (+) sign indicate attitudes or behaviors that represent attachment to society in a certain domain (e.g., trust in parliament, participation in public meetings), whereas categories with a (−) sign indicate attitudes or behaviors that represent detachment from society in a certain domain (e.g., civic disengagement, disinterest in politics). Labels with an (=) sign indicate partial societal attachment regarding a given domain. It is important to note that not all variables have intermediate or neutral categories. Table A1 (in the Appendix) presents details of the variables used and, more precisely, the categories represented in the analysis and their labels.

FIGURE 1

Multiple correspondence analysis of individuals' attachment to society

Figure 1 shows that several categories proposed to represent the different theoretical dimensions and sub‐dimensions of attachment to society are indeed strongly interrelated empirically. Categories or attributes reflecting social trust, institutional trust, political engagement, participation in public affairs and altruistic dispositions are shown to be the most clearly associated between one another and thus contribute to coherently capturing the construct of attachment to society.

The division between attachment to and detachment from society is another outstanding result shown in Figure 1. This pattern is clearly visible along the x‐axis of the graph. On the left‐hand side, we find strong associations between categories that indicate detachment from society across different dimensions. These attributes are, for instance, having a small number of friends or none at all, and social and institutional distrust in several domains (e.g., political parties, the government), all of which refer to the relational dimension. These attributes are also interrelated with categories such as disengagement with public affairs (e.g., little interest in politics), little political participation, and reduced altruistic dispositions (e.g., disagreement with giving all groups an equal chance to succeed), which refer to the political dimension. All of these categories also show associations with those representing the normative dimension, including categories that express relatively little respect for social rules (e.g., limited support for democracy, justification of violence).

On the right‐hand side, by contrast, we find strong associations between categories that entail greater attachment to society. In this case, the y‐axis captures an additional specific distinction between categories. At the top, we see strong associations between categories representing having many friends and strong social trust in fellow citizens, including social minorities (i.e., the Mapuche people, Peruvian immigrants, and homosexuals). These attributes are also interrelated with categories that indicate the most intense interest in politics and participation in collective action (e.g., attending demonstrations and participation in social movements or party politics). All of these categories are also associated with those representing altruistic dispositions (e.g., giving all groups an equal chance to succeed) and pro‐social behavior (e.g., money donation and volunteering). At the bottom right of the graph, we especially observe strong interrelations between several categories that represent strong or moderate confidence in major political institutions (i.e., the government, parliament, the judiciary, and political parties) and labor unions. This means that most of the categories representing attachment to society are strongly interrelated, although categories comprising the sub‐dimension of institutional trust might represent a slightly different kind of attachment.

In addition, it is important to note that categories that are interrelated near the point of intersection between the axes generally indicate moderate or moderately low levels of attachment to society, including intermediate levels of trust in fellow citizens and neutral altruistic dispositions. These are the most common and least distinctive patterns of societal attachment reported in responses.

In order to describe how attachment to society is linked to social class and class‐based network homogeneity, we have incorporated the last two variables into the geometric space represented in Figure 1. Note that these factors are treated as supplementary variables and, thus, rather than influencing the structure of the graph, they are only projected onto the geometric space that has already been configured. Both social class and network homogeneity are colored in black in Figure 1. In short, we can see that the x‐axis represents social class as a continuum, and the middle classes (especially the upper middle class) are firmly located on the right‐hand side of the graph where attachment to society is the rule. Network homogeneity, for its part, is notably oriented to the left‐hand side of the graph and thus fairly associated with patterns of societal detachment. Overall, these preliminary and descriptive results are in line with our first two hypotheses.

In the following analysis, we use an integrated measure of attachment to society in accordance with our MCA. As mentioned, this measure is the result of counting the most distinctive categories or attributes that represent strong attachment to society. Among other categories, we count whether individuals trust “very much” in the courts, participate in social movements, attend peaceful demonstrations, and indicate that “democracy is always preferable”.

Explanatory analyses

We first provide evidence regarding the degree of class‐based network homogeneity by social class. Specifically, we estimate class differences in the proportion of same‐class social contacts within people's social networks. Our multivariate regression analysis is presented in Table A4 (in the Appendix). Figure 2 shows the estimated proportion of network homogeneity based on that model and two main results are clear. First, there is a rather strong tendency for class homogeneity in networks. Second, however, network homogeneity is especially apparent within the upper and lower classes, with slightly more network homogeneity found within the lower classes. Our overall interpretation is, therefore, that social distance tends to be larger between the upper and lower classes, while the middle classes tend to cross hierarchical boundaries and form “bridges” between the poles of the class structure.

