Civic skill-acts, group identity, and intentions to engage in protest actions among university students in Hong Kong

Abstract This study examined the mediating role of civic skill-acts and direct associations of group identity on intentions to engage in peaceful or radical protest actions (i.e., activism or radicalism intentions respectively). A sample of 526 university students in Hong Kong was surveyed. The findings suggested that political identity complementarily mediated the relationship between joining political activities and radicalism intentions. Religious identity and ethnic/racial identity each have an indirect-only mediation to activism as well as radicalism intentions when mediated by community activities and responding activities respectively. Finally, political identity and economic identity each have direct-only mediations to activism intentions respectively. These results suggest that although group identity and civic skill-acts uniquely contribute to protest intentions, the inter-relationship is complicated by the type of group identity, civic skill-act, and protest activity studied. Recommendations for future studies are discussed in light of the findings.


Introduction
At the individual level, it has been argued that participation in social movements involves group identification (motivation) and skills gained from past involvement in civic activities (capability). These arguments derive from the intuition that individual differences in protest actions exist because certain people have a stronger identification with the aggrieved group/contentious issue than others and that certain people are more well-equipped and prepared to act than others. These two modelsthe group identification approach (Tajfel, 1978) and the resource-based approach to political participation (Brady et al., 1995) have found popular support among scholars attempting to explain participation in protest actions.
While the micro-mobilization literature has devoted much effort to explicating both prerequisite conditions to protest participation, rarely have they accounted for both components in their empirical and/or theoretical endeavors. Among the scarcity of studies that have accounted for group identification and past involvement in civic activities in predicting protest participation (e.g., Schussman and Soule, 2005), there is a lack of consideration on the heterogeneity of group identification, civic activities, and young people's vector of protest actions (e.g., peaceful or radical actions). This oversight is problematic as it conflates potentially meaningful distinctions that might exist between varying types of each component in predicting protest actions.
Similar to Brady et al. (1995), this study contends that both civic skill-acts available to the person (through past involvement in civic activities) and his/her group identification are necessary conditions for understanding intentions to engage in peaceful or radical protect actions. In the practice of protesting, certainly group identification and civic skill-acts go hand-in-hand; however, it is possible to have one aspect without the other. For instance, a person may feel an intense psychological connection with a threatened group identity but not participate in major protest activities organized by constituents who share the same identity due to a self-perceived lack of necessary skills or transferable experiences. Conversely, another person may have the necessary skills to participate in the protest activity but decides to be uninvolved because s/he feels little to no identification with the group identity shared by the movement constituents. It follows, therefore, that conceptualizing this inter-relationship is critical for understanding young people's preparedness to engage in future protest actions.
In what ways does different group identifications and past involvement in civic activities significantly associate with future intentions to engage in peaceful or radical protest actions? What are the different ways by which group identification and/or past involvement in civic activities explain the variance in both dimensions of future protest action? Empirically, this study addresses these research inquiries in the context of post-Umbrella Movement Hong Kong. The city makes for a suitable case study for three reasons. First, the city has witnessed a growing number of noninstitutionalized protest activities led by young people since 2010, and not all of these activities were peaceful and legal. 1 Rather, some protest events or campaigns, such as the widely recognized 2014 Umbrella Movement (UM), involved a variety of protest actions that ranged from legal to illegal, and from peaceful to transgressive (Cheng, 2016;Chan et al., 2017). Scholars have largely come to agree that the increased 'bottom-up' political activism, and its relationship with the emergence of more transgressive forms of protest actions, is not a transient trend but a potentially permanent fixture of the city's political culture that is suggestive of deeper changes in the cycle of contention (Cheng, 2016). Second, Hong Kong has a lively civil society that generally accommodates for the co-existence of many different types of non-governmental collective bodies, such as local, regional, and international non-governmental organizations, peer support groups, mutual aid organizations, student groups, and other interest groups. This has not only greatly expanded and transformed the city's voluntary sector, but it has also provided sympathizers of different non-political concern groups with ample opportunity to organize or participate in demonstrations, sit-ins, and/or petitions. Among other telling examples, the Hong Kong 1st July public processions, which the Civil Human Rights Front used to organize annually, 2 offered a lucid microcosm of Hong Kong's lively and diverse civil society, as it has long been animated by the involvement of groups, alliances, and formal organizations that vary in their raison d'être and demands (e.g., calls for universal suffrage among pro-democracy political parties or comprehensive employee protection measures among labor rights groups) (Ma, 2008). Finally, few studies in Hong Kong have investigated dispositional and social-psychological factors that are significantly associated with intentions to engage in protest actions (e.g., Chan 1 The cycle and dynamics of contentioninsofar as socio-political activism is concernedin post-handover Hong Kong has undergone drastic transformations, largely in response to a variety of changes in political opportunity structures, government reactions/responses, and movement radicalization (Cheng, 2016;Lee et al., 2019). Notably, with masterful use of protest event time-series analysis from 2006 to 2014 and in-depth interviews, Cheng (2016) documented astute observations of how six critical events (see Cheng, 2016: 395) inspired an evolution of 'bottom-up' activism. This new activism, characterized by decentralized organization, direct actions (e.g., adoption of disruptive forms of activism, such as building a road blockade or unlawful assemblies), and technologically mediated recruitment and mobilization strategies, broke the mold of the routinized, regulated, peaceful and legal protests that were traditionally organized by political parties or social movement organizations. The protest trend between 2010 and 2014 in particular had longer episodes of contention, increased number of prosecutions (e.g., for assaulting a police officer), and featured diverse protest repertoires on the streets (Cheng, 2016 Wong et al., 2019). To our knowledge, however, the available studies have not examined the inter-relationship between different types of group identities (i.e., beyond just an identification with a political party or stance), involvement in various types of civic activities, and intentions to engage in peaceful or radical protest actions.
