Criteria for inclusion

Search strategy

Data collection and analysis

Study selection

Following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-analyses) guidelines (Moher et al. 2009), the search and screening phases are represented in Figure 1. After removing the duplicates, the search yielded 773 unique studies which were screened using the inclusion/exclusion criteria. The majority (n = 627) of the studies were excluded at the title and abstract screening stage and a further 44 were excluded when the full text was screened. Studies were excluded because they did not fit with our definition of child/adolescent (n = 389) or our definition of health and wellbeing (n = 115), the study design criteria (n = 92) or the definition of family or community social capital (n = 73). A total of 102 articles were retained for inclusion across the health and wellbeing outcomes of the larger review; 55 of these included mental health and/or behavioural problem outcomes and represent the total sample reported here.

Description of studies

The quality appraisal ratings and key descriptive information for each of the 55 included studies is presented in Additional file 2: Table S1. The reviewers rated 37 of the studies as high quality, 17 as moderate quality and one study was assessed as being low quality.

Following data extraction the mental health and behavioural problem outcomes were grouped into four coherent categories: self-esteem and self-worth; internalising behaviours (e.g. depression and anxiety); externalising behaviours, (e.g. aggression, violence, conduct disorders and disobedience); and, composite measures of mental health and problem behaviours. Nineteen studies reported on two or more of these categories. Nine studies reported on at least one indicator of FSC, 22 reported on at least one indicator of CSC and 24 reported on both FSC and CSC (see Additional file 2: Table S1).

The majority of the studies (n = 43) were surveys, and six of these were longitudinal. Also included were eight longitudinal (Birndorf et al. 2005; Drukker et al. 2006; Drukker et al. 2010; Feldman 2010; Parcel and Menaghan 1993; Parcel and Menaghan 1994; Windle 1994; Xue et al. 2005) and one cross-sectional cohort studies (Drukker et al. 2003), one controlled trial (DuBois et al. 2002), a quasi-experiment (Bowker et al. 2010) and one qualitative study (Landstedt et al. 2009). The majority of the studies were conducted in the USA (n = 29), seven were conducted in the UK, four in Canada and four in the Netherlands. The remaining single country studies were conducted in Australia, El Salvador, Greece, Israel, Italy, Lebanon, Serbia, Sweden, Taiwan and Vietnam and one study was conducted in the UK and Canada (see Additional file 2: Table S1).

Few of the studies clearly articulated the dropout rates and in many the number of participants fluctuated across the different analyses that were conducted. To ensure consistency, in this review we report the maximum number of young people included in each study’s analyses. Samples ranged in size from 98,340 participants to 31 in the quantitative studies and in the qualitative study the sample size was 29. Two studies reported on the mental health and/or behavioural problems of preschool children, five reported on children, 34 reported on adolescents and the remaining 14 studies reported on mental health and/or problem behaviour outcomes of mixed aged groups of young people. Thirty-nine studies had mixed sex samples with the percentage of female participants ranging from 45% to 63%. One study had a male-only sample and in the remaining 15 studies, the sex of participants was unclear or not reported.

In 29 studies the ethnicity, race or nationality of the young people was not reported or it was not possible to extract this information. Nine studies described the majority participant group as Caucasian, non-Hispanic White or White and we grouped these under the single category ‘White’ for reporting. Eleven studies described the majority group as African American, Black, or non-Hispanic Black and we grouped these under the single category ‘Black’. In the remaining studies the majority participant groups were described as American Indian, Dutch, Latino, Mainland Chinese and Southeast Asian American.

Self-esteem and self-worth

Ten studies (see Additional file 2: Table S1) explored the role and impact of social capital on self-esteem or self-worth (Abbotts et al. 2004; Birndorf et al. 2005; Ciairano et al. 2007; Drukker et al. 2006; DuBois et al. 2002; El-Dardiry et al. 2012; Glendinning and West 2007; Jager 2011; Ying and Han 2008; Yugo and Davidson 2007). Four studies, all with adolescent samples, explored the role of FSC and there was evidence that parent-adolescent relationships characterised by positive communication (Birndorf et al. 2005), nurturance (Yugo and Davidson 2007) and low levels of conflict (Ying and Han 2008) were associated with higher self-esteem/worth. Moreover, there was longitudinal evidence showing that positive parent-adolescent relationships in early adolescence were associated with better self-esteem at age 17-18 years (Birndorf et al. 2005). Families assessed as being cohesive (Ying and Han 2008) and families where there was evidence of adult interest in the adolescent (Ying and Han 2008) were associated with better outcomes. This gives further support to the positive role of intra-familial relationships. In contrast, parental monitoring and control was associated with poorer self-esteem/worth (Glendinning and West 2007; Yugo and Davidson 2007).

