Insta(nt)famous? Visual self-presentation and the use of masculine and feminine issues by female politicians on Instagram

ABSTRACT

This study examines how female politicians are using Instagram to present themselves to the electorate and how this affects audience engagement. A manual content analysis was conducted to explore how female politicians, compared to male politicians (N = 40), use Instagram in terms of visual self-presentation, the use of masculine and feminine issues, and how this may lead to increased engagement (i.e., likes and comments). In total, N = 762 posts were manually analyzed. The study shows that female politicians receive more likes when they are visible in a picture compared to male politicians. It also reveals that male and female politicians both refer more to feminine issues than to masculine issues, although the use of feminine issues resulted in less likes. We find that some issues lead to more discussion amongst Instagram users, and that this differs between male and female politicians. The study sheds light on how politicians use Instagram and offers insights into how (female) politicians can use the platform to their advantage.

KEYWORDS:

• Gender
• Instagram
• Visual Political Communication
• masculine and feminine issues
• user engagement

Politicians increasingly use social media to engage with voters (McGregor et al., 2017; Meeks, 2016; Metz et al., 2019). Social media may be especially beneficial for female politicians, as it offers them a way to bypass traditional media (Ritchie, 2013; Toffoletti & Thorpe, 2018). Studies show that traditional media often portray female politicians in biased ways (Van der Pas & Aaldering, 2020). Female candidates suffer from both a stereotypical portrayal in the media and a lower amount of media attention (Ritchie, 2013; Yarchi & Samuel-Azran, 2018). This has serious consequences for them, as less media attention often results in less votes (Van der Pas & Aaldering, 2020).

Previous work found that female politicians are able to generate more public engagement compared to male politicians on Facebook (i.e., during the 2015 Israeli elections campaign). This indicates more optimism with regard to the representation of women in politics (Yarchi & Samuel-Azran, 2018). However, only a limited amount of studies investigate how female politicians use social media, and also Yarchi and Samuel-Azran (2018) urge others to investigate other social media and countries. We also explicitly try to overcome the shortcoming that previous work focused predominantly on textual elements in social media content (see, e.g., Kruikemeier, 2014; McGregor, 2018; Yarchi & Samuel-Azran, 2018). The increasingly visual nature of social media calls for new forms of self-presentation and image management (Muñoz & Towner, 2017). Words might become less, and pictures become more important online (Strand & Schill, 2019). However, female politicians might be more judged on their physical appearance (Beltran et al., 2020), indicating that female politicians may be more vulnerable on visual social media. The (growing) importance of visuals in political communication is widely acknowledged (e.g., Ekman & Widholm, 2017; Gerodimos, 2019; Hale & Grabe, 2018; Lalancette & Raynauld, 2019; Muñoz & Towner, 2017). Yet, visual aspects of political communication are still poorly understood (Gerodimos, 2019; Hale & Grabe, 2018; Schill, 2012). Moreover, in today’s information-overloaded society, visuals may be the one avenue that still has the potential to grab attention (Lilleker et al., 2019), which, given our considerations outlined above, may be especially consequential for female politicians.

A platform that distinguishes itself from other social networks through its ‘image first logic’ is Instagram (Ekman & Widholm, 2017, p. 18). In this study, we focus on Instagram for several reasons. First, Instagram is becoming increasingly important for political candidates, as voters increasingly use Instagram in their daily life (Bossetta, 2018; Filimonov et al., 2016). Second, because it is a ‘visually rich’ platform, it can be used for self-presentation and branding, such as pictures of the campaign rally or family photos. And third, Instagram is regarded to have an ‘aesthetically-oriented character’, in which mainly beautiful, carefully curated, and esthetic photos are posted (Pereira Caldeira, 2021, p. 9). This combination could make it an ‘ideal political marketing platform’ (Muñoz & Towner, 2017, p. 291). To date, only a handful studies focused on the use of Instagram by political candidates (e.g., Ekman & Widholm, 2017; Filimonov et al., 2016; Lalancette & Raynauld, 2019; Larsson, 2017, 2019; Muñoz & Towner, 2017; Russmann et al., 2019).

The current study examines how female politicians from the Netherlands and the United States use Instagram and present themselves (i.e., their visual self-presentation and the use of specific topics) on the platform compared to male politicians. Additionally, the impact on audience engagement will be examined. Audience engagement (such as likes and comments) can be considered indicators of public engagement (Metz et al., 2019, p. 5).

Audience engagement is usually understood as the sum of metrics such as comments, likes, replies, etc., depending on the platform. Yet, it is important to understand that these different components of engagement have different affordances and can serve different functions (e.g., Ferrer-Conill et al., 2021; Savolainen et al., 2020). Also empirically, it has been shown that the number of, for instance, shares, likes, and comments of political content is not influenced in the same way by the same predictors (e.g., Judina & Platonov, 2019; Larsson, 2018; Trilling et al., 2017). In the particular, liking political content has often been criticised as a means of engagement that takes so little effort that it hardly qualifies as a ‘real’ political action, as opposed to, for instance, writing a comment, which can require quite some cognitive resources and can lead to a meaningful exchange of opinions (see, e.g., Larsson, 2018).

