Protesting Parties in Europe: A comparative analysis

The article provides the first large-scale study of protest activities by political parties. The empirical analysis draws on original protest event data for 30 European countries based on semi-automated coding of news agencies. The article innovates by (a) proposing a standardized indicator for the extent to which protest and electoral politics relate to each other, (b) showing that parties’ involvement in protests differs across political contexts, and (c) mapping the profile of a typical party-sponsored event and a typical protesting party. Despite long-term trends toward differentiated modes of interest intermediation, the results indicate that a wide range of parties does protest. However, in highly differentiated contexts, the typical protesting party mirrors the outsider image of movement parties: it does not belong to a mainstream party family and has no government experience. By contrast, more strategic factors, such as opposition status, drive parties to the streets in less differentiated contexts.


Introduction
Europe's party systems are in flux, as indicated by functionalist and structuralist perspectives on party competition. The functionalists point to aspects such as declining party identification and increasing electoral volatility (e.g. Dalton and Wattenberg, 2000), which skyrocketed in countries hardest hit by the Great Recession (e.g. Dassonneville, 2018). The structuralists, by contrast, emphasize emerging cleavages and the role of challenger parties in articulating new oppositions (e.g. Bornschier, 2010;de Wilde et al., 2019;Kriesi et al., 2012). Recent work in this tradition shows how Europe's latest crises have reinforced long-term trends of part system transformation (e.g. Hooghe and Marks, 2018;Rovny and Whitefield, 2019).
However, both perspectives neglect that the key driving forces may not only have changed in programmatic terms but also in organizational form and action repertoire. That is, political parties seem to increasingly blur the lines between electoral and protest politics. Following McAdam and Tarrow's (2010) forceful call almost a decade ago, social movement scholars have thus returned to study the manifold interactions between electoral and protest dynamics (for an overview, see Hutter et al., 2019). Among the Europeanists, we can identify two main strands: On the one hand, some authors have taken up Kitschelt's (2006: 280) concept of "movement parties," defined as coalitions of activists who emanate from social movements and try to apply movements' organizational and strategic practices in the electoral arena. Studying cases from the political leftsuch as Podemos or Syriza (e.g. della Porta et al., 2017)and from the political right-such as Jobbik or the Alternative for Germany (e.g. Caiani and Císař, 2019;Pirro et al., 2019)-these scholars aptly describe the fuzzy empirical boundaries between political parties and social movements, while indicating the transitional status of such hybrid entities. On the other hand, other scholars have approached the topic from a more systemic and programmatic perspective, linking the study of issue agendas in protest and electoral politics (e.g. Císař and Vráblíková, 2019;Hutter and Vliegenthart, 2018;Walgrave and Vliegenthart, 2019). This research provides ample evidence on the close yet varying coupling of protest agendas and parties' activities during electoral campaigns and in parliament.
Missing from this emerging field of research is a large-scale comparative analysis that maps and explains the varying extent of party-movement interactions. A key reason for the absence of such large-N work is the lack of comparable cross-national data. Another one is the lack of established standardized measures to examine the phenomenon at hand. In this article, we innovate by providing a standardized indicator of the extent to which protest and electoral politics relate to each other across 30 European democracies and over 16 years (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015). More specifically, we concentrate on the extent to which political parties sponsor protest activities in the streets. Following Rucht (1998: 41), we broadly define sponsorship as (co-)organizing, taking part in and/or calling for participation in a protest event. As any standardized indicator for large-scale crossnational comparisons, the extent of party-sponsored protests comes at the cost of reducing complexity. Most importantly, the indicator emphasizes the action component over more organizational or ideational relations. 1 Nevertheless, we consider our study an important complement to the cited case studies on movement parties and the small-N cross-national comparisons of issue agendas.
We innovate, moreover, by embedding our original empirical endeavor in a theoretical framework that builds on Kitschelt's (2003) ideas of ever-more differentiated modes of interest intermediation in advanced democracies. It is under such conditions that the emergence of movement parties is seen as atypical and transitional (Kitschelt, 2006: 280). We elaborate on the scope conditions of Kitschelt's argument to explain the varying extent of party-sponsored protests across European countries. We also deduce expectations about differences in the type of sponsored protests ("what?") and the type of sponsoring party ("by whom?") depending on the level of party-sponsored protests, that is, on the level of arena differentiation. Our related research questions are as follows: To what extent do parties mobilize in the protest arena? What types of protest events do parties sponsor? What does the typical "protesting party" in Europe look like? And how do the types of events and parties differ depending on the level of party-sponsored protests?
