Representational deprivation: niche parties, niche voters and political protest

Abstract Political participation has increasingly taken unconventional forms in advanced democracies. At the same time, the traditional party systems of Europe have been upended by the recent emergence of successful niche parties. While some studies have examined partisanship and protest in tandem, the literature lacks a theoretical framework that ties niche voters to political protest. This article argues that niche voters are more likely to protest than mainstream voters, especially when the politics of governing coalitions results in niche parties being unable to represent their voters. Specifically, niche voters are more likely to protest when their party joins a coalition with a PM who is ideologically distant from the niche party’s position. Using all nine waves of the European Social Survey in Western Europe, the findings broadly support this argument. These findings have important implications for understanding conditions under which people engage in political protest, especially voters of niche parties. Supplemental data for this article can be accessed online at https://doi.org/10.1080/01402382.2022.2040237 .

In the summer of 2020, following nearly six years of the Swedish Green Party serving as a junior governing party alongside the Social Democrats, Swedish voters took to the streets to protest the inadequate response of the government at addressing climate change. Particularly, voters blame the Swedish Greens for entering the governing coalition and being unable to adequately champion a strong stance on addressing climate change. 1 In Germany, similar protests have emerged over what environmentalists view as weak climate policy from former Chancellor Angela Merkel. 2 Over the past decade, protests have taken place across Europe on a variety of issues. In much of Western Europe, these protests have been about new politics issues such as immigration during the height of the refugee crisis in 2015, the environment in Sweden in 2020, Brexit in the United Kingdom in 2016, or the mobilisation of populist interests as seen by the yellow vest protests in France in 2018.
Similarly, as we have seen an increase in the use of unconventional political acts (see Grasso 2016), a parallel trend has been large scale transformations in European party systems. Mainstream party families such as centre-left social democrats have been on the decline, while radical right and green niche parties have been on the upswing in national and regional elections throughout Europe (see De Vries and Hobolt 2020). Therefore, the goal of this article is to explore the relationship between these two trends.
Studies have primarily examined the relationship between these two trends on conventional forms of behaviour such as turnout (see Schwander et al. 2020) or examined the relationship between these parties and broader social movements (see Pirro and Gattinara 2018). Some studies have looked at partisanship and protesting, focussing on either green parties or radical right parties. For instance, Finkel and Opp (1991) apply a rational choice framework and argue that parties can generate incentives to protest by encouraging certain forms of behaviour, but their study importantly predates the aforementioned transformations in European party systems. More recent work views the party as the mobilising agent, opting to endorse protests for their own goals (Borbáth and Hutter 2021). Importantly, this work examines the conditions in which non-mainstream parties endorse protests from a party organisation perspective, not when their voters take to the streets. To fill this gap, this article explores the micro-level dynamics that drive individuals to protest through their partisanship. In what ways do niche voters and mainstream voters have different repertoires of political action? Under what conditions do niche voters differ in their political participation than mainstream voters?
I argue that voters take to the streets when they feel that they are not being adequately represented by the existing political system. Building on grievance models of political protest, I argue that the politics of governing coalitions contributes to the development of representational grievances among disadvantaged niche voters. There is a rich literature on the consequences of governing as a junior partner -a role that is often the highest a niche party is able to reach in government-for voter perceptions of the party (see Spoon and Klüver 2017), a party's ability to fulfil its campaign pledges (Thomson et al. 2017), and subsequent electoral consequences for junior partners (Klüver and Spoon 2020). By drawing on the established consequences for junior partners, this article develops a theoretical framework that ties niche voters together based on their shared representational disadvantages that drives them to protest more than their mainstream counterparts.
More specifically, I argue that voters protest when they feel that their party, or the issue they care about, is not represented by the governing coalition. When niche parties join coalitions as junior partners, they often must compromise their true positions for those of the Prime Minister's (PM) party. As a result, niche party voters, that is supporters of green, regional, or radical right parties, who care passionately about a given issue (Adams et al. 2006;Poguntke 1992) participate in unconventional ways because the issues they care about are not adequately represented by their party in the governing coalition. This effect is strongest when the PM is ideologically distant from the niche party.
In the sections that follow, I discuss the literature that connects parties to political protests and the role of deprivation in protest participation. Drawing on this work, I develop my theory of representational deprivation to explain participation in political protests among niche voters. I then test my theoretical expectations using nine waves of the European Social Survey (ESS) and find support for my expectations. Niche voters are more likely to protest, especially when the niche party joins a governing coalition as a junior partner and is ideologically distant from the PM. I conclude by discussing the implications of these findings for our understanding of political behaviour, niche voters, and coalition politics.

