1 Introduction

Community participation and sense of community (SoC) are both concerned with community member engagement and active involvement in issues that affect people’s lives and impact the larger community. In particular, the role of community participation in promoting local development, social justice and population health has been acknowledged in multiple areas of research, such as community development, community psychology and policy analysis. Research has specifically noted that community participation increases quality of life (Nussbaum 1999), enhances social wellbeing (Keyes 1998; Wandersman and Florin 2000), fosters social empowerment (Zimmerman and Rappaport 1988; Chavis and Wandersman 1990) and reinforces social capital (Putnam 2000; Wollabæk and Selle 2003). Nearly identical outcomes are associated with sense of community (SoC). Chavis and Wandersman (1990) posited that SoC can be defined as a “catalyst” for participation and community development, and research has effectively confirmed that SoC is associated with a variety of community engagement behaviors, either civic forms of community participation (Brodsky et al. 1999; Chavis and Wandersman 1990; Florin and Wandersman 1984) or conventional and non-conventional forms of political participation (Anderson 2009; Davidson and Cotter 1986; Xu et al. 2010). Hence, both community participation and SoC stand out as interrelated key factors that promote community development or actualize the capacity of communities to activate their internal human resources, solve problems and promote social empowerment.

2 Sense of Community

The literature generally agrees that SoC signifies a healthy community and exhibits an extra-individual quality of emotional interconnectedness observed in collective lives (e.g., Bess et al. 2002). In 1974, Sarason first introduced the concept of a psychological sense of community that primarily agrees with the four-dimensional model conceptualized by McMillan and Chavis (1986). This model identifies four components as crucial for the formation and development of SoC. Membership corresponds to the feeling of being part of a community; this aspect embraces the perception of shared boundaries, common history, symbols, sense of emotional safety, and personal investment in community life. Influence encompasses the individual perception of mutual influence, not only providing opportunities for individuals to participate in community life, make their own contributions, and perceive their impacts on the collective decisions and actions of the community but also heightening individual awareness that personal choices and decisions are affected by the community itself. Fulfillment of needs represents the benefits that people derive from their community membership and refers to the positive relationship between individuals and their communities to the extent that the community helps its members meet their personal and group needs. Finally, shared emotional connection unveils the sharing of common repertoires, such as history and significant events, and strengthens the quality of social ties.

The four-dimensional model proposed by McMillan and Chavis is the datum point for the majority of scholars interested in the investigation of SoC; the original SoC definition has been expanded to meet the specificities of the populations under scrutiny, such as adolescents (Cicognani et al. 2006) and people with mental illness (Townley and Kloss 2009), and to capture the attributes of various types of community, namely virtual communities (Blanchard 2008; Tonteri et al. 2011), schools (Vieno et al. 2005, 2007) community organizations (Hughey et al. 1999), and above all, territorial communities of different sizes (Prezza et al. 2001). However, regardless of the specific type of community considered, most scholars agree that SoC positively affects both the individual and the community (Long and Perkins 2007). Various authors have noted that SoC contributes to increases in the quality of life and enhances well-being and life satisfaction (Farrell et al. 2004; Prezza and Constantini 1998). Furthermore, studies investigating the effects of SoC across various contexts have highlighted positive associations with empowerment within organizational settings (Hughey et al. 2008), social cohesion (Wilkinson 2007), place attachment (Long and Perkins 2007) and sense of safety (Zani et al. 2001) within territorial settings.