Our results underscore that party-sponsored protests are an important feature of contemporary protest politics in Europe. However, the extent to which political parties are major players in the protest arena varies considerably across countries, mirroring differences in democratic history and the strength of civil society. In line with parties' incentives, we also find significant differences in the types of events sponsored by parties. Compared to the average protest in a country, they are less radical, larger, and more likely to address the so-called cultural issues or the functioning of the political system. This difference is most pronounced in highly differentiated contexts, that is, in contexts where party-sponsored protests are relatively rare. Similarly, the type of protesting party varies across contexts. In highly differentiated contexts, the typical protesting party in Europe mirrors the "outsider image" of movement parties much more than in less differentiated ones. Overall, our study adds quantitative insights on cross-national variations for a better understanding of party-movement interactions which should be incorporated into more case study-based designs.
Theoretical framework and expectations Why? The extent of party-sponsored protests across contexts While McAdam and Tarrow (2010) and the cited studies that followed in their footsteps have identified manifold interactions between protest and electoral politics, the dominant view in electoral and party research is one of separation and a division of labor. Why is this? To explain this standard view, we depart from Kitschelt's (2003) idea that we have witnessed an increasing differentiation in the patterns of interest intermediation since the end of the "Golden Age" of Western capitalism (see also Kitschelt, 2006: 278-281). According to Kitschelt (2003: 89): "The progressive differentiation of modes of collective interest mobilization and growing separation of political entrepreneurs in movements, interest groups, and parties from each other is the big story of the last third of the twentieth century in European democracies." Kitschelt (2003) explains differentiation as a product of the learning processes of political entrepreneurs and their followers, which have been underpinned by economic, social, and political-institutional changes. Two challenges faced by political entrepreneurs are of crucial importance for the argument: problems of collective action and social choice. In contrast to social movements and interest groups, political parties are portrayed as the actors who have invested the most in solving both types of problems. Parties frame their goals as long-term, durable, and encompassing programs, and they have invested most in techniques of collective preference alignment (e.g. formal rules for aggregating individual preferences into organizational purposes). In the long run, actors who have invested differently in solving the two problems are better equipped to compete in specific political arenas. Regarding the extent of party-sponsored protests, Kitschelt (2003: 97) argues, "parties focus increasingly on electoral competition, at the expense of interest group representation or social movement protest actions [ . . . ]. Social movements, finally, concentrate on public actions outside institutionalized arenas of bargaining to affect public opinion and political elites through the media." In general, the differentiation argument lets us expect that party-sponsored protests are the exception rather than the rule in contemporary European democracies. However, as we emphasize in this article, two scope conditions of the argument are important regarding cross-national variation: a long democratic history and the strength of civil society. Both conditions appear to be crucial for the establishment of functionally differentiated arenas that are populated by specialized political actors. As argued before, Kitschelt (2003) highlights a long-term process which began in the 1970s that has resulted in a twofold transformation of European party systems: a first wave driven by the mobilization of new social movements and left-libertarian parties in the 1970s and a second wave driven by the mobilization of the populist radical right since the 1990s (Kriesi et al., 2012). Based on the scholarly literature, we know that these transformations have been much less pronounced in the countries in Southern and Eastern Europe (Borbáth and Gessler, 2020;Hutter and Kriesi, 2019). Overall, we thus expect that party-sponsored protests are more likely in countries with a shorter democratic history, given that the learning processes at the core of the differentiation argument take time and depend on a democratic context. Second, closely related condition, the development of a strong and independent civil society and social movement sector relates to the other side of the equation. Even if parties withdraw from certain sites of mobilization, there needs to be other collective political organizations with the capacity to fill the void. The strength of civil society overlaps to some extent with the life span of democracy. Postcommunist countries are characterized by lower levels of participation, often seen to be the result of low levels of social and interpersonal trust originating from a totalitarian past, as well as of the non-transparent and elite-driven transition processes (Bernhard, 1996;Howard, 2003). However, a second strand of literature analyzing nonparticipatory aspects of civil society suggests the presence of influential organizations able to defend democratic institutions (Foa and Ekiert, 2017). Instead of relying on mass mobilization to gain support for their causes, these organizations use their connection to power holders to exercise leverage on the decision-making processes (Petrova and Tarrow, 2007). Both perspectives agree on the relative lack of organizations able to mobilize on the streets, which leads us to expect that party-sponsored protests are more likely in contexts where civil society participation is lower.

