Protests can involve an array of actions ranging from street marches and sit-ins to nonviolent blockades and violent attacks on police. Established action categories distinguish between normative collective action and nonnormative collective action (Wright et al., 1990). Normative collective actions are those that conform to the societal norms of acceptable and permissible collective behaviours. In open, democratic contexts these usually include petition signings, meetings with politicians, and sanctioned rallies. In contrast, nonnormative actions constitute behaviours that are outside the boundaries of current social rules and structure. These nonnormative actions can take both nonviolent forms such as sit-ins and blockades, and violent forms involving physical injury to persons or property such as arson and assault (Wright et al., 1990; see also activism vs radicalism, Moskalenko & McCauley, 2009; inside vs outside advocacy, Betzold, 2013).
Research has increasingly examined predictors of normative collective action, highlighting the important role collective identity plays in motivating this action type (van Zomeren, Postmes, et al., 2008). However, less is known regarding drivers of nonviolent and violent nonnormative action, including the influence of collective identity (Becker & Tausch, 2015; Saab et al., 2016; Shuman et al., 2021; Tausch et al., 2011). Some studies have demonstrated that the predictors of normative and violent action may differ, although the differences in psychological mechanisms that underlie these two collective action types remain unclear (Ayanian et al., 2021; Tausch et al., 2011). In particular, while politicized identity predicts engagement in collective action generally (e.g., van Zomeren, Postmes, & Spears, 2008), there are some indications that it may also predict high-risk collective action more specifically (e.g., Ayanian et al., 2021).
Scholars have suggested how group-based identities can be reinforced through participation in collective action (Becker et al., 2011). Simon and Klandermans (2001) suggested three antecedent stages that lead to politicized identity: awareness of shared grievances, adversarial attributions (where an external adversary is attributed as the cause of the group’s predicament), and involvement of society at large. In this study, we propose that the perceived social threat of mainlandization—where traditional Hong Kong values, identity, language, and freedoms are perceived to be threatened by mainland Chinese (Chan et al., 2021)—would act as a source of shared grievance. Furthermore, political mistrust may reflect an adversarial attribution directed towards the Hong Kong political system. The differential targets of perceived social threat (mainland Chinese) and political mistrust (Hong Kong political system) provide a suitable context for analysis of collective action drivers and politicized identity in the Hong Kong context, where post-Umbrella Movement activism increasingly adopted a unique Hong Kong identity situated as opposite and irreconcilable to the mainland Chinese identity (Au, 2017).
Our study makes two novel contributions to the literature. First, we examined whether social threat from an outgroup and mistrust in the political system—and their interaction—are associated with politicized identity. Second, we examined the association of politicized identity with normative and violent collective action intentions in the context of Hong Kong’s social unrest.
Identification and Collective Action
The social identity model of collective action (van Zomeren, 2013) suggests that identity is central to collective action, having both direct and indirect effects on perceived injustice and group efficacy, as well as moral conviction. Highly identified group members are more likely to engage in collective action than low identifiers (Stürmer & Simon, 2009; van Zomeren, Postmes, et al., 2008).
Furthermore, politicized identities (e.g., as a feminist or Black Lives Matter supporter) are stronger predictors of collective action than nonpoliticized identities (e.g., as a woman or American) (Ayanian & Tausch, 2016; Stürmer & Simon, 2004). A politicized identity emerges when individuals collectively seek to engage in power struggles that challenge the existing social system (Louis, 2001; Wright et al., 1990), such as through identification with a social movement (Stürmer & Simon, 2004). Politicized identification is a direct positive predictor of environmental collective action (Schmitt et al., 2019), student protest (Stürmer & Simon, 2009), and participation in gender, environmental, and religious movements (Milesi & Alberici, 2018). Furthermore, some research has demonstrated that politicized identification can influence action type, such as intentions to engage in high-risk action in repressive contexts (Ayanian et al., 2021). The current study examined the predictors of politicized identity and the association of politicized identity with both normative and violent collective action to shed further light on these relationships.
Predictors of Politicized Identity
Simon and Klandermans (2001) suggest that awareness of shared grievances is an important component of a politicized identity. Perceived social threat may act as the basis of shared grievances. Social threat is the threat arising from any group or category of persons perceived as undermining important social values, norms, and traditions and are threatening security and stability in society (Duckitt, 2006). Much research in the social identity tradition has provided evidence that higher threat perceptions elicit group-based responses. For example, Çakal and colleagues (2016) found that social identity significantly predicted collective action tendencies both directly and indirectly via perceived threats in the context of the conflict between Turks and Kurds.
