1 Introduction

Social cohesion is often recognized as a core benefit to any community system. Feeling connected and bound together contributes to a multitude of positive community attributes, such as community resilience (Jewett et al., 2021) and social progress (Borisov & Vinogradov, 2018). Alongside this, social cohesion is commonly thought to contribute to the key community behavior of volunteering. Volunteering is vital to the upkeep of communities and maintenance of social support networks, as most recently highlighted by the COVID-19 pandemic. For example, in the UK it was suggested that up to 12.4 million people (or one in four people) volunteered to provide support to vulnerable individuals (Talk Together, 2021). Despite a surge in volunteering in the UK—and elsewhere—during the pandemic, volunteering rates in the UK are now at a five-year low (Hill, 2023), and trends in social cohesion appear to suggest a decline (Schiefer & Van der Noll, 2017). Understanding the causal connections between social cohesion and volunteering may provide important insights into how these two core elements of community functioning might be sustained.

There is surprisingly limited empirical evidence that directly tests the causal relationship between social cohesion and volunteering. In this article we conduct an analysis of secondary data from two large scale longitudinal social surveys to formally test what causal relationship, if any, exists between social cohesion and volunteering. We assess the widespread assumption that social cohesion and volunteering are inherently linked. In doing so, we aim to provide an empirical foundation for the direction of investment in infrastructures for either social cohesion or volunteering, or indeed, both.

1.1 Defining Social Cohesion and Volunteering

1.1.1 Defining Social Cohesion

Social cohesion lacks a consensually agreed upon definition and conceptualization. The term is often construed and used as a broader catch-all construct (Chan et al., 2006) consisting of many related elements, such as a sense of belonging, various forms of trust (e.g., social, institutional, political) and civic engagement, among others. Being studied in several different fields, such as psychology, sociology, and political science, and through both academic and policy lenses, it is unsurprising there is no agreed definition of social cohesion, and authors have offered a variety of conceptualizations of the construct (for a full review, see e.g., Abrams et al., 2023; Horsham et al., 2023). 

Earlier work on social cohesion from Jenson (1998), and extended by Bernard (1999), conceptualized the construct as comprising five elements that each exist on a continuum: (1) belonging versus isolation, (2) inclusion versus exclusion, (3) participation versus non-involvement, (4) recognition versus rejection, and (5) legitimacy versus illegitimacy. Chan et al. (2006) refined these ideas, developing a two-by-two dimensional framework consisting of an objective versus subjective dimension and an orthogonal horizontal (cohesion between other people) versus vertical (cohesion with the state) dimension. Chan and colleagues identified nine indicators of social cohesion captured by these dimensions: trust in others, a willingness to help others (including those from other social groups), a sense of belonging or identity (all corresponding to the horizontal-subjective category), trust in public figures and confidence in institutions (vertical-subjective), social participation and volunteering (horizontal-objective) and finally political participation such as voting and the presence or absence of inter-group alliances (vertical-objective).

Subsequently, Schiefer and van der Noll (2017) collated multiple definitions of social cohesion from the literature and distilled six dimensions: social relations, identification, orientation towards the common good, shared values, quality of life, and (in)equality. Each dimension embodies sub-components; for example, social networks and trust fall under the umbrella of social relations, whilst psychological well-being and physical health come under quality of life. Abstracting from these identified dimensions, Schiefer and van der Noll (2017) concluded that the core elements of social cohesion were social relations, attachment/belonging, and an orientation towards the common good, and offered an inclusive definition of social cohesion as “a descriptive attribute of a collective, indicating the quality of collective togetherness” (p. 592).

The work of Chan et al. (2006), Schiefer and van der Noll (2017) and others focused largely on theoretical, rather than operational, definitions of cohesion, producing a gap between theory and measurement (Dickes et al., 2010). Subsequently, Bottoni (2018) operationalized cohesion based on Chan et al.’s (2006) work. Bottoni holds that social cohesion exists at micro (relations with other individuals), meso (relations with groups), and macro (relations with wider society) levels, and involves subjective (e.g., interpersonal trust, social support, openness, and institutional trust) and objective (e.g., density of social relations, civic participation, and legitimacy of institutions) components. Using data from 29 countries represented in the European Social Survey, Bottoni found cross-cultural support for the proposed model. However, there remains no clear academic or policy consensus about exactly what the component parts of social cohesion are and how they should be measured.