What? The type of sponsored protests across contexts
Parties not only decide whether to avoid or to participate in a protest but in the former case, they also need to settle what form their participation should take. Differentiation leads us to expect that party-sponsored protest events are systematically different than protest events organized by social movements or other organizations for two reasons (Somma, 2018). Firstly, from the perspective of parties as vote-or policy-seeking organizations, their fortune is determined by the dynamics in the electoral arena. Therefore, parties' participation in the protest arena should be instrumental to their electoral goals and, most likely, a function of the electoral incentives they face. Secondly, from the perspective of other organizations that mobilize in the protest arena, parties represent institutionalized politics which they challenge. Therefore, aligning themselves with parties represents a risky strategy with the potential to alienate their supporters. Given these tensions, party-sponsored protests might differ from other, typical events in the protest arena in several key dimensions.
We expect parties to sponsor events that have the potential to be attended by many participants. Consequently, they might stay away from forms of mobilization associated with disruption. Violent events, confrontations, or blockades are unlikely to gain universal recognition among the electorate of even the more radical parties. In contrast, peaceful demonstrations or petitions are less stigmatized forms of protest behavior and are potentially attended by a relatively large and often diverse group of citizens (Somma, 2018: 70). For similar reasons, party-sponsored protests might be more likely in times of increased protest mobilization. Such a context provides the opportunity for parties to "ride the wave" and benefit from a general increase in protest (e.g. Peña and Davies, 2017). It allows them to connect and build alliances with social movements and other actors who are already mobilizing on the streets. For this reason, we expect parties to sponsor peaceful events in times of increasing protest mobilization, benefitting from the opportunity to gain exposure and support from potential voters. In addition, we expect partysponsored protests to follow the electoral cycle as well. During electoral campaigns, that is, in intensified periods of competition, parties invest in mobilizing their supporters. In this context, organizing on the streets by sponsoring protest events has the potential of paying off most in terms of vote shares. Finally, taking party-sponsored protests as a phenomenon associated with the two waves of party system transformation highlighted above, we expect parties to mobilize more on cultural rather than economic issues. The ability of parties to benefit from the changing context of competition and keeping their role as the main agents of the transformation might hinge on their presence in both arenas of mass mobilization.
In addition, we expect that where political parties are the main agents of both electoral and protest mobilization, the events they sponsor differ less from protests staged by other actors. However, in contexts of high differentiation, where parties' share in protest is lower, the typical party-sponsored protest is more likely to reflect the logic sketched above: less disruptive, larger, more in sync with the electoral cycle, and more related to so-called cultural issues than non-party-sponsored events.

Who? The type of parties sponsoring protests across contexts
Another crucial element in understanding the rationale behind party-sponsored protests is the focus on which political parties are most likely to take to the streets. At this stage, it is helpful to combine Kitschelt's differentiation argument with his further elaborations on movement parties (Kitschelt, 2006;also see: Caiani and Císař, 2019;della Porta et al., 2017;Pirro et al., 2019). As emphasized initially, movement parties represent a specific organizational form characterized by the lack of investment in solving problems of social choice and collective action and, in this regard, resemble social movements both in organizational form and in action repertoire. In Kitschelt's (2006: 281) understanding, movement parties emerge as the result of a mismatch between the arena where political entrepreneurs mobilize, their organizational resources, and the narrow focus of their claim. They do not represent a durable form of mobilization, but a transitional phenomenon, most likely to appear in moments of party system transformations.