This article first discusses the group identification model and the resource-based approach to political participation. Next, this study conceptualizes the inter-relationship between group identification, civic skill-acts gained from past involvement in civic activities, and intentions to participate in peaceful and/or radical protest actions. Thereafter, background information on protests in Hong Kong and research on young people's social and political participation in the city are discussed. Finally, before discussing the methodology, results, and implications of the results, the aim and research questions of the present study will be revisited.
2. Group identification approach to studying political participation As early as the nineteenth century, with the publication of Le Bon's (1995Bon's ( /1895 seminal work on crowd behavior, scholars have understood that the force to collectively act is underpinned by the salience of a shared identification of the in-group (i.e., 'We' or 'Us') and the out-group (i.e., 'They' or 'Them'). Social group identification, as a characteristic of all persons embedded in the social web that makes up communities, is defined as 'an individual's self-concept which derives from his knowledge of his membership of a social group (or groups), together with the value and emotional significance attached to that membership' (Tajfel, 1978: 63). Scholars have argued that the ongoing multidimensional (i.e., cognitive, affective, and behavioral) personalization of social groups and evaluation of intergroup relations, particularly in terms of comparing the treatment and standings of 'other' groups to one's own social group memberships, is very much a normal part of how we see the social world, relate to one-another, and organize our social life. Among others, scores of experimental studies and field studies have demonstrated that binding with persons who share memberships with us, shielding them from out-group adversaries, and protecting the issues that our group cares about is part of our human nature (see Clark et al., 2019).
Available research has found group identities to be salient in the context of protests and, more specifically, the process of mobilization. For instance, Klandermans and Oegema (1987) proposed a four-stage action mobilization model that recognizes that a person's awareness of a 'call to action' and his/her successful self-categorization as being part of that 'call to action' precedes considerations of the person's intentions to go out or not. Put simply, if one does not feel that s/he is part of the struggle, then the potential to mobilize is not high. A particularly potent variant of these 'calls to action' are foreseeable threats to social group(s) or issue(s) that one cares about. If a person perceives an imminent threat against a group/issue that resonates with his/her social identity, then that person would be more likely to act out on behalf or in support of that group/issue. For example, in a longitudinal study of Dutch farmers' protest against the Dutch political authorities and European Union over agriculture, Klandermans (2002) found that group identification was significant in explaining the farmers' preparedness to act (i.e., intentions) as well as their actual participation in protests against changing farming practices and production. In a more recent study, using a randomized controlled trial design, Chan et al. (2021) found that self-identified Hong Kongers who were primed with a group threat (i.e., mainland China's growing influence in the territory through the implementation of a 'co-location arrangement' in a high-speed rail terminal project, which effectively facilitates an extension of Chinese law into the territory's physical space), in contrast to those who were exposed to a neutral stimulus (i.e., weather report), were more likely to express intentions of participating in contentious forms of political activism (i.e., protests and signing petitions to advocate for changing the status quo). Their thesis of mainlandization as an identity group threat to Hong Kongers was further demonstrated with results that self-identified Chinese respondents did not express similar intentions to politically participate after being exposed to the 'co-location arrangement' primer.
3. Resource-based approach to studying political participation In their seminal study, Brady et al. (1995) introduced a resource-based model of political participation. The model argued that affiliation with civic associations, such as schools or church, often served as fertile grounds for acquiring, practicing, and developing civic skills that could be later used for political participation. According to Brady et al. (1995), this process of acquiring civic skills begins during adolescence and young adulthood, particularly at home and while attending school or university. For example, being involved in high school governments or student unions in university, being involved in church, having a job, or participating in a voluntary organization. Each of these institutions present persons with opportunities to engage in activities that requiresand potentially further cultivatesskills, such as communication or organizational capacities. Brady et al. (1995) conceptualized this domain-specific measure of engagement in activities as 'skill-acts'. For example, job skill-acts could include organizing meetings or delivering public talks. These skill-acts equip the person with civic skills and competencies to not only further engage in the domain where the opportunities were first available to him/her, but also to branch out to other domains across time.