1.1.2 Social Cohesion and Social Capital

Although the present paper specifically focusses on the relationship between social cohesion and volunteering, it is important to briefly disentangle social cohesion from another closely related construct: social capital. Like social cohesion, social capital is a multidimensional construct that has often been conceptualized and defined in a multitude of ways (Cook, 2022). One of the more widely cited conceptualizations of social capital comes from Putnam (1993), who considered social capital to be the “features of social organization, such as networks, norms and trust, that facilitate coordination and cooperation for mutual benefits” (p. 36). Putnam (2001) argues that these networks and reciprocity between people have value and act as a resource that individuals and groups can utilize, which allows them to better work together and achieve collective goals.

Within social capital, three different forms are typically identified: bonding capital, bridging capital, and linking capital (Granovetter, 1973). Bonding capital refers to the networks between individuals within a group, which are usually emphasized in homogenous groups (Claridge 2018). Bridging capital reflects the networks and connections between different groups, which usually crosse the divide between homogenous groups, (Claridge, 2018). Finally, linking capital refers to the connections between individuals who belong to groups at different levels of power, such as individuals working in a community level charity networking with individuals working in the Government. Crucially, as noted by Chan et al. (2006), social capital differs from social cohesion in its scope. Specifically, social capital is primarily concerned with the networks between individuals and groups. In contrast, social cohesion typically adopts a more societal level focus, although it can also encompass individual or group level relationships (e.g., Bottoni, 2018).

1.1.3 Defining Volunteering

The definitional and conceptual issues surrounding volunteering are comparatively less complex, but also lack consensus. Volunteering is generally regarded as involving helping behaviors that are intentionally offered of one’s own free will, extended over time, and often performed on the behalf of formal organizations (Omoto & Packard, 2016; Snyder & Omoto, 2008). However, definitions differ in emphasis and scope. Some, such as that in Snyder and Omoto’s (2008) Volunteer Process Model, do not make a strong distinction between formal volunteering (conducted on behalf of an organization) and informal volunteering (occurring outside of a structured organizational context, and more akin to general helping behaviors such as collecting shopping for a neighbor; see Pearce & Kristjansson, 2019). Omitting the explicit inclusion of  informal volunteering may lead to underestimates in volunteering rates (Thomson, 2002) through failing to embrace prevalent but smaller scale micro-expressions of volunteering (Paine et al., 2010). Conceptualizing volunteering as behaviors that capture both formal and informal forms of volunteering may therefore result in more complete assessments of the construct.

1.2 The Relationship Between Volunteering and Social Cohesion

Empirical evidence that directly assesses the relationship between volunteering and social cohesion is scarce. In part, this is because many conceptualizations of social cohesion include volunteering as an element (e.g., Chan et al., 2006; Dickes & Valentova, 2012). Such conceptualizations are problematic as they result in tautological definitions, in which the consequences and antecedents of social cohesion are intertwined with the concept itself (Green et al., 2009). Indeed, there has been some recognition of this issue in the social cohesion literature already. For example, Chan et al. (2006) excluded economic factors from their conceptualization of social cohesion as these were considered an environmental precondition that may enhance feelings of social cohesion, rather than being a direct component. Schiefer and van der Noll (2017) excluded inequality from their core elements of social cohesion for similar reasons, as well as excluding quality of life and well-being, which they propose reflects an outcome of cohesive societies (see also Lalot et al., 2022).

In a similar vein, we argue that volunteering is an outcome of social cohesion rather than an integral component of it. A community in which individuals volunteer is not intrinsically cohesive (especially when volunteering is typically motivated by internal prosocial dispositions; Zischka, 2019). However, socially cohesive societies may be more likely to elicit volunteering and assistance to others within a neighborhood. In addition, the act of volunteering itself may promote feelings of identity and togetherness. Consequently, we position social cohesion and volunteering behaviors as separate entities that nonetheless may be causally linked.