According to Simon and Klandermans (2001), the second step in the politicization of collective identity is blaming an external opponent, such as an outgroup, an authority, or the system, for the grievance. Blaming the system may manifest itself in a lack of trust in the system. There is a longstanding understanding of the importance of political trust in maintaining the stability of political systems (Hetherington, 1998) and influencing acceptance of illegal behaviours (Marien & Hooghe, 2011). Mistrust reflects an individual’s doubt regarding the trustworthiness of others and, as such, can include attributions of inefficiency and incompetence (Citrin & Stoker, 2018). In this study, we tested whether political mistrust is also associated with politicized collective identity, especially when one feels under threat.
Predictors of Collective Action
When a group’s resources and/or values are threatened by an outgroup, their tendency to act collectively to protect themselves may be heightened (e.g., Çakal et al., 2016). One means of doing so might be to demand that those in power take action to safeguard the interests of the ingroup members (Mughan & Paxton, 2006; Tajfel et al., 1979). Indeed, research has found that perceived threat can increase perceptions of collective efficacy (Hornsey et al., 2015), and foster pro-environmental attitudes and behaviours under pro-environmental norm conditions (Fritsche et al., 2010). Wohl and colleagues (2011) found that when threatened with losing their unique Quebecois and Canadian identity, individuals were more likely to engage in behaviours that protect the ingroup (see also Täuber & van Zomeren, 2013). In the Hong Kong context, one study found that university students who perceived a high degree of threat from mainland Chinese counterparts were less willing to engage in social contact with them (Awale et al., 2019). Examined through the lens of intergroup threat theory (Stephan et al., 2009), this perceived social threat to Hong Kong values, identity and norms, described as mainlandization, represents a form of symbolic threat, as opposed to only realistic threat (i.e., risks to physical or material conditions).
Another source of threat during the study period may have resulted from the growing use of violence by police against protesters (Ortmann, 2015). The increased risk of experiencing violence may influence realistic threat appraisals and subsequent intentions to engage in collective action in repressive contexts, beyond the symbolic threat of mainlandization. Most studies on the predictors of collective action have been conducted in democratic societies where the legal, financial, or physical risks associated with protest engagement have been minimal (Ayanian et al., 2021). However, research on protest in high-risk contexts has indicated that repressive responses by authorities may fuel increased collective action (Ayanian & Tausch, 2016; Ayanian et al., 2021). While this line of research indicates that risk perceptions may influence protester behaviours, in this study, we conceptualize symbolic threat as a construct distinct from realistic threat, or the perceived risk related to protest participation. Our research aims to extend understanding of the predictive role symbolic threat plays in motivating collective action.
There is evidence that when perceived threat is high, low trust in the political system may increase politicized identification (Thomas & Louis, 2014). In contrast, some have theorized that political mistrust may discourage political engagement (Almond & Verba, 2016), although the evidence remains mixed (Levi & Stoker, 2000). Theorists have argued that low political trust may be more predictive of nonnormative action (Christensen, 2016). For example, Paige (1971) demonstrates in a classic study on riot participation that individuals with both low political trust and high political efficacy combinations were most likely to participate in nonnormative collective action such as rioting. More recently, Thomas and Louis (2014) showed that the preference for nonviolent action over violent action is erased when the system is seen as corrupt, while Šerek and colleagues (2017) found that lower institutional trust predicts adolescents’ readiness to engage in nonnormative actions.
Building on this past research, we expect to find an association between political mistrust and participation in violent action for greater regional autonomy and democracy in Hong Kong. Given that social threat and political mistrust have been shown to influence collective action intentions, we expect the interaction of political mistrust and social threat to predict normative and violent action intentions via politicized identification.
The Rise of Social Unrest in Hong Kong
Our research was conducted in Hong Kong in the aftermath of the 2014 Umbrella Movement and 2016 Fishball Revolution, and prior to the 2019 Anti-Extradition Law Amendment Bill Movement. The Umbrella Movement was ostensibly triggered by the demand for universal suffrage promised in the Basic Law. However, scholars and commentators attribute its precursor to the influx of mainland Chinese tourists and migrants to Hong Kong, which has escalated tensions between Hong Kong residents and mainlanders (Tiezzi, 2015). Resistance to this perceived process of mainlandization was shaped by a desire to preserve traditional Hong Kong values, rights, language, and freedoms from the perceived threat of mainland Chinese (Chan et al., 2021). In this context, mainland Chinese are a potential source of social threat for some Hong Kongers.
During the time of this study, Hong Kongers expressed considerable support for the calls for greater autonomy for Hong Kong, with support for localism and independence especially high among youth (Kwan et al., 2016). In a local poll of university students, 61% of participants indicated that they would vote for the city’s independence if given the chance (Lam & Cheung, 2016). Furthermore, the growth of the self-determination movement appeared to coincide with the decline of trust in the government and political system among Hong Kongers (But, 2013). Most politicians in Hong Kong, including its chief executive, were considered pro-Beijing and pro-establishment (Bush & Whelan-Wuest, 2017).