The seven studies that assessed the role and impact of CSC offered evidence to show that positive relationships that extend beyond the family boundaries are associated with higher levels of self-esteem/self-worth. Children and adolescents were more likely to report higher self-worth/esteem if they had access to their own support networks that include both adults (DuBois et al. 2002) and their peers (Glendinning and West 2007; Yugo and Davidson 2007). They also benefited from their parents’ networks, with better outcomes being reported in children/adolescents whose parent(s) received support from informal networks/experienced a sense of belonging and support (El-Dardiry et al. 2012).

The quality of the school the adolescent was attending was associated with positive self-esteem/worth. Adolescents who reported feeling safe at school (Birndorf et al. 2005) and adolescents who reported that they were engaged with school (Yugo and Davidson 2007) had more positive self-esteem/self-worth. However, one study that explored school quality in sub-groups of adolescents reported it to be associated with self-esteem/worth in adolescents from urban communities but not those from rural communities (Glendinning and West 2007).

There was also evidence that religiosity had a differential impact across sub-groups of adolescents. Increased attendance at religious services was associated with better outcomes for male adolescents (Birndorf et al. 2005) and weekly attendance at religious services was associated with better outcomes for adolescents who self-identified as Catholic; however, church attendance was associated with poorer outcomes in adolescents self-identifying as belonging to the Church of Scotland (Protestant) (Abbotts et al. 2004). There was no data available to explore this further in this review, but the authors hypothesise that differences in the normative behaviours of religious groups may play a role here (e.g. church attendance may be more accepted in some groups than others).

In sum, adolescents who share a positive relationship with their parent(s) and those with higher quality/quantity of social support networks are more likely to have higher self-esteem/worth. On the other hand, parental monitoring/control, which may reflect more negative elements of the parent-adolescent relationship, appears to be linked with lower self-esteem/worth perhaps reflecting the adolescents’ loss of autonomy in aspects of their own lives.

Internalising behaviours

Thirty-one studies explored the role and impact of social capital on internalising behaviours (see Additional file 2: Table S1). The specific outcomes in these studies include: depressive symptoms, anxiety and social anxiety, moods, emotions and composite scores on assessments that measure a range of these behaviours (referred to by some authors as ‘over-controlled behaviours’) (Abbotts et al. 2004; Aneshensel and Sucoff 1996; Beiser et al. 2011; Bosacki et al. 2007; Caughy et al. 2003; Caughy et al. 2006; Caughy et al. 2008; Ciairano et al. 2007; Delsing et al. 2005; Drukker et al. 2003; Drukker et al. 2006; DuBois et al. 2002; El-Dardiry et al. 2012; Fitzpatrick et al. 2005; Fulkerson et al. 2006; Glendinning and West 2007; Jager 2011; Kliewer et al. 2004; Landstedt et al. 2009; Meltzer et al. 2007; Rasic et al. 2011; Rotenberg et al. 2004; Rotenberg et al. 2005; Springer et al. 2006; Stevenson 1998; Wang et al. 2011; Windle 1994; Xue et al. 2005; Ying and Han 2008; Young et al. 2011). We also included studies that reported on suicide/suicidal ideation and self-harm.

Although explored by two studies, there was no evidence to suggest that family structure shared an association with internalising behaviours (Aneshensel and Sucoff 1996; Glendinning and West 2007). On the other hand, seven studies presented evidence to suggest that positive parent–child relationships were associated with decreased levels of internalising behaviours in children and adolescents (Caughy et al. 2008; Springer et al. 2006; Ying and Han 2008). Moreover, there was evidence that the quality of the parent–child relationship may be more important for some sub-groups of young people than others. Positive relationships were associated with better outcomes in children/adolescents living in low violence neighbourhoods (Kliewer et al. 2004), a pattern not replicated in high violence neighbourhoods, and adolescents from rural communities benefited from good relations with their parents in a way not afforded to adolescents from urban communities (Glendinning and West 2007).