The two examples Kitschelt gives are linked to the aforementioned transformation of Western European party systems: the rise of the greens and the radical right. In both cases, movement parties appeared for a brief period. As they became electorally successful, they invested in expanding their organizational basis and in developing a more encompassing programmatic appeal. One of the aims of this article is to examine the extent to which protest sponsoring is restricted to parties with programmatic, strategic, and organizational traits associated with movement parties, or whether protest is used by parties of all stripes.
Our key claim is that in the context of high differentiation-the one Kitschelt (2003Kitschelt ( , 2006 portrays-the typical protesting party comes closer to the ideal type of a movement party. First, we expect that parties challenging the mainstream in a party system in programmatic terms are the primary actors involved in protest politics. In line with Kitschelt's (2006) focus, green and radical right parties, that is, the drivers of the first and the second wave of party system transformation, are the most important in this respect. On the one hand, they are the main agents politicizing the issues related to the cultural dimension of competition in the electoral arena and extend the scope of contestation in the protest arena as well. On the other hand, these types of parties often originate from social movements. Having been excluded from the government for a relatively long period helped them maintain links with their grassroots base. Even after they were co-opted by the mainstream, their organizational features (e.g. Bailey, 2017: 139-151), as well as ideological profile, make them the likely candidates to mobilize in both arenas.
Second, we expect that, apart from adopting nonmainstream positions, factors resulting from the configuration of power in the electoral arena play a decisive role. Most importantly, parties face different strategic incentives depending on whether they are currently in government or whether they have ever been in government-the latter often termed challenger parties (de Vries and Hobolt, 2020). Governing parties are constrained by established policies and the diverse societal needs that they need to consider in their activities. By contrast, opposition parties have more leeway to respond to citizens' demands and to build broad social support coalitions. Klüver and Spoon (2016), for example, show that opposition parties respond more strongly to shifts in voters' issue priorities than government parties. Hutter and Vliegenthart's (2018) results indicate that opposition parties are more likely to respond to the signals from protesters in their parliamentary activities than government parties. Thus, we also expect that opposition parties are more likely to join forces with less institutionalized actors and to directly mobilize on the streets to challenge the government. 2 This effect should be strongest for parties without any government experience at all.
Finally, the movement party concept is also associated with a certain type of organizational model. As stated in the introduction, this is increasingly important in an era when parties seem to be in decline (Dalton and Wattenberg, 2000) and when we observe a more diverse organizational landscape-a landscape composed of parties which have abandoned the mass-party model and those that still adhere to it (e.g. Katz and Mair, 1995;Rohrschneider and Whitefield, 2012;Scarrow et al., 2017). As Katz and Mair (1995: 8) highlight, the ideal-typical mass party is considered part of civil society aiming at "breaking into the state and modifying public policy in the long-term interests of the constituency to which it is accountable." To do so, mass parties rely on their own channels of communication and adopt a bottom-up organizational approach as party elites are accountable to party members which also provide the principal resources by means of fees and contributions. In their conceptualization of "mass parties," Rohrschneider and Whitefield (2012: 118) add relations with civic associations to Duverger's (1959) original focus on a large membership base and complex organizational apparatus. Their empirical analysis of party-voter congruence indicates that, in Western (but not Eastern) Europe, parties with mass organizations are ideologically closer to partisan and independent voters than parties without such organizational features. Relatedly, we expect a positive relationship between mass-party organizations and our dependent variable of party-sponsored protests, given the stronger linkages of mass parties to other civic associations, their stronger reliance on members, and their stronger organizational capacities more generally.
To sum up, we expect the typical protest-sponsoring party in Europe to be non-mainstream, challenger, and (still) adopting a mass-party organization. As for the typical party-sponsored event, we expect to observe stronger effects in highly differentiated contexts, where the level of party-sponsored protest is lower. That is, in such a context, the typical party sponsoring protest events is expected to be a distinct entity, coming closer to the features associated with the movement party model. In contexts with less differentiated arenas, we expect that the range of sponsoring parties is more diverse in programmatic, strategic, and organizational terms.

Data and methods
Having outlined why the extent of party-sponsored protests may differ across European countries and how that may affect the type of sponsored events and the type of political party we observe in the streets, we now turn to the strategy used to test our claims. One of the main reasons why partysponsored protests have not yet been studied comparatively is the lack of large-scale protest event data sets covering multiple countries over time. For this reason, we collected an original protest event data set based on the coverage of English-language newswires. In general, protest event analysis-as a type of content analysis of media sources-has been one of the major advances in the field of protest and social movement research as it allows for quantitative analysis of protest in a cross-sectional and longitudinal setting (for a review, see Hutter, 2014).
