Finds documents with both search terms in any word order, permitting "n" words as a maximum distance between them. Best choose between 15 and 30 (e.g. NEAR(recruit, professionals, 20)).
Finds documents with the search term in word versions or composites. The asterisk * marks whether you wish them BEFORE, BEHIND, or BEFORE and BEHIND the search term (e.g. lightweight*, *lightweight, *lightweight*).
The article delves into the crucial role of social networks in facilitating protest participation in semi-authoritarian regimes. It highlights the importance of social network topology, specifically focusing on characteristics such as network size, density, closure, brokerage, and centrality. Using an empirical case study of ecological protests in Russia and data from the social network VKontakte, the research employs Bayesian structural equation modeling to test the relationships between these network features and protest participation. The study finds that larger networks, lower density, and higher brokerage positions significantly increase the likelihood of protest participation. Notably, the results also reveal a surprising negative correlation between network centrality and protest involvement, suggesting that highly central individuals are less likely to engage in protests. The findings offer valuable insights into the complex dynamics of protest mobilization in semi-authoritarian contexts, emphasizing the strategic importance of network positioning and support systems.
AI Generated
This summary of the content was generated with the help of AI.
Abstract
This study uses the case of ecological protests in the Russian Federation and the social network of 903 263 VKontakte users to investigate the unique characteristics of protesters’ social networks and assess the effects of network topology on the likelihood of participation. By applying Bayesian structural equation modeling to the data, the study finds that social network topology plays a more nuanced and complex role in predicting protest participation beyond simple exposure to information. Notably, individuals situated as brokers in lower-density and higher-closure networks exhibit a higher likelihood of participation, while central nodes participate less. The results highlight the importance of strategic positioning as brokers for increasing the likelihood of participation. The paper discusses these and other findings, contributing to a deeper understanding of the interplay between social network structures and protest participation.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Introduction
Protest participation, defined as active involvement in public demonstrations expressing dissent, grievances, or support for a particular cause or issue1, can be an important means for individuals to challenge the existing power structures and demand greater accountability. It is particularly important in semi-authoritarian regimes where traditional forms of participation (e.g., voting) face challenges such as fraud and manipulation (Frantz, 2018).
There are numerous examples of successful protests in such regimes (e.g., Amnesty International, editorial board, 2019; The Guardian, Agencies in Addis Ababa, 2018). In some cases, these regimes respond to citizens’ concerns to prevent the emergence of more organised and potentially threatening forms of dissent (Lindstaedt & Frantz, 2019). This is more likely when there is substantial public support, expressed in the number of protest participants (McAdam, 1982; Tilly, 1978).
Advertisement
Consequently, understanding the factors that influence the likelihood of protest participation has long been of interest to scholars.
This article contributes to this body of work by specifically focusing on the importance of social networks in explaining protest participation.
While the role of social networks in explaining protest participation is widely acknowledged, research directly investigating the impact of social network topology (defined as the structural arrangement of connections within a social network) on participation remains limited. Previous studies that have made substantial contributions in this field primarily focused on one of two participation mechanisms. Specifically, they suggested that social networks play a significant role in shaping protest participation by (1) providing access to recruitment (a.k.a. mobilisation) and (2) serving as sources of motivation and support for political action.
The work of Steinert-Threlkeld (2017) and Leenders (2012), for example, falls into the first category of studies2. These scholars found that peripheral nodes are especially important in mobilising larger protests (Steinert-Threlkeld, 2017), and that “peripheral” social networks facilitate protests by connecting individuals from different backgrounds and social strata (Leenders, 2012).
Advertisement
On the other hand, Putnam (2000) and McAdam and Paulsen (1993) focused on the motivations for participation, examining how social network structures influence identity, social capital, and solidarity3. These scholars emphasised that closely-knit networks promote in-group trust (Putnam, 2000), a strong subjective identification with a particular identity (McAdam & Paulsen, 1993) and other social norms that enhance the likelihood of participation.
Despite a substantial body of literature on the topic, some crucial questions remain unanswered. What characteristics define advantageous networks boosting protest participation? How do the topological features of individual networks differ between protesters and non-protesters? Are most protesters found in the periphery, acting as brokers, or among the hubs (nodes with many connections)? These questions largely remain open because answering them necessitates the comparison of properties between the individual networks of protesters and non-protesters4, the data that are difficult to collect.
Furthermore, previous studies in this field have predominantly focused on democratic contexts. However, protest participation in non-democratic settings may significantly differ. For instance, the size of protests in a non-democratic context is substantially influenced by factors such as protester and state violence. As protester violence increases, protest sizes tend to decrease significantly because engaging in protests becomes less appealing, while state violence intensifies (Steinert-Threlkeld et al., 2022). Essentially, the cost of participating in protests in non-democratic states is significantly higher compared to democratic contexts, largely due to state repression.
Given the existing research gaps, there is a need for more direct and empirical investigations into the connection between social network topology and protest participation, particularly in non-democratic settings. The rise of digital technologies has presented both: opportunities and challenges for protest participation. Social networks can serve as sources of information, motivation, and support for political action. However, they can also exacerbate socioeconomic disparities by limiting the access of individuals with low socioeconomic status (SES) to these resources.
This study aims to answer the question: What is the effect of social network topology on protest participation? In particular, the study investigates the unique characteristics of protesters’ social networks and assesses the effects of network size and topology on the likelihood of protest participation. Using an empirical case of ecological protests in the Russian Federation and the social network of VKontakte users, the study applies Bayesian structural equation modeling to test the relationships between relevant variables. The results of this study provide critical insights into the role of social network topology in accessing political decision-making.
The rest of the article is structured as follows. Section “Conceptualisation” reviews the existing literature on the relationship between social network topology and protest participation, emphasising the expected relationships between the variables of interest. Section “Data and Methods” outlines the research design, including describing the empirical case and the application of Bayesian structural equation modeling to the data. Section “Results” presents the study’s results, detailing the distinctive characteristics of protesters’ social networks and the effects of social network size and topology. The final section provides a discussion of the results, highlighting the study’s implications and suggesting future research routes in the field.
