The school class as a social network and contextual effects on childhood and adult health: Findings from the Aberdeen Children of the 1950s cohort study

Abstract

Little is known about the health consequences of the school class as a social network. The present study asked whether overall school-class structure has contextual effects on psychiatric problems in childhood and adult self-rated health. From longitudinal data on a Scottish cohort, measures of school-class structure (centralisation, degree of reciprocity and proportion of isolates) were constructed based on sociometric information. Multilevel analysis demonstrated significant effects of centralisation on both health outcomes. It is suggested that highly centralised classes are characterised by inequality, resulting in a low level of integration, with subsequent negative consequences for health.

Introduction

Previous research has indicated that there is an association between social networks and a number of health outcomes (Berkman, 1995, Cohen and Syme, 1985, Seeman, 1996, Thoits, 1995). Most of the research into social networks and health has considered adults; there are far fewer studies of the short-term and long-term consequences for health of childhood social networks. When studying children's social networks, the school class emerges as a context of central importance. Due to the amount of time that the child spends together with classmates, the school class has in fact been described as a social unit of key significance in Western cultures (Hartup, 1984). There are, to the author's knowledge, no larger-scale studies that have highlighted the overall social network structure of the school class in relation to health using multilevel modelling, allowing for quantitative comparisons between classes. For this reason, the present study adopts an explorative approach in examining the structural features of the school class and whether these are linked to the individual's short-term and long-term health. The network characteristics selected for this study are: centralisation (width of the status distribution), degree of reciprocity (share of reciprocated relationships), and proportion of isolates (share of marginalised individuals). While their equivalents at the individual level have been found to be associated with a variety of short-term and long-term health outcomes (see e.g. Gest et al., 2001), their relevance for health at the contextual level needs to be further explored.

In the present study, the school class is conceptualised in terms of a social network. The chief theoretical implication of applying a social network perspective is that it highlights patterns and structures between social actors rather than the characteristics of the actors themselves (Edling and Rydgren, 2007). One of the key assumptions within social network theory is that the structure of a network has consequences for both the individual members and for the network as a whole, that act over and above the effects of the members’ characteristics (Klovdahl, 1985). However, network structures are not static but, rather, composed by a multitude of dynamic processes. The present paper assumes that a unique social climate, based on the continuous interaction between members of the school class, emerges within the structure of the network (Dreesman, 1982, van den Oord and van Rossem, 2002). The social climate influences, but is equally influenced by, feelings of belongingness and the extent to which members identify with the group (Cota et al., 1995). This phenomenon has been referred to as ‘cohesiveness’ or ‘integration’ (Hagstrom and Selvin, 1965). The analogous meaning of the two concepts has been implied in previous theoretical work, defining social integration as “the uniting of separate entities into a cohesive whole” (Bogardus, 1958, p. 207). According to Friedkin (2004), a network is cohesive when it possesses structural conditions that produce positive membership attitudes and behaviours. In a highly integrated network, in contrast to one with a low level of integration, members are more concerned with their membership and therefore also “more strongly motivated to contribute to the group's welfare, to advance its objectives, and to participate in its activities” (Cartwright, 1968, p. 91). To sum up, the dynamics of the school class network are closely connected to the level of overall integration of network members, which in turn is expected to have consequences for these members regardless of their own specific characteristics. The present study argues that health is one of these important consequences.