While the Umbrella Movement began as a nonviolent campaign, Hong Kong pro-democracy activists have increasingly used violent action to oppose the perceived increasing interference by the Chinese central government in the city’s political, economic, and social affairs and erosion of freedoms (Ng, 2020). In the context of this social divisiveness and escalating use of violence, we examined whether perceived social threat from mainland Chinese, alongside reduced political trust, may be positively linked to the development of a politicized identity among Hong Kongers and, in turn, increase intentions to engage in violent actions as well as normative actions.
The goal of the present research was to examine the influence of social threat and political mistrust and their interaction as predictors of politicized identity (i.e., identification with the Hong Kong independence movement), and to examine their direct and indirect effects through politicized identity on normative and violent collective action. The existence of an active pro-democracy movement combined with political mistrust and perceived social threat provides a unique opportunity to study the influence of these factors on politicized identity, as well as normative and violent collective action. Results from this context would provide strong evidence for the robustness of the phenomenon; they would assist the understanding of the politicization of the Hong Kong identity in the context of the pro-democracy movement.
We argue that a high degree of politicized identity means that individuals are more willing to take actions that might entail a relatively high degree of risk, commitment, and self-investment (even extending to violent action, which was evident in the 2016 Fishball Revolution and the 2019 Anti-Extradition Law Amendment Bill social unrest). In addition, we expected both broader ingroup identity (e.g., as a Hong Konger) and politicized identity to be positively associated with normative action. This is predicted because while normative action in Hong Kong after the introduction of the National Security Laws in 2020 may have had grave consequences, during the time of these studies, normative action generally involved less effort and risk (see also Becker et al., 2011). We did not have an a priori hypothesis concerning the inhibiting (negative) association of social category identity and nonnormative action found by Becker et al. (2011), particularly given their analysis based on past participation rather than action intentions.
Using the Hong Kong pro-democracy movement as the context, across two studies, we hypothesized that perceived social threat (H1a), political mistrust (H1b), and their interaction (H1c) would be positively associated with politicized identity. In turn, we hypothesized that politicized identity would be positively associated with both normative (H2a) and violent collective action (H2b). We further hypothesized that perceived social threat and political mistrust and their interaction would have an indirect effect (through politicized identity) on both normative (H3a) and violent collective action (H3b)1. Given that the broad ingroup identity is the basis on which the pro-democracy politicized identity is formed, we also included the degree of identification as a Hong Konger in the model as a predictor of politicized identity and collective action intentions. No studies in this manuscript were preregistered; data was collected in 2016 and 2017.
Study 1 was conducted before and after a contentious election in Hong Kong for the city’s lawmakers (“The 2016 Legislative Council election”). Among the candidates were a number of pro-autonomy candidates, who were especially popular among young people (Yuen, 2016).
Data for the present study was collected in August and September of 2016. Data was collected via an online survey that was publicized through mass email and recruitment posters as well as from The University of Hong Kong through campus-wide email and the participant pool system. Ethical approval from The University of Hong Kong was obtained prior to data collection.
Four hundred and sixty participants began the survey and provided informed consent. We excluded participants who did not meet the criteria of having been born in Hong Kong or lived there for more than seven years (n = 27), the minimum duration required for permanent residency. After excluding cases that had missing data for every measured scale (n = 35), the final sample comprised 398 participants (56.0% female, 44.0% male, Mage = 20.5, SD = 4.76). Participants were primarily young adults, reflecting the principal demographic of pro-democracy protesters in Hong Kong at this time (Chan, 2016). A sensitivity analysis indicated that this sample size afforded a power of .80 for detecting effects size ρ = .14, assuming a Type-I error rate of 5% (two-sided).
Procedure and Measures
Participants completed an online questionnaire that assessed their politicized identification, identification as a Hong Konger, perceived social threat from mainland Chinese, political mistrust, and intentions to participate in normative and violent collective action.
Following Claes, Hooghe, and Marien (2012), participants were asked to rate the extent to which they trusted the “Government of Hong Kong,” “Hong Kong Police,” “Hong Kong Courts,” and “Hong Kong Legislative Council”. Responses were recorded on a seven-point scale (1 = not at all, 7 = very much), and reverse-scored. Items were averaged to form a single political mistrust measure (α = .76).
Perceived social threat from mainland Chinese was measured using an eight-item scale adapted from Duckitt (2006). Sample items included, “They seem to reject moral values that are important to me,” and “They seem to want to destroy or harm what is good in our society”. Items were rated on a nine-point scale (1 = definitely not, 9 = definitely yes), and averaged to form a single social threat measure (α = .84).