In addition to the number of likes and comments, we examine the amount of discussion a post generates to tap into the possibilities of Instagram as a social media platform that fosters deliberation. To investigate our aim, this study uses a manual content analysis on the visual and textual elements of Instagram posts, as only together, their meaning can be fully understood (Sloan & Quan-Haase, 2017).

Theoretical background and related research

Visual self-presentation of female politicians

In this section, we first focus on visual self-presentation. We start by explaining the importance of a focus on visuals in social media, we then explain how we define visuals in this study, and lastly, we discuss the visual self-presentation of female politicians.

Visuals ‘play a critical role in how members of the public perceive, understand, and evaluate politicians’, especially on visual social media platforms (Lalancette & Raynauld, 2019, p. 916). In this regard, scholars speak of the growing significance of ‘image bites’ as counterpart of ‘sound bites’ (Marland, 2012; Schill, 2012). Images are arguably more powerful than text: words matter, but visual impressions are more likely to stick because they evoke stronger emotional reactions and appeal to the subconscious (Lilleker et al., 2019; Muñoz & Towner, 2017; Russmann et al., 2019; Schill, 2012). Political images thereby have the ability to set the agenda and help politicians in building trust or promote authority and legitimacy, making them a powerful tool (Pauwels, 2019; Russmann et al., 2019; Strand & Schill, 2019).

The extant literature uses different definitions when speaking about visuals in online political communication, including visual communication (Russmann et al., 2019; Schill, 2012), visual discourse (Adi et al., 2018), visual self-presentation (Muñoz & Towner, 2017), visual framing (Grabe & Bucy, 2009), and visual branding (Strand & Schill, 2019). Following Muñoz and Towner (2017), the term ‘visual self-presentation’ is used in this study. It is important to note that the current study does not assume that (all) forms of visual self-presentation are consciously applied in posts; instead, our aim is merely to explore which forms female politicians use and how this influences audience engagement. There are also innumerable ways in which political candidates can visually self-present themselves: scholars have used up to 104 different coding categories while studying visual political communication (Adi et al., 2018). Strategies range from using a certain camera angle to wearing specific colors, sometimes carefully managed by image consultants. Despite the fact that visual analyses can be useful in many fields, a common methodological framework is lacking (Gerodimos, 2019). Based on multiple studies on visual political communication, the current study constructed its own categorization, focusing mostly on how candidates are portrayed in relation to others. Variables include distance to others, interaction with others, domination, seriousness, camera perspective, and clothing.

Physical appearances play an important role in visual self-presentation. This may be especially the case for female politicians. Previous research found that female politicians receive more traditional media coverage (television and newspaper) that was focused on their looks (physical appearance), more personal coverage about their family life, and their gender is more often mentioned compared to male politicians (Van der Pas & Aaldering, 2020). On social media, female candidates might want to counter this bias in reporting by using visuals on social media that put less emphasis on their gender or their looks: For instance, female politicians might dress up more formally than men (Lee & Lim, 2016). However, this relationship may be more complicated than it seems. Flicker (2013) argues that female politicians, in essence, cannot do it right: When dressing feminine they might be perceived as less competent, but when dressing according to male dress codes, they seem to reinforce the male cliché of politics. Pedersen (2018) connects the focus on physical appearances of politicians to personalization, arguing that the way politicians dress, implicitly refers to their personal lives or their sexuality (as this also happens in traditional media coverage; Van der Pas & Aaldering, 2020). In this regard, dressing casually might help a male candidate to present himself as more approachable, while it might be less beneficial for a female candidate. The same is true for smiling, being close to people, and for interacting with others (Verser & Wicks, 2006), also referred to as positive affect cues (smiling) and immediacy cues (proximity to others and interaction with others). These cues have been associated with femininity in earlier studies (see e.g., Carpinella & Bauer, 2021; Chen et al., 2020). Wearing more formal clothing and other visual strategies, such as domination (being at the center of attention versus being part of a group), distance to others, and looking serious, may help politicians to radiate more authority, establish legitimacy, and show the electorate that they are ‘ready for leadership’ (Verser & Wicks, 2006) and are associated more with masculinity (see, e.g., Carpinella & Bauer, 2021; Chen et al., 2020). Next to nonverbal cues, camera perspective matters. Ekman and Widholm (2017) argue that face close-ups, selfies, or half-body shots represent more intimacy, while full-body and long shots represent public distance. They found that posts from female politicians on Instagram contain more close-ups and half-body shots compared to posts from male politicians. (Extreme) closeups of female politicians have also been linked to femininity (see, Lundell & Ekström, 2008).

Up till now, there is less research available on how female candidates present themselves on social media, but Yarchi and Samuel-Azran (2018) emphasize that ‘social media provides greater opportunities for female politicians to promote themselves and improve their status in the political power play’ (p. 978). We build on this work by focusing on visual presentation on Instagram in two countries to increase the knowledge around this important research avenue.