The limited evidence that does explore the relationship between volunteering and social cohesion generally points to a positive association between the two. Pearce and Kristjansson (2019) found that perceived social cohesion was associated with both formal and informal volunteering. Lalot et al. (2022) found districts that invested in social cohesion infrastructure (e.g., community programs and initiatives) reported higher levels of volunteering and other civic participation behaviors. A key limitation of much of the evidence on the relationship between social cohesion and volunteering is that it is cross-sectional and correlational (Hill & Stevens, 2010; Miller, 2010), precluding any assessment of potential causality.

Previous research has also observed that many elements of social cohesion, such as a strong social network or a feeling of connectedness and belonging with the community, incentivize volunteering among older members of the population (Lu et al., 2021). This implies a relatively long lead time. In a longitudinal study, Nakamura et al. (2022) found that individuals with higher levels of social contact (e.g., meeting with friends multiple times a week) were significantly more likely to volunteer, an effect that was persistent for up to four years.

There is also some evidence that the act of volunteering increases perceptions of social cohesion. Theoretically, some authors suggest that volunteering contributes positively to social cohesion via the affordance of opportunities for intra and inter-community contact (Putnam, 2000; Ramsey, 2012). Experimental evidence from Dolan et al. (2021) compared individuals who volunteered as part of the UK’s NHS Volunteer Responder program (established during COVID-19 to assist vulnerable individuals) with non-volunteers. They found that volunteers reported higher feelings of social connectedness and community belonging than their non-volunteering counterparts. Likewise, Fox’s (2019) longitudinal study found that volunteering was linked to the later generation of social capital (measured as neighborhood cohesion and involvement in community associations). Taken together, the limited extant evidence suggests that social cohesion and volunteering may be bi-directionally influential.

Finally, because prior research is predominantly cross-sectional or only examines longitudinal relationships in one direction of causality, it is of interest to consider the possible duration of any causal linkages between volunteering and cohesion. Nakamura et al.’s (2022) research suggests causal effects may last, or take, several years to emerge. But the presence of cross-sectional connections suggests that causal effects may be more dynamic and could arise over shorter periods. Thus, it is of interest to consider whether cohesion and volunteering can influence one another over periods of months as well as years.

2 Method

2.1 Data Sources, Samples, and the UK Context

In this paper, we investigate the bi-directional relationship between social cohesion and volunteering across time, analyzing secondary data from two large scale, longitudinal social surveys in the UK; the UKRI Understanding Society Household Longitudinal Study (Institute for Social & Economic Research, 2022) spanning a four-year period, and the Nuffield Foundation’s Beyond Us and Them survey (Belong: The Cohesion and Integration Network, n.d.) spanning a seven-month period. Both datasets are based in the UK, which we focus on primarily for practical reasons. Specifically, there is limited empirical evidence from elsewhere that assesses both social cohesion and volunteering. We therefore focus specifically on the UK context because of the availability of these two large scale longitudinal datasets that included measures of both constructs. Compared to other European countries, the UK has middling to higher levels volunteering (European Commission, 2010; Plagnol & Huppert, 2010), although there is a similar trend of decline in volunteering across Europe (Damian, 2019). In contrast, trends in social cohesion appear much more varied across Europe with social cohesion being stable in some countries but increasing in others (European Commission, 2021).

Understanding Society is a large-scale household panel survey that has occurred annually since 2009. It surveys approximately 40,000 individuals each year, covering a diverse scope of topics which vary each year. We draw on data from waves six (2014–15), eight (2016–17), and nine (2017–18) as these contain measures of both social cohesion and volunteering. A total of 29,034 participants provided data for all three waves, of which 12,921 (44.50%) were male, 16,111 (55.49%) were female, and two did not report their gender. The mean age of participants (taken at wave six) was 39.49 years (SD = 17.58 years).

The Beyond Us and Them survey, developed and conducted by Belong and a research team including the present authors at the University of Kent, was an eight-wave survey conducted during 2020 and 2021. Surveying an average of approximately 5000 people at each wave, the survey assessed topics such as inter-group unity and division, social and political trust, and civic participation during the COVID-19 pandemic. We analyzed data from waves six (December 2020), seven (March 2021), and eight (June 2021) as these included the most consistent set of measures of volunteering and cohesion. Across these three waves there were a total of 2496 participants of which 1028 (41.19%) were male, 1454 (58.25%) were female, and 13 identified as other. The mean age of participants was 47.98 years (SD = 16.39 years).