Conceptualisation
The Importance of Social Network Structure
The rising popularity of social media platforms and their increased use for coordinating protest activities have sparked growing academic interest in the role of social networks in protest participation. Scholars such as Steinert-Threlkeld (2017); Schwenzfeier and Settle (2020); Larson et al. (2019); Lockwood (2022); Atwell and Nathan (2022) have highlighted that an individual’s position within social networks and structural characteristics of these networks significantly influence the likelihood of participation.
Much of the research in this field (e.g., Centola, 2010; Granovetter, 1973), explains this correlation by the fact that certain social network structures provide more allies, thus offering greater social support and serving as stimuli for protest participation. In other words, these studies suggest that the relationship between social network structures and participation is indirect, mediated by the availability of social support.
However, other studies (e.g., Burt, 1992; Putnam, 2000) provide a more intricate view of this relationship, suggesting that some network structures are more likely to facilitate mobilization and foster important resources for participation, such as social trust and political efficacy.
In other words, previous research indicates two key associations: A.) the overall influence of social network structure on protest participation, and B.) its indirect impact through the availability of potential allies (fellow protesters), both of which are examined in this study. Specifically, prior studies have identified the following social network characteristics that impact participation: network size, density, closure, as well as the centrality and brokerage roles of individuals. These characteristics are suggested to correlate with protest participation and with social support acting as a mediating factor.
The Effect of Network Size
Numerous studies have underscored the significance of an individual’s social network size (the number of individuals a person has within their social network) in influencing their involvement in protesting. A notable example is the research by Steinert-Threlkeld et al. (2015), who analysed nearly 14 million geolocated tweets across 16 Middle Eastern and North African countries. In this study, increased coordination of messages on Twitter corresponded to increased participation in protests the following day. Another study by Larson et al. (2019) compared the networks of participants and non-participants in the 2015 Charlie Hebdo protest in Paris. This research also demonstrated that protesters, on average, had more extensive networks and more connections with fellow protesters.
This positive correlation between personal network size and participation can be explained by two factors. Firstly, larger networks provide greater access to political information, as suggested by Granovetter (1973). Secondly, according to social capital theory (Putnam, 1993), individuals in larger networks tend to possess higher social capital, including resources like social trust and political efficacy, which positively influence protest participation (Verba et al., 1995).
Based on this, the first hypothesis is formulated as follows: the size of an individual’s social network is strongly associated with an increased likelihood of protest participation (H1a).
Moreover, Granovetter (1973) proposed that larger networks are more likely to be populated by potential allies. Therefore, beyond the direct association between network size and participation, there exists an indirect effect: larger networks are associated with more substantial support networks, further enhancing the likelihood of participation (H1b).
The Effects of Network Density and Closure
Network density and network closure are related concepts and, in a nutshell, measure the extent to which individuals in a network are connected. Network density is the number of connections in an individual’s social network to the number of possible connections in this network. Network closure can be defined as the probability that the contacts of a studied individual are connected. Higher levels of density or closure suggest that individuals in a network are more closely connected.
Previous research on the direct effects of network density and closure on protest participation is scarce. Many studies (e.g., Klandermans et al. 2008; McAdam et al. 2001; Putnam, 2000) examined network density and closure as facilitators of the spread of information and sources of motivation, which can, in turn, increase protest participation. This literature generally suggests a positive effect of social-network density and closure on participation. For example, Leenders (2012) argued that dense networks provide a sense of solidarity (motivation). Putnam (2000) suggested that tight-knit networks foster in-group trust and other social norms by facilitating sanctions. In line with this, it is expected that the likelihood of protest participation is positively associated with the density (H2a) and closure (H3a) of an individual’s social network.
Beyond the direct effect, an indirect effect has been suggested. McAdam (1986) emphasised the importance of tight-knit networks for mobilising individuals and building collective actions by providing social support. In the context of protests, high-density networks are expected to facilitate the formation of local network coalitions, which are instrumental in orchestrating high-cost activities like protesting (Centola, 2010, 2013). Essentially, tightly connected networks provide substantial social support, reinforcing individuals’ participation. Therefore, beyond the direct links between network density and network closure with protest participation, there exists an indirect influence: networks of higher density (H2b) and closure (H3b) are associated with more substantial support networks, further enhancing the likelihood of participation.
The Effects of Brokerage
A structural hole is a gap between two individuals or groups of individuals in a network.
Burt (1992) argued that networks populated by many structural holes, with few links between clusters of nodes, create opportunities for individuals to act as brokers or intermediaries between otherwise disconnected groups. Due to structural holes, information shared within clusters is often complementary rather than redundant. Burt claimed that brokers are more likely to have access to diverse sources of information and are, thus, more likely to get recruited into political activities. In contrast, individuals embedded in dense, redundant networks may encounter less novel information or political opportunities.
In general, the effects of structural holes and brokerage have been previously studied in the context of mobilisation into political actions. For example, Heaney and Rojas (2014) have found that organisations occupying brokerage positions, organisations with hybrid identities, are more successful in social movement mobilisation. Similarly, Diani and McAdam (2003), studying Italian environmentalism of the 1980s, found that brokerage played a significant role in shaping the emergence and development of new social movements.
Based on this, one would expect a direct positive relationship between an individual’s brokerage position and their likelihood of protest participation (H4a). This relationship is explained by the access to diverse information and increased mobilisation.
In addition to the direct impact, an indirect effect can be considered. While—to the best of my knowledge—there’s no existing research directly examining the correlation between the size of an individual’s support network and their brokerage position, such a relationship can be inferred. Burt (1992) has described brokers as being representative of several clusters (tightly-knit communities). When compared to nodes with low importance, brokers are more likely to have extensive support networks. In contrast, low importance nodes either belong to a single community, several small-scale communities, or may not be connected to any cohesive clusters at all, resulting in a weaker support system. In line with this, I hypothesise that beyond the direct link between a person’s brokerage position and their protest participation, there exists an indirect influence: the brokerage position of an individual is associated with a more substantial support network, further enhancing the likelihood of participation (H4b).