Data were collected with semi-automated tools in a joint effort by the ERC project "Political Conflict in Europe in the Shadow of the Great Recession (POLCON)" at the European University Institute and the SNF project "Years of Turmoil" at the University of Zurich. It is based on the coverage of 10 English-language newswires (on the data collection, see Kriesi et al., forthcoming, and Online Appendix A). The data set covers protests in 30 European countries: all European Union member states (apart from Croatia), Iceland, Norway, and Switzerland during a 16year period from 2000 to 2015. The data set covers 17,048 protest events with an organizational sponsor which involved around 268 million participants. The data set includes information on the date, the size, the form (demonstration, petition, strike, violent, blockade), the claim (economic, culturally libertarian/conservative, political in terms of a close connection to the functioning of the political system, e.g., anti-corruption rallies), and the organizer (parties, unions, other organizations) of the event. Based on the date of the event, we create two indicators to identify protest events which happened in the half a year period before a national parliamentary or presidential election or during the half-year after. We also created an indicator of "big events" to identify the events attended by more than 100,000 participants.
As stated initially, we adopt a broad definition of sponsorship referring to instances when parties (co-)organize, take part in, and/or call for participation in a protest event. Such a definition goes clearly beyond the support of the claims of protests and includes support of the action form as well. Thus, it represents a standardized measure to get closer to the extent to which protest and electoral politics are related to each other. However, it does not allow us to differentiate whether parties play a significant role in the organization of the protest event or not. As Rucht (1998: 41) highlighted some time ago, such fine-grained measures are beyond the scope of a protest event analysis based on media reports.
Overall, 13% of all protest events in our data set have been sponsored by parties. We matched each of the sponsoring parties with the ParlGov data set (Döring and Manow, 2018) to gain further information about the vote share, ideology, party family, as well as opposition/government status of each party. More specifically, we rely on three ideological scales: a general left-right scale, an economic left-right scale, and a cultural libertarianauthoritarian scale. 3 We code parties as mainstream if they were classified by ParlGov as Christian-democrat, conservative, liberal, or social democrat. To gain information on the extent to which a party organization resembles that of a mass party, we rely on an expert survey conducted by Rohrschneider and Whitefield (2012) in 2008/2013. 4 We replicate their indicator and rely on an additive index of four, equally weighting items in which experts are asked to evaluate (1) how strong the party apparatus and (2) the party membership is in determining policy (seven-point scale), (3) whether the party has a "significant membership base" relative to the other parties in the system (dichotomous), and (4) whether the party is organizationally affiliated with any interest/civil society group, including, but not limited to trade unions, business associations and church groups (dichotomous).
To measure the two contextual features introduced before, we rely on the period of democratization and the strength of civil society. We distinguish Northwestern European countries which democratized before or in the immediate aftermath of the Second World War from Southern European and Eastern European countries included in the second and respectively third wave of democratization. To measure the strength of civil society, we rely on the expert survey-based "core" civil society index of V-Dem (v2xcs_ccsi, see Coppedge et al., 2017). The measure is an aggregate index of three indicators: civil society entry/ exit, civil society repression, and civil society participatory environment. The index was designed to reflect the robustness of civil society understood as an autonomous sphere where citizens are active and free to pursue their political and/or civic goals, however conceived.

Empirical results
To begin with, Figure 1 shows the level of party-sponsored protest as a function of the context in which they take place. To do so, we show the average share of party-sponsored protests depending on the strength of civil society and democratic history. For the latter, we broadly distinguish between Northwestern, Southern, and Eastern European countries. 5 As the results show, relative to a context with strong civil society (where parties are only present in a little over 10% of all protests), the share of party-sponsored protests doubles in a context where civil society is comparatively weak, and it reaches over 20%. Regarding the timing of democratization, the results are very similar in Northwestern and Southern Europe. In both regions, parties are on average sponsoring a little over 10% of all observed protest events. Based on this result, Southern Europe passed the threshold of having a differentiated landscape of interest intermediation. In contrast, the share of party-sponsored protests is substantially higher in Eastern Europe, where parties are present in 23% of the protest events. To examine the interaction of the strength of civil society and the timing of democratization, we checked the extent to which the weakness of civil society leads to a larger share of party-sponsored protests in each region individually. The result included in Online Appendix E shows that the two factors have a separate, additive effect on the level of party-sponsored protest.