Politicized identity was measured using three items adapted from Doosje, Ellemers, and Spears (1995): “I identify with members of the Hong Kong Independence Movement,” “I see myself as a member of the Hong Kong Independence Movement,” and “I feel strong ties with the Hong Kong Independence Movement.” Items were rated on a seven-point scale (1 = not at all, 7 = very much), and were averaged to form an index of politicized identity (α = .93).
Broader Ingroup Identity
Identification with the broader ingroup was measured using the same items and rating scale used to measure politicized identity; “members of the Hong Kong Independence Movement” was replaced with “Hong Kong(er)” (e.g., “I see myself as a Hong Konger”; α = .84).
Normative Collective Action
Participants were asked to indicate how likely they were to engage in certain activities as part of a future political movement in Hong Kong. Four items were chosen based on Tausch et al. (2011): “Discussing issues on social networks,” “Signing petitions,” “Participating in discussion meetings,” and “Participating in demonstrations” (α = .83). Responses were recorded on a seven-point scale (1 = very unlikely, 7 = very likely).
Violent Collective Action
Participants were asked to indicate how likely they were to engage in certain activities as part of a future political movement in Hong Kong. Three items were chosen based on Tausch et al. (2011) and incorporated violence against property and people: “Clashing with police,” “Throwing objects such as rocks or bottles,” and “Burning objects,” (α = .90). Responses were recorded on a seven-point scale (1 = very unlikely, 7 = very likely).2
All analyses were conducted in R (R Core Team, 2013), and path analysis was performed using the lavaan package (Rosseel, 2012). We used maximum likelihood estimation with robust standard errors (the “MLR” estimator in lavaan which uses Huber-White standard errors; Huber, 1967; White, 1980) to account for potential skewness (Yuan et al., 2005). Since there was a small amount of missing data (≤ 10%), we used full information maximum likelihood estimation (FIML: Arbuckle, 1996). Because the hypothesized model was saturated, we did not evaluate its model fit.
We first tested the main and moderating effects of social threat and political mistrust on politicized identity. We then conducted a path analysis to examine the hypothesized relationships among the measured variables. We also tested the significance of the hypothesized effects using the lavaan package in R.3 Data, code, a codebook that includes all measures for both Study 1 and Study 2, and the Supplementary Tables are available on OSF (see Supplementary Materials).
Descriptive results are reported in Supplementary Table 1 and correlations among measured variables are presented above the diagonal in Supplementary Table 2. Both social threat and political mistrust were positively associated with politicized identity. Identification with the broader ingroup was also positively associated with politicized identity.
Path Analysis With Full Model
Standardized path coefficients are presented in Figure 1.
Consistent with the hypotheses, broader ingroup identity (β = .178, p = .001), social threat (β = .195, p = .006; H1a) and political mistrust (β = .235, p < .001; H1b) were positively associated with politicized identity. There was no significant interaction between social threat and political mistrust in the prediction of politicized identity (β = -.008, p = .87; H1c) or broader ingroup identity. Also consistent with our hypotheses, politized identity was positively associated with normative (β = .264, p < .001; H2a) and nonnormative action (β = .386, p < .001; H2b).
The indirect effects of perceived threat on normative (β = .052, p = .009; H3a) and violent action intentions (β = .072, p = .017; H3b), through politicized identity, were significant, as were the indirect effects of political mistrust on normative (β = .062, p = .005; H3a) and violent action intentions (β = .086, p = .003; H3b) through politicized identity.
There was no significant interaction between social threat and political mistrust in the prediction of normative action intentions (β = .015, p = .82; H3a); however, their interaction was positively associated with violent action intentions (β = .159, p = .010; H3b).
A simple slope analysis revealed that social threat was a significant positive predictor of violent action intentions at high (+1 SD above the mean) levels of political mistrust, B = .162, SE = .060, p = .007, but not at low (-1 SD below the mean) levels of political mistrust, B = -.052, SE = .055, p = .339 (Figure 2).
Broader ingroup identity was positively associated with normative (β = .166, p = .001), but not violent action intentions (β = -.029, p = .611). In addition, the indirect effects of broader ingroup identity on normative (β = .047, p = .009) and violent action intentions (β = .065, p = .010), through politicized identity, were both significant.
Study 1 examined the antecedents and consequences of identifying with the independence movement. Our hypotheses were partially supported: we found that social threat (H1a) and political mistrust (H1b), but not their interaction (H1c), were associated with politicized identity. Broader ingroup identification was also associated with politicized identity. Politicized identity was associated with both normative (H2a) and violent action intentions (H2b). However, broader ingroup identification was only significantly associated with normative action intentions (H3b), but not violent action intentions (H3c).