Niche voters and political protest
One prominent strand of the literature on the intersection of parties and protests focuses on party organisation and 'movement parties' (Kitschelt 2006). Movement parties are a type of party which are born out of social movements and have developed organisational capacities around social movements (see also Pirro and Gattinara 2018). These parties operate in both electoral and protest arenas and their goals are both to win office and mobilise voters in contentious politics. Importantly, this strand of the literature focuses on the party side intersections of parties and protests, focussing on their organisation and operational strategies, such as whether to endorse and mobilise their voters around movements (see Borbáth and Hutter 2021). Importantly, the question remains over the conditions in which supporters of certain types of parties choose to participate in political protests and whether there are heterogenous repertoires of political action at the micro-level.
Few studies have explicitly examined whether the party one supports shapes how she participates. Finkel and Opp (1991), for instance, compare partisanship and an incentives model based on public good provisions and instrumental incentives. They demonstrate that people protest largely for instrumental reasons and that partisanship does not actually lead to heightened protest activities among mainstream parties unless the party provides an incentive for the individual to protest. Interestingly, Finkel and Opp (1991) do find that green parties are the sole exception to this pattern. They find significant positive effects of feeling close to a green party even when controlling for the more powerful party-incentives model. This is not surprising, as green parties emerged from the rise of New Left social movements in the 1970s and 1980s that brought like-minded individuals around new left and environmental issues together, promoting further organisation into the electoral arena (Milder 2015;Röth and Schwander 2021). This in turn has led to them promoting protest and institutionalising grassroots democracy (Hutter and Borbáth 2019;Poguntke 1992). Overall, while the relationship between mainstream partisanship and protesting is minimal, the link between green partisans and an increased likelihood of protesting is well established.
While the literature for mainstream and green party voters and political protest is scarce, recent scholarly attention has focussed on the radical right as a unique party family. In a working paper, Amnå and Fitzgerald (2020), for example, demonstrate that radical right parties boost political engagement by appealing to voters that would otherwise be disengaged from the political process. This increased engagement has manifested itself in a heightened propensity for unconventional participation. Anduiza et al. (2019) find that those with populist attitudes are more likely to participate in non-institutionalised forms of participation such as online activism, signing a petition, and demonstrating. Similarly, Pirro and Portos (2021) find that populism mobilises individuals in non-electoral forms of participation more than non-populists, especially among left-wing populists. Several ethnographic studies of radical right groups and organisations further stress the mobilising power of the radical right at engaging voters around a common anti-establishmentism (Blee 2007;Meadowcroft and Morrow 2017;Pirro and Gattinara 2018;Teitelbaum 2020).
As the extant literature demonstrates, mainstream voters are not linked to political protest while green and radical right voters have been associated with an increased propensity to engage in protests. However, the increased likelihood of green and radical right voters to protest is often linked to different underlying motivations that ignores the conceptual similarities that green and radical right parties share as niche parties (see Meguid 2005Meguid , 2008. For green voters, prominent theories argue that they protest due to the historical legacy of green parties as movement parties. Alternatively, radical right voters are largely presumed to be more active in the non-electoral arena due to the mobilising power of the radical right in anti-institutional engagement. However, these accounts unpack the micro-level foundations of the party families as unique party families and do not address the niche similarities between the two families. This is the gap this article seeks to address by focussing on the shared representational disadvantages niche voters face.

Deprivation and political protest
Central to the study of political protest is Gurr's (1971) concept of relative deprivation which states that individuals are mobilised to protest when they feel like they are not as well off in society as they think they should be given their expectations of society. When individuals feel like they are being deprived access to resources relative to their expectations, they are radicalised and more likely to participate in political protests. Following this concept, several studies have shown that relative deprivation, or the sense that one is not doing as well as they think they should be, has been linked to a heightened propensity to support more radical political parties (Burgoon et al. 2019), heightened prejudice (Guimond and Dambrun 2002), increased identity attachments (Grant 2008), and social destabilisation (Korotayev and Shishkina 2020).
Furthermore, there is consistent evidence that relative deprivation does incentivize people to protest. For instance, Asingo (2018) demonstrates that perceived relative deprivation induces people to protest while decreasing one's propensity to vote. Similarly, Grasso and Giugni (2016) find that those that were most deprived following the economic crisis in 2010 were more likely to participate in political protests. Moreover, Grant (2008) finds that group-based relative deprivation induces migrant groups to participate in protest behaviour to overcome discriminatory barriers. In sum, the sense that one is being deprived relative to other groups within society induces people to engage in more unconventional patterns of political participation such as political protests.
Importantly, these studies focus on economic deprivation and relative deprivation based on income and material wealth. However, we know that people develop political grievances based on policy, parties, and governance. Despite this, there are few studies which connect this kind of political grievance with unconventional participation. Among those studies that do look at political dissatisfaction, evidence is mixed on whether political dissatisfaction plays a role in mobilising people to participate in political protests (Dalton et al. 2010;Stockemer 2014). However, if economic deprivation can induce participation in political protests, a sense of political deprivation should similarly encourage people to take to the streets, especially for voters who are uniquely passionate about specific policy outcomes but face structural disadvantages to their representation.