TABLE A4

Class‐based network homogeneity by social class, OLS regression analyses

	Model 1	Model 2
Control variables
Women	−0.020	−0.028**
	(0.014)	(0.013)
Age	0.001	−0.000
	(0.000)	(0.000)
Social class (ref = emerging middle class)
Poor		0.162***
		(0.037)
Precariat		0.137***
		(0.023)
Working class		0.091***
		(0.019)
Established middle class		−0.024
		(0.021)
Upper middle class		0.110***
		(0.023)
Constant	0.448***	0.384***
	(0.014)	(0.019)
Adjusted R‐squared	0.006	0.071

Note: n = 2770. Control for geographical strata is included in all models.

Abbreviation: OLS, ordinary least squares.

Standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1.

FIGURE 2

Network homogeneity by social class. Predictive estimates with 95% confidence intervals.

We now turn to the multivariate regressions of attachment to society, with the dependent variable being an integrated indicator built in accordance with our MCA (see Table 4). The interpretation of our measure is straightforward: the more attitudes and behaviors reported by individuals, the greater their attachment to society. Model 1 shows that the control variables of gender, age, having a romantic partner, political orientation and geographical strata explain 7.8% of the variance in individuals' attachment to society. In Model 2, our measure of social class is added, and the explained variance notably rises to 16.8%.

TABLE 4

Attachment to society by social class and class‐based network homogeneity, OLS regression analyses

	Model 1	Model 2	Model 3	Model 4
Control variables
Women	0.298**	0.443***	0.390***	0.326**
	(0.147)	(0.140)	(0.140)	(0.141)
Age	−0.000	0.017***	0.016***	0.015***
	(0.005)	(0.005)	(0.005)	(0.005)
Partner	−0.078	−0.232*	−0.211	−0.204
	(0.144)	(0.139)	(0.138)	(0.138)
Political orientation (ref = left‐wing)
Center	−1.255***	−1.176***	−1.202***	−1.184***
	(0.220)	(0.209)	(0.208)	(0.208)
Right‐wing orientation	−1.282***	−1.449***	−1.474***	−1.482***
	(0.241)	(0.229)	(0.228)	(0.228)
Independent	−2.291***	−2.112***	−2.099***	−2.107***
	(0.306)	(0.291)	(0.290)	(0.289)
None	−2.840***	−2.536***	−2.554***	−2.554***
	(0.196)	(0.187)	(0.186)	(0.186)
Social class (ref = emerging middle class)
Poor		−1.826***	−1.626***	−0.943
		(0.294)	(0.296)	(0.574)
Precariat		−1.237***	−1.078***	0.191
		(0.224)	(0.225)	(0.483)
Working class		−1.124***	−1.020***	0.257
		(0.210)	(0.211)	(0.471)
Established middle class		0.939***	0.906***	1.035*
		(0.265)	(0.264)	(0.590)
Upper middle class		3.085***	3.169***	4.191***
		(0.320)	(0.319)	(0.735)
Class‐based network homogeneity			−1.430***	0.934
			(0.284)	(0.900)
Social class and network homogeneity
Poor #Network homogeneity				−1.857*
				(1.127)
Precariat #Network homogeneity				−3.085***
				(1.035)
Working class #Network homogeneity				−3.170***
				(1.045)
Established middle class #Network homogeneity				−0.171
				(1.429)
Upper middle class #Network homogeneity				−2.626*
				(1.558)
Constant	10.508***	10.692***	11.281***	10.406***
	(0.226)	(0.267)	(0.291)	(0.434)
Adjusted R‐squared	0.078	0.168	0.175	0.178

Note: n = 2770. Control for geographical strata is included in all models.

Abbreviation: OLS, ordinary least squares.

Standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1.

When all variables are considered together in Model 3, the explained variance increases to 17.5%. In this model, social class appears as the most relevant predictor of societal attachment. In general, results indicate that attachment to society increases notably with social class: on average, the poor (b = −1.626, p < .01), the precariat (b = −1.078, p < .01) and the working class (b = −1.020, p < .01) present significantly lower levels of attachment to society than the emerging middle class (the reference category). Conversely, the established middle class (b = 0.906, p < .01) and the upper middle class (b = 3.169, p < .01) report significantly higher levels of attachment to society than the emerging middle class. These differences are illustrated in Figure 3, where predicted values of attachment to society according to social class are presented. Attachment to society for the poor class is approximately 7.6 points, rising significantly to 8.1 and 8.2 points for the precariat and working class, respectively. The values then climb steeply to approximately 9.2 and 10.1 points for the emerging and established middle classes, and on to 12.4 for the upper middle class. With regard to network homogeneity, our estimates reveal that this predictor is clearly significant and negatively associated with attachment to society: every 10‐% point increase in class homogeneity significantly reduces the degree of attachment to society by approximately 0.14 points (p < .01).