The increasing investigation and progressive adaptation of SoC and its measurement across settings has led some authors to challenge the transferability of SoC, arguing that it is a context-dependent construct (Hughey et al. 1999). At least two primary interconnected arguments can be expanded. Research on the conceptualization of SoC emphasizes the presence and operation of multiple senses of community (Brodsky and Marx 2001) derived from the increased complexity of individual community memberships, identities and roles. The emphasis on the multiplicity of SoC has partially shifted the focus to the examination of the relationship between sense of community, identification processes and community salience (Obst et al. 2002). Scholars have contended that the relationship between SoC and other constructs can vary across community types, as Prezza and Constantini (1998) argued with reference to the relationship between SoC and personal well-being for adolescents in a territorial community. A second significant debate considers the measurement of SoC (see Nowell and Boyd 2010 for a critical review). Several measures have been derived from the McMillan and Chavis (1986) conceptual model and adapted to assess SoC in specific community settings. The Sense of Community Index (SCI; Perkins et al. 1990) is the most popular instrument for the general measurement of SoC, regardless of the type of community investigated. Several studies have assessed the psychometric properties of such a scale (see Chipuer and Pretty 1999 for a critical review) and primarily attested to its stability across settings. Nevertheless, research has proven that the SCI fits different multidimensional structures and does not always correspond to the McMillan and Chavis (1986) four-dimensional model. Alternative solutions include the three-factor structure suggested by Long and Perkins (2003), which has been criticized for its lack of a strong theoretical grounding. The low reliability of the scale (Chavis and Pretty 1999) and its partial overlap with other constructs, such as group identification (Mannarini et al. 2012), motivate the search for additional, more adequate tools. This line of reasoning has motivated many researchers to improve the empirical measurement of SoC. New scales have been proposed to overcome the construct context-dependency and the limitations of existing measurements; these new scales include the Brief Sense of Community Scale (BSCS; Peterson et al. 2008b), the Multidimensional Territorial Sense of Community Scale (MTSOCS; Prezza et al. 2009), the Italian Sense of Community Scale (Tartaglia 2006), and a 10-item modified version of the SCI (see e.g., Obst et al. 2002; Prezza et al. 2009) with an alternative method of measurement (Brodsky 1996; Puddifoot 2003).

3 Community Participation

Different disciplines (e.g., Political Science, Social Sciences, and Community Psychology) utilize many definitions of political and social participation. Social (civic or citizen) participation has been defined as “a process in which individuals take part in decision making in the institutions, programs, and environments that affect them” (Heller et al. 1984, p. 339). Although students of comparative politics do not agree on the definition (see Conge 1988 for a review), political participation is generally referred to as an interest in political life; such interest can take many forms and can result in either indirect involvement or direct political action (Rollero et al. 2009). The most notable form is voting in elections, but political participation also includes joining a political party, running as an electoral candidate, joining a non-governmental advocacy group, or participating in a demonstration. Several classifications have been proposed by sociologists and political scientists (Almond and Verba 1963; Marsh and Kaase 1979), who have contended that political participation consists of at least two dimensions, i.e., latent participation, which includes the psychological involvement of citizens in politics (e.g., keeping abreast of political issues, discussing issues or engaging in the search for information), and manifest participation, which includes behaviors that foster direct contact with political representatives, such as holding a political office or managing political propaganda.

Ekman and Amnå (2009) recently proposed a typology of three main categories of participation in an individual or collective form: political participation, civil participation (or latent participation) and non-participation (or disengagement). Each of the three categories includes two subtypes. Political participation is either formal political participation (e.g., being a member of a political organization, voting, or running for public office) or activism, also labeled extra-parliamentary political participation (e.g., signing petitions, boycotting, or demonstrating), which may be legal, as in the examples mentioned above, or illegal (e.g., civil disobedience, sabotage, or violent demonstrations). Civil participation includes social involvement (e.g., showing interest in politics and society, identifying oneself with an ideology, or adopting an engaged lifestyle) and civic engagement (e.g., recycling, reading newspapers, or volunteering in community services and community organizations). Finally, non-participation comprises active or anti-political forms (e.g., non-voting or non-political lifestyles) and passive or apolitical habits (e.g., political passivity or non-reflected non-political lifestyles). Whereas no previously mentioned typologies distinguish between stable and transient participatory behaviors, social movement theorists such as Walgrave and Klandermans (2010) claimed that this distinction draws a crucial boundary between “participation”, a pool of nearly stable and repeated patterns of behaviors across contexts and time, and “mobilization”, a transient set of behaviors tied to a specific situation (e.g., elections, strikes or protest movements). To gauge the different forms of participation/mobilization, the majority of the empirical studies have used ad hoc scales that list behaviors. Respondents are normally invited to assess the frequency of each of the listed behaviors in a past period of time. Many of these behaviors recur across a large number of studies, yet no validated reference scale exists.