These findings support the contention that broader ingroup collective identity is insufficient for driving violent collective action. However, we found an indirect effect of broader Hong Kong identity with violent action via politicized identity. There was no inhibiting (negative) association of the broad social category identity with violent action as demonstrated in Becker et al. (2011) or Feinberg et al. (2020). The present findings highlight that the social consensus against violent protest may be breaking down in the Hong Kong context – as may be seen in other contexts around the world when social threat is increasing alongside greater political mistrust.
Our findings also indicate that individuals who perceive high social threat or high political mistrust more strongly identify with both the politicized identity and broader ingroup. However, while direct associations were found, our hypothesized interaction effect of social threat and political mistrust on politicized identity was not supported. Put differently, threat and mistrust both independently bolstered politicized identity itself, but their interaction predicted violent action intentions. This study took place several months after the 2016 Fishball Revolution, which was largely anti-mainland Chinese, and at the time of a contentious election. The main effects of political mistrust and social threat might have been heightened and thus leaving less variance for their interaction to explain. To further examine this effect, we conducted a second study in a less eventful period. We also incorporated measures of efficacy and injustice that have been shown to mediate the link between politicized identity and collective action (e.g., van Zomeren, 2013; van Zomeren, Spears, & Leach, 2008).
In Study 2, we again tested the three hypotheses concerning the associations between social threat (H1a) and political mistrust (H1b), and their interaction (H1c), on politicized identity. Following Study 1, we included broader group identity as an antecedent of politicized identity and predictor of normative and violent action intentions.
In addition, we examined group efficacy and perceived injustice. These variables have underpinned the development of the Social Identity Model of Collective Action (SIMCA; van Zomeren, 2013), which suggests that collective action can be directly predicted through politicized identity, but also indirectly via group-based anger or injustice and group efficacy (van Zomeren et al., 2012). In Study 2, we include a measure of group-based injustice, which reflects a grievance-based pathway to collective action (van Zomeren et al., 2012). Conversely, an instrumental pathway to collective action has been demonstrated through research on the role of group efficacy (van Zomeren, Postmes, & Spears, 2008). In particular, there is substantial evidence indicating that high efficacy predicts normative actions (van Zomeren, Postmes, & Spears, 2008).
Following SIMCA, we hypothesized that politicized identity (H2a), perceived injustice (H2b), and group efficacy (H2c) would predict normative action. We argue that violent action should be positively associated with politicized identity (H3a) and perceived injustice (H3b): as these may provide a legitimizing narrative and potential motivation for violent acts. However, SIMCA has relatively little empirical evidence regarding the association between group efficacy and violent collective action. Some research indicates that violent action may result from low efficacy perceptions: that is, protesters resort to violence when there is little hope of achieving their goals (Becker et al., 2011; Saab et al., 2016). Accordingly, we hypothesize that group efficacy is negatively associated with violent action (H3c). We conducted the study in 2017, at a time of relative calm.
The data collection was conducted between September and November in 2017. Ethical approval from The University of Hong Kong was obtained prior to data collection. Two hundred and thirty-three participants were recruited from the University of Hong Kong through campus-wide email and the participant pool system. They were entered into a lucky draw or given course credits as participation compensation. Participants who were not Hong Kong locals (n = 28) or failed the attention check item (n = 5) were excluded from the analysis. The final sample comprised 200 participants (68.5% female, 31.5% male, Mage = 19.3, SD = 1.78). A sensitivity analysis indicated that this sample size afforded a power of .80 for detecting effects size ρ = .20, assuming a Type-I error rate of 5% (two-sided).
Procedure and Measures
Participants were invited to complete an online questionnaire. We administered the same measures used in Study 1 to assess political mistrust (α = .76), social threat (α = .84), politicized identity (α = .94), broader ingroup identity (α = .85), as well as intentions to engage in normative (α = .86) and violent action (α = .87). In addition, we introduced the measures of perceived injustice and group efficacy in this study.
Perceived injustice was measured using a four-item scale adapted from Tausch et al. (2011): “The absence of true democracy in Hong Kong is unfair,” “The Chinese government's decision to only allow pre-screened candidates to run for Chief Executive of Hong Kong is socially unjust,” “The Chinese government’s control over the appointment of Chief Executive of Hong Kong is not legitimate,” and “The absence of true democracy in Hong Kong is justified” (reverse-coded). Items were rated on a seven-point Likert-scale (1 = strongly disagree, 7 = strongly agree) and were averaged to obtain a composite perceived injustice score (α = .74).
