Walking the line of the double bind: A cross-country comparison on women and men politicians’ self-presentations on social media

The public and the private spheres that structure modern societies are historically closely connected with masculinity and femininity. In bourgeois societies, men were assigned responsibility for the public sphere, which comprised social areas like politics and economics, associated with power. Women were associated with the private sphere comprising family, the household, personal relationships, and emotionality. Even though the boundaries between the spheres and the gender-bound responsibilities have become blurred, the patriarchal structures, traditional gender role expectations, and gender stereotypes are still influential. In a comparison of 30 countries worldwide, Williams and Best (1990) identified pancultural gender stereotypes: Characteristics such as dominant, powerful, independent, strong, and active – which are seen as characteristics of successful leadership – were across countries assigned to men by the majority. Characteristics such as obedient, dependent, anxious, soft-hearted, passive, and weak – which seem rather diametrical to successful leadership – were predominantly assigned to women.

Gender stereotypes have a normative character: They “shape what people think is appropriate behavior from men and women” (Winfrey & Schoebelen, 2019: 113), independent of individual characteristics, competencies, and qualifications. Thus, they contribute to unequal starting conditions that make it more difficult for women to start and succeed in a political career: They affect how women and men politicians are assessed by both voters and party organisations in the male-centric political sphere “in which masculine values and political values are often interchangeable” (Meeks & Domke, 2016: 896). Men politicians benefit almost naturally from this gender role congruence between masculinity and leadership. Women politicians, by contrast, are confronted with the dilemma that gender stereotypes let femininity and leadership appear incongruent (Eagly & Karau, 2002), resulting in a dilemma which men politicians do not face in the same way: They must “walk the line of the double bind by communicating they are masculine enough to lead without violating feminine gender role expectations.” (Winfrey & Schoebelen, 2019: 121)

This requires a well-thought-out self-presentation strategy, which politicians’ social media presences have become a pivotal part of. According to strategic stereotype theory (Fridkin & Kenney, 2014), politicians – independent of gender – will conform to gender stereotypes in their self-presentation when it serves their political goals, but will act against them when the stereotypes threaten to harm them. They can also choose a mixed strategy (Schneider, 2014). However, men politicians can decide more freely than women which approach to follow and benefit more easily from deliberately, well-dosed counteracting of gender stereotypes, for example, by portraying themselves as caring family fathers (Brands et al., 2021); a focus on family for women, however, is often interpreted as a lack of focus on or a lack of suitability for political office. In an interview study with Canadian mayors, Sullivan (2023) found that only women experienced a mental load on how to best present themselves with respect to gender role expectations, resulting, among other things, from comments on their appearance, which men did not receive.

Despite these seemingly clear patterns, however, it is far from clear how women and men politicians handle gender stereotypes in their self-presentations on social media (Tsichla et al., 2023). Overall, there are relatively few studies on this question, and they reveal mixed findings. However, the majority of the existing studies found rather moderate differences in women and men politicians’ self-presentations on social media (Bast et al., 2022; Brands et al., 2021; Ekman & Widholm, 2017; Farkas & Bene, 2021; Geber & Scherer, 2015; Hrbková & Macková, 2021; McGregor et al., 2017) to none (Metz et al., 2020). Only a few studies revealed relatively clear gender differences in politicians’ self-presentation (Jungblut & Haim, 2023; Tsichla et al., 2023). When it comes to Germany and Norway – the countries we investigate – research is scarce. For Norway, we only found a comparative study by Bast and colleagues (2022) on the self-presentation of populist leaders on Instagram in several European countries, including Norway and Germany. Concerning Germany, Geber and Scherer (2015) and Metz and colleagues (2020) investigated the self-presentations of German politicians on Facebook but included the politicians’ gender as a peripheral aspect. Germany was also included in a study by Jungblut and Haim (2023) on politicians’ visual self-presentation on Instagram and Twitter across 28 countries in the 2019 European elections, but they did not present results on the individual countries.

Politicians’ self-presentations are part of their strategy to generate votes, the ultimate goal of election campaigns, and thus influence voting behaviour. According to the Michigan model of voting behaviour, voting decisions are influenced by three factors: party identification as a long-term factor, as well as attitudes towards candidates and attitudes towards issues as short-term factors (Asher, 1983). Due to a dealignment process during the last decades, party identification has lost influence (Dalton & Wattenberg, 2002), while the short-term factors have become more important for voting decisions. Therefore, we focus on two strategies of (gendered) self-presentation directly related to the short-term factors: personalisation relates to attitudes towards candidates, while gender issue ownership relates to attitudes towards issues.

