Health-promotive and exclusionary mechanisms in civil society – a critical review of the empirical support for Putnam’s view of social capital

screened (n =459) Abstract excluded (n = 269) Reason 1 (n=87) Reason 2 (n=90) Reason 3 (n=92) Reason 3 Studies assessed for eligibility (n = 190) Studies excluded (n=132) Reason 4 (n= 8) Reason 5 (n= 30) Reason 6 (n= 56) Reason 7 (n= 16) Reason 8 (n= 15) Reason 9 (n= 3) Reason 10 (n=4) Studies included in review (n = 58) Identification of studies via database SocINDEX Id en ti fi ca ti o n


ABOUT THE AUTHOR
This review is part of a larger project looking at the relations between partipation in civil society and health. Åsa Kneck (. . . . . .) and Anders Kassman, associate professor at the Center for Civil Society Studies at ESBH have more specifically taken interest in sports participation among girls with an immigrant background, due to theire lower participation rate compared to other groups. We have made focus-groups interviews with immigrant parents, a survey to volunteering trainers and interviews with the sports clubs in a suburb outside Stockholm, Sweden. The results reveal that the volunteers consist of a higly homogenous group of people from a more affluent area while the participant are more heterogenous. The results also illuminates the distinctive western european characters of civil society as an institution that is not very familiar to people born and raised in countries such as Somalia, Eritrea and Syria.

PUBLIC INTEREST STATEMENT
Civil society is often seen as vital for democracy and for public health. Associations and public spaces give people a chance to meet, to discuss matters of interest and develop their social relations. In many welfare societies participation in civil society is therefore supported. Policy and research have focused on the positive functions of civil society. But civil society, and health, can mean different things for different people. In this extensive review of research covering both civil society and health we see that there sure some positive effects from taking part in civil society but there are also important side effects due to its exclusionary sides. People tend to gather with the likes of themselves and guard their groups from others. Policy interventions should be more aware of the tendencies in civil society to exclude less resourceful people from taking part.

Introduction
Childhood and youth are prolonged as life phases in modern societies worldwide to protect and socialize newcomers to become full citizens. The school system might be the most significant institution affecting the lives of children and youth. However, children and youth have more time for leisure-activities than other age groups since schooldays are short and their spare time can be used for play and recreational activities, not only just for the fun of it but also for additional socialization together with peers and adults. To promote healthy lifestyles and to prevent deviance, many states encourage after-school programs, sports organizations, associations, and other forms of participation in civil society. In his ground-breaking book "Bowling Alone" (2000), Robert Putnam placed new focus on social capital as a fundamental resource to be guarded and promoted in the North American states. In his concluding chapter, he sketched an agenda for "social capitalists" in which youth organizations and civic engagement had significant roles to play. According to Putnam, an important aspect of social capital is its promotive effect on health and well-being, at both the individual and the state levels. He even said that "As a rule of thumb, if you belong to no groups but decide to join one, you cut your risk of dying over the next year in half. If you smoke and belong to no group, it's a toss-up statistically whether you should stop smoking or start joining. These findings are in some ways heartening: it's easier to join a group than to lose weight, exercise regularly, or quit smoking" (Putnam, 2000, p. 331, italics in original).
Putnam said that the evidence was strong, but the mechanisms leading from social cohesion to health were still somewhat unclear. One theory was that networks functioned as a safety net that reduced stress and provided support if needed. Another indicated that they worked as a social control promoting positive behaviour and norms that could protect against unhealthy behaviour such as smoking, obsessive eating and other unhealthy behaviour, and a third more speculative idea was that social capital served as a "physiological triggering mechanism stimulating people's immune system to fight disease and buffer stress" (Putnam, 2000, p. 327). Regardless of which theory fits best, it is clear from Putnam's reasoning that he thought of associations as resources that individuals themselves can choose to take part in if they want to significantly enhance their health and, in his agenda for restoring social capital, as being possible to promote politically. In other words, he treated associations as politically responsive and generative of health. Health was viewed as an effect and social capital as the cause to be manipulated. In his most recent work, "Our Kids", Putnam concentrated on stratification and the widening class gap in different parts of society, and also here he described extracurricular activities as "a near-perfect tool" (p. 258) to provide mentoring and soft skills in order to close the opportunity gap and counterbalance stratification. One example he mentions is the class gap in adolescent obesity and how upper-class kids, because of their networks and access to social arenas, are less vulnerable than children from low-poverty neighbourhoods (Putnam, 2015) Although in the citations above Putnam seems to draw a more drastic conclusion than what we might be used to, his ideas seem to align well with common-sense thinking that participation in civil society promotes health. It also seems to follow the same logic as the well-established, theory that social relationships cause health, which can be traced all the way back to the classic É (1951) study of suicide. Social relationships include every type of networks, friendships, and social support between people (and animals) which of course includes participation in civil society organizations, but also much more, such as marriages, families, working life, etc. And even though the evidence speaks for a causal connection (House et al., 1988), the mechanisms explaining exactly how relationships promote health are less clear (Thoits, 2011). Since the ever-increasing importance of online activities and networking via the Internet the "dark side" of social relations have got more attention with results showing that social networking can affect well-being, self-esteem, life satisfaction and body images negatively (Webster et al., 2021). The aim of this study is to review how the relationship between civil society and health is viewed in contemporary empirical social science studies among youth. Youth is of special interest begcause parents, families, communities, the state and other stakeholders have interests in filling their spare time with something meaningful as well as investing in behaviours promoting a healthy lifestyle. This opens for a wide range of policy suggestions to promote further socialization. We attempt to answer the following questions: What aspects of civil society and health are in focus in empirical youth studies?
How is the relationship between civil society and health regarded in terms of cause, effect and correlation?
Is there a relationship between causal claims and policy suggestions in the studies?
The review will contribute to a comprehensive and up-to-date understanding of the social mechanisms connecting participation in civil society and health and it will also help disentangling in what ways policy could promote better health via support to civil society. Both civil society and health are somewhat broad concepts that sometimes seem difficult to disentangle. Associations might be the core of civil society and well-being at the heart of health, but with regard to, for instance, empowerment, trust and social capital, the divisions seem less clear. We therefore perform a first theoretical exposition of both concepts. Then, we discuss some different understandings of causation, how researchers use causal claims as metaphors in empirical research, and how causal analysis can be used for policy development in pragmatic research. Following these sections, we describe our reviewing methods and sample, and thereafter come the sections for the results and discussion where we attempt to elaborate the role of positive and exclusionary processes within civil society and the study´s theoretical and practical implications.