The Effect of Centrality
In essence, centrality, often measured as the number of direct contacts in a personal network, is closely related to the size of a personal network. Influential individuals, with larger personal networks, generally have more access to information and support: thus, mobilisation (Burt, 1992).
To a large extent, previous empirical research has focused on the effects of betweenness centrality5 (often understood as a measure more closely related to brokerage (Everett & Valente, 2016)) on participation (e.g., Diani & McAdam, 2003; Lockwood, 2022) or on the importance of central nodes for movement coordination and mobilisation (e.g., Fabrega & Sajuria, 2013; González-Bailón et al., 2011).
Literature directly studying the effect of centrality on participation presents conflicting results. For example, Atwell and Nathan (2022) found a positive correlation between centrality and donations to a local public good: where more influential individuals made larger donations. Fernandez and McAdam (1988) also found a weak positive correlation between centrality and participation in the Freedom Summer Project, highlighting that “[...] there is a tendency for the most prominent people to be participants” (Fernandez & McAdam, 1988, p. 368). In contrast, studying the surge of mobilisations in Spain in 2011, González-Bailón et al. (2011) found no associations between network position and participation, emphasising that “[...] participants do not have a characteristic network position; they are instead scattered all over the network” (González-Bailón et al., 2011, p. 4). Fernandez and McAdam (1988) also mentioned that despite a weak positive correlation between centrality and participation, there are many exceptions: peripheral individuals also often participate politically.
To summarize, previous research indicates that influential individuals might not necessarily be more inclined to engage in protests. As a result, H5a is posed as follows: there is no direct impact of an individual’s centrality on their likelihood of protest participation (H5a). Still, central individuals are expected to have larger networks, including more extensive support networks. Thus, there is an indirect influence of centrality on the likelihood of participation, as outlined in H5b: an individual’s centrality is associated with a more substantial support network, enhancing the likelihood of participation (H5b).
Figure 1 summarises the expected relationships between the variables of interest.
Fig. 1
Conceptual model of the relations among variables. Notes: A (+) symbol indicates a positive relationship between two variables
To test for the hypothesised relationships, data on ecological protests in Russia from 2018 to 2020 were collected. Russia during this period serves as a representative case of a semi-authoritarian regime. The political attitudes in Russia, such as political interest and knowledge, are similar to those in other semi-authoritarian regimes, as depicted in Fig. 2. Political interest in Russia—like in other semi-authoritarian countries—is relatively high, yet not as high as in authoritarian regimes and democracies. Russia also resembles other semi-authoritarian countries in terms of using political-information sources.
Fig. 2
Political interest and sources of information by the regime. Notes: Boxplots represent the distribution of corresponding mean values by the political regime. The dot inside the boxes is the mean value. The horizontal dotted line is the mean value in the Russian Federation. Values range from 1 (low interest/source usage) to 4/5 (high interest/source usage). Data source: “World values survey: Round seven” (Haerpfer et al., 2020)
Furthermore, akin to other semi-authoritarian governments, the Russian government employs repression (Amnesty International, editorial board, 2021) and surveillance (Human rights watch, editorial board, 2021) to control protests. This results in significantly higher costs of protest participation in Russia compared to democracies. Individuals who choose to participate face the real possibility of arrest, injuries due to police brutality, or even imprisonment (Amnesty International, editorial board, 2021), all of which deter potential participants.
Moreover, the state exercises strict control over the media and the Internet (Human rights watch, editorial board, 2021), affecting the dissemination of information and the coordination of protests. Additionally, the freedom of assembly is generally restricted, as protest organisers are required to obtain permission from authorities (Human rights watch, editorial board, 2012), often denied under the pretext of unsuitable dates or locations. Notably, authorities have been known to suggest alternative, less visible protest sites, possibly to downplay the scale of dissent (Amnesty International, editorial board, 2021).
Despite these constraints, protests remain relatively common. As illustrated in Fig. 3, participation in unconventional activities such as petition signing, e-petition signing, and protesting in Russia is more frequent than in autocracies but less prevalent than in democracies. Moreover, there are instances where protesting proves successful—for example, the protest wave studied in this article. As Lindstaedt and Frantz (2019) explained, semi-authoritarian regimes sometimes respond to citizens’ concerns to prevent the emergence of more organised and potentially threatening forms of dissent.
Fig. 3
Political participation by the regime. Notes: Boxplots represent the distribution of corresponding mean values by the political regime. The dot inside the boxes is the mean value. The horizontal dotted line is the mean value in the Russian Federation. Values range from 1 (would never participate) to 3 (have participated). Data source: “World values survey: Round seven” (Haerpfer et al., 2020)
In this context, studying protest participation in Russia serves as a representative case for understanding how protests unfold in semi-authoritarian regimes. This case allows controlling for the impact of heightened participation costs and information manipulation.
The period from 2018 to 2020 is unique for studying Russian protests. Prior to February 2022, continuous democratic backsliding raised public discontent that reached its peak in the time period covered by this study (Carnegie endowment for international peace, research group, 2022). However, in 2022, protest volume substantially reduced due to higher risks and stricter control over information shared on social media (Troianovski & Safronova, 2022).
To examine the role of a personal social network in protest participation, this study uses a network of VKontakte (VK) users informed about Shiyes protests. Shiyes protests opposed the construction of the largest landfill in northern Russia. The landfill was planned to be built near the Shiyes railway station in the Arkhangelsk Oblast. It was intended to mainly store household waste and industrial by-products from Moscow, which raised environmentalists’ discontent, and led to a series of protests from 2018 to 2020 in more than 30 regions of Russia (Protest229.ru, activists, 2019). The protests were mainly coordinated via social networking sites (SNSs) such as VK, Telegram, and WhatsApp (Poupin, 2021).
