Who Dominates the Coalition? Frame Salience and Public Support for Policy–Democracy Protest Coalitions under Authoritarian Rule

Abstract

How does the public support a coalition in which pro-democracy advocates and policy-based protesters join forces in street protests? When policy-based and pro-democracy groups protest together, they create a collective action frame that includes a policy component and a democracy component. In this article, I develop the frame salience theory, arguing that support for a policy–democracy protest coalition depends on which component of the joint frame is perceived to be more dominant. I argue that in authoritarian regimes, the policy component typically dominates the coalition because it is more accessible and available to the public. This perception shifts public support for the alliance towards the baseline level of support for the policy movement. In other words, public support for the alliance defaults to the baseline level of support for the policy movement. I find evidence for my argument using a survey experiment administered to 1,209 Vietnamese respondents. This article highlights a dilemma pro-democracy groups face: joining policy-based movements may boost support, but sustaining democracy after the protest becomes challenging.

Throughout the authoritarian world, pro-democracy advocates and policy-based movements occasionally unite in protest to challenge their governments. For example, in Malaysia, a durable competitive authoritarian regime until 2018, 23% of demonstrations between 1993 and 2018 were coalitions of policy protesters and democracy advocates.¹ Notably, the Bersih movements, which contributed to Malaysia's democratic transition in 2018, united pro-democracy advocates and civil society organizations focusing on different policy issues such as corruption, environment, health, civil rights, women's rights and economic reforms (Hooi 2020: 65). In Vietnam, a single-party regime, six out of 21 protests between 1993 and 2018 recorded in the Mass Mobilization in Autocracies Database (MMAD) included the participation of both pro-democracy advocates and policy protesters. One notable example occurred in 2016 when mass fish deaths in the central provinces of Vietnam prompted street protests. Pro-democracy advocates and environmentalists joined forces, advocating for greater government transparency and freedom of information to address the environmental issue (Nguyen and Datzberger 2018). Such coalitions are also observed in other regions. During the Arab Spring in the Middle East and North Africa, pro-democracy advocates and movements focusing on unemployment and economic issues collaborated to advocate for free and fair elections. In China, an informal collaboration between pro-democracy activists and nationalists sometimes occurs to seek democracy to defend the national interest (Weiss 2014: 229). In Egypt, pro-democracy activists sometimes linked their cause to labour issues and worker groups to garner widespread public support (Dobson 2013).

How does the public support a protest coalition between pro-democracy advocates and policy-based movements? Despite the prevalence of democracy–policy movement coalitions in authoritarian regimes, little research has examined public attitudes towards them. This is a significant oversight, given a growing body of literature that emphasizes the critical role of public opinion in the success of collective action in authoritarian regimes (Arnon et al. 2023; Truong 2024). Scholars suggest a few factors that may affect public opinion towards protest coalitions, such as the level of state repression (O'Brien and Li 2006), the resonance between the movement's demands and the public's interests and preferences (Zuo and Benford 1995), protest tactics, especially the effect of violence (Lupu and Wallace 2019), and the identity of protesters (Edwards and Arnon 2019).

In this article, I focus on one crucial aspect that can make or break a cross-movement coalition but has yet to receive much scholarly attention: the effect of coalition-based collective action frames designed to mobilize participants and attract public support (Snow et al. 1986). When a policy-based protest, such as a land, labour or environment protest, joins a pro-democracy group to seek major regime reforms to resolve the former's grievances, they devise a collective action frame combining the policy component and the democracy component. The former tells the audience which policy issues give rise to policy protesters' grievances and who is to blame. The latter prescribes democratic institutions as solutions to policy issues.

How can the public form their opinions of protest movement coalitions? I argue that public support for an alliance will depend on which component of the frame the public perceive as dominant.² Research on the psychology of framing effects emphasizes that elites' framing of policy issues can only influence public opinions if the public can understand the concepts and have regular, repeated exposure to the frames (Chong and Druckman 2007). Inspired by this research, I argue that although a policy–democracy collective action frame consists of policy and democracy components, the public will perceive one component to dominate the frame. Under both full and competitive authoritarian regimes, the policy component is expected to be perceived to dominate the policy–democracy frame for two reasons. First, policy frames are more understandable for the public, while democracy could be either abstract and confusing – especially in full autocracies – or subject to misinterpretation in competitive authoritarian regimes. Second, in authoritarian regimes, individuals have more regular exposure to policy frames than democracy frames because policy protests occur more frequently than democracy protests. This leads to the perception that the policy protest dominates the coalition, which shifts public support for the alliance to the baseline level of support for the policy protest. In other words, support for the alliance only goes as far as support for the policy protest.

To test this preregistered theory,³ I use an online survey experiment in Vietnam, randomly assigning respondents to three treatments: a hypothetical pro-democracy protest, a hypothetical land protest and a coalition in which the two hypothetical groups actively protest together on the streets to advocate for democracy to resolve the land issue. I find that when a hypothetical land protest cooperates with a pro-democracy protest, support for the coalition is significantly higher than that for the pro-democracy protest and almost the same as support for the land protest. In other words, I find that support for the coalition shifts to the baseline level of support for the hypothetical land protest. My mediation analysis suggests that the land–democracy coalition receives higher support than the pro-democracy protest because respondents perceive that the land protest, not the democracy protest, dominates the coalition.