Background
Civil society can be viewed in many ways. Here, we have started our analysis from two more descriptive ways of interpreting civil society: a) as a sphere and b) as an arena. a) As a sphere, civil society is separated from other spheres with different rationalities such as the state and the market. Here, voluntary organizations, NGOs, cooperatives and foundations are in focus. Civil society consists of the sum of all involvement, memberships and efforts made within the framework of these organizations. For youth, different types of leisure activities, especially sports organizations, attract many members and, in doing so, dominate the sphere. However, organizations with an antisocial agenda such as xenophobic or violent organizations are also included in the same sphere as more democratically governed voluntary organizations, even though they are not always emphasized in research (Anheier, 2004). b) As an arena, civil society consists of an open and semi-public space for collective deliberations. This arena is placed in-between the more private life within family and close friends and the more structured and organized existence in large public organizations. In this arena, different ideas can be confronted, debated and tested against each other. Voluntary organizations can be active in the arena, but not all of them choose to participate, and it is not only the organizations that are emphasized (Van Til, 2008, p. 208). Participation in the arena is basically voluntary and non-profit, but it is neither volunteerism nor organizational belonging that is important but, rather, the possibility for individuals to meet and exchange experiences and to discuss questions of mutual interest on their own terms (Kocka, 2006).

Health
As a commonly used concept, health has several different definitions. A well-known, broad and inclusive holistic definition of health is presented by the World Health Organization (WHO, 1946): health is a dynamic state of complete physical, mental, social and spiritual well-being and is not merely the absence of disease or infirmity. With this definition, health is considered a resource, an outcome and an intrinsic good. In a later document WHO (1948) describes health as a positive concept focusing on social and personal resources as well as physical capabilities. A holistic definition of health involves the physical, mental and social dimensions of human existence Holistic definitions can include subjective as well as objective views of health (Nordenfelt, 1995).