The Shiyes protest wave serves as a compelling example of successful mass protesting in semi-authoritarian regimes. Ecological protests offer noteworthy cases for investigation because, like political protests, they address highly visible issues that motivate people to participate. In recent years, ecological concerns have gained prominence in Russia, leading to a substantial growth in social movements dedicated to these issues (Davydova, 2021). Unlike political protests, ecological protests generally face lower state repression since these protests do not directly demand a change in power. Consequently, while the authorities monitor information shared through online channels of communication, it is not entirely suppressed.
The Shiyes protest is notable for achieving policy changes. This success underscores the movement’s effectiveness in raising awareness regarding ecological issues and mobilising large groups of people. While primarily addressing regional policy, the wave of protests was nationwide since the construction of the landfill would lead to the pollution of the White Sea, affecting a significant part of the country. The protests occupied the whole of European Russia and were strongly supported by more than 95% of the Arkhangelsk Oblast’s citizens (Britskaya, 2019).
The fact that mobilisation remained large-scale despite personal risks makes this case particularly valuable for investigation. Studying mass protests allows for the examination of protest participation across a diverse and representative cross-section of society. These protests encompass not only dedicated activists but also individuals who may not have previously been involved in political or social movements. This allows a comprehensive analysis of the factors influencing the participation of ordinary citizens.
Additionally, the protest wave’s success suggests the existence of a core group that systematically mobilises people, making it a pertinent case for studying mobilisation on social media.
VK data are used for two reasons: first, it is the largest SNS in Russia and was the most used platform for community-building during the Shiyes protest wave (Poupin, 2021). Secondly, VK allows users to create and promote events, which enables distinguishing between users who were planning to participate in the protests and those who were not.
Data Collection and Description
This study focuses on the properties of a friendship network on VK, a social networking site similar to Facebook, where users can create personal pages, connect with others, communicate publicly or privately, and form various groups. The network comprises VK users who could potentially receive information about one or more of the 10 Shiyes protests against the landfill construction held from September 23, 2018, to February 15, 2020. Users who fall into this category are defined as friends and followers of event participants, organisers, and information sharers, as well as members of groups that organised or shared information about an event.
Data collected via the official VK API include 903,263 users connected by undirected edges6 representing friendship ties, with a total of 121 882 747 links. Among informed users, only 0.23% (2041 people) participated in the protests. Each user has attributes such as membership in a group-organiser or any activist group, including 17 Shiyes groups (see Table 4). This information allowed accounting for prior activism, which was previously reported as an essential factor in explaining protest participation (McAdam, 1986; Jasper & Poulsen, 1995; Brady et al., 1999).
Participants are defined as those who self-registered for the protests, self-reported “Going” to an event.
Using self-reported registration as a participant has its strengths and limitations.
The biggest strength of this approach is that it allows clear and objective identification of participants based on their explicit actions. In contrast, surveys are typically conducted after the event, and participants are asked to self-report their participation retrospectively. Due to this, surveying may be susceptible to two biases. Firstly, recall bias may apply, as respondents may forget or inaccurately remember whether they participated in a specific event. Secondly, social desirability bias may have a higher impact when using a retrospective self-reporting approach. Respondents may falsely report participation to present themselves in a more positive light in situations where protests have been successful (as in the case of Shiyes protests).
The biggest limitation of using self-reported registration is that this approach does not account for individuals who may have participated in the protests but did not self-register or report their attendance. In regard to this, it would be desirable to use geo-location data alongside VK activity for participant identification (as in Larson et al., 2019). Nonetheless, collecting geo-location data may be considered unethical, as users may view such data as private rather than public (Townsend & Wallace, 2017)7.
Another limitation is the remaining potential for social desirability bias. Individuals may still falsely register for a protest to present themselves as politically engaged or socially responsible citizens. However, in states characterised by repression and harsh consequences for protest involvement, individuals are less likely to falsely report their participation. In such contexts, acknowledging protest involvement can put a target on individuals and expose them to risks. Therefore, the likelihood of social desirability bias may be lower due to the potential costs associated with falsely claiming protest participation.
Method
At the feature engineering stage, several variables measuring the concepts of interest were extracted from the social network’s structure (see Table 5).
Measures such as network size, density, closure, and brokerage were calculated in ego networks of one degree8. In other words, these are local measures. Network size is the number of alters in an ego network; network density is the ratio of the number of edges and the number of possible edges; and network closure is “the probability that the adjacent vertices of a vertex are connected” (Csardi & Nepusz, 2006).
Betweenness centrality was used as a measure of a node’s local brokerage. This metric quantifies a node’s importance based on its position along the shortest paths between other nodes (Freeman, 1977). Nodes with high betweenness centrality act as connectors or intermediaries, facilitating communication and interaction. While more traditional measures of brokerage, such as Gould and Fernandez’s approach (Gould & Fernandez, 1989), are valuable in situations where network groups are well-defined and known in advance, they may not be as suitable when dealing with networks where group boundaries are not clearly delineated. In such cases, betweenness centrality provides a more versatile and context-independent means of identifying brokerage roles, making it a relevant choice for characterising a node’s brokerage position (Everett & Valente, 2016).
Closeness centrality was used to measure a node’s centrality. Closeness centrality is a measure of how close a node is to all other nodes in the network, with more central nodes having a shorter average distance to all other nodes. This feature reflects how easily information can flow from one node to another, with more central nodes being those that can quickly reach other nodes in the network. In a nutshell, closeness centrality tends to emphasise the importance of nodes that are well-connected to other nodes. This feature was calculated based on the whole network of informed users.
The scale of support was quantified by the number of protesters in an ego network. The number of protesters in an ego network essentially represents the pool of potential allies that an individual can draw upon when participating in protests.