This research has important implications for democratic consolidation. Policy–democracy coalitions sometimes draw considerable public support that can topple authoritarian governments but struggle to maintain democracy afterwards (Ong 2022). Extant research suggests that whether a movement can sustain democracy after a protest movement depends on the credibility and cohesiveness of the coalition (Ong 2022), the level of polarization within the opposition movement (Nugent 2020) and the ideological and cultural congruencies among coalition members (Goldstone 2011). The organizational composition of a coalition movement can also influence democratic consolidation (Pinckney et al. 2022). In line with this research, my findings point to a dilemma that a democracy movement will face. While joining policy-based movements may generate higher support due to the dominance of policy issues, democracy may not be sustained after defeating an authoritarian government. If democracy is only secondary to policy issues, it is challenging to maintain public motivation for democracy once policy issues subside.

Finally, this article contributes to the expanding body of literature on the impacts of protest framing on public support for protests (Arnon et al. 2023; Edwards and Arnon 2019; Truong 2024). Recent research has demonstrated that in passive coalitions, where pro-democracy advocates merely endorse policy-based protests on social media platforms without active participation, public support for the policy protests diminishes (Truong 2024). In contrast, this article suggests that when pro-democracy advocates and policy protesters actively protest on the streets, the public does not withdraw their support for the policy protests and the alliance. The findings imply that when democracy activists and policy movements co-protest, the policy issues may dominate the coalition. In contrast, perhaps, when pro-democracy advocates solely endorse policy protests on their platforms (Truong 2024), the visibility of democracy issues and democracy activists' identities may overshadow the policy issues, resulting in reduced public support.

Policy-based and pro-democracy protests

In this article, policy-based protests and pro-democracy protests are categorized according to: (1) their primary goals or demands; and (2) the specific solutions sought by protesters to achieve these goals.

First, I define ‘pro-democracy protests/movements’ as those in which participants demand that ‘countries adopt or uphold democratic elections’ (Brancati 2016) and seek to expand the political and civil rights essential for conducting open and competitive elections, such as freedom of assembly, freedom of the press and freedom of speech (Coppedge et al. 2011). In pro-democracy advocates' views, democratic institutions are necessary solutions to any problems their countries face. By demanding these democratic institutions, pro-democracy groups are anti-regime movements ‘opposed not only to the government but also to the institutions that make up the political itself’ (Brancati 2016: 6). In other words, a pro-democracy movement is a maximalist campaign in that it seeks regime change (Chenoweth and Lewis 2013).

Second, policy protests are distinct from pro-democracy protests in two ways. First, they aim to achieve concrete policy objectives such as halting illegal land evictions, ensuring adequate compensation for land taken (Kerkvliet 2019; O'Brien and Li 2006), securing higher wages, improving working conditions (Fu 2017), stopping environmentally harmful projects (Steinhardt and Wu 2016) and preserving local culture (Demick 2021). Second, they aim to achieve these goals through solutions such as improvements in policy implementation or policy reforms, rather than seeking regime change. For example, to stop illegal land eviction, some policy-based movements may focus on advocating for improvements in land policy implementation at the local level (O'Brien and Li 2006). Other policy-based protests may demand the government reform national laws and policies related to land, the environment or labour rights. For instance, in 2019 many citizens in Jordan, Algeria and Iraq protested to advocate for changes in macroeconomic policies to resolve unemployment (Saab 2019). Thus, unlike a pro-democracy movement, a policy movement is not a maximalist campaign as they often demand policy changes (Chenoweth and Lewis 2013).

Policy-based movements may sometimes endorse democracy, and pro-democracy movements may discuss policy issues (Brancati 2016). For instance, Sisters in Islam, an organization based in Kuala Lumpur, Malaysia, focuses on reforming policies to address injustices faced by women under Islamic law, but sometimes endorses clean and fair elections on their website in response to calls for electoral reforms from pro-democracy groups.⁴ Whether such a movement is categorized as policy-based or pro-democracy depends on its primary demands and its main advocated solutions, especially from the public perspective. While occasionally endorsing democratic institutions, Sisters in Islam is recognized as part of a broader policy movement in Malaysia that works towards advancing the rights of Muslim women through law and policy reforms. Similarly, the Duong Noi land movement in Vietnam is primarily focused on reclaiming agricultural land and securing fair compensation, although it occasionally expresses support for democracy activists targeted by the Vietnamese government.⁵ Conversely, a protest is classified as a pro-democracy movement if its main focus is advocating for democratic institutions, even if it sometimes endorses policy issues.

Next, what constitutes a protest coalition between a democracy movement and a policy-based movement? Holly McCammon and Minyoung Moon (2015: 327) state that social movement coalitions ‘occur when distinct activist groups mutually agree to cooperate and work together toward a common goal’. Using this definition, a democracy–policy protest coalition occurs when members of visibly distinct groups of different identities, specifically policy and democracy advocates, co-protest in pursuit of a common goal. I focus on a specific goal of the coalition: advocating for democratic institutions to address policy issues (policy protesters' grievances).

This concept helps distinguish a policy–democracy coalition from a policy protest or a democracy protest in two ways. First, a coalition exists when the public knows that two or more groups of markedly different identities have come together on the streets. This ensures that a protest featuring both claims is not necessarily a coalition if it is recognized as originating from a single group or from different groups of the same nature. For example, a protest primarily demanding the cancellation of an environmentally destructive project, alongside advocating for freedom of the press, is categorized as an environmental protest when organized by either a single environmental NGO or multiple NGOs dedicated to environmental causes. Second, for this article, a coalition is considered to exist when the protest advocates for democratic institutions to address policy goals. I focus on this specific coalition for two reasons. While policy activists and pro-democracy advocates sometimes join forces to advocate solely for democracy issues, these protests are rare (Brancati 2016). Although policy movements and pro-democracy advocates may collaborate on demonstrations focused solely on policy issues without promoting democracy as a solution, this may not always align with the primary goals of pro-democracy groups who seek regime change. Typically, when these groups unite, their shared demand often revolves around advancing democracy to address policy crises. This is because such crisis moments provide opportunities for pro-democracy activists to emphasize democratic institutions, which typically receive less public attention or are quickly repressed by the state during non-crisis periods. Thus, a coalition of a pro-democracy group and a policy group may maintain the maximalist goal of regime change, but this goal is couched within a non-maximalist policy claim.