Available empirical studies have well supported this thesis. Controlling for biographical availability, political engagement, and organizational affiliations, Schussman and Soule (2005), for example, found that possessing civic skill-acts, such as writing political letters or attending a meeting, increased one's probability of participating in a protest, even if the person was not initially invited to join the mass activity. In one cross-national study, using data collected from the 1999-2002 World Values Survey on 79 nations around the world, Dalton et al. (2009) found that persons with higher education and memberships in social groups/organizations significantly predicted the adoption of non-institutionalized protest actions as a means of political engagement, independent of their level of dissatisfaction/grievance. In a more recent cross-national study, which analyzed data on civic education and citizenship among school students aged 14 from five Asian societies in east and south-east Asia, found that past community participation was significantly associated with intentions to engage in legal protest (Zhu et al., 2018). These studies support Brady et al.'s (1995) core propositions that institutions can be training grounds for civic skills that are transferable across domains, potentially predisposing persons with the prerequisite resources and confidence to engage in the political sphere of life.

Group identification, civic skills, and protest participation
Considering both sets of literature together, in the context of micro-mobilization, both concepts are complementary to one another. While group identification can account for variance in why we would be willing to participate, the resource-based approach addresses how a person might have been able to do so. It is not enough to just want to participate, for example as a result of group identification with the aggrieved. An individual must have the skills and knowledge (i.e., the know-hows) to perform the action(s) that s/he feels is needed in the protest field. Although Klandermans and Oegema (1987) did not explicitly state it in their model, the final step of the action mobilization model suggests that having the know-hows is crucial to determine whether someone actually participates or not. A person could sympathize, be targeted, and have the motivation to participate in a protest, but if s/he does not have skill-act experiences and the dispositional capabilities to match his/her cognitive-affective preparedness, then one could argue that the instance of protest action would not be observed.
Conversely, while the resource-based model of political participation does provide a theoretical basis to understand disparities in political participation, i.e., protest action varies with civic skill-acts and capacities, it does not shed light on persons' intentions or inclinations to do so. Although Brady et al. (1995) acknowledged that group identification was an important piece of the mobilization puzzle, they argued that it was trivial to actually explaining political participation beyond the natural correlation that intention has with acting. Skill-acts were argued to be more appropriate and suited for this task. It is ironic how then their demotion of group identification in the model later undercuts their capacity to offer a comprehensive explanation of how skills are transferable from non-political to political domains, with Brady et al. (1995: 273) saying, 'Those who have an opportunity to do… things in a nonpolitical setting would, presumably, be more willing and able to do them in a political context' (emphasis added). There is no doubt that Brady et al. (1995) made a substantive contribution to explaining the conditions by which people can afford to participate. However, by not accounting for group identification, they were unable to offer insights into why people would use their skills to participate, regardless if it is transferred from another domain or not. Therefore, a model that can comprehensively account for both factors of micro-mobilization would be able to provide for a richer and nuanced analysis. For instance, in comparison to those who were asked to join a protest, Schussman and Soule (2005) found that possessing civic skills was an important predictor of protest participation among those who were not asked to join, particularly if they were interested and engaged with politics. Expanding on these results, Schussman andSoule (2005: 1098) argued that participants who were politically liberal and had high civic skills were 'hyper-engaged' and were able to 'participate in protest despite never being asked to do so.' This finding instills confidence in the assertion that neither factor can supplant the other in accounting for a person's participation. Group identification with an issue or group (e.g., being politically liberal) does not displace civic skill-acts as a predictor of political participation, and neither does civic skill-acts displace the importance of group identification. Rather, they are complementary facets of human functioning in the context of micro-mobilization and should be studied as such.

Background: protests in Hong Kong
As civil society-government contentions have intensified in the last decade, non-institutional political actions, such as street protests, hunger strikes, and boycotts, have occurred with increasing regularity and in larger numbers in Hong Kong (Cheng, 2016). Alarmingly, with rising discontent over the effectiveness of peaceful responses to China's intervention and 'anti-Chinese' sentiments (Ma, 2015;Chan, 2016), the sphere of non-institutional political actions gradually expanded beyond legal, peaceful activism (Lee, 2018). This is particularly evident in the protest events and campaigns that took place between 2012 and 2016, such as the anti-parallel trading protest campaign, the 2014 UM, and the 2016 'Fishball' Revolution. In these three protests, not only did transgressive behaviors, like exerting violence or damaging public property, emerge as an expression of protest, but also the climate of opinion gradually changed from using peaceful to confrontational forms of resistance in their anti-establishment movements (Chan, 2016;Lee et al., 2019). The 2014 UM was a key critical event in this regard, as groups that embodied the stark departure from the moderate, non-violent prodemocrats became entrenched in party politics and community organizing (see Kwong, 2016). In short, as an event, the 2014 UM was a catalyst that set-in motion ripples of breakaways and departures from the once unified anti-establishment, non-violent resistance core (Cheng, 2016). Consequently, scholars have devoted greater attention to studying the changing action repertoires of activists in Hong Kong.