Turning to country differences and the relationship between parties' involvement and overall levels of protest mobilization, we examine the share of party-sponsored protests relative to the total number of events per country. As Figure 2 indicates, the relationship between the share of party-sponsored protests and the overall protest mobilization in a country is curvilinear. Parties are unlikely to dominate the protest arena in countries where protest is rare and in countries where protest is frequent. The figure also reveals the suspected regional clusters. While Northwestern European countries are diverse regarding the level of protest, the share of party-sponsored protests is relatively low. The Southern European countries are even more diverse regarding the overall level of protest-with Greece, Spain, and Italy having the highest and Cyprus, Malta, and Portugal the lowest level of mobilization-but are very similar in having an average level of party presence. The Eastern European countries have a low-to-average level of protest, but they are the ones where the share of partysponsored protest reaches its highest levels. Overall, our results indicate a limited ability for parties to ride the wave of mobilization: for parties to become the dominant actors in protest, there needs to be some level of protest mobilization, but above a certain threshold they are crowded out Figure 1. The share of party-sponsored protests in context. The level of party-sponsored protests is calculated as the share of party-sponsored protests in percent of all organized protests in a given context. The events are weighted by the countries' log population and the press agency weight-see Online Appendix A. We rely on the median value at the country level to distinguish between contexts with low and high strength of civil society-see Online Appendix C. Figure 2. The share of party-sponsored protests and overall level of protest mobilization. The level of party-sponsored protests is calculated as a share of party-sponsored protests from all organized protests in a country. The events are weighted by the countries' log population number and the press agency weight-see Online Appendix A. The symbols distinguish between Northwestern, Southern, and Eastern Europe. The curve shows a LOESS regression smoother. In Online Appendix E, Figure 2 replicates the figure using the total number of protest participants.
(also confirmed when looking at participation numbers, see Online Appendix E). We now turn to the type of protest event that parties sponsor. As previously discussed, we expect partysponsored protests to differ from other protest events, especially in a context of high differentiation (¼ low levels of party sponsorship). To examine the moderating role of a differentiated landscape of interest intermediation, we split our sample into two groups. We consider countries where the share of party-sponsored protest is above the 11% median threshold as weakly differentiated, whereas we code countries below the median as strongly differentiated (see Online Appendix C). 6 To map the characteristics of party-sponsored protests, we calculated the share of different event characteristics among all party-sponsored events (see Figure 3). In general, the results indicate that party-sponsored protests share important features across both contexts. First, most party-sponsored protests are addressing cultural demands, and only a smaller fraction of them address economic issues. Second, over 85% of them are non-confrontational or violent (with over two-third taking the form of public demonstrations). Third, party-sponsored protests are sometimes co-organized with other organizations, but in more than 70% of the cases parties are reported as the only organizational sponsor. Finally, a substantive share of the party-sponsored events falls into the category of "big events", defined as events with more than 100,000 participants. The most pronounced cross-context differences refer to the larger share of co-organized (29% vs. 18%) and big events (28% vs. 18%) in more differentiated contexts.
To what extent and how do party-sponsored protests differ from non-party-sponsored events? To answer this question, we ran logistic regression models with partysponsorship as the dependent variable. All models include country fixed effects and standard errors clustered by country. As a function of the level of party sponsorship, we split the sample between contexts of high and low differentiation, but we also conducted models with interactions. To ease the interpretation of the results, we present coefficient plots and include the regression tables in Online Appendix B.