Group efficacy was measured using a two-item scale adapted from Lee (2006). Participants were asked to rate the extent to which they agreed that “The collective action of Hong Kong people has a huge influence on public affairs,” and “The collective action of Hong Kong people can improve society.” Responses were recorded on a seven-point scale (1 = strongly disagree, 7 = strongly agree), and they were averaged to form an index of group efficacy. The inter-item correlation was r = .36, p < .001.
Using the lavaan package (Rosseel, 2012) in R, we applied path analysis to test our hypotheses. Since some of the variables exceeded the recommended limits of univariate normality (Yuan et al., 2005), similar to Study 1, we applied maximum likelihood estimation with robust standard errors (i.e., the “MLR” estimator in lavaan which uses Huber-White standard errors; Huber, 1967; White, 1980).4 Our analysis followed the sequence of Study 1. Data, code, and codebook for this study are available as Supplementary Materials.
The mean and standard deviation of the variables are presented in Supplementary Table 1. Consistent with the relative political calm at the time, independent samples t-tests revealed that participants from Study 2 in 2017 (M = 6.08, SD = 1.38) reported significantly lower perceived social threat than participants from Study 1 in 2016 (M = 6.36, SD = 1.38), t(555) = 2.27, p < .05; participants from Study 2 (M = 4.08, SD = 1.07) also reported significantly lower political mistrust than those from Study 1 (M = 4.38, SD = 1.12), t(559) = 3.13, p < .01.
Correlations among measured variables are presented in Supplementary Table 1 below the diagonal. The path model (Figure 3) is of adequate fit, CFI = .989, RMSEA = .076 (90% CI = .000, .169), SRMR = .028.
As hypothesized, broader group identity (β = .170, p = .016) and the interaction of political mistrust and perceived threat (β = .181, p = .021; H1c) were positively associated with politicized identity. Social threat (H1a) and political mistrust (H1b) were not significantly associated with politicized identity. However, simple slope analyses revealed that social threat was a significant positive predictor of politicized identity at high (+1 SD above the mean) levels of political mistrust, B = .222, SE = .105, p = .036, but not at low (-1 SD below the mean) levels of political mistrust, B = -.142, SE = .123, p = .247 (Figure 4).
Politicized identity had a significant positive association with both normative (β = .286, p < .001; H2a) and violent collective action intentions (β = .405, p < .001; H3a). There was a direct association of perceived injustice with normative collective action intentions (β = .225, p = .012; H2b), but not with violent collective action intentions (β = -.136, p = .093; H3b). In contrast, there was a negative association between group efficacy and violent collective action intentions (β = -.155, p = .029; H3c), but not normative collective action intentions (β = .005, p = .937; H2c) .5
Taken together, the results reveal significant indirect effects of broader ingroup identity on normative (β = .044, p = .033) and violent collective action intentions (β = .075, p = .014), through politicized identity. Similarly, the indirect effects of the interaction of political mistrust and social threat on normative (β = .047, p = .048) and violent collective action intentions (β = .080, p = .048), through politicized identity, were significant.
In Study 2, we tested our original model with additional variables from SIMCA, namely perceived injustice and group efficacy. Our hypotheses were partially supported. Unlike in Study 1, we did not find a significant association of perceived social threat (H1a) and political mistrust (H1b) on politicized identity. However, our results yielded a significant interaction between the two predictors, suggesting that social threat and political mistrust play an important role in identity politicization (H1c). Those who felt both highly threatened by mainland Chinese influences and did not trust the political institutions of Hong Kong were more likely to identify with the independence movement, which in turn contributed to more conventional protest but also to violent protest choices.
Also similar to Study 1, the broader identity of Hong Konger was positively associated with politicized identity and normative collective action but not associated directly with violent action. Politicized identity had a significant positive association with both normative (H2a) and violent collective action (H2b). While we did not find positive associations between group efficacy and normative collective action (H2c), consistent with Tausch et al. (2011), group efficacy (itself associated with lower mistrust) was negatively associated with violent collective action (H3c). Perceived injustice was positively associated with normative collective action (H2b), however, contrary to our hypothesis, no association was found with violent collective action (H3b).
Across two studies, we examined the extent to which perceived social threat from an outgroup and political mistrust predicts politicized identity and intentions to engage in normative and violent collective action. Using Hong Kong’s 2014 Umbrella Movement and the 2016 Fishball Revolution as a context, Study 1 found a small but significant association of perceived social threat (H1a) and political mistrust (H1b) on politicized identity. Study 2 found a small but significant interaction effect of perceived social threat and political mistrust on politicized identity (H1c). As hypothesized, politicized identity was associated with both normative (H2a) and violent collective action intentions (H2b) across both studies. In addition, Study 1 found the hypothesized indirect effect from perceived social threat and political mistrust to normative and violent collective action through politicized identity. Study 2 incorporated SIMCA measures and found the hypothesized indirect effect from the interaction of perceived social threat and political mistrust through politicized identity to normative (H2a) and violent collective action (H2a).