What affects public attitudes towards democracy–policy protest coalitions? Research on public opinion towards protests in authoritarian regimes provides some clues. It indicates that using violent tactics tends to decrease public support for the protests and increase support for government repression (Edwards and Arnon 2019; Lupu and Wallace 2019). Protest framing can also impact support, with opposition to the status quo potentially undermining it (McLeod and Detenber 1999), while alignment with political values and public interests can boost support (Zuo and Benford 1995). Additionally, the government's response to protests (Frye and Borisova 2019; Pan and Siegel 2020) and the identity of protesters (Edwards and Arnon 2019) also shape public attitudes towards protests.

These factors are undoubtedly important in determining public attitudes towards protest coalitions. However, when pro-democracy activists and policy-based movements co-protest on the street, the joint collective action frames are crucial pillars. The following section provides a theory of how such frames affect public opinion on the policy–democracy protest coalitions.

Theory: Frame salience and support for policy–democracy protest coalition

When two groups of markedly different natures cooperate for a common goal, they need to devise a joint frame that aligns with the identity of each movement's members and appeals to the potential supporters of separate movements (Benford 1993; Deana and Quadagno 2009). Thus, the joint collective action frame should be broader than single movement members' frames but must convey each member's identity (Mayer et al. 2010). In general, joint collective action frames have two characteristics. First, they help render events or occurrences meaningful for the public by highlighting the problem, identifying who is to blame and suggesting potential solutions (Snow et al. 1986). Second, they reflect the combined identities and priorities of the members of the coalition of movements (Mayer et al. 2010).

When a policy-based movement cooperates with a pro-democracy group, a coalition-based collective action frame, including the identities of both groups, is created. Such a frame tends to have one component focusing on the origin of the policy issues that led to the policy protest and one that emphasizes that democratic institutions promoted by democracy activists are needed to resolve the policy issues. For example, the 2011 Yemeni uprising was composed of a broad-based coalition between groups protesting about topics related to the economy and corruption and pro-democracy groups that advocated for free and fair elections to resolve these issues (Brancati 2016). The joint frame emphasized that the grievances of policy-based protesters resulted from high unemployment rates and poverty and that democratic institutions were required to resolve these problems.

Because individuals not involved in the movement may have limited resources – such as time, knowledge and information to assess the issues the coalition raises thoroughly – they may rely on a shortcut or cue to form their opinions (Iyengar 1989; Mason 2015). Research on the psychology of framing effects suggests that whether elites' framing influences public opinion depends on the availability and accessibility of the frame (Chong and Druckman 2007). Availability refers to the extent to which the public understands the concept in the frame (Eagly and Chaiken 1993; Higgins et al. 1982), while accessibility refers to the extent to which individuals are regularly exposed to the frame (Zaller 1992). Inspired by this research, I argue that how the policy–democracy frame influences public opinion will depend on which component of the joint frame is more available and accessible to the public. I contend that the public is more likely to understand and have regular exposure to the policy frame than the democracy frame. Thus, they may perceive that the coalition is about a policy rather than a democracy issue, leading them to form their attitudes towards a policy–democracy movement coalition based on their opinions towards the policy-based movement.

The comprehensibility of policy protest frames

Scholars have long suggested that whether a frame of complex issues affects public opinion depends on the extent to which the public understands the concepts the frame touches upon (Eagly and Chaiken 1993; Higgins et al. 1982). Following this logic, the component of the joint frame that is more understandable to the public will dominate the coalition. Between the policy and democracy components, the public in authoritarian regimes should find the former more understandable.

First, the concepts of democracy, free speech, free press and free and fair elections could be highly abstract and confusing or subject to interpretations for citizens in authoritarian regimes. In single-party regimes where opposition parties are banned, unfamiliarity with any liberal democratic values and institutions makes the concept of democracy highly inaccessible. The absence of opposition parties allows the ruling party to define democracy in a way that benefits itself, often demonizing the notion of liberal democracy characterized by free and fair elections. In Vietnam, democracy is often referred to as ‘democratic centralism’, a term that is highly confusing. The government defines Vietnam's democracy as a system where people elect representatives to run the government on their behalf, and policies ‘are soundly established upon the summation of experience and opinions of all party members’ (Vu 2021). Similarly, Chinese citizens express a high level of confusion about the definition of democracy (Ma 2013). These examples show that the popular definition of liberal democracy is highly abstract and inaccessible in single-party regimes.

In competitive authoritarian regimes where multiparty elections are held regularly but the playing field is uneven and biased towards the ruling party (Levitsky and Way 2010), the concept of democracy can be subject to misinterpretation. Many citizens in such regimes believe they already live in a liberal democracy (Letsa 2019). To convince citizens that their country is genuinely democratic, the ruling party often emphasizes regular elections and multiparty competition as norms. It also highlights the opposition parties' perceived inability to generate economic development as a reason why these parties struggle to win elections (Letsa 2019). Therefore, in competitive autocracies, citizens tend to overestimate the level of democracy, allowing autocrats to avoid providing more democratic reforms because citizens perceive their demand for democracy as already fulfilled (Yeung 2023). Thus, the ruling party's ability to persuade citizens they already live in a democracy makes the pro-democracy frame of a policy–democracy coalition less resonant with the public.