However, it is still poorly understood what factors associate with intentions to engage in peaceful or radical protest actions. As we see it, there are three shortcomings in the current body of literature that needs to be addressed: first, scholarship on Hong Kong protests has yet to fully embrace the rich diversity of social groups/issues that people identify with and act on behalf of. For example, among the notable few that broached the subject of civic skills, group identification, and activism in Hong Kong, Chan et al. (2017) constrained the scope of group identification to national/local identification (i.e., Hong Konger identity and mainland Chinese identity). Similarly, Chan (2017) operationalized group identification as a unidimensional measure of identification with the pro-democracy movement (i.e., a movement identity). Despite its semi-democratic institutional setup, Hong Kong has a vibrant civil society that animates protest events and campaigns on behalf of a variety of issues and groups (Ma, 2008). Among other social protests, the emergence of grassroots activism (Lai, 2018), environmental activism (Lee, 2015), religious activism (Chan, 2015), and LGBT activism (Chan, 2018) in the past two decades lend evidence to the vibrancy of Hong Kong's civil society activism. Therefore, expanding the scope of groups/issues that one could possibly identify with can help unfold nuances in the relationship between group identification and protest participation with greater clarity and precision. Second, there is insufficient focus on the role of involvement in non-political and political civic activities to understanding political participation. While there is little doubt that past protest involvement predicts future protest activism (Moskalenko and McCauley, 2009), and that skill-acts gained in past protest activity involvement can be used in future protest activities (Van Dyke and Dixon, 2013), any design that neglects the potential impacts of involvement in non-political civic activities potentially overestimates the importance of past protest participation in the development of requisite skill-acts for future activism. This is what Brady et al. (1995: 277) referred to as the 'locus of development problem.' The frequency of practicing skill-acts in a particular domain does not necessarily imply that they were developed in activities of that particular domain. It is possible that the skill-acts practiced in a political activity during adolescence, for example, originated in another type of nonpolitical activity. Thus, focusing on the independent effects of involvement in non-political and political civic activities on political participation could help expand the field of inquiry beyond political skill-acts, potentially making way for more robust models that accounts for non-political and political skill-acts. Third, there are only a few studies that have clearly delineated peaceful protest actions from radical protest actions and investigated intentions to engage in both forms of actions as outcome variables (e.g., Wong et al., 2019). Majority of the studies have either focused on intentions to engage in peaceful protest actions (Chan, 2017;Chan et al., 2017) or examined attitudinal support (instead of behavioral intentions) for radical protest actions (Lee, 2018).

About the study
Our research aims to shed light on the following two inquiries: in what ways do different group identifications and civic skill-acts significantly associate with future intentions to engage in peaceful or radical protest actions? What are the different ways by which group identification and/or civic skill-acts explain the variance in both dimensions of future protest action? Based on prior research, the following three hypotheses are formulated: H1: Civic skill-acts is positively associated with activism and radicalism intentions.
H2: Group identity is positively associated with activism and radicalism intentions.
H3: Civic skill-acts differentially mediates the relationship between group identity and activism and radicalism intentions.
Our research departs from previous work in four important ways. First, we focus on a broader set of groups that people identify with, rather than solely focusing on identification with political groups (e.g., a political ideology or party). Second, instead of aggregating different forms of prior civic activity engagement (or skill-acts) into one index (Schussman and Soule, 2005), focusing on activity engagement in political domains only (Van Dyke and Dixon, 2013), or using attitudinal (instead of behavioral) scales to measure civic engagement in political and non-political domains (Torney-Purta et al., 2008), we examine actual past involvement in four types of civic activities and treat them as unique subsets of civic engagement. Third, being mindful of the expanding protest repertoires among young people, this study focuses on intentions to engage in legal protest activities as well as more radical forms of protest activities. Finally, this study explores the relationship between group identification (as the independent variable), prior involvement in civic activities (as the mediator), and intentions to engage in peaceful or radical protest actions (as the outcome variable) structurally in a mediation model. Special focus is placed on examining the different types of mediation at work and understanding what the results mean for how certain group identifications and skill-acts influence one's vector of protest actions in the future.
As citizen self-mobilization is becoming increasingly evident in Hong Kong (Lee, 2015), this article focuses on young people's civic skill-acts, instead of their organizational affiliations, as proxies for involvement in political and non-political domains. The advantage with this arrangement is that explanations offered by the former (i.e., skill-acts) for micro-mobilization does not exclude the potential impact of the latter (i.e., organizational affiliations), as most community and political activities in Hong Kong tend to be organized by larger entities but not all of them may require membership as a prerequisite for involvement in their activities.

