Political Polarisation on Gender Equality: The Case of the Swiss Women’s Strike on Twitter

2.1 The Use of Social Media by Social Movements Promoting Gender Equality

There is currently a consensus that social media are widely used in social movements as they serve for fast information diffusion and for raising attention to specific social and political claims (Soares and Joia 2015). Consequently, social media have emerged as a key venue for political debates. Social media, especially Twitter, serve as a place to engage in civic-related activities, notably by using viral capabilities such as #hashtags, which are often referred to with the concept of hashtag activism (Xiong, Cho, and Boatwright 2019). In the case of feminist movements, the topic of the present study, Dixon (2014) traced the different ways that hashtag feminism has been enacted, such as through the sharing of personal experiences and the challenging of dominant discourses. In our study, we used the case of the women’s strike movement in Switzerland, which also developed specific hashtags to promote fast information diffusion and raise attention to the social and political claims of the movement (e.g., #Frauen*streik in German and #Grevedesfemmes in French).

Social media platforms constitute an indispensable tool for social movements to organise their actions and mobilise public opinion around particular claims to promote social change (Eltantawy and Wiest 2007; Poell et al. 2015). While social media have been a major tool for spreading equality claims and actions since the MeToo movement in 2017 (Modrek and Chakalov 2019), a recent study also showed that levels of modern sexism among the American mass public did not respond to the rise of MeToo (Archer and Kam 2020). One reason to this might be that gender equality related issues are typically a topic on which political polarisation is strong, especially on social media where the progress achieved in women’s rights and gender equality has become the target of a backlash driven by anti-gender users and right-wing populists (Wallaschek et al. 2022). Despite this, we still have limited knowledge about how the scope of ideologies and content from social media discussions can influence the reach of social movements’ claims in public opinion.

2.2 How Social Media Content Surrounding Social Movements Connects to Public Opinion

The details of social movements engagement on social media with the public discourse in society are of utmost importance as the goal of a social movement is to bring concerns to the forefront of the political agenda. In this view, the fact that social media users are typically unrepresentative of the general public (Mellon and Prosser 2017) does not mean that social media content is unrelated to what the public thinks, notably because influential user groups also have the potential to influence public opinion (Weeks, Ardèvol-Abreu, and de Zúñiga 2017).

Figure 1 illustrates the conceptual framework underpinning our study. It is inspired from the study of Gordon (2015). In Gordon’s model, social movements rise from the development of a community of interest in response to a set of underlying grievances stemming from the elaboration of an agenda. Until a call to action takes place, following an event raising the claims of the social movement, the community is likely to either lie silent or even fade away. Likewise, a call to action that is not supported by a community consists of an individual cause, not a social movement. The outcome of the movement is likely to depend on the success of the mobilisation and can consist of short-term, as well as long-term, changes. Although the model appears quite linear, there are many moments where the development of the social movement can be stopped, re-oriented or cancelled either by the community of interest itself or by external factors.

Figure 1:

Conceptual framework underpinning the study (inspired from Gordon 2015).

This model enables us to better situate our study in regard to the different stages of a social movement. In the present paper, we only focus on the elements referring to the mobilisation on social media and the type of actors involved in the online conversations. However, in addition to Gordon’s model, we consider that the mobilisation and call for action surrounding a social movement (in our study, the women’s strike) are embedded in a broader context. In this context, public opinion and political polarisation towards gender equality questions prevail and are generally measured with opinion surveys. While social mobilisation around the women’s strike happened on the streets, in newspapers, and in official communication, it also took place on social media, which is the focus of the present study.

To understand the interrelation between offline opinion and online discourse, it is important to identify the actors who mobilise on social media, both supporting and opposing the strike. Indeed, if social media are a now a widely used communication tool to mobilise supporters around common grievances and to implement calls to action, opponents of social movements also use social media for developing counter-protests and for spreading their counter-arguments or programs (Gordon 2015, p. 19). Hence, in a first step, we aimed to identify the user groups who mobilised on Twitter surrounding the women’s strike.