4 SoC and Community Participation: Moderators and Mediators

Studies on the relationship between SoC and community participation have analyzed many variables as covariates, moderators or mediators. Socio-demographic variables, such as income, education (Hayghe 1991; Verba et al. 1995), length of residence (Schiff 1990; Verba et al. 1995) and age often serve as variables that reinforce the association between SoC and participation. Some studies have stated that the role of age may be curvilinear (Pillemer and Glasgow 2000), suggesting that the relationship between SoC and participation would be weak for young and elderly people and strong for adults. Nonetheless, the effect of age cohorts would vary according to the form of participation in which individuals are engaged (Putnam 2000). Findings on gender are controversial; some research highlighted differences between men and women (Chambre 1984), while other studies showed no difference (Fischer et al. 1991). In addition to socio-demographics, some authors have considered context variables such as neighborhoods and churches (Huckfeldt and Sprague 1995) and workplaces (Mondak and Mutz 2001, 2002; Mutz and Mondak 2006) as moderators of the SoC-participation relationship. Scholars have also considered psychosocial variables, such as social control (Rollero et al. 2009), social ties (Liu and Besser 2003), well-being (Itzhaky and York 2000), empowerment (Peterson et al. 2008b) and community capacity (Bowen et al. 2001). Research on SoC and political behavior has focused on the interplay between SoC, political efficacy and trust in institutions. Anderson (2010) demonstrated that social forces, such as community, exert positive and significant effects on internal and external political efficacy and personal trust in political institutions, regardless of income, age, gender, and education. Moreover, this author emphasized that SoC increases participatory behaviors; this effect is mediated by political efficacy and trust in politics. The cited study supports the notion that SoC, community participation and political efficacy are interconnected and suggests that “the prevalence and density of kinship, friendship, and acquaintanceship networks and the level of participation in community-based organizations fosters the emergence of collective efficacy, or solidarity and mutual trust (social cohesion) among community residents combined with shared expectations for social control-related action” (Browning et al. 2004, pp. 506–507).

5 Study Goals

Despite evidence attesting to the association between SoC and community participation, the strength of this relationship is still unknown, and the direction of such a relationship is not obvious. The majority of the empirical studies have considered participation as a dependent variable, but theoretical approaches have assumed the existence of a circular relationship between these two variables: SoC enhances active citizen participation, which in turn reinforces SoC. Moreover, studies that have analyzed the association between SoC and participation have utilized diverse types of participatory behaviors to consider differences in commitment, duration and organizational features (i.e., ranging from long-term engagement in neighborhood associations to short-term campaigning or voting). Large variability exists in the measures utilized. This heterogeneity does not help to establish whether the association between SoC and participation is constant across a wide range of participatory behaviors (including mobilization behaviors), contexts and populations or conclude whether SoC plays a role in community competency and responsibility. Based on this summary, the goal of the current study was the utilization of a systematic review to test the stability and strength of the relationship between SoC and participation, specifically the association of SoC with civic and political participation. The review additionally intended to verify whether this relationship was influenced by specific characteristics of the populations under scrutiny, such as nationality, age and gender. Questions raised include the strength and stability of the SoC-participation relationship, variations in the relationships according to different forms of participation, and the influences of age, gender and nationality on the SoC-participation relationship. These questions were addressed with a meta-analysis of studies published in leading international journals in the fields of psychology, sociology, political and social sciences.

6 Method

6.1 Selection of Studies

“Sense of community” and “participation” were entered as multiple queries utilizing the keywords, descriptors, and terms included in the subject and the abstract. The following databases were probed: EBSCO, PsycINFO, PsycARTICLES, Scopus, Web of Knowledge, Social Services Abstracts, Sociological Abstracts and Worldwide Political Science Abstracts (see Table 1). The search included peer-reviewed journals in all databases and peer-reviewed journals and conference papers in Scopus. This procedure resulted in a list of 742 publications, 483 of which were excluded as duplicates. The abstracts of the remaining 259 publications were examined by three independent judges, all of whom agreed to exclude 153 records that reported qualitative studies or did not include measures of SoC and participation. As a result, a list of 106 publications from 1980 to 2012 was obtained (Table 2). A full-text examination of the 106 publications led to the further exclusion of 81 items (68 did not report measures of SoC and participation; two utilized inaccurate measures for participation; 12 did not report the data required to calculate the effect size; and one reported the same data as a previous study by the same author), resulting in a final list of 23 publications.

Table 1 Databases, queries and publications retrieved
Table 2 Meta-analysis eligible publications

Only SoC that applied to territorial communities was considered. Publications that focused on organizational (No. 67) or virtual (No. 88, 105) SoC were excluded. Studies that considered forms of participation that were neither civic nor political, such as participation in college programs (No. 30, 31, 91), web-related activities (No. 47, 88), recycling interventions (No. 48) or drug prevention programs (No. 65) were additionally excluded.

The flow diagram in Fig. 1 shows the described process.