In stark contrast, the policy component of the policy–democracy frame is more straightforward and understandable to a layperson, regardless of whether they live in a full or competitive authoritarian regime. For example, exposed to a frame related to a land dispute, an individual can picture in concrete terms what happens (i.e. villagers have their land taken by local authorities, and the government does not provide sufficient compensation) even though they may not have a deep understanding of land issues and policies.

Most contemporary authoritarian regimes derive legitimacy from promising economic growth, efficient policies and the provision of public goods (Gerschewski 2023). In these regimes, citizens' priorities often centre around economic issues such as wages, homeownership, land and environmental concerns. Consequently, the public prioritizes tangible policy matters over abstract democratic ideals. As a result, policy protests gain more attention and comprehension among the populace, rendering them more dominant.

Precisely because the policy frame is more understandable to the public than the democracy frame, regime dissidents who advocate for highly abstract issues such as democracy, free speech and free press typically link these concepts with issues that align with the public's interests, beliefs and experiences (Snow et al. 1986). Evidence shows that even when individuals are not committed to democratic values, they still take to the streets if a particular component of the protest frame – such as corruption or antipathy for the incumbent – makes sense to them, rather than the democracy component (Beissinger 2013; Ong 2022).

My argument does not claim that the public in democracies universally understands democracy frames better than policy frames. In competitive contexts with diverse and opposing frames, the effect of framing depends largely on individual characteristics such as political knowledge and motivation (Chong and Druckman 2007).

The accessibility of policy protest frames

Scholars argue that a frame influences attitudes and opinions more effectively with regular exposure (Nabi 2003; Zaller 1992). These scholars suggest that individuals are influenced by the frame they hear most often without conscious evaluation of the content. Inferred from this logic, the component of the policy–democracy coalition frame to which the public is more frequently exposed has a stronger effect on public attitudes towards the coalition. It is crucial to note that repeated exposure to a frame does not guarantee its appeal. An individual may grow weary of a frame encountered frequently. Exposure is likely to shape their opinion if it is intense enough to evoke either attraction or fatigue. An individual under authoritarian rule is more likely to be exposed to policy frames than democracy frames.

First, in authoritarian regimes, people are more often exposed to policy-based protests than protests advocating for democracy. Policy protests typically navigate the boundaries of legality within an authoritarian framework (O'Brien and Li 2006). In countries such as Vietnam and China, central authorities often tolerate protests that target local officials' legal transgressions to some degree (Kerkvliet 2019) because controlled protests can help autocrats gather information about local affairs and local authorities (Lorentzen 2013).

On the other hand, democracy protests are much less likely to occur than policy-based protests because they are either repressed or restricted to only educated urban citizens (Brancati 2016). Because the public under authoritarian rule is much more likely to encounter policy-based protests than democratic protests, they encounter the frames of policy protests more frequently than the other.

Second, autocrats attempt to emphasize that grassroots disputes are related to policy issues, not democracy issues. Indeed, authoritarian governments allow the discussion of policy issues to a certain extent. Because autocrats have strategic incentives to keep their citizens happy (Svolik 2012), they allow citizens to voice opinions on policy issues (Chen and Xu 2017) and even to complain about poor government services and policies so that they can figure out how to optimize performance (Botero et al. 2013). Although policy-based protests are often painted negatively, authoritarian governments publicly discuss the relevant policies (Kerkvliet 2019). What matters is that they emphasize that these disputes are purely related to policy issues, not to democracy issues.

The analysis above suggests that: (1) the policy component of the policy–democracy frame is more understandable to the public than the democracy component; and (2) the public is much more likely to encounter policy protest frames than democracy protest frames. For these reasons, when a policy frame is combined with a pro-democracy frame, people pay more attention to the former, which leads to the perception that the policy frame is more salient and dominates the coalition. Because the public thinks the alliance is about a policy rather than a democracy issue, support for the coalition defaults to the baseline level of support for the policy protest. The causal mechanism can be summarized as:

Exposed to a policy–democracy protest coalition → Perception about which protest is dominant → Support for the coalition

This leads to the following hypotheses:

Hypothesis 1: In an authoritarian regime, support for a coalition between a policy protest group and a pro-democracy group defaults to the baseline level of support for the policy protest.

Hypothesis 2: Support for the coalition defaults to the baseline level of support for the policy protest because respondents perceive the alliance as focused on the policy rather than the democracy issue.

Vietnam context

I test these hypotheses in Vietnam. Vietnam is a relevant and interesting case for several reasons. First, since the mid-1990s, Vietnamese citizens have taken to the streets to raise concerns about various policy issues, including land, labour, nationalism and the environment (Kerkvliet 2019). Policy-based protests have become prominent and contentious political issues in contemporary Vietnam. Second, while pro-democracy movements have challenged the dominant role of the Vietnamese Communist Party to a certain extent, they typically act in isolation and do not receive high levels of public support (Kerkvliet 2019).

Third, it is evident that pro-democracy activists have attempted to build coalitions with or to endorse policy protests (Kerkvliet 2019). Vietnamese policy-based protesters also sometimes seek assistance from independent lawyers and journalists, typically pro-democracy activists, to help them file complaints and spread their grievances to the public (Kerkvliet 2019). While still rare, informal protest coalitions between policy-based movements and pro-democracy activists have sometimes occurred in the last ten years. For example, pro-democracy activists and nationalists joined forces to protest Chinese aggression in Vietnamese territory multiple times during the 2010s (Nguyen 2021). Similarly, pro-democracy activists and environmentalists protested against the mass fish deaths in coastal provinces in 2018 and the bauxite project in the Central Highlands in 2009 (Nguyen and Datzberger 2018). What is striking is that despite the prominent element of democracy in these protest coalitions, they attracted significant public support and participation.