The results of the regression analysis in Figure 4 support the expectation that party-sponsored protests are more likely to center on cultural and political issues than Figure 3. Characteristics of party-sponsored protest events. We distinguish low differentiation contexts with a high share of party protests, from high differentiation contexts with a low share of party protests based on the national level median value-see Online Appendix C. The events are weighted by the countries' log population and the press agency weight-see Online Appendix A. economic ones. Also, political parties are likely to organize protests alone, and these protests take the form of demonstrations or petitions rather than strikes (for details, see Online Appendix B). In line with our expectations, we find a more clear-cut picture in highly differentiated contexts, where party-sponsored protests are less prominent in the protest arena. In such contexts, party-sponsored protests are also less confrontational or violent than non-party-sponsored events, and they are more in sync with the electoral cycle. More specifically, our results suggest that political parties are more likely to take it to the streets shortly before and after Election Day in highly differentiated contexts, whereas we find no statistically significant electoral cycle effects in less differentiated contexts. We also modeled the likelihood of party-sponsored protests as a function of the overall flow of protest mobilization over time. Similar to the cross-national pattern in Figure 2, we observe a curvilinear relationship in countries with low differentiation (see the results in Table 1 of Online Appendix B). In differentiated contexts, by contrast, partysponsored protests seem to develop independently from the overall dynamics of protest mobilization. We take these results as evidence of our two key expectations: partysponsored protests are systematically different from other types of protest events, and the differences we expected regarding form, issues, and timing are stronger in a context of high differentiation (¼low levels of such partysponsored events).
Turning to the type of political parties that are sponsoring protests, we again start with some descriptive findings. This time we calculated the share of different party characteristics among all party-sponsored events. 7 Figure 5 shows that most protests are sponsored by parties that do not belong to the political mainstream in terms of their party family or which have no government experience. 8 Importantly, in contexts of low differentiation, 44% of all party-sponsored protest events are organized by mainstream parties, whereas in highly differentiated contexts, their share only reaches 33%. The share of parties with executive experience also drops from 26% to 20%. Moreover, more than 80% of all party-sponsored events are being organized by parties in opposition.
The descriptive findings tend to support our expectations of the typical protesting party in Europe. However, Figure 4. Impact of event and context characteristic on party-sponsorship-logistic regression. Coefficients plotted as odds ratios. Standard errors clustered by country. Observations are weighted by the countries' log population number and the press agency weight-see Online Appendix A. We distinguish low differentiation contexts with a high share of party protests, from high differentiation contexts with a low share of party protests based on the national level median value-see Online Appendix C. The model also includes overall protest, overall protest squared, length of government, and country fixed effects. Following the advice of Bolsen and Thornton (2014), we calculate 84% confidence intervals which in case they do not overlap indicate statistical significance at the p < 0.05 level. See the corresponding regression in Table 1 of Online Appendix B.
we had to construct a party-level data set to systematically compare the involvement of different parties across contexts. To do so, we calculated the number of times a particular party sponsored an event during a cabinet. 9 We find that a large number of parties sponsor protest: overall, there are 266 parties from the 30 countries under scrutiny which at some point from 2000 to 2015 protested. Unfortunately, only 137 of them were included in the expert survey on party organizations. The estimates presented in Figure 6 are based on this subset of the sample, but the effects for the other variables remain the same for the larger sample of 266 protesting parties (see Online Appendix D). To explore the variation of the types of parties protesting, we ran OLS regression models with country-fixed effects and standard errors clustered by country. The units of analysis are political parties and the dependent variable is the number of events a party sponsored during the term of a government. We also controlled for the duration of the government. Figure 6 presents the results graphically (for the regression tables, see Online Appendix B).
Our key expectation regarding the profile of protesting parties referred to their resemblance to movement parties, that is, they belong to the green or radical right party family, are in opposition, without government experience, and adopt a mass-party organization. In general, the regression results in Figure 6 reinforce our descriptive findings. We find that parties are more likely to take it to the streets if they belong to non-mainstream party families and are in the opposition. At the same time, the results underline that the effects for government experience and mass-party organization only point to the expected direction in more differentiated contexts. In less differentiated contexts, political parties with a wider variety of organizational models tend to sponsor protest events, and having a history of government participation even shows a small positive effect. Interestingly, the opposite holds when looking at the effects of the current configuration of power. It turns out that opposition status has an even larger positive effect on protest sponsorship in less differentiated systems. The finding points to the importance of strategic factors rather than the more structural features associated with movement parties in explaining which parties are likely to take it to the streets in such a context.