Data
This study analyzes data collected through self-report in-class surveys. Ethical clearance and approval from the Human Subjects Ethics Sub-Committee of the researcher's home university were received prior to data collection. To recruit survey respondents, we first reached out to faculty members at various tertiary institutions in Hong Kong, explaining the goals of the project and protocol for survey administration. Contingent on their approval for providing access and assistance, one or two members of the research team would attend the lecturer's class and, upon being invited to the front of the class, deliver a short briefing about the project. During the briefing, we emphasized voluntary participation and the measures employed to protect their identity. Every briefing ended with a statement that they were free to withdraw at any time with no adverse consequences and we thanked them for their participation in advance.
Between October 2015 and February 2016, a total of 547 pencil-and-paper surveys were collected from three UGC-funded tertiary institutions. On average, the survey took 20 min to complete. Due to missing data in key variables of interest, the final sample used is 526 university students. All respondents were aged 18 or above.

Youth inventory of involvement
The Youth inventory of involvement (Pancer et al., 2007) measures youth's involvement in community and political activities in the past 12 months. The 30-item measure is composed of four subscales, each of which captures the respondents' involvement in a specific genre of activities. The four subscalesand their respective computed Cronbach's alpha scoresare: political activities (Cronbach's α = 0.86) (eight items; e.g., 'You joined a protest march, meeting or demonstration'); community activities (Cronbach's α = 0.75) (seven items; e.g., 'You helped organize neighborhood or community events'); responding activities (Cronbach's α = 0.64) (five items; e.g., 'You participated in a school academic club or team'); and helping activities (Cronbach's α = 0.85) (10 items; e.g., 'You volunteered at a school event or function'). All of the subscales used a 5-point Likert scale (ranging from 0 = You never did this to 4.00 = You did this a lot) to assess the frequency of involvement in activities. Subscale scores are computed by averaging the total score by the number of items. Examined by subscale, a low mean score indicates less involvement, whereas a higher mean score indicates greater involvement. Moskalenko and McCauley (2009) established the activism-radicalism intentions scale (ARIS) as a way to measure the strength of an individual's intention to engage in activism (i.e., legal activism or peaceful protest actions) or radicalism (i.e., illegal activism or radical protest actions) on behalf of a group that s/he feels closest to. The scale has two main parts. The first part uses one item to ask the survey respondent to identify which group s/he cares about most. Available options include Religious (e.g., Catholic, Muslim), Single-issue (e.g., environmental, human rights),

Activism-radicalism intentions scale
Economic (e.g., grassroots, unemployed), Political (e.g., pro-democracy, pro-establishment), Ethnic/ Racial (e.g., Chinese, Indian, Indonesian, White, Filipino, other Asians), and Gender (e.g., gay/lesbian, bi-sexual, women). The respondent can select only one of the six options. The second part uses an 8-item scale to measure respondent's intentions to engage in peaceful or radical activism on behalf of the group they identified in the first part. Both subscales (i.e., activism and radicalism intentions) have four items each. A 4-point Likert scale (ranging from 1.00 = Very unlikely to 4.00 = Very likely) is used to capture the strength of the intentions to engage in peaceful activism (Cronbach's α = 0.82) or radical activism (Cronbach's α = 0.83) (Moskalenko and McCauley, 2009). A sample item from the 4-items activism intentions subscale is 'I would donate money to an organization that fights for my group's political and legal rights.' A sample item from the 4-items radicalism intentions subscale is 'I would participate in a public protest against oppression of my group even if I thought the protest might turn violent.' Subscale scores are computed by averaging the total score by the number of items. A low mean score (i.e., below 2) indicates weak intentions, whereas a high mean score (i.e., greater than 2) indicates strong intentions.

Socio-demographics
Age, sex (male coded as '1'), and year of university study (Year 1 coded as '1' to Year 4 or just graduated as '4') are included as control variables. Survey respondents' year of study was controlled for in this study as the more years one spends in tertiary education, the more opportunities s/he has to participate in (non-)political activities (which, in effect, would influence his/her civic skill-acts).

Data analysis strategy
Analysis of the survey data was performed with SPSS 23.0. For doing mediation analysis specifically, the PROCESS v.30 macro for SPSS was used. Following the advice of Zhao et al. (2010), apart from the indirect path (i.e., a × b), information on the unstandardized regression coefficients for paths a, b, and c ′ and the significance test of the indirect association are provided. The total effect (c) is also provided. In addition to detailing the findings, the type of mediations uncovered, based on Zhao et al.'s (2010: 201) classification, will also be reported in order to enrich discussions of the results in subsequent sections. For all mediation models, the significance of the indirect path will be determined by performing bootstrapping at 5,000 iterations and examining whether the confidence intervals (CIs) includes 0 or not (Preacher and Hayes, 2004). If the CI includes 0, then the indirect path is nonsignificant (i.e., mediation is not established). Alternatively, if the CI excludes 0, then the indirect path is significant and mediation is established.