Representational deprivation
Building on the concept of relative deprivation, I develop what I term the representational deprivation thesis, which argues that voters will engage in protests when they have political grievances based on the inability of the party they support to adequately fulfil its promises and represent its voters. Representational deprivation focuses on political grievances as opposed to economic deprivation which is central to relative deprivation. When a group perceives that it is not receiving the kind of representation compared to its expectations given the group's status in government or society, the members participate in political protests to remedy the perceived representational gap.
Representation plays an important role in advanced democracies, and we know that voters respond to how represented they are by the political system. Spoon and Klüver (2019) show that as mainstream parties have converged on the middle, voters have switched their votes to parties that better represent their preferences punishing the mainstream parties for no longer representing them. Similarly, Bakker et al. (2020) find that voters do care about the proximity of political parties to their own individual preferences. The greater the ideological divergence between a voter and their political party, the more dissatisfied they are with the political process (Bakker et al. 2020). Along these lines, Blais et al. (2017) find that voters are sensitive to how well they are represented by electoral outcomes. When a party does not receive seats that are proportional to its vote share, voters become less satisfied with democracy (Blais et al. 2017). Plescia et al. (2020) further confirm that voters do value proportionality and prefer representative political systems. In short, voters care about how well the electoral system and party platforms represent them, and they become dissatisfied when they feel unrepresented.
As this work demonstrates, a voter's satisfaction with democracy and the political system is shaped by how well they are represented by that political system (Bakker et al. 2020;Blais et al. 2017;Plescia et al. 2020). However, we know from work on grievance theories that dissatisfaction alone is not a sufficient condition to drive citizens to join political protests. As Gurr (1971) argues in relative deprivation, individuals must not only feel dissatisfied, but they also must feel deprived of something they feel they should otherwise have -the difference between what is and what is expected. When voters support parties, they expect that party to represent the voters' interests. However, electoral rules, coalition politics, and other institutionalised forces of consensus politics can lead voters to not receive the representation they expected when supporting the party. When parties are forced to compromise on their agenda, voters are deprived the representation they expected to receive when they voted in an election. This representational deprivation, wherein voters perceive the political system as not representative of their voices as much as they expect it to, should lead voters to take to the streets.

Niche party voters, governing coalitions, and representational deprivation
Under what circumstances then should we expect representational deprivation to form and drive people to participate in political protests? In this article, I argue that niche voters are the most likely to demonstrate the underlying logic of representational deprivation, especially when their party joins a governing coalition. Niche parties are defined by their specific issue emphases (Meguid 2005(Meguid , 2008. Importantly, extant work on these parties demonstrates that they attract policy-oriented voters (Adams et al. 2006;Ezrow 2008;Spoon 2011). When these parties moderate their positions and sacrifice some of their policy appeals, they lose votes (Adams et al. 2006;Ezrow 2008) requiring them to balance their agenda between pure-policy and pure-vote seeking moderation unlike other parties in the party system (Spoon 2011).
However, niche issues are often unaddressed in the political or party system. Central to Meguid's (2005) definition of a niche party is that they champion issues that are outside the traditional domains of contestation. Further, Meguid (2005) demonstrates that the fate of niche parties is not within their own control and is instead determined by whether mainstream parties choose to discuss the issue or ignore it. Additionally, De Vries and Hobolt (2020) argue that challenger parties, of which niche parties can also be considered until they join a government and then they are no longer considered 'challenger' , are able to emerge when their issue has gone unaddressed, and they are able to innovate on that issue. Importantly, while there is a debate in the literature surrounding the precise definition of niche parties, there is general conceptual agreement among scholars that niche parties emphasise and compete on specific, often non-economic issues, that are largely ignored by their competition (Bischof 2017;Meyer and Miller 2015;Wagner 2012). At their core, niche parties represent non-traditional issues which are often unaddressed, or not prioritised, by mainstream parties. In short, niche voters care passionately about specific policy but are unrepresented by the existing political system which often under-emphasizes their issue. Thus, my first hypothesis is the following: H1: Niche party voters are more likely to engage in protests than mainstream party voters.
According to H1, niche voters should be predisposed to participate in political protests because they are voters that care about specific policy that is often unaddressed by the political system. Then, according to the representational deprivation thesis, they should be more likely to protest when they develop an expectation that is not met. The best condition to capture this logic is when the party enters a governing coalition. Prior work on political protest has established that citizens are less likely to protest when their party is in government (Torcal et al. 2016). However, this assumes that governments are static and all parties in government face the same consequences. This assumption is especially problematic for niche parties who are often only able to enter coalitions as junior partners. The party may hold cabinet posts, but it does not have as much policy influence as the majority party ). There is a large literature that demonstrates that serving in government often forces junior partners to compromise on their true stance to align with the majority party's goals as junior members of the coalition (Brommesson and Ekengren 2019;Fortunato and Adams 2015;Sagarzazu and Klüver 2017). Extant work has subsequently shown that this compromise leads to increased voter misperception of the junior party's true position (Adams et al. 2016;Falcó-Gimeno and Fernandez-Vazquez 2020;Fortunato and Adams 2015;Spoon and Klüver 2017) and electoral losses for junior parties in subsequent elections disproportionate to expected incumbency losses (Greene et al. 2021;Hjermitslev 2020;Klüver and Spoon 2020). Importantly, Fortunato (2021) finds that the compromise required of coalitions leads voters to view the partners as failing to represent their voter's interests and harms the credibility of the representational linkage between the party and its supporters.