At the same time, studies have shown that the reliance on social media by these different actors to express their views can lead to increased polarisation (Quattrociocchi, Scala, and Sunstein 2016). Therefore, we aimed to assess the extent to which the views expressed by the politically engaged users on social media reflected trends prevailing in the broader context. We further looked at the extent to which social media increased polarisation trends observed in public opinion surveys and in traditional political surveys. Finally, to better grasp if social media can be used to advance social movements to promote its messages and to achieve its goals, we took advantage of the textual nature of social media data to explore the views and arguments of each actor surrounding the women’s strike. This online content can also be compared to open-ended survey responses from citizens concerned about gender equality related issues. In the next sections, we will develop our research questions and the respective hypotheses in further detail.

2.3 Assessing the Congruence Between Social Media Discussion and Public Opinion Regarding Gender Equality

Not only has the topic of gender equality become central to scientific literature, but it has also been prevalent in political agendas and parliamentary debates across the world (Hooks 2000). As activists, politicians and other groups seek to find an agreement on gender equality policy measures, gaining knowledge about the overlap between online and offline opinions around gender equality is key to better grasping the opportunities for social change.

The study of Scarborough (2018) takes a step in this direction by demonstrating to what extent social media data, specifically tweets, can be used to account for gender equality attitudes. The main premise of the study states that if tweets about feminism deal with issues that are central to gender relations, then they should capture the same underlying dimensions as those opinions measured through gender attitude surveys. In a similar vein, other studies have pointed to difficulties in the identification of opinions expressed online, notably the supportive and opposing views about gender equality. For instance, Kirkwood et al. (2018) were unable to differentiate between very polarized pro-, neutral-, and anti-feminist views in discussions on Twitter, especially because of content-specific challenges pointing to the difficulty of extracting relevant tweets and of precisely classifying the diversity of sub-topics. Both studies, however, place little emphasis on the actor groups involved in gender equality discussions. Furthermore, by considering sentiment (or tonality) as the main content feature to be correlated with surveyed opinions, they say little about how social media discursive content reflects what the broader public thinks.

In the framework of public opinion surveys, Baldassarri and Park (2020) found that the U.S. population is moving towards more progressive views on a host of issues – from LGBTIQ+ rights to gender roles and sexual behaviours (see also Eisner, Spini, and Sommet 2020 in the Swiss case). However, the authors warn that, contrary to public opinion dynamics on economic and civil rights, the above-mentioned issues can less clearly be described in terms of increased issue partisanship. Wallaschek et al. (2022) found a similar trend towards the support of gender equality on social media. The authors investigated the users’ engagement and the content of debates about gender equality in tweets about the 2021 International Women’s Day in Germany, Italy, and Poland. They showed that social media users and discussions were predominantly supportive of gender equality, as users engaged with the value of gender equality mainly in an acclamatory fashion. They also showed that political and societal actors exhibit high levels of online engagement. These studies demonstrate that actors and content dominant in social media discussions are also mainly in favour of gender equality measures.

In line with the findings outlined above, and to answer our first research question (i.e., to what extent does the involvement of politically active users reflect trends from public opinion?), we expect to find a high congruence between the online involvement and the offline support for gender equality in terms of political ideology (Hypothesis 1).

2.4 Assessing the Extent to Which Social Media Leads to More Polarisation on Gender Equality than Trends Prevailing Offline

Notwithstanding the closing of the gap on gender equality measures between progressive and conservative positioning, there remain important divergences between political elites on gender policy. This can lead to important polarisation between politically involved actors and the wider public, which could be further heightened on social media. Therefore, our second interest is to investigate the extent to which political polarisation surrounding gender equality is increased on social media.

With respect to gender equality, polarisation can be conceived in terms of positional dynamics relying on rhetorical moral arguments (De Wilde and Zürn 2014; Roggeband 2018). For instance, Kantola and Lombardo (2020) conducted a qualitative analysis of populist interventions in EP plenary debates on gender equality in the European Parliament and found a variety of radical right opposition strategies to gender equality, mainly drawing on old and traditional gender imaginaries packaged in novel populist ways. Their findings reflect previous studies on political elites’ opinion polarisation displaying similarly extreme opinions about connected topics, such as sexual minority rights (Wojcieszak 2010) and homosexuality (Munro and Ditto 1997). Furthermore, social media play an important role in the polarisation of the political debate on gender equality. For instance, Russell et al. (2020) showed that hyper-partisanship in Parliament extends from the legislative process into politicians’ social media strategic communications. Social media are thus useful as they can cover a large spectrum of political positions that underlie the topic of gender equality, thus also providing a platform for the backlash against the ideas and goals of feminism (Lawrence and Ringrose 2018).