While there are few theories which explicitly link school-class structure with health outcomes, there are a number of parallels with what we know about adult networks and health. Berkman and colleagues have developed a theoretical framework for improving our understanding of how social networks may influence health outcomes (see Berkman and Glass, 2000, Berkman et al., 2000). They start by embedding social networks in a larger social and cultural context which conditions the extent, shape and nature of the network structure. Social networks are then assumed to influence social and interpersonal behaviour through the provision of social support, social influence, social commitment and attachment, and access to resources and material goods. These psychosocial mechanisms subsequently impact health through a variety of behavioural, psychological and physiological pathways. Applying this reasoning on the present study, it is likely that the school-class structure reflects certain conditions at the school level, which in turn are affected by such factors as its location, the teacher/student ratio, the average level of school performance, and so forth. These contextual conditions will influence, for example, the size and composition of the class: prerequisites that would furthermore have an impact on school-class dynamics. As previously mentioned, the overall level of integration which is established within the context of the school class is assumed to have an impact on the amounts of social support, influence, and engagement among the students, as well as on informational flow and access to resources. Arguably, a low level of integration could suggest that the school class is poorly functioning in terms of students not helping each other during lessons or with homework. One may also expect higher levels of negative peer pressure and comparison processes. On the contrary, members in highly integrated school classes would have a strong social identity and enjoy a positive climate. Concerning the links between school-class structure and adult health, various pathways may exist. Firstly, depending on how well integrated (or cohesive) the school class is, it would either generate (more or less) positive or negative sets of behaviours, expectations, ambitions and resources among the students (cf. Östberg and Modin, 2007). These psychosocial mechanisms are then assumed to influence subsequent health behaviours (e.g. smoking, exercise or alcohol consumption) and psychological factors (e.g. sense of well-being, self-esteem, or coping effectiveness). They may also give rise to physiological responses (e.g. allostatic load or immune system functioning). Through these pathways, health may be directly affected. It could also indirectly affect health via, for example, adult circumstances in terms of socioeconomic career and labour market experiences. Secondly, classroom structure could influence the types of network that the individual seeks out in adulthood: in all probability, similar structures with similar consequences for health. Thirdly, the overall structure of the school class affects students’ current health, which in itself has implications for their later health. Most likely, it is a combination of these various pathways and mechanisms. A final point concerns health as an outcome. In the present study, it is argued that the network structure of the school class is not related to any specific type of health outcome; an assumption that is based on the various types of health-related pathways that are considered (i.e. behavioural, psychological and physiological). Thus, a broad definition of health will be applied, encompassing both psychiatric problems in childhood and subjective health status in adulthood.

Multilevel modelling enables us to distinguish between classroom-level and individual-level variation, whereby the contextual effects of the school class structure on health may be captured (Sellström and Bremberg, 2006). There is, however, limited empirical evidence in this area of health-related research. There are a few studies that have focused on variations in health at the school level and found relatively small variations: 1% in well-being (Konu et al., 2002), 2.5% in psychological health (Karvonen et al., 2005), 2.6% in depressive symptoms (Goodman et al., 2003), and 0.6% in psychosomatic health (Modin and Östberg, 2009). Moreover, in a study by Karvonen et al. (2005) it was concluded that a poor class spirit was associated with a higher degree of health complaints at both the individual level and at the school level. Others have examined the variation in health at the school-class level. These studies found 11% variation in psychological health (Van den Oord and Rispens, 1999), 1.9% in psychosomatic health (Modin and Östberg, 2009) and 5.6% in health complaints (Torsheim and Wold, 2001). Although the empirical evidence about the contextual effects of the school class on health is scarce, two conclusions may be drawn. Firstly, the variation in health at the school-class level seems to be greater than the variation that may be ascribed to the school level. Focussing on the contextual effects of the school class on health is thus a relevant approach. Secondly, no studies of the long-term contextual effects of the class structure on health seem to have been conducted, which highlights a gap of knowledge in this specific area of research.

There is less consensus on how to define, operationalise and measure network structures at the classroom level than at the individual level (Cillessen, 2007). Brisette et al. (2000) have proposed social network analysis as a useful tool. In social network analysis, which is chiefly based on sociometry, nodes (e.g. people) and ties (e.g. relationships between people) are the central analytic entities. In order to outline the structure of a social network, the research population is asked sociometric questions such as “Whom would you choose as a friend?” or “Who would you best like working with?” While several statistical measures based on this information may be calculated in order to quantify the structure of social networks, the nature of the data material also has implications for the choice of network measures. The school class is a type of network that has ‘exogenously defined boundaries’, which indicates that group membership is well defined (Butts, 2008), but also that the size and composition of the network is not defined by the individuals in the network themselves. In some studies, individuals are allowed an unlimited number of nominations while in other studies, such as this one, it is restrained. This will undoubtedly affect all measures based on nominations. The network characteristics included in the present study (centralisation, degree of reciprocity and proportion of isolates) were chosen based on three considerations: firstly, they were all assumed to reflect different dimensions of overall social integration in the school class. Secondly, they were compatible with the design of the data being used and thirdly, their equivalent measures at the individual level had previously been linked to health. The measures are discussed in greater detail below.