Personalisation of politics describes a state and a development through which individual politicians are or become more central in politics, news coverage, and citizens’ perception of politics (Van Aelst et al., 2012). While not a new phenomenon, personalisation is central to politicians’ communication on social media. It is a communication strategy to reduce complexity of political issues, which resonates well with the brevity of communication on social media. Moreover, the basic idea of social media is to provide a platform on which individuals can share information about themselves and their personal lives, which shapes both users’ expectations on what kind of content they receive there and the selection criteria of the algorithms. Politicians must, to a certain degree, adjust their communication to these rules (McGregor et al., 2017; Tsichla et al., 2023).

Van Aelst and colleagues (2012) distinguished between two forms of personalisation in news: individualisation and privatisation. Individualisation “concerns a focus on individual politicians as central actors in the political arena, including their ideas, capacities and policies” (Van Aelst et al., 2012: 204–205). While often still focusing on substantive political issues, visibility shifts from parties or governments to individual politicians, for example, by demonstrating a political leader’s professional competence (Karlsen & Skogerbø, 2015). Privatisation, by contrast, “implies a shift in media focus from the politician as occupier of a public role to the politician as a private individual, as a person distinct from their public role” (Van Aelst et al., 2012: 205). Indicators are, for example, a focus on personal characteristics and on the private lives of politicians. From a normative perspective, privatisation is often considered problematic, since it turns away the focus from the political sphere. With the stereotypical dichotomy between public (“male”) and private (“female”) sphere in mind, individualisation is closer to male stereotypes, while privatisation matches better with female stereotypes.

This differentiation has been used in some studies on the self-presentation of women and men politicians on social media (e.g., Brands et al., 2021). For Germany (Haßler et al., 2023; Metz et al., 2020) and Norway (Enli & Skogerbø, 2013), it has been shown that individualisation is relatively common, while privatisation happens only rarely, but without focusing on gender differences in politicians’ self-presentation. However, given the poor state of research, we formulated a research question for our comparison between Germany and Norway:

RQ1. To what degree did women and men politicians use individualisation and privatisation elements in their self-presentations on social media?

Another focus in existing research on politicians and gender stereotypes is the association of certain political issues with women and men politicians. Following the traditional dichotomy of public and private spheres, both genders are stereotypically considered more competent to handle certain issues: Men are considered more competent in relation to policy issues that are strongly associated with leadership (Brands et al., 2021), for example, economy and finance, (national) security, crime, terrorism, and defence, while women are associated with “soft”, “compassion” policy issues related to caring, for example, health, family, social policy, education, environment, and culture (Lee & Lim, 2016; Winfrey & Schoebelen, 2019). This so-called “gender issue ownership” (Herrnson et al., 2003) is often reflected in which ministries are headed by men and women. “Male” issues are often perceived as societally more important – including by the voters and the news media (Zulli, 2019) – which brings along larger budgets, more societal power (Cardo, 2021; Lee & Lim, 2016), and higher news visibility (Zulli, 2019).

If politicians choose a gender issue ownership strategy, they address issues “matching” their gender. However, it can also be promising to address “male” issues as a woman and vice versa, for example, due to proven personal competence for certain topics, to add new facets to their profile to become more relevant to new target groups, or since their party is considered more competent to handle issues (party issue ownership; Petrocik, 1996) which are “contradictory” to the politicians’ gender (Fridkin & Kenney, 2014). Moreover, it can be strategically advantageous to address issues that voters are particularly concerned about (riding the wave; Ansolabehere & Iyengar, 1994), even if they “contradict” a politician’s gender. A recent example thereof are environmental issues: Given that this was by far the most important issue in the 2021 national election campaign both in Germany (GLES, 2023) and in Norway (Institute for Social Research, n.d.), there may have been something to gain by addressing this “female” topic, even as a man. The state of research on gender issue ownership in politicians’ self-presentations on social media is mixed (Brands et al., 2021; Cardo, 2021; McGregor et al., 2017), and studies on Germany and Norway are missing. This led to our next research question:

RQ2. To what degree did the topics addressed by women and men politicians in their social media campaigns reflect gender issue ownership?

How political leaders manage to walk the line of the double bind is affected by the framework conditions under which they work and communicate. These are shaped by both social media platforms and countries.