Taken together, this study will explore which forms of visual self-presentation are used by female politicians on Instagram. In the methods section, we will outline how these diverging conceptualizations can be operationalized into one original codebook, which we will then use to address multiple research questions. To start with, we ask:

(RQ1) Which forms of visual self-presentations do female politicians use on Instagram and

(RQ2) to what extent is this different compared to male politicians?

Lastly, we are interested in the effects of self-presentation on audience engagement. Previous research found that Facebook use by female politicians generated more engagement (i.e., likes and shares) compared to male politicians. However, they received the same number of comments and participants in the discussion (Yarchi & Samuel-Azran, 2018). However, due to the lack of specific research, we pose the following research questions:

(RQ3) (a) Does the use of specific forms of visual self-presentations lead to more audience engagement, and (b) does this depend on the gender of the politician?

Political issues: stereotypically masculine and feminine issues

When studying the ways in which (female) politicians present themselves online, we also need to understand political issues or topics politicians bring forward (Yarchi & Samuel-Azran, 2018; Druckman et al., 2010; Lalancette & Raynauld, 2019). Addressing specific issues can help politicians in shaping a certain image or personal brand (Ekman & Widholm, 2017; Lalancette & Raynauld, 2019; Russmann et al., 2019). It is often thought that female politicians are better at dealing with ‘feminine issues’ and male politicians with ‘masculine issues’ (Van der Pas & Aaldering, 2020). Specifically, masculine and feminine issues are stereotypical issues that are usually owned by the corresponding gender (Huddy & Terkildsen, 1993a, 1993b; Lee & Lim, 2016; Meeks & Domke, 2016). Feminine issues include ‘softer’ topics like education, family, healthcare, and arts, whereas examples of masculine issues are taxes, guns, business, and terrorism (Lee & Lim, 2016). The mentioning of certain issues can have political consequences. Earlier studies have shown that masculine issues are more often covered by the traditional news media, next to being associated with leadership (Zulli, 2019). At the same time, female politicians are more likely to talk about feminine issues and are, interestingly, more likely to win elections if they do so, whereas the same is true for male politicians and masculine issues. This mechanism is referred to as ‘gender issue ownership’ (Herrnson et al., 2003, p. 247; Huddy & Terkildsen, 1993a, 1993b; Lee & Lim, 2016). It implies that voters perceive women as more capable of handling political issues, such as human rights, welfare, and the environment, whereas men would be more competent to deal with the economy, foreign affairs, or the military (Lee & Lim, 2016). Politicians can use this to their benefit: Herrnson et al. (2003) found that women who brand themselves as women are more likely to win elections, as they can exploit voters’ gender stereotypes. This can thus be beneficial for politicians’ campaigns and might actually already be taken into account. Some research shows that politicians seem to mention gender-congruent issues on social media more often than issues that do not correspond with their gender. A study on Twitter use by Hillary Clinton and Donald Trump showed that Clinton mentioned feminine issues almost twice as much as masculine issues, while the opposite was the case for Donald Trump (Lee & Lim, 2016). If we look at a meta-analysis on traditional media coverage, however, Van der Pas and Aaldering (2020) find some, albeit inconclusive, evidence for a gender bias. This study will examine whether the dynamic of gender issue ownership also exists on Instagram.

When we look at research on digital media, scholars found differences in the issues raised on social media. Male politicians are more likely to discuss security issues, economy, and welfare on Facebook (Yarchi & Samuel-Azran, 2018). Female politicians are more likely to attack their opponent, and they deal more often with gender issues on Twitter and Facebook (Evans & Clark, 2016; Yarchi & Samuel-Azran, 2018), and present themselves on Facebook (Yarchi & Samuel-Azran, 2018). No difference was found in the presentation of personal issues on Facebook (e.g., private lives, families; Yarchi & Samuel-Azran, 2018). This latter finding is in contrast to McGregor et al. (2017), who found less personalized communication on social media by female candidates compared to male candidates (Facebook and Twitter). In general, researchers focusing on social media platforms, note that politicians, in their online communication, steer users towards personalization – especially in the case of a visually minded platform like Instagram (Ekman & Widholm, 2017). However, political personalization studies also analyze politicians’ references to policy issues or ‘professional’ self-personalization (McGregor et al., 2017; Metz et al., 2019). These are posts containing references to specific political issues, such as foreign relations or taxes, instead of posts focusing on emotional self-personalization or personal issues. These studies found that most posts in their sample contained professional self-personalization (Metz et al., 2019) and that politicians posted around twice as much about policy issues than about personal issues (14.9% as opposed to 7.9%, McGregor et al., 2017).

In general, less evidence exists on the specific policy issues politicians mention in their posts or tweets. This is a paucity in the literature. From a democratic viewpoint, citizens should be informed about specific political issues and topics for the electorate, as it might help them to understand which topics politicians care about. We ask:

(RQ4) Which political issues do female politicians refer to on Instagram and

(RQ5) to what extent is this different compared to male politicians?

Lastly, we examine the effects of specific issues on audience engagement and thereby take gender into account. As research on the engagement of audience is lacking here as well, we pose two research questions:

(RQ6a) Does the use of specific political issues lead to more audience engagement and discussion among users, and (b) does this depend on the gender of the politician?