An individual who upholds many relations with other network members is more influential, has better access to information, and greater opportunity for spreading this information (de Nooy et al., 2005). Here, the term ‘centrality’ or ‘point-centrality’ is often used (see e.g. Freeman, 1979). Having high centrality at the individual level reflects a kind of social dominance based on pro-social behaviours (Gest et al., 2001). On the contrary, low centrality suggests that the individual takes a lesser part in the social life of the network. In previous studies on school children and health, centrality has commonly been conceptualised in terms of ‘peer status’ (measured as the number of nominations received from classmates in different respects), referring to the degree to which the individual is an accepted, integrated and respected member of the group (Östberg, 2003). Associations have been found between peer status and various short-term health outcomes, for example malaise (Östberg, 2003) and depressive symptoms (Hecht et al., 1998, Kiesner, 2002). Moreover, the health consequences of peer status do not appear to be restricted to childhood: previous studies have found associations with self-rated health and limiting longstanding illness (Östberg and Modin, 2007) as well as in-patient care for mental, behavioural and life-style related diseases (Almquist, 2009) in adulthood. Centrality at the contextual level is called ‘centralisation’ or ‘graph-centrality’ (Freeman, 1979). In the present study, it refers to the width of the distribution of social positions in the school class: highly centralised networks include a few individuals who are highly central, whereas in less centralised networks, most individuals are centred near the mean. The present paper argues that a wide status distribution, i.e. high centralisation, has a great degree of inequality (the network is dominated by a few people whereas others are marginalised). Hence, if the network relations are focused through a small number of individuals the network is more vulnerable to segmentation and fractionalisation, which would result in a low level of integration (cf. de Nooy et al., 2005, Markovsky and Lawler, 1994). Moreover, in networks with diffuse patterns of friendship and influence most students perceive themselves as high status, whereas in networks with pronounced hierarchies only students who actually have high status see themselves as high status. According to Schmuck (1966), the consequences in the latter type of network involve a decrease in the overall level of integration and support. Previous studies have additionally demonstrated that highly centralised school-class structures are typical for bullying and victimisation networks (Vermande et al., 2000).

At the individual level, reciprocity has traditionally been seen as friendship between individuals (Rubin et al., 1998), although this is not necessarily always the case (i.e. two individuals may choose each other based on criteria other than friendship). Nevertheless, having friends contributes to the development of the self-system and pro-social skills (Bagwell et al., 1998). Children with friends are also more socially out-going, have a higher self-worth and display fewer depressive symptoms than children without friends (Bagwell et al., 1998). In sum, having many reciprocal relations seems to be positive for health (Nangle et al., 2003). At the contextual level, ‘degree of reciprocity’ refers to the level of mutual relationships in the school class. This has previously been considered as an indication of social integration (Lott, 1961, Moreno and Jennings, 1937). Its influence on health is, however, less obvious. In the present study, it is tentatively hypothesised that a high degree of reciprocity may be linked to a high level of pro-social behaviour, which would increase the level of social integration in the school class and thereby be beneficial for health.

It is just as important to recognise the lack of relations in a network as the existence of relations (Edling and Rydgren, 2007). Most networks contain members who are ‘isolates’. At the individual level, isolation is a type of exclusion that forms an interpersonal source of solitude. In the case of the school class, it indicates that the individual is being left out of peers’ activities either by indirect actions or by direct refusals (Gazelle and Ladd, 2003). The risks to health of being socially isolated may be compared to the risks associated with smoking, obesity and high blood pressure (House et al., 1988). Less is known about the consequences of the ‘proportion of isolates’ at the school class level. Since isolated children play a lesser part in the social life of the school class, a high proportion of isolates in the school class would hypothetically reduce the overall level of social integration.

The aim of this study is to investigate whether the structure of the school class (centralisation, degree of reciprocity and/or proportion of isolates) is associated with child and adult health: (1) Is there any variation between school classes in (a) minor psychiatric disorder in childhood, and/or (b) self-rated health in adulthood? (2) Is school-class structure linked to (a) minor psychiatric disorder in childhood, and/or (b) self-rated health in adulthood?