At the meso-level, platforms have differing characteristics – for example, the technical features and the usership of the platforms (Kreiss et al., 2018) – and they can facilitate different campaign strategies. A current study by Larsson and colleagues (2024) – which, however, does not deal with the self-presentation of political leaders – has shown that Norwegian political parties employed different campaign strategies on Facebook, Instagram, and Twitter in 2021. Concerning the platforms we compare here, Facebook is more text-heavy (Magin et al., 2024), while visuals are more central on Instagram (Brands et al., 2021; Ekman & Widholm, 2017). Following previous scholarship, such a visual focus might contribute to a stronger focus on the private sphere on Instagram (Kreiss et al., 2018). Concerning usership (see Table A1 in the Appendix), both platforms are more popular in Norway than in Germany, which might make them more important campaigning tools in Norway. Instagram has younger users than Facebook, which makes it attractive in election campaigning for reaching age groups that are more difficult to reach via more traditional communication channels. While Facebook is more popular than Instagram independent of users’ gender, Instagram is more popular among females than among males, particularly in Norway (Newman et al., 2021). However, it is unclear how these differences relate to the strategic self-presentation of women and men politicians, which led to our next research question:

RQ3. To what degree did the self-presentations of women and men politicians differ between Facebook and Instagram?

Despite many similarities, however, both countries differ in one respect which might affect how women and men politicians present themselves, and it is thus pivotal to our research interest: Gender equality is more pronounced in Norway (as shown in Table 1) for both structural indicators (e.g., a smaller gender pay gap, less women working part-time, and more men taking parental leave) and gender norms that people express: The gender social norms index which is calculated based on seven questions from the World Values Survey about gender equality regarding politics, education, economy, and physical integrity shows that 63 per cent of Germans did not show biases in any of these indicators in 2017–2022. In Norway, this share was almost equally high already in 2005–2009 (the last wave from which data are available), and it can be assumed that it has further grown since then. Concerning politics, Norway has both a longer tradition with women heads of government and a higher current proportion of women various political offices.

Even though both countries show a high degree of gender equality in international comparison (World Economic Forum, 2022), the differences between them are highly relevant for our research interest since they both reflect and shape gender role expectations. They can also affect politicians’ self-presentations since politicians and their campaign teams do not only orient toward voter expectations but are also (consciously or unconsciously) influenced by gender role expectations themselves. However, since these mechanisms are complex and far from monocausal, we formulated the following research question:

RQ4. To what degree did the self-presentations of women and men politicians on Facebook and Instagram differ between Norway and Germany?

We answer our research questions by means of a standardised content analysis of all posts published on the official public Facebook and Instagram accounts of all party leaders or frontrunners whose parties made it into the parliament after the election (Norway⁽¹⁾: 4 female, 5 male; Germany: 3 female, 6 male) during the last four weeks before (and including) election day (Germany: 30 August–26 September 2021; Norway: 17 August–13 September 2021). We focus on political leaders since they are widely perceived as prominent politicians by the population, which makes their self-presentation particularly influential, potentially. The sample consists of 1,448 posts (see Table 2). The post content and metadata were collected using the CrowdTangle API. For coding the posts, the coders opened them on the respective platform via the URL.

This study builds on the larger, comparative project “Digital Election Campaigning Worldwide (DigiWorld)”, which investigates the use of social media in election campaigns in more than 30 countries using a joint English codebook.⁽²⁾ Due to reasons of practicability, our coding included the entire text of each post and the first image or the (first minute of the) first video included in it. While other studies in this field employed automated methods (e.g., Enders et al., 2022; Sandberg, 2022), we coded the materials manually, which entails one decisive advantage: Automated procedures such as topic modelling and natural language processing are based only on text but neglect visual materials, even though the latter are central to fully understanding the meaning of social media posts. For example, they often contain content elements not covered in the text. Moreover, visuals are important for catching user attention and being favoured by the social media algorithms, which increases posts’ visibility on the platform. That only 2 per cent of all posts in our sample did not include any visuals reflects their centrality.

All categories were coded independent of one another. Thus, as many categories as present could be coded per post (for category descriptions, see Table A2 in the Appendix):

Politicians: The 18 political leaders as owners of the sites from which the post originated were coded automatedly when collecting the data. For our analyses, we recoded this category into a binary category differentiating between women and men politicians.

Separate national coder teams (Germany: 6 coders; Norway: 3 coders) were intensively trained on the joint English codebook, allowing for cross-country comparability. Since the coding in both countries took place at different points in time, and the materials were in different languages, we could not perform cross-country reliability tests, but we performed inter-coder reliability tests within each country (Germany: 150; Norway: 140 – randomly selected posts). The researchers from both countries were closely involved in the development of the codebook. In addition, to ensure cross-country comparability, they exchanged views regularly during the coder training. The codebook builds strongly on a codebook we used in the previous project “Campaigning for Strasbourg (CamforS)”, which achieved satisfactory cross-country reliability for the issue categories (the other categories used here were not included in that codebook).