Research design

To test these hypotheses, I employ an online survey experiment administered to 1,209 respondents in Vietnam. I hired TGM Research to recruit the respondents through local Vietnamese market research firms with proprietary survey panels. The survey used the Qualtrics platform and was collected in January 2022. To mitigate the concerns that the sample is skewed towards a particular population, I set quotas for groups that are typically underrepresented in online survey experiments, such as women, older residents, respondents without a college degree, and respondents living outside Hanoi and Ho Chi Minh City – the two biggest cities in Vietnam.

Since studying public opinion towards protests under authoritarian rule is highly sensitive, several measures were taken to ensure participants' safety and comfort in expressing their views. First, using an online survey experiment offers some advantages. Studies show that respondents tend to answer survey experiments honestly (Mummolo and Peterson 2019), and conducting surveys online potentially mitigates concerns of preference falsification, as respondents remain anonymous to enumerators (Mutz 2011). Second, clear instructions were provided at the beginning of the survey, emphasizing the confidentiality and anonymity of the response, the voluntary nature of participation and the absence of consequences for opting out of the survey early and expressing differing viewpoints. Third, as shown below, hypothetical protests reduce the issue sensitivity because respondents were not asked to provide opinions on a real-world event. These measures cannot entirely rule out the possibility of preference falsification; as shown below, it is evident that this may not drive the results.

Regarding the survey instrument, before the treatment, all respondents were first asked several questions on their demographic characteristics such as age, education level, occupation, profession, income, political interest, national pride, trust in the government and willingness to take risks. After being asked questions that measure these covariates, respondents were randomized into one of three protest frames.

The policy protest frame group read the following information:

Recently, villagers in Dong Xuan commune, Dong Ha province, protested outside the Commune People's Committee against the commune's decision to reclaim 100 hectares of agricultural land and convert it for an industrial project.

Protesters highlighted that commune officials, in cahoots with investors, acquired the land illegally by not compensating farmers properly, according to the 2013 Land Law.

I used a hypothetical land protest because of the sensitivity of land issues in Vietnam. Although this land protest is hypothetical, it represents a typical land dispute frame in Vietnam. Benedict Kerkvliet (2019) illustrates that land protesters in Vietnam commonly portray land disputes as local policy issues related to illegal evictions and inadequate compensation by local authorities. In addition, while Vietnamese citizens have raised concerns about different policy issues, land is the most contentious political issue.

The pro-democracy protest is framed as follows:

Recently, a group of pro-democracy advocates organized a protest in Dong Xuan commune, Dong Ha province, to advocate for significant regime reforms such as democracy, more representation, pluralism, citizens' participation in policymaking and more press freedom.

The purpose of this treatment group is to examine public support for pro-democracy activists acting alone without linking their causes with policy-based movements and policy issues. This treatment is designed to be as realistic as possible. First, in Vietnam, democracy, pluralism, citizens' participation and freedom of the press are the demands that pro-democracy advocates typically focus on. When pro-democracy dissidents protest alone or do not link their causes with policy issues, their demands are highly abstract and broad. For example, in July 2008, a group of Vietnamese pro-democracy advocates protested by hanging a banner around 3 metres long over a Hanoi bridge upon which they demanded democracy (United Press International 2009). Many activists seek radical demands such as multiparty elections (Kerkvliet 2019). However, because these issues are highly sensitive, including them in the survey may appear illegal. As such, I include other aspects of democracy that the regime is more tolerant of. Second, one potential limitation of this group is that demands of different levels of extremity are put together. For example, calling for more citizen participation may be more moderate than calling for democracy and freedom of the press, which may affect public opinion differently. However, because research on Vietnamese politics suggests that different democratic demands do not influence public attitudes in significantly different ways (Truong 2024), I incorporate all of them into the treatment to enhance the statistical power of the experiment.

The land–democracy coalition frame read the following information:

There was a protest in Dong Xuan commune, Dong Ha province.

Dong Xuan villagers teamed up in a coalition with a group of pro-democracy activists to protest and advocate for significant regime reforms such as democracy, more representation, pluralism, citizens' participation in policymaking, and more press freedom to address a land issue that Dong Xuan citizens are faced with.

Villagers in Dong Xuan commune, Dong Ha province, protested outside the Commune People's Committee against the commune's decision to reclaim 100 hectares of agricultural land and convert it for an industrial project. Protesters highlighted that commune officials, in cahoots with investors, acquired the land illegally by not compensating farmers properly, according to the 2013 Land Law.

This design closely follows the concept of a policy–democracy coalition discussed earlier. First, the treatment makes it clear that two visibly distinct movement groups of different identities are protesting together, including villagers in Dong Xuan commune whose policy grievances focus on illegal land eviction, and pro-democracy activists who advocate for democratic institutions. Second, the two distinct groups protest together to advocate for democracy to address the policy (land) issues.