Since niche parties always join governing coalitions as junior partners, they take on all of the costs of governing as a junior partner in conjunction with their policy-oriented base which punishes moderation (see Adams et al. 2006). As a result, these are parties with a voting bloc that cares passionately about specific policy entering coalitions in which they are often forced to compromise on pledges they have made to those policy-oriented voters. Indeed, research shows that niche parties join governing coalitions when they can achieve policy goals (Heller 2002) but as junior partners, they fulfil significantly less of their campaign pledges than the PM party (Thomson et al. 2017). As a result, joining a governing coalition sends a signal to their voters that their policy interests are going to be addressed, setting an expectation of a certain level of representation. When the niche party, which is often the primary champion of these specific, under-emphasized policy goals as discussed above, then fails to deliver on those policy goals in the manner their voters expected given the campaign promises and its status in government, its voters will feel that they are not being represented relative to their expectations.
This sense of deprivation should be especially likely to emerge when their party joins a governing coalition with a majority party that is ideologically divergent from the party's positions. We know from the literature on the consequences of governing as a junior partner that the junior partner moves itself on policy to endorse the majority party's position (Brommesson and Ekengren 2019;Sagarzazu and Klüver 2017) and that the PM's party sets the legislative agenda for the government (Green-Pedersen et al. 2018). The further that the niche party must move to align itself with the majority party and compromise on its positions, the more its voters may feel that they are not being represented as they expected by the party. Furthermore, we know that voters adjust their perceptions of the junior partner to align with the PM's party (Fortunato and Adams 2015;Spoon and Klüver 2017); however, this adjustment is dependent on the distance between the governing parties. Falcó-Gimeno and Fernandez-Vazquez (2020) demonstrate that the further the coalition parties are from one another, the more voters adjust their attitudes about the junior partner. Partnerships between ideologically compact parties do not influence a voter's perceptions of the junior partner as much as partnerships between ideologically distant parties (Falcó-Gimeno and Fernandez-Vazquez 2020). Therefore, niche party voters whose party enters government with a PM ideologically close may not feel like they are being deprived of representation. Instead, niche voters whose parties govern with a majority party that forces the niche party to move further from its true stance would be more inclined to protest as they do not perceive themselves to be represented by their party or the government. 3 My central hypothesis, which directly tests the representational deprivation thesis is the following: H2: Niche party voters whose party is in government are more likely to protest as the difference between their party's position and the Prime Minister's position increases compared to niche opposition voters.

Data and methods
In order to test these expectations, I rely on the ESS which is a cross-national survey administered every two years and uses the same questions across each wave, allowing for cross-national analysis of European citizens through time. 4 I compile a pooled dataset of nine waves of the ESS with 242,229 individuals in 15 West European countries from 2002 to 2018. 5 I focus on only West Europe and do not include East Europe because of important differences in the levels of democratisation, development of the party system and the radical right, and the function of coalitions between the two regions which suggests the representational linkage between voters and their parties is different enough between East and West Europe that East Europe requires separate, additional theoretical and methodological work. 6

Dependent variables
The dependent variable in this study is whether an individual participated in a lawful public demonstration in the last 12 months prior to being surveyed. 7 This question is included in the ESS's political participation battery and respondents can either indicate that they did participate in a lawful protest in the last year or that they did not participate in such an activity. 8 Across all 15 cases from 2002 to 2018, 8.15 percent of respondents have participated in lawful demonstrations while 91.85 percent of respondents have not participated in this form of political activity.

Niche voter
In addition to the protest participation question included in the ESS, respondents are also asked which party they voted for in their last national election. Respondents are then able to select from a list of parties that corresponds to their respective party system to indicate which party they voted for. As opposed to using closeness to a specific party (see Finkel and Opp 1991), I focus on the respondent's vote choice as it produces a larger sample of individuals. 9 The key independent variable for this study is whether a respondent voted for a niche party. Following Meguid's (2005Meguid's ( , 2008 definition of a niche party, a niche voter is coded as a voter of either a green, radical right, or regionalist party. 10 Importantly, the recent work to precisely categorise niche parties finds that continuous measures designed to more precisely classify these parties closely correspond to Meguid's (2005) classification strategy based on party family (Meyer and Miller 2015). A full list of the parties considered niche and mainstream is included in Table A2 in the online appendix. 11 In the sample, 15.15 percent of respondents are coded as niche voters while 84.85 percent voted for parties classified as mainstream across all 15 countries. 12 Remaining individuals who did not support either a mainstream or niche party, such as special issue parties or the Scandinavian centre parties, 13 or who did not indicate their vote choice were coded as missing, restricting the comparison to either a niche voter or a mainstream voter.