Taken together, our findings shed light on factors that may influence identity politicization. In Study 1, perceived social threat and political mistrust were associated with politicized identity, while in Study 2 their interaction was associated with politicized identity. Both variables indirectly predicted both normative and violent collective action intentions via politicised identity. However, while there was no evidence that these variables directly affect collective action intentions (cf., Šerek et al., 2017), they demonstrate the importance of threat and mistrust to group status as antecedents of collective action (Wohl et al., 2011; Yustisia et al., 2020). In this context, we suggest that social threat may function as a type of shared grievance (Simon & Klandermans, 2001), which, when channeled via powerful politicized identities (van Zomeren, Spears, et al., 2008), increases a groups’ tendency to act collectively in protection. To the extent that the results of Study 2 generalise, it is when citizens lose faith in the political institutions or systems that the indirect paths of threat to politicised identity to violence are most likely to emerge. The path from mistrust to lower efficacy to higher violence found in Study 2, while unexpected, is somewhat consistent with the hypothesised finding, both highlighting that when authorities are seen as unresponsive and untrustworthy, processes of politicisation and radicalisation accelerate (Louis et al., 2020). These independent, and the possible moderating, effects of mistrust (Study 2), which may buffer the emergence of violent action under threat, both deserve future attention.
However, significant pathways between perceived threat and political mistrust were not consistent across both studies. The significant interaction of social threat and political mistrust on violent action in Study 1 was not replicated in Study 2. Rather, Study 2 showed an indirect relationship via politicised identification. Study 1 was undertaken during a time of high conflict salience and corresponding social unrest, whereas Study 2 was conducted during a period of relative calm. This is reflected in the data, with significantly lower average social threat and political mistrust perceptions for Study 2 participants (Supplementary Table 1). As engaging in violent collective action can be costly (e.g., with physical and legal consequences), we speculate that the perceived cost might have increased over time when more activists were prosecuted. This might have contributed to the lowering of the association between these variables. Further studies are needed to examine the outcome of this interaction effect in different contexts.
Turning to the role of identity, our results indicate that politicized identity is a predictor of both normative and violent collective action intentions. This was supported in both studies, extending previous work examining the role of politicized identity (e.g., Milesi & Alberici, 2018; Stürmer & Simon, 2009) by applying it within a pro-democracy movement and specifically demonstrating its association with violent collective action intentions. We also examined the association between ingroup and politicized identity, finding that broader ingroup identity was positively associated with politicized identity in both studies. Across both studies, we found that ingroup identity (as a Hong Konger) was only significantly associated with normative collective action intentions, and not violent collective action intentions. These findings build on work by van Zomeren and colleagues (2011) in their study on Hong Kong and Mainland Chinese, who demonstrate that identification with a disadvantaged group was the relevant group identity predictor of collective action tendencies.
In Study 2, we also included SIMCA measures of group efficacy and injustice (van Zomeren, Postmes, & Spears, 2008). As noted above, perceived injustice was associated with normative (but not violent) collective action intentions. Our injustice measure referred to the absence of democratic rights and legitimacy in Hong Kong, which may not have been connected to a narrative legitimizing violence (e.g., Webber & Kruglanski, 2018). Indeed, while violence escalated during the study period, opinion polls indicated general opposition to the protests (Yuen & Cheng, 2017), and many groups remained committed to the principle of non-violence (Ng & Kennedy, 2019). The present data may indicate that, as in Moghaddam’s (2005) staircase model of political violence, the presence of injustice may be necessary but insufficient to motivate violent collective action directly.
Conversely, efficacy was negatively associated with violent (but not normative) collective action. Our group efficacy measure sought to align with Hong Konger concerns regarding mainlandization; namely, the ability of Hong Kongers to influence public affairs and improve society. The limited political participation opportunities and the lack of concrete steps toward democratization or increase in autonomy resulting from the peaceful protests may help explain why this group efficacy measure was not associated with normative collective action. In this context, the null association of efficacy with normative collective action would be interpreted as showing that such action may be seen as symbolic rather than instrumentally linked to effecting change (Louis et al., 2020).
The findings suggest that those who intend to engage in violent collective actions may have lost belief in the power to effect change via normative action tactics, while still being highly driven to engage in collective action behaviours (Ayanian et al., 2021; Saab et al., 2016). Furthermore, violent collective action entails a higher degree of risk and commitment than normative action (e.g., Becker et al., 2011). It may be that those with higher politicized identity have more strongly internalized the norms of the movement (Thomas & McGarty, 2009) and thus are more willing to accept the risks that arise from violent action, while still doubting its efficaciousness.