Method

This study uses a quantitative content analysis of Instagram posts by Dutch and American politicians. We are mostly interested in which social media strategies are used by political elites. Posts were collected over a period of ten months, from September 2018 until June 2019. Whereas a majority of studies focuses on campaign periods when studying social media content (see, e.g., Metz et al., 2019; Strand & Schill, 2019), this study will focus on a longer period of time in which a post- and pre-election period is included for both countries.¹ Given the state of permanent campaigning politicians are in nowadays, periods outside of elections are becoming increasingly important to investigate (Metz et al., 2019; Vasko & Trilling, 2019). An important condition is to focus on countries where social media adoption is high (by elites and citizens) and female politicians are using Instagram on a regular basis. Moreover, this study should be viewed as explorative as it is mainly focused on differences between female and male politicians and less between countries. Earlier studies found that politicians from both countries are very social media oriented (Jacobs & Spierings, 2016; Tromble, 2018), although their political systems are quite different, with a multi-party system in the Netherlands and a two-party system in the US. At the same time, both are western democratic countries and provide similar contexts for female politicians.

Sample

To select a subsample of politicians for the manual coding, lists of politicians from both countries were compiled. This included the cabinet and all members of congress and the House of Representatives in the US, and the cabinet and all members of de Eerste Kamer, Tweede Kamer and European Parliament in the Netherlands. For each country, the ten male and ten female politicians with the highest number of followers were included in the final sample, adding up to a total of 40 politicians (see Appendix A, Table 1). Only politicians with more than 100 posts were included in the final sample. Following these criteria, all politicians in the sample were active and engaged Instagram users. In total, 20 posts of each politician were gathered, using the first and last post of each month for analysis.

The unit of analysis is a single Instagram post, which includes the visual as well as the caption (i.e., a textual description of a post that can be added by Instagram users). When a post consisted of multiple images, only the first one was included for analysis. Videos were excluded from the study. All posts were collected simultaneously using Python to ensure that changes over time (e.g., an increase of likes or comments) could not affect the results. In total, 762 Instagram posts were collected and manually coded. For each post, a link to the post, the caption, number of likes, and number of comments were collected.

Following Yarchi and Samuel-Azran (2018) and Metz et al. (2019), control variables used in the analyses were politician’s age (M = 49.65, SD = 13.47), number of total posts (M = 829.23, SD = 908.35), ideology (M = 1.53, SD = 0.72), number of followers (M = 879,720.48, SD = 2,571,277.2), presidential candidacy for 2020 (M = 0.18, SD = 0.38), and current president/prime minister (M = 0.05, SD = 0.22). We additionally controlled for the total number of posts shared in the sample period (M = 203.33, SD = 199.02). It is important to note that the sample is not evenly distributed when it comes to ideology: due to the low number of followers, no female Republicans are included in the final sample.

Sample description

The sample consists of 20 male and 20 female politicians, with an equal distribution across countries and gender, but an unequal distribution across ideology. Because of the sampling method, there is a bias towards candidates that are more popular on Instagram. A majority of male US candidates is Republican (eight out of ten), whereas all female candidates are Democrat, which may be explained by the fact that women are underrepresented in the Republican party. In the Dutch sample, a majority of female politicians also belong to parties on the left side of the spectrum. The average age of politicians in the sample is 49.65 years old. Male politicians were slightly older (M = 51.85, SD = 14.27) than female politicians (M = 47.45, SD = 12.59), and there was a considerate age difference between Dutch politicians (M = 43.90, SD = 8.90) and politicians from the US (M = 55.4, SD = 14.94). The youngest politician in the sample was 30 years old, the oldest was 79, both were female politicians from the US.

Operationalization

As Gerodimos (2019) states, ‘one of the most notable gaps in the literature of social research methods is the lack of coding frameworks for the content analysis of visual data’ (p. 64). Therefore, it is not uncommon to use ‘an original coding framework’ (Adi et al., 2018). Hence, original variables were developed for the purpose of this study. The variables distance to others (M = .16, SD = .37), interaction (M = .75, SD = .44), domination (M = .47, SD = 0.50), and seriousness (M = 1.87, SD = 0.56) were based on the typology by Verser and Wicks (2006) and were modified accordingly to suit the Instagram context. The variables of interaction and domination were coded as present (1; ‘candidate is interacting with other people’, ‘candidate dominates the photo and forms the center of attention’) or absent (0; ‘candidate is not interacting with other people’, ‘candidate appears more among a group and does not dominate the picture’). The variable ‘distance to others’ was also based on Verser and Wicks’ (2006) operationalization, using the coding categories ‘candidate appears close to other people and could be physically touched’ versus ‘candidate appears far from others’. These three variables were coded as dummies, just as the variables ‘others’ (0; ‘candidate is alone’ versus 1; ‘candidate is photographed with other people’) and ‘politician visible’ (visible (1) versus not visible (0)). For the variable ‘seriousness’, we followed Verser and Wicks (2006) by distinguishing between ‘candidate appears serious’ and candidate looks playful, cheerful or smiling. Additionally, we added a third category for cases in which the candidate’s face was not clearly visible (‘not clear’).