FIGURE 3

Attachment to society by social class. Predictive estimates with 95% confidence intervals.

As we expected network homogeneity to influence an individual's attachment to society to different degrees according to social class, the testing of interaction effects is an important part of the analysis. Model 4 introduces an interaction term between social class and class‐based network homogeneity and shows that associations between network homogeneity and attachment to society are not uniform across social classes. Specifically, it shows negative interaction effects between social class and network homogeneity for the precariat (b = −3.085, p < .01), the working class (b = −3.170, p < .01), the upper middle class (b = −2.626, p < .1) and the poor (b = −1.857, p < .1) when compared to the emerging middle class (the reference category). Additional tests—not presented here—that compare class coefficients inform significant and negative interaction terms for the precariat and the working class when compared to the established middle class. They also indicate that the association between network segregation and attachment to society does not differ significantly between the upper middle class and the lower classes. Figure 4 plots the interactions, based on Model 4, and generally demonstrates that the most consistent effects are found for the precariat, the working class, and the upper middle class, but not for the (established and emerging) middle classes. For the precariat and working class, for instance, the level of attachment falls from approximately 9.5 to 7 points—a 25% reduction.

FIGURE 4

Attachment to society by social class and network homogeneity. Predictive estimates with 95% confidence intervals.

In other words, our results indicate that attachment to society among the upper middle class is high by default, and that increasing network homogeneity slightly reduces this pattern. The intermediate classes show an average attachment to society, and network homogeneity does not noticeably change this result. For the lower classes, attachment to society is relatively low and network homogeneity further reduces this tendency, especially among the precariat and the working class.

Additional analyses

In addition to our main analyses, we conducted several OLS regressions in order to provide a more detailed look at particular dimensions and sub‐dimensions of attachment to society theorized. We created several dependent variables by considering the variables included in our ACM (see Table A1). Our main analyses are based on multivariate regression models for the four main dimensions. The dependent variable that represents the cultural dimension is the average of two items: “I feel proud to be Chilean” and “I identify with Chile”. The other three main variables are created by disaggregating our primary dependent variable. In order to represent the relational dimension, for instance, we counted 10 attributes: (i) having 6 or more friends; expressing (ii) that “most people can be trusted” and trust towards (iii) Mapuche people, (iv) Peruvian immigrants and (v) homosexuals; and trusting in (vi) the government and parliament, (vii) the courts, (viii) political parties, (ix) labor unions, and (x) the police. This dependent variable therefore ranges from 0 to 10. We also complement these analyses by addressing specific sub‐dimensions highlighted in our theoretical background (e.g., social trust, political participation).

Figure 5 displays the predicted values from models that comprise the cultural and relational dimensions. Regarding the cultural dimension, we see that people from the lower classes have higher national identification than those in the upper middle classes, and network homogeneity especially reduces this sense of belonging for those at the top.

FIGURE 5

Separate analyses for the cultural and relational dimensions of attachment to society according to social class and network homogeneity. In order to facilitate the analyses, we used a dummy variable representing network homogeneity in the interaction term. The value ‘1’ indicates having more than 50% of similar‐class contacts in a social network. Predictive estimates with 95% confidence intervals.

With respect to relational aspects, results show that upper‐class individuals are more attached to society than their middle‐ and lower‐class counterparts. Network homogeneity clearly reduces societal attachment, and this effect is particularly strong for the lower classes. Further analyses presented in the Appendix complement this result by showing that social class is important for all the domains of the relational dimension, while network homogeneity diminishes, in particular, the number of friends and trust in major institutions (see Figure A1). For the lower classes, network homogeneity especially reduces institutional trust.

FIGURE A1

Separate analyses for the relational dimension of attachment to society according to social class and network homogeneity. Social trust is a count variable that includes whether people (i) consider that ‘most people can be trusted’ and (ii) express trust towards Mapuche people, (iii) Peruvian immigrants, and (iv) homosexuals. Institutional trust is also a count variable that includes the six institutions described in Table A1. In order to facilitate the analyses, we used a dummy variable representing network homogeneity in the interaction term. The value ‘1’ indicates having more than 50% of similar‐class contacts in a social network. Predictive estimates with 95% confidence intervals.