Causation
For a long period of time, the relationship between cause and effect has been under debate in attempts to explain human action (Bernert, 1983). A major reason for questioning causal claims in the social sciences is their deterministic presumption. Even if we might not yet have found the right research designs and methods to detect it, the relationship between cause and effect must be compulsory to be meaningful. The search for covering laws, which dominated in the past, has been replaced by probabilistic, statistical correlations, and analysis has been refined by elaborating with necessary and sufficient causes (Mackie, 1965;Mahoney, 2008). However, analytical and methodological sophistication has not silenced the quest for the driving force that connects cause and effect. The idea of causality as a generative process places social mechanisms in focus for causal analysis (Goldthorpe, 2001). A social mechanism is a theoretical, analytical concept that focuses on the tie between cause and effect, the "nuts and bolts" between cause and effect, to borrow the words of Jon Elster (1989).
For a long period of time, randomized controlled tests (RCT: s) or experiments have been apprehended by many as especially strong designs for causal analysis because they make it possible to ascertain that the cause precedes the effect, that cause and effect correlate and that rival explanations can be controlled for (Shadish et al., 2002). In line with common sense and especially appealing for evaluation and policy development, the RCT: s is designed to test the effects of known causes. The possibility to manipulate the causes is of highest importance for policy development. If research can find causes behind desired effects that are possible to manipulate, the translation to policy suggestions should be quite straightforward (Shadish et al., 2002).
In parallel with technical and statistical sophistication, empirical research has come to abandon the deterministic idea implicit in causation and instead started to think about causes as raising the probability of different effects, or it has even abandoned speaking of causes at all, replacing causation with statistical correlation. Following pioneers such as Lazarsfeld, multivariate analysis of cross-sectional data was developed. This made it possible to analyse causal chains and path models with latent variables, illustrating causal nets in entire populations or, more commonly, samples thereof, strengthening the external validity (Bernert, 1983). Cohort studies and longitudinal designs are examples in which correlations are possible to analyze over time, in this aspect facilitating causal analysis (MacLeod et al., 2004).
The search for social mechanisms linking cause and effect has been described as an example of middle-range theorizing and as an alternative complementing the widespread variable-centred theorizing exemplified by causal modelling (Hedström & Swedberg, 1996).
In an empirical review of the "naturally occurring" usage of causal claims in American sociology, three models were found: "the Initiating, which credits a forceful object that starts an unfolding causal process; the Conditioning in which force is located in an environment to which affected entities are exposed; and the Probabilistic, which also credits a forceful object, but with associative instead of initiating effect" (Vaidyanathan et al., 2016, p. 21, italics in original).
The discussion of causation shows that the same phenomenon that makes causation attractive in explanative terms, the binding force between cause and effect also makes it complicated to apply to human actions, including policy development. RCT: s can be used to show how people react to certain manipulations, and cross-sectional designs can be used to analyse how populations act in real life; however, both designs have shortcomings when analysing abstract phenomena such as health and civil society.

Method
An integrative review was conducted to capture the complexity of knowledge regarding the relation between civil society and health. The method was considered appropriate, allowing both quantitative and qualitative research to be integrated in the review, and it was described as suitable for summarizing the existent research, defining concepts, and reviewing theory (Whittemore & Knafl, 2005). Initially, through discussions, the authors identified the scope and aim of the review. Inspired by Whittemore and Knafl (2005) a five-stage approach consisting of problem identification (presented in the background section), the literature search, data evaluation, data analysis and presentation was adopted and performed. In line with Kirkevold (1997) a broad and diverse sampling was pursued, and studies from different parts of the world and within different contexts were included.

Literature search
In collaboration with a librarian, a search strategy was developed solely for the SocINDEX database. Initially, the authors presented a list of keywords related to civil society and health based on the theoretical and empirical literature. The key-words were then translated by the librarian into subject terms in SocINDEX (appendix 1). The search string involved limitation to young persons with the SocINDEX terms: children (0-12 years), teenagers (13-18 years), youth (13-25 years), adolescence (between puberty and adulthood) and young adults (18-25 years), publications only in English and without time limit for publications.
The developed search string that can be seen in appendix 1 resulted in 477 titles. After several steps (see Figure 1, flow diagram), 190 full-text articles were assessed for eligibility, and following another round of assessment of the full papers the selection resulted in a total of 58 articles that met the inclusion and selection criteria: (1) they described some aspect of civil society and health as well as the relationship between them, and (2) they utilized quantitative, qualitative and/or mixed methods to assess the outcomes of the relationship between civil society and health. We thereby excluded many articles that might be of indirect interest to the review discussing close topics that could be relevant. But that was a prize we judged worth paying for keeping up stringent inclusion criteria.