Following Flicker (2013) and Verser and Wicks (2006), clothing was operationalized by coding for formal (versus less formal) clothes (M = 2.17, SD = 1.03). According to Flicker (2013) the dark suit (or pantsuit for women) is still the ‘globalized dress code and symbol for management, politics and masculinity’ (p. 206). Instead of coding solely for formal versus informal clothes, three separate coding categories were created, thereby distinguishing between a full suit, formal to casual clothing (not a full suit, but still formal), and informal or casual clothing. Lastly, camera perspective was added (M = 1.35, SD = 0.48), which was operationalized as close-up versus full body shot or long shot of the candidate, the latter showing more distance. For all these variables, only the visual was taken into account during coding, and not the caption. Following Metz et al. (2019), considering the complexity of coding visuals the codebook included elaborate instructions and multiple (visual) examples for each variable and the related categories.

The typology of Lee and Lim (2016) was followed to code for masculine and feminine issues. Feminine issues coded for are children and education (N = 129, 16.9%), health (N = 62, 8.1%), rights (N = 102, 13.4%), family (N = 79, 10.4%), immigration (when referring to the rights of immigrants; N = 6, 0.8%), welfare (N = 21, 2.8%), poverty (N = 9, 1,2%), environment (N = 65, 8.5%), animal rights (N = 14, 1.8%), and arts (N = 16, 2.1%). Masculine issues coded for are taxes (N = 10, 1.3%), guns (N = 7, 0.9%), military (N = 30, 3.9%), foreign affairs (N = 79, 10.4%), immigration (when referring to border control, N = 13, 1.7%), economy (N = 42, 5.5%), business (N = 24, 3.2%), national defense (N = 19, 2.5%), science and technology (N = 6, 0.8%), and terrorism and crime (N = 26, 3.4%). For each issue we coded for present (1) or absent (0) in the post. All the issues were elaborately explained to ensure that the first and second coder included the same subjects for each issue. The visual and caption of a post were both taken into account.

Following Metz et al. (2019), the dependent variable of audience engagement was operationalized as the numbers of likes and comments for each post as retrieved by our Python script. We also investigated the amount of discussion a post generated. To do so, the number of original comments and the number of comments that are placed below these original comments were captured. We wrote a Python script to gather the number of original comments. The total numbers of comments are displayed on Instagram and were copied directly from the platform at the same time as the original comments were gathered.

Reliability

All coding was performed by the first author. A subsample of 10.4% was coded by a second coder to test for intercoder reliability. The second coder was a graduate student at the University of Amsterdam and was trained by the first author. Before coding, the codebook was elaborately explained to the second coder, after which some minor changes were made to the codebook by adding more explanations for some of the coding categories that were not entirely clear yet to the second coder. Krippendorff’s α was calculated using SPSS version 25.0. Since some of the issues hardly or never occurred during intercoder reliability testing or showed too much overlap with other issues, some of them were combined into one variable. These were the feminine issues of poverty and welfare, and the masculine issues of business and economy, next to national defense and terrorism and crime. Except for the issues national defense, terrorism and crime (α = .66), welfare and poverty (α = .79), and foreign affairs (α = .71), all variables had α-values above .80 (see Appendix A, Table 2). The α-values for the issues of foreign affairs and welfare and poverty are considered reliable as they are above .70. Since an even distribution in masculine and feminine issues had to be maintained and an α of .66 is on the edge of being acceptable, the variable national defense, terrorism and crime was not excluded from the sample.

Analyses

As each politician posted multiple posts, the observations were not independent. Therefore, we used STATA to estimate multilevel negative binomial regression models. All models included the independent variable of interest as well as all control variables mentioned before. The numbers of likes and comments on social media are typical examples of count data (see, e.g., Trilling et al., 2017): All values are positive integers, and hence have a lower bound (zero) but no upper bound. Negative binomial regression models are designed to estimate such regression problems. We report the incident rate ratios (IRRs) with their corresponding 95% confidence intervals. IRRs are very straightforward to interpret: for each one-unit increase of an independent variable, the expected value of the dependent variable has to be multiplied by the IRR. Therefore, IRRs <1.0 represent a negative effect and IRRs >1.0 represent a positive effect. We also examined whether certain elements in posts led to more discussion amongst Instagram users. In order to do this, we looked at the proportion of comments that are placed below other comments out of the total number of comments. When a post received less than ten comments, it was taken out of the analysis, since posts with only few comments showed misleading proportions. For example, when a post only received two comments and one was a comment placed below the other comment, the percentage of discussion was 50%. After selecting only posts containing 10 comments or more, we ran linear regressions with the same control variables as mentioned above, with a log transformed variable of the comments (10+) added as a control variable.