Conclusion
In this article, we presented the first large-scale comparative analysis of parties' involvement in protest politics, covering 30 countries and the years 2000-2015. In doing so, we aimed to contribute to the literature on party system change. This literature highlights the transformations that party organizations and systems have witnessed since the 1970s, and it argues that we are faced with an ever-more complex party landscape. However, it tends to give short shrift to parties' activities outside the electoral arena. More specifically, we contributed to the emerging scholarly literature on party-movement interactions in three respects: First, by proposing a standardized indicator for the extent to which protest and electoral politics relate to each other; Figure 5. Type of parties sponsoring protests. The share of protests sponsored by parties of different type is calculated as a percentage of all party-sponsored protests organized on a context of low respectively high differentiation. We distinguish low differentiation contexts with a high share of party protests from high differentiation contexts with a low share of party protests based on the national level median value-see Online Appendix C. The events are weighted by the countries' log population number and the press agency weight.
second, by showing that parties' involvement in protests differs across political contexts; and third, by mapping the profile of a typical party-sponsored event and a typical protesting party.
Using parties' relative presence in protest as a standardized indicator of the relationship between electoral and protest politics, we were able to test and qualify Kitschelt's (2003Kitschelt's ( , 2006 argument about increasingly differentiated landscapes of interest intermediation. Overall, we find that a relatively diverse set of parties does take it to the streets. We also highlight that parties' protest activities differ across contexts in terms of its level and character. The differentiation of interest intermediation is a development specific to established democracies with a strong civil society. In contrast, parties are more dominant outside of their home arena in Eastern Europe, where democratization happened later and participation in civil society organizations is lower. In highly differentiated contexts, where parties are less significant protest sponsors, party-sponsored events are predominantly peaceful demonstrations, often wellattended, co-organized, concern cultural issues, and are in sync with the electoral cycle. Although a relatively large variety of parties protests in this context as well, the typical protesting party mirrors the outsider image of movement parties (Kitschelt, 2006): it does not belong to a mainstream party family and has no government experience. Therefore, the typical protesting party we find in this context corresponds to the profile of parties described in the literature as challenger parties (e.g. de Vries and Hobolt, 2020) and as main agents of cleavage transformation (e.g. Kriesi et al., 2012).
In less differentiated contexts, where parties are more likely to be dominant actors in protest politics, the type of party-sponsored events and the character of protesting parties are more diverse. The events are more likely to be confrontational, often rather small, and organized by the parties alone. Parties in less differentiated contexts mobilize when there is some protest on the streets, but protest activity has not yet reached its peak. Their presence also appears to be fairly independent of the electoral calendar. Protesting parties in this region are more often from the political mainstream, with some governing experience, but currently in opposition. In this regard, parties tend to follow a more strategic logic in deciding when to mobilize on the streets than the outsider image of movement parties would suggest. Figure 6. Impact of party characteristics on sponsorship of protest events-OLS regression. Standard errors clustered by country. Observations are weighted by the countries' log population number and the press agency weight-see Online Appendix A. We distinguish low differentiation contexts with a high share of party protests from high differentiation contexts with a low share of party protests based on the national level median value-see Online Appendix C. The model also includes overall protest, overall protest squared, length of government, and country fixed effects. Following the advice of Bolsen and Thornton (2014), we calculate 84% confidence intervals which in case they do not overlap indicate statistical significance at the p < 0.05 level. See the corresponding regression in Tables 2 and 3 of Online Appendix B.
We believe that by showing what type of protest events parties sponsor, which type of parties choose to protest, and under what conditions, the article provides a useful model for reconnecting party and social movement research. In a time of high volatility, decreasing turnout, and growing mistrust in parties, protests and social movements are ever-more seen as a potential source of democratic innovation. By showing the blurred boundaries between arenas of interest intermediation, our study highlights the interdependence and strategic use of protest mobilization by political parties. While parties may be able to channel the demands of the street in institutionalized decision-making processes, their presence threatens to appropriate the claims of protest movements and to narrow the space for radical political alternatives. At the same time, protest sponsorship might add to the resilience of political parties, providing them with one foot on the street and another in parliament. However, future research should account for the strong crosscontext variations revealed by our study when interpreting specific cases of party-movement interactions and their consequences.