Figure 6 illustrates the predicted values from models that comprise the political and normative dimensions. With regard to political dispositions, analyses indicate that social class is, again, extremely important. Individuals from the upper middle class report much stronger political commitment than their middle‐ and lower‐class counterparts. Network homogeneity markedly diminishes political attachment, especially for upper middle‐class individuals. When looking at dependent variables representing the specific sub‐dimensions in the political domain, social class is particularly relevant to interest in politics, political participation (e.g., political parties, labor unions, demonstrations) and pro‐social behavior (volunteering and donations) (see Figure A2 in the Appendix). Network homogeneity reduces all these dispositions. For the upper middle class, network homogeneity especially reduces political participation, and for the lower classes, interest in politics.

FIGURE 6

Separate analyses for the political and normative dimensions of attachment to society according to social class and network homogeneity. In order to facilitate the analyses, we used a dummy variable representing network homogeneity in the interaction term. The value ‘1’ indicates having more than 50% of similar‐class contacts in a social network. Predictive estimates with 95% confidence intervals.

FIGURE A2

Separate analyses for the political dimension of attachment to society according to social class and network homogeneity. Political participation is a count variable that includes the eight instances described in Table A1. Pro‐social behaviour is also a count variable, counting whether people are involved in voluntary work and charity. In order to facilitate the analyses, we used a dummy variable representing network homogeneity in the interaction term. The value ‘1’ indicates having more than 50% of similar‐class contacts in a social network. Predictive estimates with 95% confidence intervals.

Finally, we address the normative dimension. Figure 6 illustrates that normative dispositions are higher at the top of the class structure. Normative attachment is reduced by network homogeneity, and this particularly holds true for the lower classes. We also disentangle the normative dimension through further analyses (see Figure A3 in the Appendix), which indicate that social class increases respect for all social rules, while network homogeneity especially reduces support for democracy and non‐aggressive behavior.

FIGURE A3

Separate analyses for the normative dimension of attachment to society according to social class and network homogeneity. Justification of violence is a count variable that includes whether people offer justification for the three behaviours described in Table A1. In order to facilitate the analyses, we used a dummy variable representing network homogeneity in the interaction term. The value ‘1’ indicates having more than 50% of similar‐class contacts in a social network. Predictive estimates with 95% confidence intervals.

Let us summarize by answering our general research questions. First, what is the association between social class and attachment to society? Our results consistently show that the level of attachment to society—the number of dispositions (e.g., trust in fellow citizens, interest in politics, participation in associations) reported by individuals—strongly and significantly increases with social class. Additional analyses indicate that higher class positions result in greater attachment to society in almost all dimensions and sub‐dimensions, with the exception of the sense of identification expressed by individuals. This provides clear support for Hypothesis 1.

Second, to what extent is class‐based network segregation associated with societal attachment above and beyond social class? The general pattern is that network homogeneity strongly reduces individuals' attachment to society, even when controlling for social class. This occurs in several dimensions and sub‐dimensions of attachment to society, especially the relational, political and normative dimensions. As such, Hypothesis 2 is supported.

Third, does the association between network segregation and attachment to society vary with social class? Network homogeneity reduces attachment to society to a greater degree for individuals from the upper and lower class, especially within the working class and the precariat. For the lower classes, network homogeneity especially reduces the relational, political and normative dimensions of attachment to society, most notably, institutional trust, interest in politics, support for democracy, and non‐aggressive behavior. For the upper middle class, network homogeneity especially decreases the cultural and political dimensions of attachment to society, in particular, political participation. These are intriguing differences that provide general support for Hypothesis 3 and will be discussed in the following and final section of the paper.

CONCLUSIONS AND DISCUSSION

This paper contributes to the knowledge on the association of social class and network homogeneity with attachment to society. We focused on Chile, a country known for its high economic inequality and low social cohesion. Our comprehensive assessment of attachment to society covered four dimensions and several sub‐dimensions. The categories representing the different dimensions and sub‐dimensions theorized were found to be strongly interrelated, and this allowed us to build an integrated measure of attachment to society for use in our main analyses. Taking the results together, we can draw the following conclusions.