Lastly, as a control, information about users’ prior activism was captured by a binomial variable. This variable flagged users’ memberships in activist groups, inter alia the 17 Shiyes groups on VK. Here, 1 indicated that a user is a member of at least one activist group, and 0 indicated no prior activism. Accounting for potential differences in political engagement and mobilisation between individuals with varying levels of prior activism was a primary goal of including this variable in the analysis.
The outcome, participation in protests, was measured as a binomial variable, where individuals were coded as either having participated in at least one protest during the observation period or not.
Bayesian structural equation modeling (BSEM) was used to analyse the data. BSEM allows for modeling complex relationships between variables while accounting for measurement error and uncertainty. Figure 4 illustrates the structure of the model, which consisted of several equations. As Fig. 4 shows, participation (a binomial variable) was expected to be influenced by prior activism (included in the model as a control), network structure, and the scale of support. Prior activism and network structure were also expected to affect the scale of support, while network structure was expected to be impacted by prior activism.
Fig. 4
Graphical representation of the model fitted to the data. Notes: Prior activists are expected to acquire more advantageous networks than those who have not engaged previously (Tindall et al., 2012); prior activists are also more likely to have larger support systems Brady et al. (1999); prior activists are expected to be more likely to participate again (McAdam, 1989)
This model was estimated using the Markov Chain Monte Carlo (MCMC) algorithm in the R package ‘brms’ (Bürkner, 2021). The model was fit to the data, and posterior distributions of the regression coefficients and other model parameters were obtained.
The regression with the binary outcome participation had a logit specification; and all numerical variables were standardised prior to the analysis.
BSEM allowed examining the direct and indirect effects of the predictor variables on participation and the relationships between the predictor variables.
Results
As expected, prior activism is a significant predictor of all: individual social network features, the scale of support, and participation in protests. A simple test of the hypothesis that prior activists have more extensive support networks showed positive evidence (the results of the Welch two sample t-test are as follows: \(t = -12.71\), \(df = 37282\), \(p = 6.35e-37\)). Specifically, the mean scale of support in the “prior activist” group is 1.45, contrasting with 1.13 in the “non-activists” group (see Table 1).
Table 1
Welch two sample t-test results
Variable
Mean in group
t
df
p–value
Sig.
Non-activist
Activist
Network size
272.25
215.67
20.15
42088
7.70e–90
***
Network density
0.277
0.201
75.25
41268
< 2.2e–16
***
Network closure
0.205
0.114
152.55
45885
< 2.2e–16
***
Node brokerage
4.79e–08
8.76e–08
-10.85
38563
2.20e–27
***
Node centrality
6.12e–04
5.90e–04
37.24
38146
3.72e–298
***
Scale of support
1.13
1.45
-12.71
37282
6.35e–37
***
Groups: “prior activists” and “non-activists”. Significance levels: *** correspond to p-values < 0.001
Moreover, “prior activists” are also more likely to participate in protests again: out of all 2041 participants, 67% (1370 people) fall into the “prior activist” group.
Regarding individual social network characteristics, prior activism is associated with smaller networks of lower density and closure, higher brokerage and decreased ego centrality (see Table 1).
Participation in protesting is associated with somewhat similar characteristics of an individual social network. Brokers in networks of lower density and closure are more likely to participate (see Table 2). For example, the mean network closure in the “non-participants” group equals 0.202, compared to 0.082 in the “participants” group (Welch two-sample t-test results: \(t = 61.13\), \(df = 2032\), \(p < 2.29e-16\)). Participation is also positively associated with the scale of support, with the mean scale of support in the “participants” group at 4.53, contrasting with 1.13 in the “non-participants” group (see Table 2). However, node centrality is negatively associated with participation—akin to prior activism—indicating that more central nodes are less likely to participate.
Regarding network size, the likelihood of participation positively correlates with this network property, with “participants” having, on average, more extensive networks (496 nodes) compared to “non-participants” (269 friends).
Table 2
Welch two sample t-test results
Variable
Mean in group
t
df
p–value
Sig.
Non-participant
Participant
Network size
269.35
496.17
-10.45
2043
5.86e–25
***
Network density
0.274
0.121
48.74
2029
< 2.2e–16
***
Network closure
0.202
0.082
61.13
2032
< 2.2e–16
***
Node brokerage
4.82e–08
6.59e-07
-9.75
2040
5.65e–22
***
Node centrality
6.11e–04
6.02e-04
5.24
2041
1.81e–07
***
Scale of support
1.13
4.53
-13.20
2007
3.35e–38
***
Groups: “participants” and “non-participants”
Similarly, the scale of support is more extensive in larger networks of lower density and closure, as well as among nodes with higher brokerage (as shown in Table 3). Furthermore, central nodes exhibit larger support systems, just as expected. However, node centrality displays one of the weakest correlations with the scale of support (Pearson correlation: 0.04). In contrast, network size and node brokerage emerge as the most strongly correlated features (with Pearson correlation coefficients of 57% and 52%, respectively), suggesting that brokers in bigger networks have larger support systems.
Table 3
Pearson correlation results
Predictor
Pearson correlation
CI
df
p–value
Sig.
Network size
0.566
[0.565; 0.568]
901614
< 2.2e–16
***
Network density
− 0.100
[ − 0.103; − 0.098]
901614
< 2.2e–16
***
Network closure
− 0.030
[ − 0.032; − 0.028]
883507
1.45e–178
***
Node brokerage
0.517
[0.516; 0.519]
901614
< 2.2e–16
***
Node centrality
0.040
[0.038; 0.042]
901614
5.2e–316
***
Dependent variable: the scale of support
However, the reported statistics do not take into account the complexity of the relationships between the variables. In particular, individual network structures are expected to be affected by prior activism; the scale of support by prior activism and individual network features; and participation in protesting by all of the mentioned variables. The estimated Bayesian SEM, accounting for these relationships, is shown in Figs. 5 and 6.