Party-PM divergence
In order to measure the representational deprivation hypothesis (H2), I rely on the left-right (RILE) position of the party's manifesto through the Comparative Manifestos Project (CMP). I rely on the RILE position because it serves as a useful proxy for a party's broader ideology and captures a party's position on the dominant dimension of competition in Europe (Marks and Steenbergen 2002). 14 Through the hand-coding of manifestos into quasi-sentences classified under different policy areas, the CMP provides comprehensive cross-national data on the policy positions of political parties. The RILE score from the CMP data is calculated using the percent of the total quasi-sentences in a manifesto coded under the CMP codebook classified as left or right, with the difference between the two percentages being the overall RILE score. Ranging from −100, which represents the most left-leaning position, to 100, which represents the most right-leaning position, these values were then rescaled to a 0-10 scale to ease interpretation. The divergence score is then calculated by taking the absolute value of the difference between the RILE position of the party one voted for and the RILE position of the PM at the time of being surveyed (see Spoon and Klüver 2019). 15 Higher values indicate there is a greater divergence between the party that one voted for and the PM's party, with lower scores indicating greater congruence. 16 These divergence scores range from 0 to 4.565, with a mean divergence score of 0.676.
Then, using data from ParlGov, I use responses to the question in the ESS which asks which party a voter supported in the last national election to code whether they are government or opposition voters. I code those whose party was in government at the time of being surveyed as supporting a government party (1). Those whose party was in the opposition when they were surveyed are coded as having voted for an opposition party (0). These assignments are not based on the cabinet that formed immediately after an election and are based on the governing status of the party at the time of the survey. If an individual supported a party that initially joined a government after an election, but left prior to that individual being surveyed, they are considered an opposition voter. This produces a sample where 47.64 percent of the respondents supported an opposition party while 52.36 percent supported a governing party.

Controls
In order to isolate the effect of being a niche voter on an individual's propensity to protest, I include several controls to account for existing accounts of political protest. First, I add one's self-placement on the RILE scale. Extant work has demonstrated that right-leaning voters, or higher scores on the RILE scale, should be less likely to protest (Dalton et al. 2010). Second, I control for whether an individual holds an extreme ideology, which I consider those who reported themselves to be between zero and two or eight and ten on the RILE scale. Extremists should be more likely to protest (Opp et al. 1995). Third, I control for both satisfaction with the government and trust in the parliament. As satisfaction and trust increase, or as grievances decrease, the likelihood of protest should be lower (Stockemer 2014). Next, I account for the years of education a respondent had, their household income, and the number of hours they work a week. As they become more educated, have a higher income, and have more time (work less), the more likely they should be to have participated in a protest (Brady et al. 1995). Then, I control for the level of civic participation of the individual by using responses to two questions in the ESS asking individuals if they worked with a party, action group, or other organisation. Those who indicated that they had worked with any group are coded as 1 and I expect these individuals to be more likely to protest (Giugni and Grasso 2021). Lastly, I control for demographic factors including age, age squared, and gender. As age increases, the likelihood of protesting should increase (Grasso 2016). At the country-wave level, I include the unemployment rate using data from the OECD and the WorldBank. Based on grievance models, higher unemployment should be associated with a higher propensity to protest (Grasso and Giugni 2016). I also account for the levels of income inequality using the GINI coefficient 17 and the GDP per capita of a country. As the inequality of a country increases, I expect the propensity to protest to increase. Meanwhile, I expect the propensity to protest to decline as the country's wealth increases. To account for other coalition factors, I control for the number of other parties that are in the coalition. I expect this to lead to an increased propensity to protest as the larger the coalition, the more compromise is needed to govern (Leiserson 1968). To adjust for party-system factors, I include the overall representativeness of the political system through the effective number of parties in the legislature (ENPP). As the representativeness of the political system increases, I expect the propensity to protest to decrease (Plescia et al. 2020). Summary statistics for all variables are in Table A4 of the online appendix. 18 Since individual responses are nested within country-wave contexts, it is necessary to account for this hierarchical structure of the data through a multi-level model. Failure to do so may overrate the significance of the estimated effects by deflating standard errors and inflating the type I error rate (Steenbergen and Jones 2002, 219-220). As the outcome variable is binary with respondents either protesting (1) or not (0), I thus estimate a multi-level logistic regression with random effects at the country-wave level. Additionally, the random effects are at the country-wave level, but respondents are both nested in countries and waves independent of one another. Therefore, it is important to nest the country-wave level within the country level (Fairbrother 2014). However, given the small number of countries in the sample (15), using country random effects may produce biased estimates and so I rely on country fixed effects. I also use wave fixed effects as country-wave contexts vary within each wave as well. 19 Table 1 presents the results of all models to test each of the hypotheses. Model 1 includes all independent variables and controls without any interactions for H1. Model 2 then adds a triple interaction between voting for a niche party, voting for a governing party, and the divergence between the party one voted for and the PM. Since two of the three variables are dichotomous, the triple interaction effectively subsets the data and allows me to estimate the effect of divergence for each of the four possible combinations: opposition mainstream, governing mainstream, opposition niche, and governing niche. Importantly, all niche parties in this sample were junior partners in their respective coalitions. Therefore, when interacting niche and government, I am effectively capturing the difference between being in the opposition or joining the government as a niche junior partner. A full list of niche governing parties is available in Table A3 of the online appendix.