Following this concept, while the protest coalition differs from the policy protest by seeking democratic institutions to address land issues, both protests ultimately aim to resolve land issues. One concern is how observers perceive the coalition's emphasis on democracy without interpreting the land issue as dominating the protest. It is important that the coalition treatment is designed to be as realistic as possible. In Vietnam, pro-democracy groups sometimes work with policy protesters to propose democratic solutions during significant policy crises. While such coalitions aim to address policy issues, their emphasis on democratic institutions as solutions is often conspicuous and quickly draws government attention, resulting in the arrest of protesters and negative portrayals in state media. These protests, from the perspectives of the government and political observers, become highly politicized beyond mere policy grievances, evolving into direct challenges against the regime. Despite the contentious nature of advocating for democracy, these protests frequently attract widespread public support. For instance, during the 2016 environmental crisis involving mass fish deaths in Vietnam's coastal provinces, pro-democracy advocates seized the moment to join aggrieved residents in demanding government transparency and freedoms to address the environmental issue. This protest coalition garnered substantial public support despite its prominent democracy-focused narrative. In such protest coalitions, the government may see democracy as the central issue. The coalition treatment aims to examine whether the average citizen also sees the democracy component when it is advocated to address a policy issue.

Another concern of the protest coalition frame is that respondents may perceive the frame component they read first to be more salient. To address this concern, I randomized the order of the second and third paragraphs in the land–democracy coalition group. This ensures that frame order does not affect which component is perceived to be more prominent.

Following Truong (2024), I used two dependent variables which measure respondents' support of the protests and respondents' perception of the justification of the protests.

Protest support: To what extent do you support the protest you have just read? This variable is measured on a six-point Likert scale: 1 = Strongly oppose, 2 = Oppose, 3 = Somewhat oppose, 4 = Somewhat support, 5 = Support, 6 = Strongly support.

Protest justification: To what extent do you think the reason why protesters took to the street was justified? This variable is measured on a six-point Likert scale: 1 = Strongly unjustified, 2 = Unjustified, 3 = Somewhat unjustified, 4 = Somewhat justified, 5 = Justified, 6 = Strongly justified.

The theory suggests that support for the coalition moves to the baseline level of support for the policy protest because the public perceives that the policy protest frame dominates the coalition. In other words, the hypothetical land component of the alliance is perceived to be more salient than the pro-democracy component. To probe this mechanism underlying respondents' support for the protest, I asked and randomized the order of the following questions:

In your opinion, what is the salient issue of this protest? This variable has two values: land issue or democracy issue.

To what extent do you agree this protest is mainly about a land issue? This variable is measured on a six-point Likert scale: 1 = Strongly disagree, 2 = Disagree, 3 = Somewhat disagree, 4 = Somewhat agree, 5 = Agree, 6 = Strongly agree.

Table 1 shows the treatment conditions and the number of respondents in each condition. Figure 1 shows the distribution of the two outcomes. Both have a relatively normal distribution, suggesting that respondents are equally willing to express their support for and opposition to the hypothetical protests. Respondents equally perceive the root cause of the protest to be justified and unjustified.

Table 1. Experimental Conditions

Note: The table shows the different treatment groups. The hypotheses predict that support for the frame that combines land and pro-democracy protests will not be significantly different from the land protest frame, but significantly different from the pro-democracy frame.

Figure 1. Distribution of Protest Support and Protest Justification.

First, I examine whether the three treatment groups are balanced across important covariates asked before the treatments. Appendix 1 in the Supplementary Material suggests that the three treatment groups are not significantly different in most covariates except for national pride. Because national pride could impact public support for protests, I will add covariates in the regression models in the robustness check.

To check whether respondents pay attention to the treatment texts, I asked respondents to recall the name of the hypothetical province where the protest occurred. Appendix 2 shows that most respondents successfully identified the name of the hypothetical protest. Most respondents recognized that the protest happened in the hypothetical province of Dong Ha.

Results

Hypothesis 1 states that support for the land–democracy protest coalition defaults to the baseline level of support for the land protest. We should expect that the mean support for the coalition protest is not significantly different from the land protest while substantially different from the pro-democracy protest. In the primary analysis, all respondents are included in the sample, irrespective of passing the manipulation check to avoid bias in the results (Aronow et al. 2019). Figure 2 plots the means of three protest groups.

Figure 2. Means of Protest Support/Perceived Protest Justification across Treatment Frames.

Note: These figures show the means of support for the hypothetical protests (95% confidence interval).

The first thing to notice is that support for the pro-democracy protest frame is significantly lower than that for the land protest frame (p < 0.001). While the difference is relatively small, it is substantively meaningful. Respondents in the land protest expressed some support for the protest (mean response of 3.8, closer to 4, which is ‘Somewhat support’), while respondents in the pro-democracy frame somewhat opposed the protest (mean response of 3.3, closer to 3, which is ‘Somewhat oppose’). The pro-democracy protest is also perceived to be significantly less justified than the land protest (p < 0.001). Respondents in the land protest frame perceived the protest as somewhat justified, estimating a mean response of 4.0 on a six-point scale. With a mean response of 3.4, respondents in the pro-democracy group viewed the protest to be close to being unjustified.

More importantly, although support for the pro-democracy protest when it operates independently is lower than support for the land protest, linking the pro-democracy protest with the land protest increases support for the coalition to the level of support for the land protest. In other words, support for the combined protest moves to the baseline level of support for the land protest. The mean values of both dependent variables are almost the same across the land protest and the land–democracy coalition. There are no significant differences (p < 0.67) in terms of expressing support for the protests between the land group (mean response: 3.8) and the coalition group (mean response: 3.7). Similarly, the two groups do not significantly differ (p < 0.85) regarding perceived protest justification. Both groups view the protest as somewhat justified, estimating a mean response of 4.0 for the land group and 3.9 for the group that combines land and democracy issues. These results suggest that Hypothesis 1 is supported. Given the sensitivity of democracy and protests in Vietnam, preference falsification is possible. However, similar results between the pure policy treatment group and the coalition group, both of which contain policy information, indicate that any preference falsification does not drive the treatment effect. Because democracy is a sensitive issue, if preference falsification drives the results, we might anticipate respondents concealing their true preferences for democracy and showing less support for the coalition protest that involves the sensitive element.