Fig. 5
Coefficient estimates. Notes: The diagram presents coefficient estimates for the relationships outlined in the hypotheses. The effects of “prior activism” are also estimated in the same BSEM as a control. As the coefficient estimates for the effect of prior activism do not directly address the hypotheses, they are presented separately in Fig. 6. Paths represented by plain arrows have a BF \(\ge 10\), indicating substantial evidence for the hypothesised effect
In many instances, the relationships depicted in Fig. 5 align with the hypotheses. Specifically, H1, H3, and H4 are supported by the data. As anticipated, individuals with larger networks are more likely to possess extensive support systems and participate in protests. The direct effect of network size on protest participation is 0.276, with a confidence interval (CI) of [0.231; 0.321], while the indirect effect amounts to 0.047 (calculated as \(0.449\times 0.104\), as shown in Fig. 5).
The total effect of network closure is even more pronounced. The direct effect of this variable on protest participation equals 0.724 (with a CI of [0.490; 0.967]), while the indirect effect is 0.002. Node brokerage also has a positive effect on protest participation, with the size of the direct effect equal to 0.049 (CI: [0.039; 0.061]) and an indirect effect of 0.027. In other words, individuals serving as brokers in networks characterised by higher closure are more inclined to participate in protests.
These results can be shown in two Figures visualising the differences between the networks of a randomly selected participant (Fig. 7) and non-participant (Fig. 8). In Fig. 7, the ego clearly has a broker position and is connected to other brokers in the network. The network of the ego is also quite big compared to Fig. 8. Meanwhile, in Fig. 8, the ego is low on brokerage. Moreover, out of nine direct (one degree) alters, only two have broker positions, while the remaining seven are also low on brokerage. All-in-all, Fig. 7 is a brokerage network where broker removal leads to fragmentation or the formation of new, smaller sub-networks.
Fig. 7
Individual social network of a participant. Notes: The nodes are scaled by the betweenness centrality; the graph represents an ego-network of 2 degrees; the degree of the ego equals 9
Individual social network of a non-participant. Notes: The degree of the ego equals 9; the non-participant is represented by the node with 9 outcoming links in black
The effects of network density and node centrality are contrary to the expected. Users in networks of lower density have larger support systems and are more prone to participation. The direct and indirect effects of network density on protest participation are negative, with the coefficients at -3.299 (CI: [− 3.573; − 3.034]) and − 0.011, respectively. Similarly, there is a substantial direct negative effect of node centrality on protest participation, amounting to − 230.774 (CI: [− 248.820; − 212.937]), along with an indirect effect of -1.836. Consequently, H2 and H5 are not supported by the data. In high-density networks, more central nodes are less likely to participate in protests.
Regarding the control variable, prior activists are more likely to acquire larger support systems (CI: [0.121; 0.135]) and brokerage positions (CI: [0.140, 0.159]), as shown in Fig. 6. Meanwhile, networks of prior activists are smaller (CI: [− 0.096; − 0.078]), less dense (CI: [− 0.358, − 0.343]), and lower on closure (CI: [− 0.503, − 0.485]). At the same time, prior activism is associated with lower node centrality (CI: [\(-8e-05\); \(-1e-06\)]), which is negatively associated with participation. Additionally, as expected, prior activists are more likely to participate in protesting (CI: [3.476; 3.637]).
Lastly, the studied predictors showed to have a high level of explanatory power, accounting for 58% of the variation in the outcome, the likelihood of protest participation.
In summary, out of the five hypotheses, three, H1, H3, and H4, are supported by the data. Discrepancies between the expected paths (Fig. 1) and the observed associations in the data (Fig. 5) emerge as H2 and H5 fail to find support. These unexpected results are discussed in Sect. 5.
Discussion
The analysis showed that individual network structure is an important predictor of protest participation in Russia, as an example of semi-authoritarian regimes. Key network attributes, including size, density, closure, node brokerage, and centrality, exhibit substantial and statistically significant effects on the likelihood of protest participation. In many instances, these effects align with the expectations in Fig. 1.
Among the hypothesised relationships, only one—the influence of network size—has been directly studied previously, with numerous works (e.g., Gil de Zúñiga et al. 2012; Shahin 2016) providing evidence for a positive effect of network size on the likelihood of participation. This current study aligns with these findings, indicating that larger network size is associated with larger support systems and a higher likelihood of protest participation. The direct positive effect of network size has been previously attributed to increased exposure to political information, leading to greater motivation.
Concerning the impact of other variables, prior studies have not directly delved into the role of network structure in explaining participation in protests. Instead, the focus has often been on the influence of network topology on the spread of information on the Internet as a predictor of participation (e.g., Burt, 1992; Granovetter; 1973). This study, however, aligns with the main ideas of Burt (1992): it reveals that brokers in networks with structural holes are more likely to participate in protests. The results demonstrate that nodes with larger brokerage in networks of lower density and higher closure are more prone to participation. Networks featuring higher closure and lower density consist of tightly-knit clusters with fewer connections, essentially networks with structural holes. The analysis indicates that brokers in such networks not only are prone to participation but also have larger support systems.
There are at least two possible explanations for the link between brokerage in networks with structural holes and participation. Firstly, brokers in these networks may feel a heightened sense of urgency and personal responsibility to act. Individuals in tightly-knit networks may experience greater social pressure to conform and be more likely to rely on others for action (Atwell & Nathan, 2022). In contrast, having enough support, brokers may be more willing to take leadership roles or participate in a protest. Additionally, brokers in networks of lower density may encounter a broader range of information and perspectives, potentially motivating them to take action.
Secondly, brokers may possess more resources, including higher levels of education and income, and connections to other influential individuals in the network. This aligns with the previous findings of Menon et al. (2020), who found high socioeconomic status to be associated with larger networks of low density. More resources enable brokers to overcome barriers to participation, thereby enhancing the likelihood of protest involvement.