Further supporting the positive association between family relationships and internalising behaviour outcomes, children and adolescents who assessed their relationships with other family members as high in justice (i.e. fairness) and trust (Delsing et al. 2005), those who were part of a cohesive family (Ying and Han 2008) and those from families that frequently had meals together (Fulkerson et al. 2006) had better internalising behaviour outcomes. In contrast, reports of parental monitoring in two separate studies were inconsistent; one reported a positive association with adolescents’ internalising behaviours (Ying and Han 2008) and the other reported a negative association (Glendinning and West 2007).

Eleven of the included studies explored the role of support networks. There was evidence to suggest that children and adolescents with access to wider social networks (i.e. a higher number of friendships) (Rotenberg et al. 2004) and higher quality social networks (e.g. friendships low in hostility) (Beiser et al. 2011; Windle 1994) had fewer internalising behaviours than children/adolescent with smaller or poorer quality social networks. Again, some sub-groups of children/adolescents may benefit more from social support networks than others. For example, preschool children living in affluent neighbourhoods had fewer reported internalising behaviours if their primary caregiver reported knowing their neighbours, on the other hand, in impoverished neighbourhoods not knowing neighbours was associated with better outcomes for preschool children (Caughy et al. 2003). Moreover, peer support was associated with fewer internalising behaviours in adolescents from rural communities but this was not replicated in urban communities (Glendinning and West 2007).

There was evidence to suggest that schools and neighbourhoods with higher quality environments offered children and adolescents protection in relation to internalising behaviours. Cohesive neighbourhoods (Kliewer et al. 2004), neighbourhoods low in hazards (Aneshensel and Sucoff 1996) and neighbourhoods high in other indicators of social capital were associated with lower internalising behaviours. Only one study reported a negative association between neighbourhood quality and internalising behaviours; adolescents who perceived that adults in their neighbourhood imposed too many constraints on them reported higher levels of internalising behaviour (Glendinning and West 2007). Although the authors did not explore this further, it might be hypothesised that while control over adolescent behaviour (e.g. anti-social behaviour) may improve the quality of the neighbourhood in the eyes of adult residents this may not be perceived as such by adolescent residents.

In sum, children and adolescents with more positive relationships with other family members and who have wider and higher quality networks that extend beyond the family, either directly with their peers or indirectly through their parents’ networks, have fewer reported internalising behaviours. Living in a higher quality neighbourhood is also associated with better child and adolescent mental health outcomes. That said, it is important to note that social support networks may not benefit all children equally. For example, in impoverished communities, better outcomes are reported for children whose primary caregiver reported knowing fewer of their neighbours. The authors hypothesise that mothers who are able to manage adversities in their impoverished neighbourhood, perhaps because they can access other assets, may need less social capital to support healthy development in their children (Caughy et al. 2003).

Externalising behaviours

Twenty-four studies (see Additional file 2: Table S1) explored the role and impact of social capital on externalising behaviours and these included measures of: aggression; anger; violence; lying; conduct and oppositional defiant disorder symptoms (negative, short tempered, defiant, argumentative, disobedient and hostile behaviour towards adults and authority figures); and composite scores on assessments that measure a range of externalising behaviours (referred to by some authors as ‘under-controlled behaviours’) (Abbotts et al. 2004; Aneshensel and Sucoff 1996; Bearinger et al. 2005; Caughy et al. 2003; Caughy et al. 2006; Caughy et al. 2008; Champion et al. 2008; Ciairano et al. 2007; Delsing et al. 2005; Drukker et al. 2003; Drukker et al. 2010; DuBois et al. 2002; El Hajj et al. 2011; Fulkerson et al. 2006; Jager 2011; Johnson 1999; Kingston et al. 2009; Kliewer et al. 2004; Meltzer et al. 2007; Oman et al. 2005; Smith and Barker 2009; Springer et al. 2006; Stevenson 1997; Windle 1994). The evidence available to assess the role and impact of family structure on externalising behaviours was limited to two studies and only one of these found an association; living in a one-parent household was predictive of increased oppositional defiant disorder symptoms (Aneshensel and Sucoff 1996). There was, however, evidence to demonstrate that positive relationships between parents and their adolescent/child were associated with reporting of fewer externalising behaviours (Caughy et al. 2008; Kliewer et al. 2004; Springer et al. 2006). Moreover, the parent-adolescent relationship appears to be particularly important for those from a one-parent household (Oman et al. 2005).