However, notwithstanding the increased opportunities to express political opinions online, social media can also amplify the phenomenon called the “spiral of silence” (van Aelst et al. 2017), which consists of self-censorship behaviours on the part of politicians or citizens who do not express their opposing views about a topic (Noelle-Neumann 1984). This silence can be due to the fact that individuals perceive a majoritarian public consensus (Sunstein 2017) or simply because the topic does not trigger enough of their attention (Lasorsa 1991). On social media, this self-censorship behaviour can lead to an increased polarisation, notably as certain views will become inflated at the expense of other opinions (Dubois and Szwarc 2018). Additionally, the effect of filter-bubbles (or echo-chambers) is likely to lead to a fragmentation of (more extreme) opinions towards political issues (Zuiderveen Borgesius et al. 2016).

These studies thus point to the fact that social media serve as a tool for conservative politicians to voice their opposition to gender equality measures. These studies also suggest that there are dominant ideologies surrounding the gender equality debate, thus leading to political polarisation on gender related issues. In line with these findings, and to answer our second research question (i.e., to what extent is political polarisation surrounding gender equality increased on social media?), we expect to observe increased levels of polarisation of the online debate on gender equality compared to trends observed through the lens of opinion surveys (Hypothesis 2).

2.5 Identification of Gender Equality Opinions in Social Media Content

In addition to the focus of political polarisation on gender equality issues, we also aim to investigate the extent to which social media claims have the potential to echo what the broader public thinks. This is paramount to understanding the extent to which social media mobilisation can help social movements build long standing support in public opinion.

In line with the idea that social movements have the ability to connect with public opinion, Mirbabaie et al. (2021) investigated how specific user groups participated over the course of the MeToo debate in 2017 and 2019. Drawing from the theory of connective action, they found that the framing of and the attention to the movement were spread in different ways according to actor groups – namely, the starters and the maintainers. Overall, the authors found little variety in the content of online discussions, although they pointed to different underlying motives, ranging from self-serving and branding intentions to calls for attention and action.

Their findings echo the results from Baik, Nyein, and Modrek (2021) by showing how difficult it is for social movements to garner new adherents. The authors concluded that, although the movement they analysed did increase awareness and participation among those already sympathetic to the movement, it might not have enlisted new supporters. This is mainly because social media are used by user groups to promote their ideas through different communication strategies. For instance, direct supporters have an interest in promoting the event, whereas politicians point to fewer collective goals and highlight specific aspects of the debate which resonate with their own agenda.

The studies outlined above remain, however, in the framework of social media research without mobilising other data sources, such as opinion surveys, to assess the potential of social media messages to impact public opinion on gender equality. The study of Adams-Cohen (2020) proposes to address the question of causality in the domain of same-sex marriage. More precisely, it uses Twitter data and machine-learning methods to analyse the causal impact of the Supreme Court’s legalization of same-sex marriage at the federal level in the United States on political sentiment and discourse towards gay rights. Results showed that there was a relatively stronger negative reaction in public opinion towards same-sex marriage in states where the Court’s ruling produced a policy change as compared to that of other states. Nonetheless, this study is also not able to rely on survey data to benchmark its findings.

In our study, we propose to take advantage of the textual nature of social media texts and of open-ended survey questions to look at the extent to which ideas and views expressed online by users involved in discussions surrounding the strike are similar to open-ended answers of citizens concerned with gender equality. Based on the literature, and to answer our third research question (i.e., what does the content of social media data tells us about the potential of social media to promote the claims of the social movement?), we expect to observe a continuum between calling for attention, on the side of the strike organisation committee and of its followers, and discussing concrete policy measures, on the side of politicians and citizens (Hypothesis 3).

3.1 Data Collection: Original Tweets from Different Actor Groups

In the present study, we focus on the 2019 women’s strike in Switzerland. The strike consisted of demonstrations in the country’s major municipalities and revolved around the issues of equal pay, recognition of unpaid care work, and governmental representation. It followed the first 1991 Swiss women’s strike, which was organised 10 years after the Swiss population’s acceptance of the constitutional article on the equality between women and men.

To construct our corpus of tweets about the women’s strike covering the period from January to December 2019, we retrieved social media texts from the social media platform Twitter using the Application Programming Interface. Our data collection strategy was based on both seed user profiles and search queries.