Results
I find strong support in all the models that niche voters are more likely to protest than mainstream voters, confirming H1. In Model 1, the positive coefficient for being a niche voter is statistically significant and associated with a higher propensity to protest over being a mainstream voter. Based on Model 1, niche voters have a 6.35 percent chance of protesting while mainstream voters have a 4.88 percent. Niche voters are 1.47 percent more likely to protest. The significant constitutive term in Model 2 provides more confidence in this finding.
Turning to H2, the results presented in Model 2 lend support to the representational deprivation thesis. Based on the coefficients in Model 2, Figure 1 presents the predicted probability of protesting broken down by which type of party a voter supported and whether that party is in the opposition or in the government. 20 The distribution of divergence scores is presented on the bottom of each plot. The solid lines present the predicted probability of protesting as ideological divergence increases, while the dashed lines represent 95% confidence intervals. Among opposition niche and mainstream parties, the divergence between the party and the PM's party does not have a significant effect on their propensity to protest. However, among governing parties, divergence has a positive and significant effect for both types of voters, supporting the argument that representation in the governing coalition mobilises individuals to protest. When there is complete ideological convergence between the party and the PM, as presented by the coefficient for the interaction between being a niche voter and being a governing party voter, opposition niche voters are more likely to protest. For niche voters, there is a 6.63 percent chance they protest when their party is in the opposition as opposed to a 4.02 percent chance when they are in the government. For mainstream voters, there is a 4.94 percent chance they protest when they are in the opposition and a 4.44 percent chance when they are in the government. 21 However, once there is ideological divergence between the party and the PM, the propensity to protest among governing voters increases as divergence increases and surpasses opposition voters. At a divergence score of roughly 1.113 between the party one voted for and the PM, governing niche voters become more likely to protest than opposition niche voters. Governing mainstream voters overtake opposition mainstream voters at a divergence score of roughly 0.692. These results demonstrate that when voters feel their party is not being represented well by the governing coalition and the PM, they are more likely to protest when their party is in government. Their expectations for what they will receive by supporting a party in government are not being met because of the ideological compromise required of their party to be a part of the coalition as a junior partner.
This effect is most pronounced among governing niche party voters, where the propensity to engage in protest increases from 4.02 percent when there is no divergence between the party and the PM to 20.29 percent when the divergence score is equal to 3.5. This is an increase of 16.27 percent. Among mainstream voters, the predicted probability to vote only increases to 8.18 percent when the divergence score is equal to 3.5 from 4.44 percent under perfect convergence, which is only a 3.74 percentage point increase and is 12.11 percent lower than for governing niche voters. This is not a trivial difference. 22 To put these results in context, Germany serves as a useful example to demonstrate the relationship between divergence in the governing coalition and a niche voter's propensity to protest. The governing arrangement following the 2017 elections was a coalition government between the Christian Democratic Union (CDU/CSU) and the Social Democratic Party (SPD). The two niche parties in Germany -the Alliance 90/Green Party and the Alternative for Germany (AfD) -were in the opposition. Under that arrangement, Green Party voters had a 7.23 percent chance of protesting while AfD voters had a 6.82 percent chance of protesting. If either party were to have entered the government with the CDU/CSU, the likelihood that Green Party voters protest increases to 11.71 percent and the likelihood for AfD voters decreases to 5.77 percent. Based on manifestos from the 2017 elections, the Green Party and the CDU/CSU have a divergence score of 2.234, which is above the average divergence score (1.113) where governing niche voters overtake opposition niche voters as more likely to protest. Meanwhile the AfD and the CDU/CSU have a divergence score of 0.734, which is below the inflection point. As this example demonstrates, with low levels of divergence between the party one voted for and the PM, opposition niche voters are more likely to protest. However, ideologically distant coalitions such as a hypothetical CDU/CSU-Green coalition results in Green Party voters being more likely to protest if the Green Party were to enter government under a centre-right PM. 23 These results are especially telling when compared to the predicted probabilities of protesting based on the traffic-light coalition between the SPD, Green Party, and Free Democratic Party (FDP) that formed following the 2021 elections. With the SPD holding the chancellorship, Green voters have a 6.68 percent chance of protesting had they remained out of government, while the radical right AfD has a 7.32 percent chance of protesting as an opposition party. Importantly, these probabilities are not substantively different from the probabilities that Green or AfD voters protest as opposition parties to the CDU/CSU in the previous coalition. For opposition niche voters, the distance between the party and the PM does not significantly impact their propensity to protest. However, in the new coalition, the Greens have joined government with an ideologically close PM. The divergence between the Greens and the SPD is 0.182 based on the 2021 manifestos, and the propensity for Green voters to protest decreases to 4.4%. Had the radical right been included in the governing coalition, the divergence between the party and the PM would be 2.536 and the propensity to protest among its voters would have nearly doubled to 13.41 percent. As these results demonstrate, the effect of governing status on the propensity of niche voters to protest is conditioned by the party's ideological distance from the PM.