Results

Visual communication styles

The first research question addresses how female politicians visually self-present themselves on Instagram: Which forms of self-presentations do they use? The data show a lot of similarities between genders (see Table 1; for specific differences between countries, see Appendix A, Table 3b). Politicians were visible in the majority of their posts (63.1%), with no major differences between genders. Politicians were mostly photographed with other people (77.7%) instead of alone. In these pictures, politicians were interacting with others (74.5%) and close to others (84.6%), although politicians still dominated the picture in 46.8% of all cases. They were furthermore mostly photographed from up close (65.5%) as opposed to a full body or long shot and were predominantly smiling (66.9%).

Table 1. Use of visual self-presentation divided by gender (in percentages).

	Female politicians, in %	Male politicians, in %	p-value
Politician visible in post	63.8	62.4	0.701
Others in picture	79.7	75.6	0.285
Distance from others	17.0	13.6	0.355
Interaction with others	74.0	75.1	0.800
Domination	47.4	46.2	0.782
Looking serious	20.8	25.7	0.208
Smiling	70.8	62.6	0.057
Close-up	65.2	65.8	0.890
Clothing: full suit	9.6	56.7	0.000***
Clothing: formal to casual	41.6	23.4	0.000***
Casual clothing	28.4	13.0	0.000***
Clothing: not visible	20.4	6.9	0.000***

Note: Chi-Square (χ²) Tests, *p < 0.05, **p < 0.01, ***p < 0.001.

In response to the second research question (RQ2), no significant associations were found between gender of the politician and the use of the above communication styles, with the exception of clothing: male politicians more often wore a full suit (χ² (1) = 122.004, p < .001), whereas female politicians more often wore formal to casual clothing (χ² (1) = 18.076, p < 0.001) or casual clothing (χ² (1) = 17.194, p < .001). Regarding countries, politicians from the US interacted more often with others in photos (80.7%) than Dutch politicians (67.8%) (χ² (1) = 8.289, p < .05); they also wore a suit more often than Dutch politicians (χ² (1) = 5.240, p = .000). Other aspects of self-presentation showed no significant differences between countries.

Some forms of visual self-presentation increase the level of engagement (e.g., likes or comments) for a post. To answer the third research question (RQ3a), several negative binomial regressions were performed with likes and comments as the dependent variables, with visual self-presentation styles as independent variable, and the control variables: gender, country, age, ideology, number of posts (in total and in the sampling period), number of followers and following, current president/prime minister, and presidential candidate for 2020. Several forms of visual self-presentation did not significantly increase or decrease the level of engagement. Clothing, distance (vs. closeness) to others, and whether a politician smiled or looked serious had no significant effects on engagement levels.

However, whether or not a politician was shown on a picture, had an effect on the number of likes: when a politician was visible, a post received 1.13 times as many likes as when no politician was visible (IRR = 1.133, p = 0.031, 95% CI [1.011, 1.268], see Table 2). The presence of others had a significant negative effect on the number of likes, meaning that when a politician was photographed with others (instead of alone), the number of likes decreased. The presence of others also had a negative effect on the number of comments. This indicates that posts in which politicians are photographed alone generate more engagement. The same was true for interaction: If a politician interacted with others in a post, it received significantly less likes, though the effect was not significant for the number of comments. Although smiling or looking serious did not significantly affect engagement levels, there was a considerate negative effect on the number of likes when the face of a politician was not visible. The same was true for the effect on the number of comments. These results imply that Instagram users engage more with pictures in which a politician is clearly visible, which is also reflected by the result that close-ups received more engagement than full-body or long shots of politicians. Whether or not a politician appeared dominantly on a picture did not have a significant result on the number of likes, but it did lead to significantly more comments.

Table 2. Summary of negative binomial multilevel models predicting the number of likes and comments for Instagram posts.

	Likes IRR [LCI, RCI]	Comments IRR [LCI, RCI]
Controls
Gender	4.045*** [2.035, 8.043]	6.751*** [2.851, 15.983]
Country	0.033*** [0.015, 0.071]	0.045*** [0.017, 0.118]
Age	0.961** [0.934, 0.989]	0.972 [0.938, 1.008]
Ideology	0.766 [0.485, 1.210]	0.751 [0.422, 1.335]
Running for president	1.862 [0.594, 5.837]	1.252 [0.300, 5.222]
Current president	1.163 [0.176, 7.674]	0.609 [0.058, 6.442]
Number of posts sampling period	1.003 [0.999, 1.007]	1.003 [0.998, 1.008]
Number of posts total	0.999* [0.998, 1.000]	0.999 [0.998, 1.000]
Number of followers	1.000*** [1.000, 1.000]	1.000** [1.000, 1.000]
Following	1.000 [0.999, 1.001]	1.000 [0.998, 1.001]
Visual self-presentation
Politician visible	1.133* [1.011, 1.268]	0.871 [0.721, 1.052]
Politician visible * gender	1.146*** [1.169, 1.830]	0.972 [0.659, 1.433]
Visibility of others	0.743*** [0.631, 0.875]	0.543*** [0.422, 0.697]
Interaction with others	0.784* [0.649, .946]	0.918 [0.669, 1260]
Face politician not visible	0.595*** [0.462, 0.765]	0.390*** [0.264, 0.575]
Camera perspective	0.761*** [0.654, 0.885]	0.625*** [0.492, 0.794]
Camera perspective * gender	1.771* [1.090, 2.876]	1.790* [1.061, 3.020]
Distance * gender	3.026** [1.369, 6.686]	1.586 [0.714, 3.525]
Dominance	0.906 [0.776, 1.059]	1.297* [1.039, 1.617]
Political issues
Feminine issues	0.855** [0.763, 0.957]	0.785**[0.655, 0.941]
Family * gender	1.267 [0.872, 1.842]	1.775* [1.011, 3.118]
Environment	0.717** [0.586, 0.898]	0.654** [0.476, 0.898]
Arts	0.565** [0.394, 0.811]	0.943 [0.389, 2.287]
Guns	0.514* [0.300, 0.880]	0.567 [0.267, 1.197]