The general pattern is that social class is a main predictor of attachment to society. Individuals from the upper middle classes are more strongly attached to society than their middle‐ and lower‐class counterparts. Our findings are in line with long‐standing arguments made by several scholars (e.g., McClosky & Schaar, 1965; Meier & Bell, 1959) and with more contemporary empirical examinations concerned with attachment to society, whether in terms of social and institutional trust, socio‐political participation, or other aspects (e.g., Alesina & La Ferrara, 2002; Li et al., 2003; Mckenzie, 2017; Verba, 2003). As suggested, there are several pathways along which this may occur, including the potential sense of control and entitlement that members of the upper classes may feel due to the extensive availability of socio‐economic resources that they enjoy. For lower‐class individuals, the situation is clearly different. We argued that they are more likely to sink down below a certain level of subsistence and to develop strong feelings of alienation (e.g., powerlessness, meaninglessness, self‐estrangement). Overall, this indicates that social class is of continuing relevance in contemporary Chile and influences the bond between the individual and society.

While previous studies provided some indications of the prevailing class divides in attachment to society, which are clearly complemented in this research, knowledge on the ways through which class interactions affect societal attachment has been rather limited. We developed a theoretical framework that discusses the potentially detrimental influence of class‐homogeneous networks on attachment to society, while suggesting that network homogeneity in essence represents the lack of knowledge about and experiences of the lives of others. Our results largely support this proposition. We found considerable homogeneity in networks, especially at the edges of the class structure, and, more importantly, we showed that homogeneity in networks clearly diminishes attachment to society overall. More detailed analyses indicated that this occurs especially with respect to the relational (e.g., number of friends and institutional trust), political (e.g., interest in politics and political participation), and normative (e.g., support for democracy) dimensions of attachment to society. As suggested, there may be several mechanisms that explain this process, including the incorporation of narrow identities and class prejudices. Whatever the cause, the present study serves to demonstrate that class homogeneity in networks clearly makes societies weaker.

That said, we ultimately showed that the negative association between class‐based network homogeneity and attachment to society only applies to individuals from the upper and lower classes, although it is notably stronger for the precariat and the working class. For the upper middle class, we noted that network homogeneity especially reduces the sense of national identification and political participation. It may be that upper middle‐class people with class‐based homogeneous networks share a more (selective) parochial belonging that provides them with optimistic perspectives of the lives of others and reduces the sense of responsibility that is needed for collective action. For the lower classes, we found that network homogeneity, in turn, particularly exacerbates institutional distrust and political apathy while diminishing respect for social rules. As proposed, people from segregated lower‐class segments probably lose their faith in political authorities and institutions because they have been excluded even further from wider society.

We would like to highlight that this novel attempt to examine the extent to which social class and class homogeneity in networks are associated with attachment to society is not without limitations. We acknowledge three shortcomings of our study. First, survey data often under‐represent individuals from the highest classes. This issue may be more relevant in Chile than in other contexts because the most privileged tend to live in gated communities or exclusive buildings where access is limited or even prohibited, especially in Santiago, the Chilean capital. This may explain why we did not find a stronger negative association between network homogeneity and attachment to society for the highest social classes. Second, we used cross‐sectional data, so no inference about causal relations could be made. In other words, we were unable to conclude that class‐based network homogeneity is the cause of lower attachment to society. This relationship may also work the other way around, although we think it is less plausible to argue that the degree of attachment to society influences choice of network members who belong to the same social class. Finally, although the study tested the relationship between social class, network segregation, and attachment to society while drawing on different mechanisms, we have not focused on the actual testing of these explanatory principles. We advocate future research that tests such processes more directly and unravels explicit pathways towards level of attachment to society.

All in all, the present study provides information on individuals' attachment to society and class‐based network homogeneity, and more generally on social cohesion in a highly unequal context. As such, it could provide researchers with useful insights into what might happen in other countries where inequality has been rising. We found that network homogeneity is observable across all social classes, but especially at the edges of the social structure, meaning that intermediate classes form a “bridge” between the upper and lower classes and represent an important pillar of social cohesion. While attachment to society is strongly anchored in class positions, results support the proposed argument that class segregation in networks also influences attachment to society. Class‐based network homogeneity is negatively associated with attachment to society overall, and this is especially the case among the lower classes. This suggests that the role of cross‐class relationships in facilitating societal attachment is at least as important as that of traditional values and ideologies that are usually spread across societies in order to promote social cohesion. In the end, it is not the ideas, but the relationships that make it possible. We thus conclude that enhancing social cohesion in Chilean society would require not only more equal opportunities but also changes to the social settings in which social class segregation is (re)produced.

CONFLICT OF INTEREST

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.