For the mediation analysis, the independent variable is social identification, operationalized as which group s/he cares about most (i.e., choosing between religious, single-issue, economic, political, ethnic/racial, and gender). Dummy variables were created for each group. The mediating variables are involvement in political activities, community activities, responding activities, and helping activities (i.e., civic skill-acts). The dependent variables are activism intentions and radicalism intentions. Age, sex, and year of study are included as control variables in the mediation models. The mediation model is shown in Figure 1.
It should be noted that the non-random sampling strategy employed and cross-sectional nature of this study limits its ability to yield generalizable and causal results. As such, the forthcoming results should be interpreted as being correlational.

Descriptive statistics
Mean, standard deviations, and range statistics of the key variables of interest for the survey sample are shown in Table 1. In the sample, the mean age of the respondents was 21.31 (S.D. = 2.16), and the age range was from 18 to 26 years old. Males were a minority in the sample (n = 168, 31.9%). Majority of the respondents were incoming freshmen starting out their first year in university (n = 269, 51.1%).
On average, the respondents were involved in at least one civic activity in the past 12 months. Helping activities had the highest mean score of 1.71 (S.D. = 0.81), while involvement in political activities had the lowest mean score of 0.79 (S.D. = 0.76). Regarding group identification, most of the respondents selected the single-issue group as being the one they care about most (n = 187, 35.5%). Following this was identification with the political group (n = 102, 19.4%). The religious group was the least endorsed of the six options (n = 44, 8.4%). In the sample, the mean score for activism intentions is 2.69 (S.D. = 0.57) and radicalism intentions is 2.18 (S.D. = 0.65). On average, thus, the respondents had slightly strong intentions to participate in activism or radical activities in the future.
Examining social identification groups by civic activities and activism/radicalism intentions, Table 2

Correlation analysis
The zero-order correlations revealed that all of the civic activities positively correlated with activism intentions (ranging from r = 0.26*** to r = 0.33***) and radicalism intentions (ranging from r = 0.15** to r = 0.36***). With regard to the correlation between civic activities and group identities, significant correlations were only found between religious identity and community activities (r = 0.10*), political identity and political activities (r = 0.10*), and ethnic/racial identity and responding activities (r = 0.09*). Single-issue identity, economic identity, and gender identity did not significantly correlate with any of the four civic activities. Also, helping activities was the only civic activity that did not significantly correlate with any of the six group identities. Finally, only political identity significantly correlated with both activism intentions (r = 0.14**) and radicalism intentions (r = 0.15**). While non-significant in association with radicalism intentions, economic identity negatively correlated with the mental determination to engage in activism in the future (r = −0.13**). Table 3 shows the results from testing the indirect associations of different social identities on activism intentions via involvement in different civic activities. Based on Zhao et al.'s (2010: 201) classificatory diagram, the mediation analyses revealed that there were two cases of indirect-only mediation and two cases of direct-only non-mediation at 95% CI. Starting with the indirect-only mediation results that were significant at P < 0.05, we found that community activities significantly mediated the relationship between religious identity and activism intentions (β (ab) = 0.0424; 95% CI [0.0020, 0.0923]). The direct association (β (c ′ ) = 0.0040) was found to be non-significant (P = 0.9643). The indirect association, i.e., involvement in community activities, accounted for 91.4% of the total effect of religious identity on activism intentions. Ethnic/racial identity also had a positive indirect association on activism intentions, although through involvement in responding activities (β (ab) = 0.0524; 95% CI [0.0019, 0.1073]). The direct association (β (c ′ ) = −0.1145) was found to be non-significant (P = 0.1411). Next, regarding the direct-only non-mediation results yielded, as the label suggests, none of the indirect paths were found to be significant. Rather, in all four equations with different mediators, it was found that economic identity had a significant negative direct association on activism intentions (β (c ′ ) ranging from −0.1869 to −0.2231 and all significant at P < 0.01), while political identity had a significant positive direct association on activism intentions (β (c ′ ) ranging from 0.1643 to 0.2163 and all significant at P < 0.01).  Note. Controlling for age, sex, and year of study. Bootstrap is based on 5,000 resamples. When the CI does not include zero, there is a significant indirect effect. ***P < 0.001, **P < 0.010, *P < 0.050. Note. Controlling for age, sex, and year of study. Bootstrap is based on 5,000 resamples. When the CI does not include zero, there is a significant indirect effect. ***P < 0.001, **P < 0.01, *P < 0.05. Table 4 shows the results from testing the indirect associations of different group identities on radicalism intentions via involvement in different civic activities. The mediation analyses revealed that there were two cases of indirect-only mediation and one case of complementary mediation at 95% CI. Starting with the indirect-only mediation results that were significant at P < 0.05, we found that involvement in community activities significantly mediated the relationship between religious identity and radicalism intentions (β (ab) = 0.0508; 95% CI [0.0025, 0.1085]). The direct association (β (c ′ ) = −0.1104) was found to be non-significant (P = 0.2697). Ethnic/racial identity also had a significant positive indirect association on radicalism intentions, but through involvement in responding activities (β (ab) = 0.0453; 95% CI [0.0019, 0.0947]). The direct association (β (c ′ ) = −0.0569) was found to be non-significant (P = 0.5293). Uniquely, political identity had a significant direct association (β (c ′ ) = 0.1526, P < 0.05) and indirect association on radicalism intentions through involvement in political activities (β (ab) = 0.0499; 95% CI [0.0012, 0.1044]). 3 In this case of complementary mediation (β (c ′ ) = 0.2026, P < 0.01), it was found that political identity accounted for a greater share of the total effect (75.3%) than the indirect path (i.e., through involvement in political activities) (24.6%).