Note: IRRs with 95% confidence intervals in brackets. Values <1 indicate a negative effect, values >1 indicate a positive effect. IRRs = incidence rate ratios. For clarity reasons, we combined all separate models into a single table, and we only show the significant interactions.

*p < .05. **p < .01. ***p < .001.

Next, we examined whether different modes of visual self-presentations are evaluated differently for male and female politicians (RQ3b). A positive and significant interaction effect was found for gender and whether or not a politician was visible on a picture for the number of likes (IRR = 1.146, p = .001, 95% CI [1.169, 1.830], see Table 2). Since the IRR is larger than 1, it indicates that female politicians receive more likes when they are present in a picture in comparison to male politicians. The effect was not significant for the number of comments (p = .115). Furthermore, a positive and significant interaction effect was found for gender and distance, indicating that female politicians received more likes when they appear far away from other people in a photograph, instead of appearing close to them. Lastly, a positive interaction effect was found for gender and camera perspective. This means that posts by female politicians in the sample received more likes and comments when they shared a full-body shot or long shot as opposed to a close-up. No other significant interaction effects were found between gender and visual communication styles, implying that the use of these styles seem to have roughly the same effects on engagement levels for male and female politicians.

Masculine and feminine issues

To answer the fourth research question, we examined how often political issues occurred in posts by female politicians. Some issues were mentioned more often than others (see Table 3). The most frequently used issues were children and education (16.9%), rights (13.4%), family (10.4%), and foreign affairs (10.4%). To answer our fifth research question, we found that, overall, male and female politicians referred more to feminine issues than to masculine issues: 52.6% of posts by female politicians contained one or more references to feminine issues, this was the case for 38.9% of all posts by male politicians. Male politicians referred a bit less often to masculine issues (32.2%), whereas the difference for women was far greater, with 18.1% of posts containing at least one masculine issue. Notable differences between male and female politicians were found for the use of the feminine issue of rights (female politicians: 19.1%; male politicians: 7.3%), and the masculine issue of foreign affairs (female politicians: 6.9%; male politicians: 14.1%). When examining issue-use by country, the differences are more notable than for male versus female politicians (see Appendix A, Table 5b). Politicians from the US refer much more to rights (17.4% versus 9.3% amongst Dutch politicians) and family (16.6% versus 4.0% amongst Dutch politicians), while Dutch politicians refer more to the environment (13.0% versus 4.2% amongst US politicians) and foreign affairs (13.3% versus 7.5% amongst US politicians). Significant differences were furthermore found for the feminine issues of children and education, immigration, welfare and poverty, animal rights, and arts. This implies that politicians from the US and the Netherlands refer to different feminine issues in their posts. Regarding masculine issues, other significant differences that were found are for the issues military and national defense, crime and terrorism.

Table 3. Percentages of posts containing masculine and feminine issues divided by gender.

	Total	Female politicians	Male politicians	p-value
Feminine issues
Children and education	16.9	17.6	16.2	0.610
Health	8.14	9.7	6.5	0.106
Rights	13.4	19.1	7.3	0.000***
Family	10.4	9.7	11.1	0.530
Immigration	0.79	1.3	0.3	0.117
Welfare & Poverty	3.5	3.3	3.8	0.727
Environment	8.5	8.9	8.1	0.685
Animal Rights	1.8	3.3	0.3	0.002*
Arts	2.1	2.8	1.4	0.162
Masculine issues
Taxes	1.3	0.5	2.2	0.045*
Guns	0.9	1.0	0.8	0.762
Military	3.9	2.3	5.7	0.016
Foreign Affairs	10.4	6.9	14.1	0.001**
Immigration	1.7	0.0	3.5	0.000***
Business & Economy	7.9	6.6	9.2	0.190
Science & Technology	0.8	1.3	0.3	0.117
National defense, crime and terrorism	5.5	3.8	7.3	0.036*

Note: Chi-Square (χ²) Tests, *p < 0.05, **p < 0.01, ***p < 0.001.