Given the nature of these behaviours, it is perhaps surprising to note that only one of the included studies explored the association between parental monitoring and externalising behaviours and failed to find one (Smith and Barker 2009). However, there was evidence demonstrating that positive relationships between children/adolescents and their extended family were associated with better outcomes. Children and adolescents from families that were high in feelings of trust and justice (Delsing et al. 2005) and cohesion (e.g. more frequently ate meals together) (Fulkerson et al. 2006) had lower levels of externalising behaviours. In contrast, adolescents living in high risk neighbourhoods reported increased suppression of anger when extended family support was higher. The authors suggest this demonstrates that the family has an important role to play in moulding anger suppression in adolescents and, we surmise, that this may be context-specific (Stevenson 1997). Families in neighbourhoods where the risk of violence and/or conflict is high are likely to transmit different messages to young people about appropriate behaviours than families living in neighbourhoods where the risk is low.

There was mixed evidence relating to the association between social support networks and externalising behaviours. In one study, preschool children living in areas with high levels of poverty were reported to be at increased risk of displaying externalising behaviours if their primary caregiver reported higher levels of social support from their neighbours. In contrast, preschool children from more affluent areas were less likely to display externalising behaviours if their primary caregiver reported having social support from neighbours (Caughy et al. 2003). For adolescents, increased quantity and quality of social networks was associated with increased lying and disobedient behaviours in one study (Ciairano et al. 2007) and increased reporting of fighting in another (El Hajj et al. 2011). However, a number of other studies reported that social support networks offered adolescents protection against some externalising behaviours (e.g. fighting, delinquency and anti-social behaviours) (Champion et al. 2008; Oman et al. 2005; Windle 1994).

Also associated with externalising behaviour outcomes was the quality of a child/adolescent’s school and neighbourhood environment. Children and adolescents who attend a higher quality school and/or live in higher quality neighbourhoods are less likely to display externalising behaviours (Aneshensel et al. 1996; Bearinger et al. 2005; Springer et al. 2006).

In sum, in the context of externalising behaviours FSC offers the most consistent protective role for children and adolescents. In the context of CSC a number of studies reported risk relationships and in other studies social capital was protective for some externalising behaviours but not others. Consistent with internalising behaviours, caregivers from impoverished neighbourhoods who reported knowing few of their neighbours also reported better outcomes for their children. As noted above, this may be because these caregivers have access to assets other than social capital that enable them to deal with the demands of their environment and support healthier development in their children (Caughy et al. 2003).

Composite internalising and externalising behaviours

Thirteen of the included studies (Dorsey and Forehand 2003; Dufur et al. 2008; Feldman 2010; Galboda-Liyanage et al. 2003; Harpham et al. 2006; Maynard and Harding 2010; Maynard and Harding 2010; Newman 2007; Parcel and Menaghan 1993; Parcel and Menaghan 1994; Parcel and Dufur 2001; Slee and Murray-Harvey 2007; Wen 2008) (see Additional file 2: Table S1) explored the role and impact of social capital on internalising and externalising problem behaviours measured as a single outcome on a scale such as the difficulties sub-scale of the Strengths and Difficulties Questionnaire (Goodman 1997). Nine of the studies were cross-sectional surveys and they reported on mixed sex samples across the various age groups.

Family structure was assessed by five studies and the evidence suggested that young people who lived in a two-parent family were less likely to have internalising/externalising problems (Galboda-Liyanage et al. 2003, Wen 2008). There was stronger evidence, in six studies, that positive parent–child relationships were protective against internalising/externalising problems in children and adolescents (Feldman 2010; Maynard and Harding 2010; Parcel and Dufur 2001; Wen 2008). There was inconsistent evidence for the role of parental monitoring with one study reporting a negative impact of control for adolescents (Maynard and Harding 2010) and another reporting monitoring to be positive for children and adolescents (Parcel and Dufur 2001). Total FSC, assessed using a composite measure, was also associated with better child/adolescent outcomes (Dorsey and Forehand 2003; Dufur et al. 2008).