2.2 Measures

2.2.1 Understanding Society

The volunteering and social cohesion measures from Understanding Society were not simultaneously present across all waves. Wave six of the survey contained both measures of volunteering and social cohesion. However, wave eight only contained measures of volunteering and wave nine only contained measures of social cohesion. The measures used for each construct are described below.

2.2.1.1 Social Cohesion

Social cohesion was measured using Buckner’s (1988) Neighborhood Cohesion Instrument. This consists of eight items that assess three sub-components of social cohesion: neighboring (‘If I needed advice about something I could go to someone in my neighborhood’; ‘I borrow things and exchange favors with my neighbors’; and ‘I regularly stop and talk with people in my neighborhood’), psychological sense of community (‘I feel like I belong to this neighborhood’; ‘The friendships and associations I have with other people in my neighborhood mean a lot to me’; ‘I would be willing to work together with others on something to improve my neighborhood’; and ‘I think of myself as similar to the people that live in this neighborhood’), and attraction to the neighborhood (‘I plan to remain a resident of this neighborhood for a number of years’). Items were measured on a 5-point Likert scale ranging from 1 (Strongly Disagree) to 5 (Strongly Agree).

2.2.1.2 Volunteering

Volunteering was assessed with a single binary response item: “In the last 12 months, have you given any unpaid help or worked as a volunteer for any type of local, national or international organization or charity?”. Responses were coded as 0 = “No” and 1 = “Yes”.

2.2.2 Beyond Us and Them

In contrast to the Understanding Society survey, the Beyond Us and Them survey employed a wider set of items to assess several components of cohesion. We selected for analysis those that most closely represented Chan et al.’s (2006) subjective dimensions of horizontal and vertical social cohesion.Footnote 1

2.2.2.1 Horizontal Cohesion

Horizontal cohesion was measured using three survey items and one composite index score. Survey item one was ‘How much do you feel that you belong to your neighborhood?’, akin to the ‘sense of belonging or identity’ component of Chan et al.’s (2006) model. Item two was ‘Would you say that most people in your neighborhood can be trusted?’, akin to social trust. Item three was ‘How much do you feel a responsibility to try to improve your neighborhood?’, akin to a willingness to help others. Each of these items was measured from 1 (Not at all) to 5 (Very much so). As these items did not include an element of a willingness to help people from ‘other’ social groups that is present within Chan et al.’s model, we additionally included a composite score of feelings towards people from various social groups that was rated on a feeling thermometer from 1 (Very cold feelings) to 11 (Very warm feelings). The feeling thermometer assessed participants’ feelings towards several groups, specifically: wealthy people, poor people, older people, young people, legal immigrants, asylum seekers, black people, Muslim people, seasonal workers, and illegal immigrants.Footnote 2 Responses were aggregated into a standardized (z-score) index of people’s general openness towards people from different social groups.

2.2.2.2 Vertical Cohesion

Vertical cohesion was measured using five items. Three of these items assessed the political trust element of Chan et al.’s (2006) model: ‘Most members of the UK Parliament are honest’; ‘I trust my local member of parliament to represent the interests of all communities across the constituency’; and ‘Politicians are mainly in politics for their own benefit and not for the benefit of the community’ (Reverse coded). Two items assessed the confidence in political and social institutions component of Chan et al.’s model: ‘I believe the UK Government is handling the causes and consequences of the pandemic competently’ and ‘I believe my local council (i.e., town or city or district) is handling the causes and consequences of the pandemic competently.’ All items were measured on a five-point Likert scale, from 1 (Strongly disagree) to 5 (Strongly agree).

2.2.2.3 Volunteering

Volunteering was measured using a question that asked participants whether or not they had voluntarily engaged in any of several forms of pro-social action during the last month. From this measure, we created a binary variable for participants who had (1) or had not (0) volunteered.