Data evaluation
To obtain a first understanding of the concepts and context of the data, a matrix based on the included studies was arranged involving the following areas: the context and country of the study, the year of the study, the journal of publication, the research design, the sample size, key concepts and the dimensions of health and civil society.

Data analysis
In this stage, the data were further categorized, summarized and integrated, aiming for a new synthesis based on the primary sources. In the categorization phase, the relationships between participation in civil society and health were classified as (1) civil society as a determinant factor for health, (2) health as a determinant factor for participation in civil society (outcomes related to participation in civil society) and (3) a circular relationship between health and participation in civil society. Civil society was divided based on the views of sphere" and "arena" and health based on the dimensions of "physical", "mental" and "social". In a following phase, a critical analysis of the pattern developed from the categorization was conducted (cf. Whittemore & Knafl, 2005) and is presented in the results and discussion sections. See also appendix 2 for a summary of the articles main themes, methods and context.

Results
We used no time limit in our search string, but the first article included in the review was written in 1997 and the search updated to cover May 2021. Figure 2 shows that there was a slight increase in the number of articles over the period, spiking in 2006 and in 2018 and a more stable level between 2012 and 2015; however, since there are quite a few articles/year, there is a risk of random effects in the figure.
There were seven very large studies with over 10.000 sampled individuals; another nine studies had more than 1.000 participants, 22 had 100-999, and the rest were based on samples fewer than 100 persons. In some studies, it was very unclear what age group was being studied, but most noted that the teenage years were in focus; a couple of studies were more interested in young adults, and one study therefore included participants from 18 to 45 years of age.
Overall, it was most common to research mental health and civil society as an arenaas diplayed in table 1. Well over two thirds of the reviewed articles were interested in mental health and twothirds in civil society as an arena. Another overall observation was that participation in civil society was most often researched for its health effects rather than vice versa. It was not very common to study the selection effects within civil society or whether different organizations attracted certain populations that could be healthier than others, even though some articles were interested in these questions.
As seen in table 2, health was often operationalized as subjective answers to questions of wellbeing, sometimes combined with other more objective indicators of risk behaviour such as smoking, substance use, stress or eating disorders. Only a few studies used objective measures of different diseases as indicators of unhealthy living. The operationalisations of participation in civil society were more diverse, and objective measures such as membership in certain organizations, being part of a network or a self-help group or internet use were more common. Subjective indicators were less common, but when they were used, they were often formulated to measure aspects of trust, cohesion, collective efficacy or neighbourhood social capital. Table 3 shows that approximately two-thirds of the articles were primarily based on quantitative methods, preferably some type of cross-sectional design, whereas seven of the 58 reviewed  articles used a longitudinal design, which is often regarded as one of the preferred choices when investigating causal relationships.
Although only 14 of the 58 articles were based on research specifically designed to enable causal analysis, such as controlled experiments or longitudinal designs, causal claims were very common in the abstracts and discussion and conclusion sections. In table 4 it can be seen that it was most common to use conditioning causal claims such as "influenced by" or "contributes to", whereas the more assuming, initiating causal claims such as "determine" or "produce" were only used in approximately half of the articles' abstracts, discussions and conclusions. We could also find many more ways of articulating the causal claims than prior research. We could position most of them as initiating, conditioning or probabilistic, but we also separated out interactive claims and social mechanisms as two separate ways to articulate causal relations as displayed in table 5.
Although not giving significant chi-square statistics at the 95% level, there were tendencies indicating that articles based on designs more suited for causal analysis (longitudinal and experimental), initiating causal claims and policy proposals were made more often.
Many of the articles mainly treated civil society as a health promotive arena. This means that our broad sampling procedure put other forms of civil society participation at focus, besides formal organizations. More loosely coupled networks and meeting places seem to be more at focus for research, which might be taken as a first indication that civil society is a complex entity and hard to reach for policy of different sorts. We also found that only very few articles framed their findings as civil society studies. They covered topics such as volunteering, youth participation, sports clubs, self-help groups, mentoring programs, networks, and online communities, and they studied empowerment, efficacy, trust, social capital, socialization and engagement. However, they rarely discussed their results in relation to theories of civil society. Neither was health as a general concept much discussed in the articles; instead, different types of well-being were described as a desirable outcome. To determine what was meant, we often had to investigate the questionnaires, items or psychometric measures used. This might be less of a problem for the published research articles than for the theoretical thinking around health and civil society. Theoretical propositions must be developed in close contact with empirical studies to obtain gains in clarity, comprehension, and relevance.