Evidence from three studies points to children/adolescents benefiting directly and indirectly from social support networks; directly, through their own networks, (Newman 2007; Wen 2008) and indirectly, through their parents’ networks (Harpham et al. 2006). Attendance at religious services (Parcel and Dufur 2001; Wen 2008), attending a school with a higher quality environment (Parcel and Dufur 2001) and living in a neighbourhood with higher levels of safety (Dorsey and Forehand 2003) were all associated with fewer general internalising/externalising problems.

In sum, positive relationships that exist within the family and those that extend out into the community are associated with better outcomes when internalising and externalising behaviours are assessed as a composite. Children and adolescents also seem to benefit from the structural support that comes in the form of higher quality schools and neighbourhoods.

Mental health and behavioural problems – synthesis

Summary of results

The primary aim of this integrative systematic review was to identify, analyse and synthesise empirical evidence on the association between family and community social capital and mental health and behavioural problems in children and adolescents. In doing so we assessed evidence from 55 studies making this the largest and most comprehensive systematic review in this field. In addition, this is, to the best of our knowledge, the first review to focus specifically on the mental health/behavioural problems of children and adolescents.

The large body of evidence included in this review supports the conclusion that both FSC and CSC are important in the context of children and adolescents’ mental health and behavioural problems.

Strengths and limitations

The strengths and limitations of this review exist at two levels: in the review process itself and the individual studies. In terms of the review process, while the extensive heterogeneity in the outcome measures prevented us from performing a meta-analysis, by adopting an integrative approach with a robust analytical process we were able to synthesis data from a large number of studies that employed a range of different research designs. Moreover, with the exception of one, the included studies were rated by the reviewers as being moderate to high quality which strengthens the conclusions that can be drawn from the synthesised results. We do, however, acknowledge that relevant literature may not have been identified. For example, journals are known to favour papers that report statistically significant results meaning that studies failing to identify significance may be under-represented. Our success in capturing studies was also dependent upon adequate indexing of papers within the databases; however, as outlined above, we did take measures to ensure that our search strategy was as robust as possible.

In terms of the individual studies, social capital is a multifaceted concept whose dimensions function in various directions (Morrow 1999; Stone 2001), the lack of an agreed definition and little uniformity in its measurement across the studies made synthesising the evidence challenging. However, we employed well-defined study-specific inclusion/exclusion criteria to ensure relevant data were captured in an objective fashion. It is important that future research defines and operationalises social capital in a more consistent and robust manner enabling a clearer understanding of its relationship to important outcomes and assisting comparisons across studies.

Other reviews exploring social capital and mental health have included children and adolescents but considered them alongside adults (Almedom 2005; De Silva et al. 2005) and, as we note above, this means that the definitions of social capital applied in these previous syntheses of data may not be adequate for the child/adolescent group (Morgan 2011; Morrow 2001). We sought to overcome this criticism of previous work by adopting a definition of social capital that was theoretically pluralistic but informed by previous research to ensure that it encapsulated aspects of social capital relevant to young people. For example, given that school is integral to the lives of children and adolescents and their families, and in keeping with the work of Coleman (1988) and Ferguson (2006), we included this in our definition of community social capital.

Moreover, as noted in the Background section, while current directions in health and social policy place emphasis on the promotion of an assets based approach, the majority of research in the field of mental health/behavioural problems in children and adolescents is conducted at the negative end of the spectrum, focusing on the onset and treatment of disorder. That said, while there was limited availability of positively framed evidence, we sought to ensure that, where possible, the conclusions we drew were framed within an assets based approach.

Despite the limitations that we highlight, the large body of evidence in this review means that we have been able to demonstrate conclusively that FSC and CSC are both associated with mental health and behavioural problems in children and adolescents; however the cross sectional nature of many of the studies prevented us from drawing firm conclusions about the direction of these associations. While it is assumed that social capital can be an asset that supports better outcomes, it is equally plausible that families with a child/adolescent who has a mental health problem (including behavioural problems) are limited in their ability to access and mobilise social capital. This review also highlights a lack of qualitative evidence. Future qualitative research is essential if we are to develop a theoretical framework that articulates the mechanisms through which social capital works to effect health and wellbeing, the various circumstances in which this occurs and how family and community social capital interact and mediate outcomes.