Robustness check

Recall that I randomized the order of the land frame and the pro-democracy frame to ensure that respondents' perceptions are not affected by which frame they see first. I rerun the main analysis for each group. Appendix 3 represents the results of respondents who read the land frame first in the coalition protest. The results remain. Among those exposed to the land frame first, support for the land–democracy protest coalition is significantly higher than the pro-democracy protest and not different from the land protest. Appendix 4 represents the results with respondents reading the pro-democracy frame first in the coalition. Similarly, support for the coalition protest among those exposed to the pro-democracy frame first is not different from the land frame and significantly different from the pro-democracy group.

To address concerns about multiple testing and model noise when analysing two dependent variables separately, I rerun the analysis using an averaged index of support for protests and perceived protest justification as the outcome variables, with randomized frames and identities as the independent variables. Appendix 5 demonstrates that the results remained consistent.

Appendix 1 shows that groups are imbalanced in the level of national pride, an important factor that could affect support for protest. I examine whether this imbalance can affect the result by running an OLS regression with support for the protests and the perception that the protests are justified as the outcomes and dummy variables for the randomized frames as the independent variables. I also include covariates in Appendix 1 in the model. As Appendix 6 shows, results do not change when accounting for the potential lack of balance of covariates across treatment groups.

Another concern is that the results may be affected by the inclusion of respondents who did not pay attention to the treatments. Despite recent cautions against exclusively testing participants who pass the manipulation check (Aronow et al. 2019), I omit those who fail to identify the hypothetical province where the protests occurred correctly. As Appendix 7 shows, the results remain even after dropping those who did not respond to the treatment.

Because my theory is based on the extent to which the public (1) finds movement frames understandable and (2) has regular exposure to the frames, these characteristics may be different across subgroups of the population. To test potential heterogeneous effects, I interact protest frames with education, residence, political interest, party membership and trust in the government. Appendix 8 shows that the main results remain. Importantly, none of the interaction terms is significant, suggesting that opinions towards protest frames do not depend on respondents' demographic characteristics.

My sample does not represent the Vietnamese population regarding party membership. While about 23% of respondents are members of the Communist Party, only 5% of the population are party members. Party members tend to be more negative and sceptical of pro-democracy protests, which may affect their support for the coalition. I rerun the analysis, dropping all party members and including all the covariates. Appendix 9 shows that the results remain even after dropping party members.

Mechanism

Why is support for the land–democracy coalition not significantly different from that for the land protest? Hypothesis 2 states that support for the joint protest defaults to the baseline level of support for the policy protest because respondents perceive that the joint protest is about land rather than democracy issues. The causal mechanism can be summarized as follows:

Information about the land–democracy coalition → Perception about land as a dominant issue → Support for the coalition

Descriptively, we would expect that more respondents in the land–democracy group choose land as the central issue of the protest when asked which issue dominated the protest. As Figure 3 shows, in the group that combines land and pro-democracy frames, most respondents perceive that the joint protest is about land rather than democracy. Whether respondents read the land protest or the democracy protest first, they equally perceive the joint protest to be about a land issue. This may explain why support for the joint frame moves to the baseline level of support for the land protest.

Figure 3. Distribution of Protest Issues.

To formally test this mechanism, I conduct a causal mediation analysis outlined by Kosuke Imai et al. (2011). I use the question asking respondents how much they agree that the protest is about a land issue to measure the mediating variable. Figure 4 suggests that the treatments (protest frames) affect respondents' perception of whether the protest is mainly about a land issue. The land protest and the land–democracy protest equally consider the land issue as the salient issue that protesters raise. It makes sense that respondents in the pro-democracy frame are significantly less likely to perceive that the pro-democracy protest is mainly about a land issue. However, many respondents in the pro-democracy protest tend to think that the land issue dominates even though land is not mentioned in the protest frame. One potential explanation is that land disputes are popular and contentious political issues, frequently receiving endorsements from pro-democracy activists. Thus, some respondents may have associated democracy issues with land issues.

Figure 4. Effect of Protest Frame on Perceived Salient Protest Issue.

Note: This figure shows the means of treatment groups regarding their perception of whether the protest is about a land issue (95% confidence interval).

I use the ‘mediation’ package in R to implement the mediation analysis (Tingley et al. 2014). I conduct two OLS regressions. The first regresses the impact of the protest frames on the mediating variable and covariates in Appendix 1. The second regresses the two dependent variables on the protest frames, the mediating variable and covariates. Then, I assess the direct effect and the degree to which the effect is mediated through the mediating variable. I focus on comparing the pro-democracy protest and the land–democracy coalition. This is because there are no differences between the land protest and the land–democracy protest in terms of protest support and perceived protest justification. In other words, this section examines whether respondents in the land–democracy coalition express higher support for the protest than those in the pro-democracy protest because they perceive the land frame dominates the coalition. Table 2 shows the results.

Table 2. Mediated Effects of Protest Frames through Perception of Protest Issues

Note: The mediator is the extent to which the protest is about a land issue on a six-point Likert scale from ‘Strongly disagree’ to ‘Strongly agree’. This analysis compares the pro-democracy protest (control condition) with the land–democracy coalition (treatment). CI = confidence interval; ACME = average causal mediation effect. All decimals are exactly as shown in R, with no additional rounding.