Summary of results
The empirical results found in the articles may be summarized as follows, starting with those elucidating a promotive relationship, then some mixed results, and, finally, some studies that analysed excluding mechanisms within civil society.
Based on a large (n = 8.152) longitudinal database, two researchers could conclude that participation in organized sports during adolescence resulted in more positive self-beliefs one year later and more physical activity and better subjective health six years later. Positive self-beliefs worked as a mediator between sport participation and health in young adulthood. The positive effects were not due to physical activity, instead the researchers hypothesized that it was the membership in a sports team that provided a supportive social context that  (Vaidyanathan et. al., 2016:21). A British study found that, in particular, girls from more privileged backgrounds had positive experiences from taking part in a government-funded choir program aimed at stimulating local social capital (Hampshire et al., 2010). However, other participants reported that they risked their position in other networks and that they felt excluded by the choice of repertoire and other less visible mechanisms, indicating possible stratifying effects of this type of arts project. A project among Hispanic youth in USA found significant gender differences in the complex relations between family cohesion, social support, community participation and substance use. Community participation, and civil society, were more important for the males because females were kept closer to the family (Lardier et al. 2018). An evaluation of a self-help group for adolescent outpatient substance abuse treatment revealed selection effects and mixed results for different groups of youth but concluded that, at least for some of them, participation might be beneficial for the recovery process (Gangi et al., 2012). Two other evaluations showed that intergenerational mentoring and peer support networks can have positive complementary effects in attempts to reach specific high-risk groups among youth (Nicholas et al., 2009;Rogers et al., 1997). A British evaluation showed that lifestyle sports, exemplified by mountain biking, contain both exclusionary and inclusionary mechanisms in parallel, leading to uneven effects among the participants. Some, white males, were empowered, whereas others further from the core were excluded (King et al., 2015).
A North American longitudinal study (n = circa 2.000) found that without interventions, adolescents with poor health tended to form smaller local networks and occupy less central positions than their healthier peers. Those with poor health were stigmatized (Haas et al., 2010). Another longitudinal study (n = 1.050) from the USA found that youth with eating disorders and depressive symptoms tended to self-select into certain sports activities that might contribute to worsening their problems (Fay et al., 2013). A third longitudinal study from the USA (n = 1.820) found that depressed youth had fewer friends and smaller and more homogenic networks as an effect of both their own homophilic selection and withdrawal from peers, resulting in a negative spiral leading to increasingly more isolation (Schaefer et al., 2011). A very large and repeated (n = 289.672) crosssectional study from USA found that children with chronic health conditions were more likely to never attend religious worship services. Especially, children with conditions that impede communication and social interaction were most likely to never attend. These correlations were stable over time (Whitehead, 2018).