We calculated reliability indicators using Holsti’s formula and Brennan & Prediger’s kappa. Since we relied on binary codings and many categories appeared sparsely, Brennan & Prediger’s kappa provided a robust result. Among the chance corrected indicators for reliability, Brennan & Prediger’s kappa appears to be better suited than Krippendorff’s alpha for evaluating the reliability of skewed variables like ours (Quarfoot & Levine, 2016). Reliability was satisfactory for all variables used in our analyses (see Table A2 in the Appendix).

For Figures 1–4, grey indicates the degree to which a specific condition was not found for each variable investigated below. We start with individualisation as measured by references to professional competencies. Figure 1 shows that these were more common in Germany than in Norway on both platforms. As indicated by percentages, a stronger effect of gender can be discerned in Germany than in Norway: In Germany, men referred to professional competencies more often than women on both platforms, in line with gender stereotypes – more than 2.5 times as often on Facebook and 5 times as often on Instagram. In Norway, by contrast, we found only slight differences between women and men politicians in this regard.

Moving from individualisation to privatisation, Figure 2 details the percentages of posts that featured at least one referral to home and family. We hardly found any differences between women and men politicians in this respect for Facebook in both countries, as well as for Instagram in Germany. However, the reported percentages indicate that Norwegian women politicians referred clearly more often to home and family on Instagram than the men.

Figure 3 shows the results for referrals to personal preferences. For Facebook in both countries and Instagram in Germany, we found similarly low percentages for both women and men politicians. In Norway, however, women politicians referred to personal preferences twice as often as male politicians – a clear difference.

When it comes to private background stories, Figure 4 shows that these were hardly used in Germany at all, independent of gender. In Norway, the candidates used them more often, particularly on Instagram. The clearest gender difference we found is that Norwegian women politicians used private background stories on Instagram more than three times as often as men politicians.

Now we turn to the topics addressed by women and men politicians. Figure 5 shows the percentage of the posts by women and men politicians from both countries on the two platforms which addressed the policy issues we look at. Referring to gender issue ownership, the upper panels show the “female issues” and the lower panels show the “male issues”. Figure 5 bundles all three comparative dimensions we look at:

Gender: Women politicians are represented by triangles, men politicians by squares.

The larger the triangles and squares, the higher the number of posts containing the respective issue. Where on the x-axis a triangle or square is located indicates the percentage of posts containing the respective issue.

When it comes to the results, Figure 5 shows that the campaigns in both countries focused relatively strongly on certain issues while neglecting many others, as shown by the larger squares and triangles for these issues. Among the three most addressed issues were one “male” issue (economy/finance) and two “female” issues (environment, social policy), which might seem surprising against the background that “female” issues are traditionally considered least important in politics. However, which topics are addressed in election campaigns also depends on the current situation, and in both countries, environmental issues were high on the agenda at the time of the election campaigns (GLES, 2023; Institute for Social Research, n.d.). Thus, politicians of both genders seem to have set on this “hot issue”, following the “riding the wave strategy” (Ansolabehere & Iyengar, 1994). The finding on social policy might have methodological reasons: Our social policy category included, among others, “labour”, in which case the gender-specific assignment is less clear.

To find out how far the politicians’ posts were shaped by gender issue ownership, we compare how often women and men politicians addressed female and male issues. However, Figure 5 does not show any clear, continuous patterns. The only issue which was completely in line with the expectations of gender issue ownership was economy/finance: Men addressed this issue clearly more often than women in both countries on both platforms. For four topics – two “female” (families/children, civil rights/gender) and two “male” (domestic policy, defence policy) – we find hardly any gender differences. One probable reason is the very low number of posts addressing these issues at all.

For most issues, however, we find mixed results. Some differences between women and men politicians are in line with the expectations of gender issue ownership: The “female” issue of environmental policy was addressed more often by women than men on Facebook in Germany. Norwegian women politicians addressed educational policy and cultural policy more often than men politicians. Concerning “male” issues, the only clear gender-specific difference besides economy/finance is that Norwegian men politicians addressed foreign policy on Instagram more often than women politicians.

In contrast, we also find clear differences that run counter to what gender issue ownership expects: Concerning “female” issues, German men politicians addressed education and cultural policy more often than women politicians on Instagram. In Norway, social policy (on both platforms) and environmental policy (on Instagram) were addressed more often by men than by women politicians. Concerning “male” issues, Norwegian women politicians addressed energy policy more often than men politicians on Facebook. Overall, however, we find only low to moderate indication of gender issue ownership.