Concerning the user profiles, we extracted all tweets emitted from the seed user profiles. This included the main organisation committees of the women’s strike that had a Twitter account. We also collected all tweets from the committees’ followers (as it was in June 2019), as well as tweets from political accounts (candidates of the October 2019 federal elections, elected politicians, and national political parties) referring to the women’s strike between January 2019 to June 2019. For the followers, political accounts, and trade unions, we then filtered out tweets that did not explicitly refer to the women’s strike by using a list of search queries. The list of search queries read as follows: ‘.*womenstrike.* | .*frauen.*streik.* | .*feministi.*streik.* | .*femstreik.* | .*frauen.*strassen.* | .*frauen.*mobilisier.* | .*grève.*femmes.* | .*greve.*femmes.* | .*grève.*féministe.* | .*femmes.*grève.* | .*femmes.*greve.* | .*femmes.*rues.* | .*femmes.*mobilis.* | .*femmes.*manif.* | .*14.*juin.* | .*14.*juni.*’.

We also included tweets that contained specific search queries stemming from the most common hashtags found in the collected data. This strategy enabled us to make sure that the queries were precise enough to collect tweets related to the specific event of interest, but not too large as to include unrelated tweets. The search queries read as follows: ‘frauenstreik | 14juni | femstreik | feministischerstreik | fstreik | femstreik | frauendemo | frauenbewegung | grevedesfemmes | 14juin | grevefeministe | femmesengreve | grevefeministe’. The hashtag #womenstrike was not used as it returned tweets mostly unrelated to the event of interest. We kept only tweets not emitted from the above-mentioned groups (organisation committees, followers, politicians, candidates, and trade unions).

We did not include retweets in our sample because we are interested in the content of original tweets and to what extent different actor groups contribute to the elaboration of the content of social media discussions. Our final corpus contained 41′062 tweets and 15′919 unique Twitter accounts.

3.2 Manual Annotation of Tweets

To address our first research hypothesis, we conducted a manual annotation of the entire list of Twitter profiles that contained a description. We coded for the type of accounts that were not already included in the list of seed accounts. We thus added the following additional categories: foreign politicians, media, and journalists (national and foreign), activists (national and foreign), other political users (e.g., national and foreign ambassies, governmental departments), organisations with feminist-related aims (national and foreign), other users tweeting actively (more than five tweets) about the strike, and other unlabelled users. These categories were heuristically found to be encompassing enough to describe our sample of top users. The category for followers only indicates users which are not labelled in any other user categories. Finally, the category for trade unions does not differentiate between national and foreign organisations.

We also labelled the profiles to specify whether they are Swiss, foreign, or unknown accounts. The Annex 2 in the Annexes provides more information about the distribution of accounts according to geolocation. This annotation was also done for every profile and shows that most users included in our sample stem from Switzerland (27%), France (16%), Germany (14%) and other European countries (11%).

3.3 Correlation Between Social Media and Survey Data According to Political Affiliation

To address our second research hypothesis, we aimed to identify the party affiliation of Swiss politicians on Twitter. To do so, we manually specified the party affiliation of the Swiss political accounts. This enabled us to correlate the online attention with the offline support to gender equality along partisan leaning. In order to achieve this, we relied on survey data to measure politicians’ and citizens’ opinions about women’s rights. The obtained survey scores are reflected along a left-right political continuum (e.g., the party affiliation of politicians and of citizens who declare a party affiliation). These scores are also correlated to the prevalence of Twitter discussions about the women’s strike for each political affiliation.

Concerning politicians, three items from the Swiss part of the 2019 Comparative Candidate Survey ask politicians about women’s rights. We used the following items measured on a 5-points Likert scale: ‘Women should be given preferential treatment when applying for jobs and promotions,’ ‘The government should take measures to reduce differences in income levels,’ and ‘Women should be free to decide on matters of abortion’. From these items, we built a mean score for gender equality support by political affiliation. The mean of the score is 3.4 with a standard deviation of 0.8. Concerning citizens, two similar items were selected from the wave 22 of the Swiss Household Panel survey. The items were measured on a scale from 1 to 10 and read as follow: ‘Gender: Women in general penalized’ and ‘Gender: In favour of measures.’ We also built a mean score for gender equality support by political affiliation. The mean of the score is 5.8 with a standard deviation of 1.1.