Fig. 1
figure 1

Flow diagram

6.2 Data Extraction

The effect size was calculated with the Comprehensive Meta-Analysis® software (CMA version 2; Borenstein et al. 2000). The following information was extracted from each study reported in the selected publications: (a) year of publication, (b) nationality of the sample, (c) data utilized to calculate the effect size (correlation coefficients or group means), (d) type of participation (i.e., civic, political or general, when the scale measures included both forms), (e) SoC measures utilized, (f) participation measures utilized, (g) sample size, (h) percentage of women in the sample, (i) population from which the sample was drawn (adults, >18 years; adolescents, 14–19 years; college students; military; immigrants; mentally disabled), (j) average age of the sample, and (k) age range. Some articles included more than one study or more than one measure (No. 5, 25, 26, 37, 41, 66, and 70; see Table 2). All the studies included were based on cross-sectional data.

6.3 Statistical Methods

The relative effect size of each study was extracted for each of the relationships between SoC and participation. Therefore, some studies were included more than once, such as studies that contained sub-group comparisons or multiple measures of participation or analyzed the relationship between SoC and both civic and political participation. Cohen’s d was computed for each study (or subsample of a study). The most frequent type of effect size was based on the correlation coefficients. According to Cohen’s (1988) criteria, d < .10 is equivalent to a small effect, d = .25 to a moderate effect and d = .40 to a large effect (numbers are considered in their absolute value).

In our case, the effect size aimed to test the strength of the relationship between SoC and participation. CMA computed 95 % confidence intervals (CI) around the point estimate of an effect size. The Q statistic was utilized to test the homogeneity of the specific set of effect sizes and the significance of moderators (Borenstein et al. 2000; Mullen 1989). CMA utilizes two different models to calculate the overall effect size; the fixed-effect model assumes a common effect size in all studies (true effect size), while the random-effect model assumes a normal distribution of the effect sizes. The random-effect model considers both the variance within each study and the variance between the studies (between-studies variance, T2 [tau-squared]). The weights of each study in the random-effect model are more evenly distributed and less likely to affect the final results. The random-effect model was preferred because the studies were heterogeneous and quite different from each other.

Additionally, the Duval and Tweedie’s (2000a, b) “trim and fill” method was developed to estimate the potential publication bias. Finally, subgroup analyses were performed to test the differences between the political and civic forms of participation and the types of samples, nationalities, sample sizes and participation measures that were utilized. Finally, meta-regressions were performed to explore the roles of age, sample size and percentage of women as possible moderators of the SoC-participation relationship.

7 Results

7.1 Descriptive Characteristics of the Studies

From the 23 scientific papers (published from 1990 to 2012), 34 primary studies were extracted (see Table 3), all of which utilized self-report questionnaires. Fifteen (44.1 %) primary studies were conducted in the USA, 10 (29.4 %) in Italy, 3 (8.8 %) in Israel, 2 each (5.9 %) in Spain and Iran and 1 each (2.9 %) in Australia and the UK. Consideration of the type of data utilized to calculate the effect size indicated that 29 studies (85.3 %) reported correlation coefficients, and 5 studies (14.7 %) reported group means. Civic participation was investigated in 18 studies (52.9 %), political participation in 12 studies (35.3 %) and both forms in 4 studies (11.8 %). Table 4 shows the measures of SoC and participation utilized in the studies. The most frequently utilized measure of SoC was the SCI developed by Perkins et al. (1990; 35.3 % of the studies), while the majority of studies developed ad hoc scales to measure participation (85.3 %). Data on the type of sample showed that 22 studies (64.7 %) were on adults, 5 (14.7 %) on adolescents, 4 (11.8 %) on college students and 1 (2.9 %) on immigrants, military personnel and mentally disabled individuals. The average age of the samples was 35.48 years (SD = 14.66). The average percentage of women was 56.86 (SD = 14.09), and the average size of the samples was approximately 549 subjects.

Table 3 Publications included in the meta-analytic review
Table 4 SoC and participation measures used in the studies

7.2 SoC-Participation Relationship

Table 5 shows the overall results of the meta-analysis based on the correlation coefficients displayed in the studies. Both the fixed model (r = .36; sig. = .00) and the random model (r = .27, sig. = .00) showed a significant, positive and moderate correlation between SoC and participation in a heterogeneous set of studies: Q(33) = 2083.77, p = .00; I2 = 98.42. The procedure “one study removed” was utilized to control the impact of specific single studies, especially studies with very large samples. The analysis confirmed that no study significantly influenced the overall result. This outcome was particularly relevant to the Liu and Besser (2003) study, which utilized a very large sample and elementary measures. When this study was removed, the general correlation coefficient changed to r = .26 (p = .00). Table 6 shows the results of the subsamples meta-analysis and the meta-regressions. No significant difference in relation to SoC and either civic participation (r = .27) or political participation (r = .26) emerged, while the correlation was slightly higher when the two forms were considered together (r = .31). Consideration of the type of sample indicated that the SoC-participation correlation was significant when adults (r = .31) and specific targets (r = .38) were involved (i.e., military personnel, immigrants and mentally disabled individuals) and not significant in the studies with adolescent and college students samples. Moreover, a significant correlation emerged for studies conducted in the USA (r = .32) and other nations (r = .35), but no significant correlation was found for studies conducted in Italy and Israel or studies that recruited small (r = .32) or very large samples (r = .41). Finally, no substantial difference emerged between studies that utilized validated measures of participation (r = .29) and studies that developed ad hoc scales (r = .27). The results of the meta-regression analysis showed that only the age of the sample had a significant, though mild, influence (β = .02) on the SoC-participation relationship.