Concerning the impact of centrality, the results reveal an unexpected negative relationship between centrality and protesting, indicating that more central users are less likely to engage in protests. Potential explanations include heightened caution among central users due to increased visibility and potential risks. These users may have more to lose if they are targeted by authorities or face negative consequences for their actions.
An alternative explanation is that influential users, with their extensive exposure to information, might perceive planned events as much bigger acts. Consequently, they may be more likely to free-ride on the efforts of others, expecting the benefits of collective action without contributing themselves. In general, in semi-authoritarian regimes, higher visibility of protests can attract increased government repression, including heightened police presence and brutality, potentially raising perceived risks for central nodes.
Additionally, central nodes don’t necessarily have larger support systems; in fact, centrality is linked to reduced support in the network. This underscores the significance of the quality, rather than the quantity, of connections for increased participation likelihood. Overall, the role of network structure is more nuanced and complex than merely facilitating exposure to political information.
Lastly, prior activism is linked to reduced network size and an elevated brokerage position, lower centrality, decreased density and closure within these networks. Simultaneously, prior activists have larger support systems and exhibit a greater propensity for participation. This suggests that, while prior activists may decrease the quantity of connections, the connections they do establish are strategically positioned. The lower size, density, and closure in a network imply that they are less likely to be part of tightly-knit online communities. Meanwhile, their larger support systems indicate that their direct contacts consist of other politically active individuals. Coupled with the other findings discussed earlier, these results—once again—underscore that strategic positioning as brokers is a crucial factor explaining protest participation.
This study has several limitations. Firstly, the study focuses on protest participation in Russia as a typical case of semi-authoritarian regimes. To generalise the results to other semi-authoritarian countries, more cases should be studied.
Secondly, the analysed network of VK users may not represent the whole population of protesters in terms of age, education, income, and other demographic factors (e.g., those people who don’t use social media).
Thirdly, the study design is cross-sectional, meaning that causality cannot be established. Some participants might have altered their network structure after joining protests, potentially becoming more influential brokers. It is possible that the reverse relationship took place, although significant changes in the network structure are unlikely, given that the relationships between users are friendship ties rather than conversation/thread ties on discussion forums.
Lastly, the study relies on self-reported data, such as users’ protest participation, which poses challenges in comprehensively capturing the full spectrum of protesters and is partially subjected to social desirability bias. Specifically, individuals prioritising privacy and those with high security considerations may not have reported their participation in the protests, thus, potentially being underrepresented in the dataset. Conversely, some individuals may have falsely registered for a protest to present themselves as socially responsible citizens. These limitations may have resulted in an analysis that partially underestimates or overestimates the true extent of protest participation.
Future research endeavours should address the limitations of this study and delve deeper into the role of network structure in predicting protest participation in semi-authoritarian regimes. More cases should be studied to test the generalisability of the findings to other semi-authoritarian countries. Employing mixed-method designs and gathering demographic information can offer insights into the intricate relationships between socioeconomic status, network structure, mobilisation, and participation.
To establish causality and explore dynamic relationships, future studies should adopt longitudinal research designs. This will help to determine whether changes in network structure lead to changes in protest participation or vice versa.
Moreover, researchers could explore alternative data sources and methods to overcome the limitation of self-reported participation. Using geo-location data alongside social media information allows for the identification of protesters and a more comprehensive study of the entire population.
Declarations
Conflict of interest
The author declares that there are no conflicts of interest regarding the publication of this manuscript.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
where \(g_{ij}\) is the total number of shortest paths between vertices i and j and \(g_{ivj}\) is the number of those shortest paths which pass though vertex v (ego)
topology_c
Node centrality calculated as Google PageRank (Brin & Page, 1998) in the full network consisting of 903 263 nodes
prior_activism
Flag indicating the ego node membership in at least one of the activist groups, calculated as the maximum across columns group_\(X_1\), group_\(X_2\), ..., group_\(X_{17}\), group_organiser, saw_information_in_group
While certain online activities are commonly labelled as protest participation, in this article, this term specifically refers to involvement in demonstrations, marches, and other forms of collective action that necessitate physical presence.
More generally, these are political communication, information, and social influence studies. This literature centres on the mobilisation process and identifying recruiters, rather than protesters.
One notable study that directly compared the individual networks of protesters and non-protesters is the work by Larson et al. (2019), who analysed tweets from over 90 million users to investigate participation in the Charlie Hebdo protest. This study primarily focused on comparing two groups—protesters and non-protesters—by analysing metrics such as the number of reciprocated ties and the number of in-group links to other protesters. In essence, the scholars mainly examined the sizes of individual networks, which represent only one topological property of these networks. Consequently, the questions mentioned earlier remain unresolved.
Betweenness centrality is a measure used to quantify the extent to which a node (in this case, a user) in a network lies on the shortest paths between other nodes (Freeman, 1977). Shortest paths represent the most direct routes between pairs of nodes (users) in a network. A node with high betweenness centrality has a significant influence over the flow of information, resources, or interactions between other nodes in the network.
An undirected edge between two users A and B indicates a bidirectional connection. In this example, a friendship tie is established when one user sends a friend request and the other user accepts it.
All data collected in this study are publicly available social-media data. The data access to which was restricted by users (e.g., membership in “closed” groups) were not collected.
An ego network of one degree refers to the network structure centered around a single node (ego), encompassing the ego itself, its direct neighbors (alters), and the connections between these neighbors.