3.4 Pre-processing Steps for Data Cleaning

Switzerland is a multilingual country with German and French being the most represented languages (Italian and Romansh are the other two national languages). To preserve the most authentic content of discussion, we did not translate the tweets into a single language. We nevertheless conducted several pre-processing steps. For instance, we filtered out URLs and characters that are not natural language texts. We also filtered out stop words, which are words that provide no information towards the analysis. We further split concatenated words (e.g., WomenStrike becomes women strike) and we lowercased the text. All further typos, misspellings, and slang terms remained intact. We then lemmatised the text using the library udpipe (Wijffels, Straka, and Straková 2018) for the programming language R.

3.5 Unsupervised Text Representation

Correspondence and cluster analysis of our corpus of tweets were used to investigate the theme and opinions surrounding the strike to address our third research hypothesis. To explore the content of the tweets and how opinions relate to the different actor categories, we analysed the co-occurrence of words in tweets, extracting shared semantic regions via correspondence analysis. To do so, we used the library FactoMineR (Husson et al. 2013) from the R language. Correspondence analysis can be understood as principal component analysis for categorical data. It is used to discover structure in textual data (D’Enza and Greenacre 2012; Morselli, Passini, and McGarty 2021). Correspondence analysis works as an unsupervised bag-of-words approach where the words are projected on a factorial space such that the proximity between words indicates a higher association (or a shared semantic meaning). Correspondence analysis calculates the contributions of each word to the inertia of a factorial axis, showing how each word contributes to identifying the axis. Hence, words that are projected further from the centre of the axis provide a higher contribution.

For our analyses, the data-term matrix was aggregated by the user categories. Using a graphical representation on a two-dimensional space, we projected the Twitter vocabulary on the correspondence analysis space to visualise ideas and opinions associated with each user category. As such, user categories that appear closer on the graph share a similar vocabulary and set of ideas. To account for the polarisation between left and right-oriented accounts, we created one category for each political orientation. Furthermore, to look at the extent to which what was said on Twitter is representative of what the lay audience thinks, we added open-ended responses from the Selects survey respondents to the items asking about the first and second ‘most important issue facing Switzerland’. The pre-processed (same steps as the Twitter data) respondents’ vocabulary was used as supplementary rows and was not used for the definition of the principal dimensions. Their coordinates were predicted using only the information provided by the performed correspondence analysis on the active vocabulary from the Twitter accounts.

4.1 Description of the Corpus

Table 1 displays the involvement of the different actor groups in our corpus of tweets. It provides the number of accounts, the number of tweets, and the tweeting frequency of the different user groups involved in online discussions about the women’s strike. In total, almost 16′000 unique accounts took part in the online discussions about the strike with an overall tweeting frequency of three tweets. Table 1 shows that the organisation committees were the most active users when considering the tweeting frequency. They are followed by their followers, Swiss political parties, and other involved users that tweeted more than five times. Table 1 also shows that Swiss political actors (parties, elected politicians, political candidates) emitted 11% of all collected tweets. When looking at the distribution of these political accounts according to political ideology (see Annex 1), we see a domination of the left in terms of number of accounts (55% of the Swiss political accounts) and tweets (75% of the Swiss political tweets). However, political accounts from the right were not absent from the online debate.

In complement to Table 1, Figure 2 displays the number of original tweets about the women’s strike emitted by actor groups over time. We see that organisation committees’ followers, other top users, and the media or journalists were essential actors that generated original content about the strike. Political candidates formed the third most prolific group. We also notice two peaks in the collected data pointing to two major events, namely International Women’s Day in March and the Women’s Strike in June.

Figure 2:

Prevalence of tweets by important actor groups over time.

4.2 Congruence Between the Online Involvement and the Offline Support for Gender Equality in Terms of Political Ideology

Figure 3 provides four panes displaying the relationship between the political leaning (x-axis) and the support for gender equality measures (y-axis). The panes are organised so that the offline and online patterns are compared horizontally, and so that the citizen and political patterns are compared vertically. The upper left pane describes the relation between citizens’ left-right positioning (x-axis) and citizens’ support for gender equality (y-axis). The lower left pane displays the same relation for politicians. Both of these left panes are solely based on survey data. The upper right pane includes the relative tweeting frequency of Swiss political actors active on Twitter (x-axis) in relation to citizen support for gender equality (y-axis). The lower right pane includes the relative tweeting frequency of Swiss political actors active on Twitter (x-axis) in relation to politicians’ mean support for gender equality (y-axis). Both of these right panes combine the salience of Twitter discussions about gender equality on social media (x-axes) and the opinions towards gender equality measured in survey data (y-axes).