Table 5 General meta-analytic results
Table 6 Meta-analysis for subsamples and meta-regressions

Finally, a cumulative meta-analysis based on year of publication and sample size (Table 7) was performed to estimate the potential publication bias, based on Egger et al. (1997) linear regression method (t value = 1.76, df = 32) and Duval and Tweedie’s (2000a, b) “trim and fill” method (zero studies “trimmed”, r = .27). The results showed no publication bias.

Table 7 Cumulative meta-analysis based on year of publication and sample size

8 Discussion

In accordance with the literature on the relationship between SoC and participation, the findings confirmed the association between the two constructs and revealed that the magnitude of association is moderate, nearly the same size for civic and political forms of engagement, and quite stable across the studies included in the meta-analysis. The analysis of the population characteristics only showed a relevant effect for age, the sole variable that slightly impacted the association between SoC and participation. The results emphasized that the significance of the association between the two constructs is only true for some of the targeted samples, namely adult populations and highly specific targets such as mentally disabled individuals and immigrants, and some of the nationalities from which the samples were drawn.

Hence, the study results overall showed that when people are involved in civic forms of engagement, protest activities, public deliberation, political campaigning or voting, they also show high levels of SoC. Though the nature of the data does not allow to make any causal inference, and therefore it cannot be assumed that SoC promotes the social and political engagement of citizens, the findings confirmed that feelings of membership, interpersonal sharing and emotional connection, i.e., the psychological dimension of being part of a territorial community, do entertain a significant relationship with active participation, and that this relationship should not be understated in community actions, planning and policies. Also the moderate magnitude of the SoC-participation association indicated that other variables besides SoC are possibly crucial in fostering social and political action. This unsurprising finding is consistent with the most established explicative models elaborated by collective action theorists, who have identified a pool of correlates and predictors of participation both at the individual and collective level (for a review of the psychosocial models, see van Zomeren et al. 2008). While research on SoC and research on political collective behavior have hitherto proceeded separately, this study suggests that an integration of theories and models from both areas might be fruitful.

Another interesting finding of the current meta-analytic review lies in the variations undergone by the SoC-participation binomial across different phases of an individual’s life. These variations indicated that the connection between SoC and community engagement seems to be established in the adult population, not in adolescents and young adults. This outcome agrees with studies that highlighted how SoC increases as individuals reach the central and late stages of their life cycles (Prezza et al. 2001) and somewhat supports investigations concluding that the relationship between SoC and participation is likely to be weaker for young people and the elderly and stronger for adults (Pillemer and Glasgow 2000). The results are consistent with studies that emphasized how adolescents identify their peer groups as the most subjectively important group, rather than the community (Albanesi et al. 2007). Finally, the observed variability across countries suggests that the SoC and community engagement might be more closely interconnected in cultural contexts that assign a special value to active civic and political involvement, indicating that cultural effects can account for this variability. Though not completely unexpected, the general indication emerging from these findings reinforces the need to consider the specificity of the targets of policies and public actions.

In broader terms, this review argues that SoC should be considered as a context-dependent construct whose transferability across settings, populations and measurements is not only inherent, but potentially objectionable (Nowell and Boyd 2010; Hughey et al. 1999). Specifically, SoC seems to be a context-dependent concept at least to the extent to which target characteristics, such as age, nationality and other specific features, are included in the general definition of “context”.

However, final considerations need to be taken with great caution. Indeed, we cannot exclude that, given the disproportion between the adult samples and the adolescent samples, and the US samples and the other nations’ samples, a lower correlation between SoC and participation among adolescents and non-US populations could be due to chance. Further limitations of our results are due to the limited number of studies included in this meta-analysis, the lack of validated measures to assess participation, and the impossibility of considering the moderating effects of any psychosocial variable because too few studies were retrieved for each of the possible moderators.