A growing body of research indicates that the specific nature of efficacy perceptions may be important in untangling the complex relationship between efficacy and normative and violent collective action. For example, Ayanian and Tausch’s (2016) examination of the role of efficacy in repressive contexts found that an individual’s participatory efficacy – their belief in the importance of their own contribution – was a significant predictor of action intentions (as opposed to political efficacy). Study 2 utilized a measure of group efficacy assessing collective action’s ‘influence on public affairs’ and ‘improving society’. Further studies may consider incorporating specific measures directly related to the specific goals of the movement (e.g., increasing democratic rights for citizens), as well as individual and participatory efficacy measures.
Taken together, these findings suggest fruitful avenues for future research on the antecedents and consequences of politicized identity. We suggest that these findings warrant future research examining causal effects of ingroup identity, politicized identity, and nonnormative actions. This research will help identify pathways to and from violent collective action, and the context in which beliefs about efficacy and injustice become encapsulated within politicized identities rather than independent predictors in their own right.
Although our analyses were guided by established theory and findings from prior research, the cross-sectional nature of our data does not allow us to establish causal relations between variables. It is also possible that the causal relationship between threat and politicized identity as well as mistrust and politicized is bi-directional in nature. It is therefore important for future studies to corroborate our findings with longitudinal and experimental data. Similarly, our group efficacy measure in Study 2 was at a lower level of specificity than other measures such as threat and political mistrust (Fishbein & Ajzen, 1974). Future studies should consider replicating findings while including equivalent measures across variables.
It is also important to note that our student sample may in general reflect a more anti-establishment, distrusting, and radical profile than the general population (Chan, 2016). In addition, different findings across the two studies may reflect sample differences. Participants in Study 1 reported higher levels of perceived threat and lower levels of political trust. There was also a temporal difference between the two studies; Study 1 was conducted over a year prior to Study 2, during which much had changed in the political climate. Moreover, the present study measured intentions to participate in collective action instead of actual participation. Although previous research has found that behavioral intentions are a good proxy of actual participation in collective action (Moskalenko & McCauley, 2009), future studies should examine the predictive power of our current model with respect to actual behavior.
Finally, our analysis of the predictors of violent action was grounded in the situational context of Hong Kong, where protests increased the use of violence in response to political and police repression (Ng, 2020). As highlighted by Drury (2020), protest participants can experience identity changes, particularly when police actions are perceived as illegitimate, which can then motivate further collective action participation (Stott et al., 2021). Further research could consider examining data gathered during actual protest events as well as comparative work analyzing the drivers of short-term responsive violent action with longer-term use of violence as a strategic collective action choice.
The present study makes a number of important contributions to the literature. Our findings are of particular interest given that research on the antecedents of nonnormative collective action primarily focuses on collective action within well-functioning democratic societies and on relatively low-risk, nonviolent action (Ayanian & Tausch, 2016). Given that violent collective action is not usually widespread (Moskalenko & McCauley, 2009), our studies help increase our understanding of when widespread violent collective action is most likely to erupt. Specifically, these findings point to the role of perceptions of social threat and political mistrust: when mistrust is high, citizens may become more likely not only to mobilize under threat, but also to radicalize. As such, activities such as increasing repression of a disadvantaged group while restricting access to political power or allowing the political system to erode may help fuel increased violent responses.
These results suggest that proponents of violent tactics could seek to increase perceptions of social threat and political mistrust to help build politicized identity, and in turn, increase support for violent collective action. But this tactic may also bring risks. Our study did not investigate the wider effects of increasing perceptions of social threat. Other research suggests increasing threat perceptions could exacerbate polarization between disadvantaged and advantaged groups and affect societal-wide responses such as increased support for authoritarianism (Fritsche et al., 2012). Conversely, authorities seeking to manage increasingly violent social movements might profit from considering the problem of mistrust, not just by seeking to manage citizens’ perceptions, but also by seeking to earn their trust.
The Hong Kong pro-democracy movement provides a unique opportunity to examine how normative protest can be transformed into violent collective action. By examining the antecedents of politicized identity, political mistrust, and perceived social threat, and exploring the links to violent collective action, the present research contributes to understanding the rise of violent mass movements, and the factors that might fuel them. Our findings are consistent with the view that increasing individuals’ politicized identity through manipulating perceptions of social threat and political mistrust may increase engagement in both normative and violent collective action. Conversely, when mistrust is low, threat may not only be less likely to mobilize politicized citizens, but also less likely to lead to violent collective action. Given the widespread use of violence in many past and present social movements, better understanding the processes which may influence the use of violent collective action is important both for social movement activists and their opponents. The question of how mistrust decreases (or increases) over time is a critical direction for future research and theorizing.