Turning to the controls included in the models, self-placement on the left-right scale has a negative impact on the likelihood of protesting. This confirms prior accounts that the more right leaning an individual is, the less likely she is to engage in political protests. Extremists are more likely to participate in political protests as anticipated. In line with expectations, the more satisfied an individual is with the government, the less likely they are to protest. Interestingly, trust has a significant positive relationship with the propensity to protest, suggesting that the more one trusts parliament, her propensity to protest increases. This would challenge the grievance accounts that those who distrust institutions are more likely to protest. 24 Education has the expected effect of increasing the propensity to protest. Age has a positive effect, meaning that as one ages, they become more inclined to protest. However, the significant squared age term signifies that this relationship is an inverted U-shape wherein those who are middle-aged are the most likely to protest. Lastly, having participated in a party or other organisation increases the likelihood of protesting as anticipated. The only significant country-wave control is the ENPP which has the expected negative effect. The more representative the political system, the less likely one is to engage in political protests.

Conclusion
To summarise, the results presented in this article lend support to the broader theoretical argument that those who are deprived representation are those most likely to engage in political protest. When there is greater divergence between the party one voted for and the PM, the likelihood of governing party voters, especially governing niche party voters, to engage in protests increases. Once there is divergence between the niche party and the PM's party, the propensity for niche governing voters to protest increases and surpasses the propensity of opposition voters to protest. Since the PM sets the agenda and is perceived to have the most policy influence Green-Pedersen et al. 2018), governing alongside an ideologically distant PM sends a signal to niche voters that they are not being adequately represented by the party they voted for in the coalition. These results support other findings that the distance between voters and the leaders of government increases their propensity to protest (Torcal et al. 2016). This sense of representational deprivation drives them to take to the streets and protest for their issues.
Importantly, this work sheds light on how different types of voters interact with their political contexts to develop their repertoires of political action. Mainstream voters share the same underlying responses to the politics of governing coalitions. They are also more likely to protest when their party is ideologically distant from the PM's party. However, the effect is much stronger for niche voters. When divergence between a governing party and the PM is at 3.5, there is a 12.11 percent greater likelihood that a niche voter protests over a mainstream voter. Just as niche parties have to behave in unique ways to survive as small parties (see Spoon 2011), niche voters have to behave differently than mainstream voters to overcome their representational disadvantages.
This work also contributes to our understanding of the impacts of coalitions on junior parties and their voters. In the debate over whether joining coalitions helps or hurts niche parties and satisfies niche voters, this research provides some mixed evidence. Joining a coalition with an ideologically close partner may help the party by achieving policy goals with little compromise that satisfies the party's voters, as shown by the increasing likelihood of protesting for opposition voters, especially in ideologically compact coalitions. However, joining a coalition with an ideologically distant partner drives the party's voters to protest the government in which the party is a part due to a sense of representational deprivation formed when the party is unable to deliver on its promises.
While the theory of representational deprivation has offered some insight into what explains the circumstances under which niche party voters will protest, future work should build on these findings and continue to unpack the conditions in which different voters engage in political protests. Most importantly, we currently lack the necessary data to directly test the representational deprivation mechanism. Future work should aim to test this mechanism more directly. Additionally, future work should aim to unpack if there are lingering effects of the patterns observed in this study. Are voters more likely to protest even if their party is no longer in governing coalitions after being in representationally deprivational arrangements? Lastly, this article has focussed on niche voters, but it will be important to unpack if the mobilising power of feeling representationally deprived can be applied beyond niche voters to minorities and other disadvantaged groups.