Atwell, P., & Nathan, N. L. (2022). Channels for influence or maps of behavior? A field experiment on social networks and cooperation. American Journal of Political Science,66(3), 696–713.CrossRef
Brady, H. E., Schlozman, K. L., & Verba, S. (1999). Prospecting for participants: Rational expectations and the recruitment of political activists. American Political Science Review.,93, 153–168.CrossRef
Brin, S., & Page, L. (1998). The anatomy of a large-scale hypertextual Web search engine. Computer Networks and ISDN Systems, 30 (1), 107-117. Retrieved from https://www.sciencedirect.com/science/article/pii/S016975529800110X (Proceedings of the Seventh International World Wide Web Conference)
Centola, D. (2010). The spread of behavior in an online social network experiment. Science,329(5996), 1194–1197.CrossRef
Centola, D. (2013). Homophily, networks, and critical mass: Solving the start-up problem in large group collective action. Rationality and Society,25(1), 3–40.CrossRef
Csardi, G., & Nepusz, T. (2006). The igraph software package for complex network research. InterJournal, Complex Systems, 1695. Retrieved from https://igraph.org
Diani, M., & McAdam, D. (2003). Social movements and networks: Relational approaches to collective action. Oxford University Press.CrossRef
Everett, M. G., & Valente, T. W. (2016). Bridging, brokerage and betweenness. Social Networks,44, 202–208.CrossRef
Fabrega, J., & Sajuria, J. (2013). The emergence of political discourse on digital networks: The case of the Occupy movement. arXiv. Retrieved from arxiv:1308.1176
Fernandez, R. M., & McAdam, D. (1988). Social networks and social movements: Multiorganizational fields and recruitment to Mississippi Freedom Summer. Sociological Forum,3, 357–382.CrossRef
Frantz, E. (2018). Authoritarianism: What everyone needs to know. Oxford University Press.CrossRef
Freeman, L. (1977). A set of measures of centrality based on betweenness. Sociometry,40(3), 35–41.CrossRef
Gil de Zúñiga, H., Jung, N., & Valenzuela, S. (2012). Social media use for news and individuals’ social capital, civic engagement and political participation [Journal Article]. Journal of Computer-Mediated Communication,17(3), 319–336.CrossRef
González-Bailón, S., Borge-Holthoefer, J., Rivero, A., & Moreno, Y. (2011). The dynamics of protest recruitment through an online network. Scientific Reports,1(197), 357–382.
Gould, R. V., & Fernandez, R. M. (1989). Structures of mediation: A formal approach to brokerage in transaction networks. Sociological Methodology,19, 89–126.CrossRef
Granovetter, M. S. (1973). The strength of weak ties. American Journal of Sociology,78(6), 1360–1380.CrossRef
Haerpfer, C., Inglehart, R., Moreno, A., Welzel, C., Kizilova, K., Diez-Medrano, J., . . . others (2020). World values survey: Round seven - country-pooled datafile. Madrid, Spain & Vienna, Austria: JD Systems Institute & WVSA Secretariat.
Heaney, M. T., & Rojas, F. (2014). Hybrid activism: Social movement mobilization in a multimovement environment. American Journal of Sociology,119(4), 1047–1103.CrossRef
Jasper, J. M., & Poulsen, J. D. (1995). Recruiting strangers and friends: Moral shocks and social networks in animal rights and anti-nuclear protests. Social Problems,42(4), 493–512.CrossRef
Klandermans, B., van der Toorn, J., & van Stekelenburg, J. (2008). Embeddedness and identity: How immigrants turn grievances into action. American sociological review,73(6), 992–1012.CrossRef
Larson, J. M., Nagler, J., Ronen, J., & Tucker, J. A. (2019). Social networks and protest participation: Evidence from 130 million Twitter users. American Journal of Political Science,63(3), 690–705.CrossRef
Leenders, R. (2012). Collective action and mobilization in Dar’a: An anatomy of the onset of Syria’s popular uprising. Mobilization,17, 419–434.CrossRef
Lindstaedt, N., & Frantz, E. (2019). Democracies and authoritarian regimes. Oxford University Press.
Lockwood, S. J. (2022). Protest brokers and the technology of mobilization: Evidence from South Africa. Comparative Political Studies,55(4), 628–656.CrossRef
McAdam, D. (1982). Political process and the development of black insurgency, 1930–1970. University of Chicago Press.
McAdam, D. (1986). Recruitment to high-risk activism: The case of Freedom Summer. American Journal of Sociology,92(1), 64–90.CrossRef
McAdam, D. (1989). The biographical consequences of activism. American Sociological Review,54(5), 744–760.CrossRef
McAdam, D., & Paulsen, R. (1993). Specifying the relationship between social ties and activism. American Journal of Sociology,99(3), 640–667.CrossRef
McAdam, D., Tarrow, S., & Tilly, C. (2001). Dynamics of contention. Cambridge University Press.CrossRef
Poupin, P. (2021, Apr.). Social media and state repression: The case of VKontakte and the anti-garbage protest in Shies, in Far Northern Russia. First Monday, 26(5).
Protest229.ru, activists. (2019). Results of the single day of protest 22.09.2019. Retrieved from https://protest229.ru
Putnam, R. D. (1993). What makes democracy work? National Civic Review,82(2), 101–107.CrossRef
Putnam, R. D. (2000). Bowling alone: The collapse and revival of American community. Simon & Schuster.
Shahin, M. (2016). The relationship between social capital and political participation. Mediterranean Journal of Social Sciences, 7 (2 S1).
Steinert-Threlkeld, Z. C. (2017). Spontaneous collective action: Peripheral mobilization during the arab spring. American Political Science Review,111(2), 379–403.CrossRef
Steinert-Threlkeld, Z. C., Chan, A. M., & Joo, J. (2022). How state and protester violence affect protest dynamics. The Journal of Politics,84(2), 798–813.CrossRef
Steinert-Threlkeld, Z. C., Mocanu, D., Vespignani, A., &, et al. (2015). Online social networks and offline protest. EPJ Data Science,4, 19.
Tilly, C. (1978). From mobilization to revolution. Random House.
Tindall, D., Cormier, J., & Diani, M. (2012). Network social capital as an outcome of social movement mobilization: Using the position generator as an indicator of social network diversity. Social Networks,34(4), 387–395.CrossRef
Townsend, L., & Wallace, C. (2017). The ethics of using social media data in research: A new framework. In The ethics of online research. Emerald Publishing Limited.