2.3 Analytic Strategy

For each dataset, we first conducted a confirmatory factor analysis for the social cohesion measure to assess both the validity and temporal invariance of the measure. Measurement invariance assesses whether the measures capture the same constructs similarly at each time point, and consists of testing several increasingly constrained models to assess whether different components of a measurement model are equal across time. Broadly, there are four models to be tested in measurement invariance (Mackinnon et al., 2022): configural, metric, scalar, and residual invariance. Configural invariance assesses whether the basic structure of the model is consistent across time. Metric invariance assesses whether the representativeness of each item remains consistent over time. Scalar invariance assesses whether mean scores on each item remain consistent over time and finally, the most restrictive model of residual invariance assesses whether the influence of external factors on the model are equal over time. Each progressive model is taken as stronger evidence of the invariance of the measure over time. Crucially, it is necessary to establish invariance to accurately measure changes in a variable over time. The absence of measurement invariance means that the appearance of change (e.g., mean levels of a variable increasing over time) may be due to changes in the interpretation of the measure rather than reflecting true change. Although residual invariance indicates a fully invariant measure, many authors consider the requirement of residual invariance too stringent for practical use, as constraining external influences to be equal is not necessary for assessing mean levels of change (Putnick & Bornstein, 2016; Vandenberg & Lance, 2000).

We then conducted a longitudinal cross-lagged structural equation model in which social cohesion and volunteering predicted each other across time. We focus on a longitudinal analysis in order to facilitate causal conclusions about the relationship between volunteering and social cohesion. Although other methods, such as experimental research designs, allow for stronger inferences of causality (as they allow for the control of extraneous influences), they are not applicable for the present analysis of secondary data. Given the dearth of causal evidence, we utilize the longitudinal analysis as an important first step in providing causal evidence of the relationship between social cohesion and volunteering.

3 Results

3.1 Confirmatory Factor Analysis and Measurement Invariance

3.1.1 Measurement Invariance

To assess the robustness of the social cohesion models in the Understanding Society and Beyond Us and Them datasets we first tested for measurement invariance across waves. For brevity we summarize the outcomes of the measurement invariance analysis below. A full technical description of the analysis is available in the supplementary materials. All models were estimated using the lavaan package in R (Rosseel, 2012).

The measurement model of social cohesion for the Understanding Society dataset is displayed in Fig. 1. For this dataset, we concluded that the model was residually invariant. For determining model fit, we utilize the common fit indices of the comparative fit index (CFI), the root mean square error of approximation (RMSEA), and the standardized root mean square residual (SRMR). The residual invariance model showed good fit to the data, χ2(119) = 13903.95, p < 0.001, CFI = 0.936, RMSEA = 0.064, SRMR = 0.047, and the CFI and RMSEA metrics did not substantially differ from the less restrictive scalar model (ΔCFI = 0.001, ΔRMSEA = 0.002; see Chen, 2007; Cheung & Rensvold, 2002; Rutkowski & Svetina, 2014). Consequently, we concluded that the social cohesion measure was fully invariant across time. The factor loadings for the final residual invariance model are displayed in Table 1.

Fig. 1
figure 1

Measurement invariance model for social cohesion (Understanding Society). Note: ovals represent latent factors and rectangles represent observed variables. In all invariance models, the residual covariance between observed items was constrained to be equal across waves. This model achieved residual invariance in which the factor loadings, item intercepts, and residual variances of items were constrained to be equal across waves

Full size image
Table 1 Factor loadings for the residual invariance social cohesion model (Understanding Society dataset)
Full size table

The proposed measurement model of social cohesion for the Beyond Us and Them dataset is presented in Fig. 2. For this dataset we concluded that the social cohesion measure was metrically invariant. The metric invariance model showed good fit to the data, χ2(318) = 1511.93, p < 0.001, CFI = 0.972, RMSEA = 0.039, SRMR = 0.056, and the CFI and RMSEA metrics did not substantially differ from the less restrictive configural model (ΔCFI = 0.002, ΔRMSEA < 0.001). However, the more restrictive scalar model provided significantly worse fit than the metric model, Δ χ2(18) = 690.20, p < 0.001, and there was a sizable change in CFI (ΔCFI = 0.015). We conservatively concluded that the model was not scalar invariant. We therefore adopted the metric invariant model for analysis. The factor loadings for the final metric invariance model are displayed in Table 2.