Notes
1. See https://www.politico.eu/article/sweden-greens-struggle-greta-thunbergshadow/ 2. See https://www.aljazeera.com/economy/2020/1/15/green-activist s-launch-lawsuit-against-merkels-climate-plan 3. Since it could be argued in both directions that either being in the opposition or government could lead to representational deprivation, the key consideration is the conditioning effect of divergence between the party and its PM. 4. Question wording is available in Table A1. 5. The countries included are Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, the Netherlands, Norway, Portugal, Spain, Sweden, and the United Kingdom. Not all countries participated in all nine waves of the ESS, with some countries only participating in certain rounds. Every wave in which a country participated is included. Despite Belgium having effectively two separate party systems, it is treated as a single system as there is one government. 6. As a result of this case selection, my results are constrained by characteristics of the West European context. When East Europe is included in the models, the results are robust; however, the results are not robust in models of just East European countries which suggests there is something different about the representational linkage in East and West Europe. Future work will need to be done to apply the concepts generated in the West European context elsewhere considering aspects of different regions that make them different from West Europe. 7. This 12-month window means that voters may have had changes in governing coalitions during this period, restricting the inferences that can be drawn about the role of the governing coalition on participation. As a robustness check, I run the same models restricting the sample to only those whose coalition is at least 1 year old, ensuring the length of the coalition covers the time window for participation in Table A6. Results are substantively the same. 8. Respondents are also asked about whether they have engaged in a political boycott. Table A7 presents results using boycott participation as an alternative form of unconventional participation. The results in Table A7 are the same for H1, but different for H2. Table A8 estimates a multi-level linear model where the dependent variable is a count of the number of unconventional acts an individual participated in. The results in Table A8 are substantively the same for H1, but not for H2. 9. Respondents are also asked to identify which party they are closest to. As a robustness check, I run the same models using one's partisan identification rather than vote choice. The results are presented in Table A9 of the online appendix and are substantively the same.
10. Adams et al. (2006) include the radical left in their niche classification. I also code voters according to this definition but present the results using Meguid's (2005) definition as there is a consensus that green, radical right, and regional parties are niche parties (Bischof 2017;Meyer & Miller 2015;Wagner 2012) while there is some dispute about the use of left-right extremism as a defining characteristic (Wagner 2012). As a robustness check, the models including radical left voters are presented in Table A10 of the online appendix. The results for H1 are the same while the results for H2 are not. Furthermore, models which include the radical left with mainstream parties to compare niche parties to non-niche parties supports H1, but not H2. This all suggests that the radical left is an interesting actor and future work should aim to unpack the radical left in a more focused way to study the relationship between their role in government and their voters. 11. Importantly, this assumes a static classification of niche parties. Some research suggests that 'nicheness' is better understood on a continuum (Meyer & Miller 2015) and that we need to better understand how parties can become niche or mainstream under different circumstances (Meyer & Wagner 2013). Future work should unpack the relationship between a parties' nicheness and its governing status. 12. Mainstream parties include social democrats, Christian democrats, conservatives, and liberal parties. Classifications were made using Spoon and Klüver (2019) and the Comparative Manifestos Project. 13. As a robustness check, I run the same models with these centre parties included as mainstream parties. The results are substantively the same. 14. Importantly, the RILE dimension in European politics serves as a useful heuristic for voters when evaluating their political system, especially when evaluating coalition politics (Fortunato et al. 2016). Therefore, it is sufficient to rely on the RILE dimension as this is the ideological dimension that voters are evaluating the coalition on. Future work should aim to unpack the multidimensionality of representational deprivation. 15. As a robustness check, I also calculate a divergence score between the party's left-right position and overall government position not including the party one voted for if it was a member of the governing coalition. The results using this measure of divergence are not significant, suggesting the most important divergence value is between the party one supported and the PM. This supports other work which demonstrates that voters attribute the position of the government to that of the PM's party and rely on simple heuristics over complex calculations ). 16. To rule out the possibility that what I am measuring is how radical a party is, with more radical parties having greater divergence scores, I also estimated all models weighting my divergence scores on the average position of the party-system. Results are substantively the same. 17. For years with no reported GINI score by either the WorldBank or the OECD, I use the year prior to the survey year as a proxy since GINI coefficients change gradually. 18. All country-wave variables are assigned based on the year of the survey wave. The only exception is Italy's 2004 wave, which was actually administered in 2006. These respondents are assigned values for Italy in 2006. Italy did not participate in the 2006 wave of the ESS.
19. As a robustness check, I run the same models without fixed effects in Table A11 of the online appendix. I also run the same models nesting the country-wave level in country-level random effects in Table A12. All results are substantively the same under both alternative specifications.
20. Table A5 presents the full predicted probabilities in Figure 1. 21. One important avenue for future work is how voters react based on the prior governing experiences of their parties. Evidence from preliminary tests suggests that voters whose party is in government for the first time are not impacted by party-PM divergence, but that experienced governing voters do become more likely to protest as divergence increases. 22. None of the findings presented in this paper are being driven by one particular party family. Results in Table A13 indicate that models which only include regional, radical right, or green voters do not produce significant effects on any of the main IVs. This indicates that the observed patterns apply to niche voters as a whole and are not an artifact of any one group of voters. 23. Importantly, the data used in this study pre-dates the rise of the Green party in Germany when it was clearer that the German Greens were a niche party. Future work following the 2021 federal elections should evaluate whether we can still consider the German Green Party a niche party. 24. Unpacking this counterintuitive finding extends beyond the scope of this study, but future work should unpack the effect of political trust on protest behavior. Results are robust when I run the models without political trust as a control.