Conclusion
Individual participation in political actions or protest activities does not just happen. It is a highly complex decision-making process that involves accounting for factors that are external and internal to the decision-maker. A subset of social movement scholarship, which focuses on micro-level factors, has investigated dispositional and social-psychological factors that explain protest participation. This article brings together two well-recognized factors: group identification and civic skill-acts, to examine their associative significance in relating to intentions to engage in legal or illegal protest actions (i.e., activism and radicalism respectively). The present study tested these relationships in a mediation model and with an original dataset of 526 university students in Hong Kong. Supporting H1, our results revealed that political and non-political skill-acts were positively associated with both activism and radicalism intentions. This provides confirmatory evidence for the thesis that skill-acts are related to protest participation (Brady et al., 1995). The effects of group identity on activism and radicalism intentions, however, were less linear. Not all group identities significantly associated with protest intentions. Additionally, while some group identities associated with protest intentions in a positive direction, an opposite direction was observed for other group identities. Thus, H2 was partly supported.
Partially confirming H3, we show evidence that group identity and civic skill-acts differentially relate to activism and radicalism intentions. In most cases of significant mediation pathways, non-political skill-acts mediated the relationship between group identification and activism as well as radicalism intentions. This was found to be evident in the religious identity-community activities-activism (or radicalism) intentions and the ethnic/racial identity-responding activitiesactivism (or radicalism) intentions models. In these four mediation models, identification with the respective groups associated with greater involvement in civic skill-acts, which then facilitated higher levels of activism (or radicalism) intentions.
In one exceptional case of mediation in our findings, both the indirect pathway and the direct associations were significantly associated with greater radicalism intentions in the same direction (i.e., positive sign). In the complementary (or partial) mediation of the political identity-political activities-radicalism intentions model, we found that, contrary to the implications afforded by the indirect-only mediation models, higher levels of radicalism intentions were explained by the direct association (i.e., political identity) to a greater extent than through political skill-acts. This generally coincides with previous studies' findings that persons with civic skill-acts or prior involvement in civic activities and identification with a politicized group (e.g., being a 'Hong Konger' during the 2014 UM) significantly related to greater intentions to participate (and actual participation) in protest activities (Klandermans, 2002;Chan et al., 2017). Given limitations in the research design and observational data, our study is not in a position to draw causal claims and explore counterfactual outcomes. However, with more sophisticated designs (e.g., randomized controlled trials), it would be interesting to test what would have happened had the political group identified with been pro-Chinese vs Hong Konger, keeping the domain and level of civic skill-acts constant.
Two models were found to have significant direct associations and no significant indirect association (i.e., direct-only non-mediation). In both cases, although in different directions, the group identity predicted activism intentions; political identity directly predicted greater activism intentions, while economic identity predicted decreased activism intentions. Considering available scholarship on social identity and protest readiness, as well as the specific context of Hong Kong, the direct relationship between political identity and readiness to participate in protest activities is not surprising. However, what is puzzling is the negative relationship between economic identity and activism intentions. Despite its affluence, Hong Kong has an alarming income disparity and a history of collective actions that call for greater social protection policies and welfare for the economically deprived (Ma, 2008). Hence, one would expect a positive association between caring about issues surrounding economic well-being and activism intentions. One probable explanation here is that there are plenty of unions, working groups, and non-governmental organizations working on these issues in Hong Kong already. We speculate that, while the subsample had the group identification, they lacked opportunity or willingness to participate because of situational circumstances around the existing organized advocacy for economic well-being. Future studies should consider capturing strength of group identification and prior behavioral involvement in (organized) activities aligned with the group identification as potential third variables to account for in their models.