The results show that the treatment impacts support for protest and the perception that the protest is justified nearly entirely through respondents' perception of what the protest is about. In other words, support for the land–democracy coalition is higher than that for the pro-democracy protest because respondents perceive that the alliance is about a land issue. The average causal mediation effect (ACME), which is the effect of the protest frame on respondents' support for the protest and respondents' perceived protest justification that goes through their perception of what the protest is about, is positive and significant. Put simply, compared to the pro-democracy protest, respondents in the land–democracy coalition are more likely to perceive that the land issue dominates the protest, increasing their support for it or their perception that the protest is justified. The indirect effect is relatively large. The indirect effect is 0.4337 on protest support, at 0.31 standard deviation of the dependent variable. The ACME of 0.4029 on perceived protest justification is at 0.30 standard deviation of the dependent variable.

It is important that once I have accounted for the impact on the perception of the salient protest issue, the direct effect of the protest frame on support for the protest (−0.0299) and perceived protest justification (0.0673), which does not go through the mediating variable, is no longer statistically significant. In other words, once respondents' perception of the salient issue is considered, the treatment has nearly no direct effect on protest support or views towards protest legitimacy.

However, these findings are obtained under the sequential ignorability assumption (Imai et al. 2011). To produce an unbiased estimate of the ACME, we must assume there are no unobserved covariates before and after treatments are assigned, which confound the relationship between respondents' perception of which frame is salient (the mediator) and support for the protest or protest justified (the outcomes).

As such, I perform a sensitivity analysis to check whether the sequential ignorability assumption is violated. The purpose of conducting a sequential ignorability analysis is to examine the degree to which a correlation between the errors of the mediation model and the errors of the outcome model would impact the estimated mediated effect. Appendix 10 shows that, for both dependent variables, the confidence intervals of the ACME contain 0 when ρ equals 0.2 and 0.4, and the ACME is 0 for a ρ value of 0.3. Thus, the estimated mediation effects are sensitive within these values of ρ. Because the values are relatively not so close to 0, the perception of whether a land issue dominates the policy–democracy coalition may not be highly sensitive to a potential correlation of error terms.

In general, the results show that when a pro-democracy frame is combined with a land dispute frame, people think that the coalition is about a land issue, or in other words, the land frame dominates the coalition. Furthermore, this effect mediates an overall increase in support for the land–democracy coalition and perceived coalition justification (to the baseline level of the land protest). While a single mechanism design cannot rule out other possible mechanisms, the results suggest that perceptions about salient issues are important in mediating the relationship between protest frames and public attitudes toward protests.

One could argue that the land issue is more dominant in the coalition group because the policy treatment is more concrete, specific and detailed, while the democracy treatment is highly general and abstract. However, in a real-world scenario, when acting alone, pro-democracy activists typically focus on abstract post-materialist values, while policy protests focus on specific policy issues. Because democracy issues tend to be abstract and open to individual interpretation, they are less understandable and accessible to average citizens compared to policy issues. Thus, when democracy issues are combined with policy issues, the democracy issues become less dominant.

An alternative interpretation of the results is that support for the policy–democracy coalition is higher than for the democracy protest because the public views democracy as a solution to the land issue. When pro-democracy advocates act alone, the public may not see a reason to support democracy. However, when exposed to a contentious policy issue, they may support democracy as a solution. To rule out this interpretation, I asked respondents about the need for democracy in Vietnam. If the interpretation is correct, respondents in the land protest and land–democracy protest should be more likely to support democracy than those in the democracy protest. As Appendix 11 shows, there are no significant differences in support for democracy across the three treatments.

Conclusion

When policy-based protests and pro-democracy groups unite to advocate for regime reforms to resolve policy issues, they create a coalition-based collective action frame that includes both policy and democracy components. I argue that the policy component, being more available and accessible to the public, influences public attitudes towards the coalition more than the pro-democracy component. Using an online survey experiment administered to 1,209 internet users in Vietnam, I find that support for a hypothetical land–democracy coalition is almost the same as support for the land protest and is significantly higher than that for the democracy protest. My mediation analysis indicates this difference is because respondents perceive the land protest as dominating the coalition.

My findings suggest that the public plays an important role in sustaining democracy. While pro-democracy advocates can bring people to the streets by linking their causes with relatable and comprehensible issues, they obviously will encounter enormous challenges in sustaining democracy when the public is not committed to democracy issues. This is a dilemma that pro-democracy activists face when building coalitions with other movements.

One question is whether, in these protests, policy protesters and pro-democracy activists actively collaborate behind the scenes or democracy activists show up when policy protests occur. From the public's perspective, it should not matter because when democracy advocates and policy protesters attend demonstrations, the public should see both policy and democracy frames.

If what garners public support for a democracy–policy coalition is the policy issue, what benefits do pro-democracy activists seek in collaborations with other movements? I argue that building coalitions with policy protests still benefits pro-democracy dissidents.

First, cooperation allows pro-democracy activists to promote democratic values to the aggrieved population. Collaboration with policy protests helps regime dissidents reach a wider audience beyond their urban and educated constituencies. Second, activism can spark more activism, gradually influencing policy, shaping social identities, changing public opinions and boosting political engagement over time (Mansbridge 1986; McCann 1994; O'Brien and Li 2006). Thus, persistent cooperation with other movements may change the public's perception of democracy issues.

Should policy protesters collaborate with pro-democracy advocates? The results indicate that such cooperation offers multiple benefits to policy protesters. The public supports policy protesters in an active coalition with pro-democracy advocates. This helps protesters tap into democracy dissidents' resources, including connections with political elites, interpretation of legal documents, and international support and funding.