Theoretical discussion
Putnam focused on participation in associations as an important fertilizer for social capital. Participation in civil society was theorized as a way to create social capital, giving better health. In fact, Putnam argued that social capital determined health, regardless of socioeconomic position. He stated that even with more risk factors for, e.g., heart attack than the general population, people with great social capital run lower risks of dying than others. He concluded that not only did healthy, health-conscious and privileged people tend to live longer but also there was an independent effect of social engagement that made a difference (Putnam, 2000). Even in his latest book when discussing the widening class differences in the USA he treated participation in extracurricular activities, such as girls and boy scouts, sports, etc., as a small but important piece to start resolving class differences (Putnam, 2015). However, we could find no clear support for these conclusions in our review. In stark contrast, we found that civil society is highly dependent on its context and not that easy to govern in wanted directions. In the cases where civil society seems to have the clearest promotive effects on health it tends to be the more organized versions where youth-adult relations are purposefully and carefully cultivated in order to support positive development and control excluding tendencies.
Although Putnam noted that social capital could be an asset for both individuals and groups, he mainly focused on it as a collective asset. Whereas prior social capital researchers had tried to use social capital as a theoretical concept linking the macro and micro levels in action theories (Bourdieu, 1986;Coleman, 1990;Lin, 2001), Putnam had a more practical interest in promoting policy development. Following Coleman, he viewed the social norms upholding the networks' reciprocity and trust as an important part of the definition of social capital (p. 19). Taken together, these lines of reasoning led to a highly instrumental view of civil society as promoting health via cohesion, trust and social control, which was envisioned in Putnam's agenda for social capitalists and in which the socialization and education of youth were of the utmost importance.
People who organise can gain social capital but they may also try to secure its accumulation by excluding those who they see as less productive (Portes, 2000). This might be more salient in more exclusive organisations. Associations focusing on personal, material individual or status group identity are more likely to be exclusive, where other associations such as hobby-, humanitarianand environmental-organizations tend to be more inclusive and foster trust, responsibility and an outward-orientation (Griesshaber & Geys, 2012). Participation in civil society and different networks contributes to social control both from peers and adults, which may promote healthier lifestyles for those who are included. Therefore, social control seems in most cases to be the most obvious candidate of the three mechanisms hypothesized by Putnam for stimulating better health among youth. However, it may also alienate youth from the very same groups that would most need change. Civil society has exclusionary effects that we should not dismiss and which already Pierre Bourdieu described accurately when introducing networks (social capital) as one form of capital, besides cultural and economic capital. All forms of capital are accumulated resources that are convertible into other forms of capital and therefore protected by its owners (Bourdieu, 1986). Members of every network, organization or group safeguard their social capital by inviting resourceful members and excluding those judged as less contributive.
Social control might be a side effect but not the main reason for many young persons to join a club, and it is also likely that youth are aware of this and therefore self-select into networks and groups that seem to align with their own interests. Even if young person themselves may lack interest in social control it might be a more obvious interest for their parents. From James Colemans perspective civil society and its organisation can be seen as natural extensions of families, or in more modern societies, as purposive structures aimed at socialising youth into responsible adults (Coleman, 1990).

Practical conclusions
Causal explanations are attractive because they are conclusive and predictive and can give clear directions in practical matters. If we could find the cause behind a desired effect, then it might be possible to manipulate it via different types of interventions. However, as we found in the review, the causal claims made in empirical social science studies are not always that conclusive, and researchers often make policy suggestions grounded in considerations other than strict causal analysis. A closer examination of the uses of causal claims reveals that very few made direct conclusions from their own results via initiating causal claims to policy suggestions. The closest that we could find comes from a cross-sectional, ecological study of health among children in five Nordic countries (Berntsson et al., 2006).
Causal explanations place high demands on the possibility of separating cause and effect, which is often complicated when working with abstract concepts such as civil society and health. Neither civil society nor health was thoroughly discussed as important concepts in the articles. Most articles analysed the relation between taking part in networks, associations or other organisations and some form of health, often in the form of well-being. Following Putnam and others, social capital was often analysed as the link between civil society and health, but the operationalisations reveal that different concepts often float together and make causal analysis doubtful.

Final words
An important strength of this review is its breadth and systematic collection of current, peer-reviewed international empirical research studies. With help from a professional librarian and colleagues, we could develop search strings covering most aspects of health and civil society. However, our search strategy also comes with some limitations. We could only review studies from one database with articles written in English and following often very strict journal formats, excluding books and more theoretically oriented texts. As a result, almost half of the reviewed articles were performed in the USA and a little over 10% in Great Britain, with their specific civil society and welfare mix, which does not align in all aspects with the rest of the world (see appendix 2).
To conclude, we would like to note that empirical research seems to confirm that there is a positive relationship between participation in civil society and health for many young persons, even though there seems to be a significant risk of circular reasoning in empirical research. There are also examples of notable exclusionary mechanisms not so often discussed in civil society research that we should not forget. Online networking seems more often to be connected to risk behaviour (Webster et al., 2021). The homophilic selection process into different groups tends to have excluding effects if not guarded against. The positive aspects of participation in civil society may take time to develop and may not be as easily politically promoted as Putnam thought. It might even be easier to quit smoking than to join a group, at least if it is better health that one is after. Civil society offers no quick fix for health problems because it is likely not more open for equal participation than other parts of society.