The significance of civic skill-actsregardless of domainon associating with activism and radicalism intentions deserves further explanation. One potential explanation for the strong relationship between civic skill-acts on peaceful or radical protest actions is the diversification of protest tactics that was observable during the 2014 UM and thereafter. During the 79-day civil disobedience campaign, apart from the use of distinctly political skills (e.g., creating online petitions to advocate for political change), a variety of action tactics from non-political domains were also evident among attendees, ranging from organized donation drives to tutoring young protesters and delivering first aid (Yuen, 2015). Attendees were not excluded from participation by lack of resources nor membership; rather, civic skills nurtured from non-political institutions were flexibly transferred to the protest space. Following the 2014 UM, protest tactics increasingly broke out of the mold of peaceful, non-violent activism (Cheng, 2016;Lee et al., 2019). The diversification, innovation, and steady radicalization of protest tactics culminated in a rich vibrancy of protest actions during the 2019-2020 Anti-Extradition Bill Movement (AEBM). As Lee et al. (2019) discussed in their richly descriptive study on the first few months of the 2019-2020 AEBM, various protest repertoires and tactics were employed by protesters of the movement. Among others, the actions included distinctly political civic skill-acts, such as strikes and class boycotts, as well as skill-acts nurtured in other domains, such as composing music or creating art installations (Lee et al., 2019). Insofar as the actions advanced claim-making, citizens were encouraged to be 'climbing mountains together, making your own effort'innovating and experimenting with different strategies and civic skill-acts (Lee et al., 2019: 24). With the central role that digital platforms played in sustaining the 2019-2020 AEBM, it is worth exploring how different civic skill-acts were deployed not only in offline contexts but also online (e.g., publishing relevant graphic designs on social media platforms). Furthermore, in the context of ideological and tactical radicalizations that were brewing as clashes persisted and triggering events unfolded (e.g., the 21st July 2019 incident, when thugs violently attacked commuters in a train station in Yuen Long) (Lee et al., 2019), future studies may consider unpacking how intentions to engage in radical protest actions (and actual involvement in illegal or violent actions) relate not only to social psychological changes in intergroup relations and in-group identity (i.e., a relationship advanced by identity-action frameworks such as the elaborated social identity model, see Stott et al., 2020), but also skill-act experiences and dispositional capabilities. By accounting for both aspects, we would be in a better position to understand how an individual's cognitive-affective inclination to participate in radical protest actions (e.g., because of tribalistic impulses to protect in-group members and engage in violence against out-group persons) corresponds with their capacities do so.
This study is not without limitations. First, because the data were non-randomly sampled and cross-sectional in nature, we were not in an ideal position to make causal claims regarding the inter-relationships between group identification, civic skill-acts, and intentions to participate in peaceful or radical protest actions. Randomized experiments and/or longitudinal studies are needed to probe the causal effects of group identification on peaceful or radical protest intentions, and examine the causal mediation effects of civic skill-acts in the relationship (see Imai et al., 2010 for guidance on causal mediation analysis). Second, due to the relatively small subsample sizes for some of the social identification groups (e.g., religious identity), caution is warranted when interpreting the results. Future studies with larger subsample sizes are needed to estimate the magnitude of the effects for each group identity and domain-specific skill-acts on intentions to participate in peaceful or radical protest actions. Third, while the list of groups/issues for group identification provided by ARIS is comprehensive, it could be argued that 'localism' does not completely fit into any one of the groups/issues. Since the 2014 UM, localism has emerged as a rather unique group/issue that demands the integration of native 'Hong Kong' cultural elements (e.g., Cantonese mannerisms) and democratic civic dimensions of identification (e.g., anti-authoritarianism) into a unified local identity, which is distinctly contrasted against the ethno-cultural-political elements of what it means to be 'Pan-Chinese' (see Chow et al., 2019or Veg, 2017 for more detailed discussions). As such, if a participant identifies with issues relating to localism, s/he may not be able to definitively choose any one of the six groups (e.g., while it could be thought of as 'political,' it could also be identified as fitting with the 'ethnic/racial' option). Future studies using ARIS in Hong Kong are recommended to diversify the item options by including 'local identity' or 'localism' as a choice. For example, it could be used to explicate how 'localism' group identificationin comparison to other selected group identities on the listinfluences respondents' willingness to participate in peaceful and/or radical protest actions. Finally, as Brady et al. (1995) aptly suggested, people participate in civic life if they can (i.e., resources like time, money, and civic skill-acts), if they want to (i.e., identification with a group with shared interests), and if somebody asked (i.e., recruitment networks). While this study did account for resources and group identification, it did not consider recruitment networks. Future studies should consider including variables that capture this construct so that more sophisticated models can be built.
Overall, our findings support the theoretical position that it is necessary to account for both civic skill-acts and group identity in understanding cognitive commitments to participating in legal or illegal activism. While group identification does play an important role in relating to intentions to engage in activism or radicalism, most of these effects were mediated by prior involvement in civic activities. The patterns varied by type of group identified with and the civic skill-acts gained. Future investigations are recommended to account for both factorsin its political and non-political variationsto develop comprehensive explanations of intentions to engage in legal and/or illegal protest actions. Also, buttressing the claim that civic skill-acts are transferable across domains, we provide evidence that non-political domain skill-acts are related to intentions to engage in the political realm (Brady et al., 1995). Qualitative inquiries on how and why particular skill-acts are transferred and used in the political realm, as well as the relevance of group identification in these processes, would help enrich our understanding of young people's activism. Conflict of interest. The authors declare none.