Fig. 2
figure 2

Measurement invariance model for horizontal and vertical cohesion (Beyond Us and Them dataset). Note: items are numbered in the order they appear in text. Ovals represent latent factors and rectangles represent observed variables. In all invariance models, the residual covariance between observed items was constrained to be equal across waves. This model achieved metric invariance in which the factor loadings of items were constrained to be equal across waves

Full size image
Table 2 Factor loadings for the metric invariance social cohesion model Beyond Us and Them dataset)
Full size table

3.2 Longitudinal Analysis

3.2.1 Understanding Society

To assess the relationship between volunteering and social cohesion within the Understanding Society dataset we estimated a structural equation model which included the full residual invariance measurement model as described above, with the addition of cross-lagged pathways between the latent social cohesion variables and the binary volunteering items over time. Due to differences in the items measured at different time points, social cohesion and volunteering measured at wave six was cross-lagged onto cohesion measured at wave nine and volunteering measured at wave eight. Participants’ age, gender, and race (dichotomized as White vs. Non-White) were included in the model as covariates.Footnote 3 The structural component of this model is displayed in Fig. 3. The overall model provided good fit to the data; χ2(193) = 20026.34, p < 0.001, CFI = 0.95, RMSEA = 0.07, SRMR = 0.06. As shown in Fig. 3, the cross-lagged paths between volunteering and social cohesion were significant at all waves.

Fig. 3
figure 3

Longitudinal cross-lagged relationships between social cohesion and volunteering (Understanding Society dataset)

Full size image

3.2.2 Beyond Us and Them

To assess the relationship between horizontal cohesion, vertical cohesion and volunteering in the Beyond Us and Them dataset, we similarly estimated a structural equation model. This included the metric invariance measurement model as described above, as well as the structural model in which the cross-lagged pathways between horizontal cohesion and volunteering and vertical cohesion and volunteering were estimated.Footnote 4 Participant’s age, gender, race (dichotomized as White vs. Non-White), political orientation, subjective socioeconomic status, and income were included in the model as covariates. The structural component of the model is displayed in Fig. 4. The overall model provided marginally adequate fit to the data, χ2(533) = 7180.894, p < 0.001, CFI = 0.87, RMSEA = 0.08, SRMR = 0.08.

Fig. 4
figure 4

Longitudinal cross-lagged relationships between horizontal and vertical cohesion and volunteering (Beyond Us and Them dataset)

Full size image

As shown in Fig. 4, the cross-lagged pathways between horizontal cohesion and volunteering (and vice versa) were significant across all waves. However, the relationship between vertical cohesion and volunteering was unidirectional: only the pathways from volunteering to vertical cohesion (shown with solid lines) were significant, and not the other way around (shown with dashed lines).

In summary, we find that in the four-year Understanding Society data there is consistent evidence that social cohesion predicts subsequent volunteering, and vice versa. The two effects appear to be of comparable size. In the seven-month Beyond Us and Them data, we find that horizontal cohesion predicts volunteering and vice versa; however, whereas volunteering predicts vertical cohesion, vertical cohesion does not predict volunteering. Moreover, the strength of relationships involving horizontal cohesion are substantially larger than those involving vertical cohesion.

4 Discussion

In this paper we sought to provide empirical evidence for the causal relationships between social cohesion and volunteering. Through a secondary analysis of data from two large scale social surveys in the UK, we found consistent evidence for a bi-directional, causal relationship between social cohesion and volunteering. Specifically, people who felt more cohesive were more likely to volunteer at later time points, and people who volunteered were more likely to feel socially cohesive later in time. However, this bi-directional relationship seems to apply only to horizontal forms of social cohesion. Vertical cohesion, which reflects cohesion with institutional authority, had a unidirectional relationship to volunteering. More concretely, people who volunteered felt more trusting of politicians and confident in government institutions at later time points, but these feelings of trust and confidence did not in turn affect future volunteering.

It is plausible that this unidirectional effect may be due to the differing scopes of volunteering and vertical cohesion. Specifically, volunteering is usually focused on local community issues, whereas vertical cohesion (as operationalized here) concerns trust in national institutions and Government. There appears little reason to assume that trust in national institutions would impact the likelihood of volunteering in local community contexts. It may be the case that more localized levels of vertical cohesion (e.g., trust in local councils) may exhibit a bi-directional relationship, but the present data indicate that nationally focused levels of vertical cohesion do not impact volunteering.