Figure 3:

Relationship between online and offline gender equality opinions.

Figure 3 allows us to address our first research hypothesis about the distribution of political ideologies in relation to the support of gender equality measures. More concretely, we assess whether there is a congruence between the online involvement and the offline support for gender equality in terms of political ideology. The upper left pane shows that citizens with a left-leaning orientation have a more positive attitude towards gender equality measures compared to citizens with a right-leaning orientation. The lower left pane displays a similar pattern for politicians that responded to the survey. With respect to Twitter conversations, the upper right pane shows that politicians with a left-leaning position were more involved than politicians with a right-leaning orientation. Overall, social media discourses from politicians reflect the pattern survey data from their potential electorate. Here, citizens and politicians with a leftist orientation are clearly the most favourable towards gender equality measures and the most involved on social media.

Figure 3 also enables us to test our second hypothesis according to which we should observe increased levels of polarisation of the online debate on gender equality compared to trends observed through the lens of opinion surveys. This polarisation can be observed on both the upper and lower right panes of Figure 3, where there is a clear segregation between politicians with a left-leaning compared to the remaining politicians with either a centrist or a right-leaning orientation. The EDU (Eidgenössisch-Demokratische Union in German or Federal Democratic Union in English) is an outlier in comparison to the other rightist parties as it demonstrates a high relative tweeting frequency. This can be explained by the fact that the EDU made several actions to voice its opposition towards the legitimacy and the usefulness of the strike (see the “a rose for you” leafleting campaign, where the EDU aimed at thanking the women who would not go on strike). Overall, we observe that social media increases the polarisation between left and right-leaning politically involved actors compared to opinion surveys.

We rely on the correspondence analysis displayed in Figure 4 to test our third research hypothesis according to which we expect to observe a continuum between calling for attention, on the side of the strike organisation committee and of its followers, and discussing concrete policy measures, on the side of politicians and citizens. Figure 4 thus enables us to observe if online discussions are along a continuum between calling issues to attention and discussing concrete policy measures. The obtained two-dimensional space reveals the structure from the vocabulary employed by the actor groups. The final data-term matrix is based on 22′341 German tweets and 550 open-ended survey answers. The matrix includes 2′723 lemmatized terms which are either nouns or adjectives. Figure 4 shows the projection of the terms on a factorial space with the active user categories (in red) and the passive user category from survey respondents (in brown). The terms were automatically translated in English using deepL and the translation is given after the “_” on Figure 4.

Figure 4:

Graphical representation of the correspondence analysis.

The first dimension explains 18.3% of the variance. On the negative side of the axis, it includes terms such as “justice”, “assembly”, “consent” (lower left quadrant), and “principle” (upper left quadrant) in relation to terms such as “patriarchy” and supportive actions or movements (e.g., “collective”, “flyer”). On the positive side of the axis, it includes terms such as “council of states” and “understanding” (lower right quadrant), and “regulation” (upper right quadrant) in relation to policy issues about taxpayers, childcare, and wage discriminations. Therefore, this axis seems to refer to a “normative-representative” continuum going from the defence of women’s rights through norms and actions to the implementation of policymaking in the political arena.

The second dimension explains 15.1% of the variance. On the negative side of the axis, the figure includes terms such as “chinderchübu” and “gängeviertel” referring to public and open projects (lower left quadrant), and “albanieen” or “saxon” referring to debates about foreign aspects on Twitter (lower right quadrant). On the positive side of the axis, the figure includes terms such as “spfrau” and “parental leave initiative” referring to Swiss political initiatives in the framework of gender equality (upper right quadrant), and “strass” and “industry group” or “demonstrators” referring to important supporters and stakeholders actively taking part in the Swiss women’s strike (upper left quadrant). Therefore, this axis seems to refer to a “foreign-national” continuum going from the reference to foreign projects to the concrete Swiss mobilisation for the defence of women’s rights.

The first axis differentiates between the strike organisation committees, their followers, and organisations with gender-related aims on the left side, and political actors on the right side. Accounts from trade unions, media, activists, and other users are grouped in the centre. The second axis differentiates between the different political leanings of political accounts. The survey respondents are located close to the Swiss political accounts, at an equal distance from left and right-oriented accounts.