Overall, the present findings are consistent with earlier evidence that social cohesion and volunteering are linked. However, previous research supporting the relationship between social cohesion and volunteering (e.g., Lalot et al., 2022; Pearce & Kristjansson, 2019) has largely been correlational in nature (Hill & Stevens, 2010; Miller, 2010). Causal evidence, such as that from longitudinal research designs (e.g., Fox, 2019; Nakamura et al., 2022), has been far more limited and restricted either to specific samples (e.g., elderly populations) or to exploring only unidirectional pathways (e.g., volunteering to social cohesion). The present research addresses these limitations of earlier research by explicitly assessing the bi-directional relationships between volunteering and social cohesion, contributing a critical extension to previous efforts by supporting the plausibility of the contention that social cohesion and volunteering are indeed causally related.

The present findings also highlight the need for a more precise and comprehensive conceptualization of social cohesion. Previous theoretical models of social cohesion have positioned volunteering as a constituent part of cohesion (Chan et al., 2006). However the present study challenges that conceptualization and demonstrates the importance of positioning social cohesion and volunteering as separate entities in order to study the relationships between the two. For example, our approach enabled us to identify that horizontal and vertical forms of cohesion have different relationships with volunteering. The present evidence also reinforces previous calls for research to conceptualize social cohesion in a manner that differentiates it from its antecedents and consequences (e.g., Schiefer & van der Noll, 2017).

Identifying the causal relationship between social cohesion and volunteering can also help inform investment strategies in these areas. For one, the causal relationship should affirm confidence in existing government investments into social cohesion and volunteering infrastructure (Abrams et al., 2020). The bi-directional relationship indicates that investment in either social cohesion or volunteering initiatives will be beneficial in promoting more cohesive and inclusive societies. However, given that volunteering additionally promoted feelings of vertical cohesion, it is likely that investment that supports volunteering will have broader benefits beyond cohesion within local communities.

A further important conclusion from the present research is that efforts to support either social cohesion or volunteering can yield gains in both a relatively short time frame (a few months) and a longer one (a few years). This means that fundholders can have confidence that investments in promoting strategies to support social cohesion, volunteering, or both will ‘pay off’ within a politically relevant timeframe as well as the longer term.

Notwithstanding the strengths and contributions of this research, the present work also has some limitations. First, the measures available for study were limited, requiring us to explore different operationalizations of cohesion across the two datasets. Although the models are conceptually similar and show consistent findings, there are important differences between the two sources of evidence. Most notably, the Understanding Society model did not include an element of vertical cohesion, and our having to use different operationalizations of cohesion reflects the broader limitation in the literature that there remains no consensual definition or conceptualization of cohesion, and consequently little consensus in its operationalization (Horsham et al. 2023). Additional research is needed to unify and standardize the measurement of social cohesion in order to support progression of research in the field.

Secondly, the social cohesion measure in the Beyond Us and Them data was only metric invariant, which limits confidence in its stability over time. However, this may be inconsequential as the constraining of item intercepts in the scalar model accounts for grand mean level changes over time, and does not affect evidence for the relationship between constructs. Consequently, the metrically invariant scale was appropriate for the purpose of the current study. Additionally, the lack of a fully invariant measure of social cohesion may be expected due to the COVID-19 context in which the Beyond Us and Them data was collected. During this period several elements of social cohesion, such as political trust, were more volatile (Davies et al., 2021) and therefore some level of variation in the mean levels of social cohesion constructs over time was to be expected.

Finally, whereas the items used to measure cohesion were quite multi-faceted and involved a range of items, those of volunteering were more limited. This may mean that the strength of relationships between volunteering and cohesion have been somewhat underestimated. Despite any such limitations of measurement it remains clear that these relationships do exist and persist over time.

5 Conclusion

The present research provides an important basis for the argument that there are causal relationships between social cohesion and volunteering. The findings affirm previously held assumptions in the field that these relationships may be bi-directional. They therefore provide reassurance that investment in both social cohesion and volunteering infrastructure and initiatives will be beneficial in bolstering communities. Moreover, the evidence suggests that these effects can be observed both over shorter periods (months) and in the longer term (years). Therefore, this present evidence contributes a substantial first step in establishing an evidential basis for policy. Further conceptual work is needed to establish a broadly accepted definition and approach to measuring social cohesion. Further empirical work is also needed to understand how social cohesion, volunteering, and their relationship may be underpinned, attenuated or augmented by other societal and individual factors.