Surprisingly, not one political issue leads to more audience engagement (see RQ6a). However, some issues did lead to less likes or comments: references to the environment lead to less likes (IRR = 0.717, p = 0.001, 95% CI [0.586, 0.877]) and less comments (IRR = 0.654, p = 0.009, 95% CI [0.476, 0.898]). References to arts and guns only generate less likes (arts: IRR = 0.565, p = 0.002, 95% CI [0.394, 0.811] and guns: IRR = 0.514, p = 0.015, 95% CI [0.300, 0.880]). Thus, the majority of issues did not lead to a significant increase or decrease in audience engagement. Regarding gender and audience engagement (RQ6b), only one issue leads to more audience engagement: references to the issue of family lead to significantly more comments for female politicians, as opposed to male politicians (IRR = 1.775, p = 0.046, 95% CI [1.011, 3.118]). Interestingly, the use of one or more feminine issues in general in a post leads to significantly less audience engagement (for likes (IRR = 0.855, p = 0.007, 95% CI [0.763, 0.957]) as well as comments (IRR = 0.785, p = 0.009, 95% CI [0.655, 0.941])), whereas the effect is not found for the use of masculine issues. We also found a small effect for the use of masculine issues on audience engagement depending on country: Dutch politicians receive less comments compared to US politicians when sharing a post containing one or more masculine issues (IRR = 0.055, p = 0.000, 95% CI [0.021, 0.147]).

RQ6a also addresses whether the use of specific political issues leads to more discussion amongst users, and whether this depends on the gender of the politician. Linear regression showed that the issues rights, environment, taxes, and science and technology lead to significantly more discussion amongst Instagram users (see Table 4). The issue family led to significantly less discussion. The other issues had no significant effects on the amount of discussion between users. Regarding gender (RQ6b), the results showed a significant positive interaction effect for masculine issues and gender, indicating that posts by female politicians containing references to one or more masculine issues generate more discussions than posts by male politicians containing masculine issues. We also found a positive interaction effect for gender and immigration (when referring to the rights of immigrants), foreign affairs and national defense & terrorism, issues that generated more discussion when referred to by female politicians. The other issues showed no significant interaction effect, indicating that the gender of a politician did not influence the amount of discussion that these issues generated.

Table 4. Summary of the regression analyses that investigate the impact of issues and gender on discussions.

	Β	SE	t
Feminine issues	2.389	1.492	1.60
Masculine issues	1.867	1.694	1.10
Feminine issues * gender	1.066	2.994	0.26
Masculine issues * gender	9.031*	3.617	2.50
Children and education	−.461	1.924	−0.24
Children and education * gender	3.560	3.947	0.90
Health	3.984	2.869	1.39
Health * gender	−4.897	5.731	−0.85
Rights	5.397*	2.184	2.47
Rights * gender	3.662	4.435	0.83
Family	−4.520*	2.218	−2.04
Family * gender	−.619	4.463	−0.14
Immigration (feminine)	5.457	7.632	0.71
Immigration (feminine) * gender	37.998*	18.888	2.01
Welfare & poverty	3.314	3.484	0.95
Welfare & poverty * gender	−2.192	7.029	−0.31
Environment	9.729**	2.819	3.45
Environment * gender	−6.946	5.790	−1.20
Animal rights	5.869	5.885	1.00
Animal rights * gender	13.482	18.091	0.75
Arts	4.582	7.629	0.60
Arts * gender	−2.056	15.672	−0.13
Taxes	16.318**	6.026	2.71
Taxes * gender	27.988	17.883	1.57
Guns	9.265	6.403	1.45
Guns * gender	.412	12.990	0.03
Miltary	2.186	3.419	0.64
Military * gender	6.551	8.106	0.81
Foreign affairs	−.856	2.470	−0.35
Foreign affairs * gender	13.177*	5.775	2.28
Immigration (masculine)	.726	4.821	0.15
Immigration (masculine) * gender^a
Business & economy	5.079	2.713	1.87
Business & economy * gender	5.273	5.796	0.91
National defense & terrorism	2.143	2.874	0.75
National defense & terrorism * gender	13.315*	6.185	2.15
Science & technology	25.176*	11.943	2.11
Science & technology * gender	−23.818	23.875	−1.00
Constant	−5.232	6.821	−0.77
N	553	553	553
R−squared of control variables	0.242	0.242	0.242

Note: Unstandardized OLS regression coefficients. *p < .05. **p < .01. For gender, male served as the reference category in the analyses. For clarity reasons, we included all separate models into a single table – each test contained one interaction effect with control variables. ^aFemale politicians did not mentioned this issue.

Discussion

The purpose of the current study was to investigate how female politicians use Instagram and present themselves on the platform compared to male politicians, and to assess the impact of this usage on audience engagement. Doing so can help to understand whether social media can help female candidates in presenting themselves to the electorate. Our study is among the first to examine political communication on Instagram and to take the effects on audience engagement into account. It provides insights into politicians’ use of Instagram divided into two different aspects: visual self-presentation and the use of political issues. Both come with different theoretical and societal implications. First of all, the study sheds light on how female and male politicians visually present themselves on Instagram, reflecting clear patterns regardless of gender and country. Overall, politicians in the sample predominantly showed the use of positive affect cues and immediacy cues such as smiling and interaction, which express closeness and intimacy (Verser & Wicks, 2006). This implies that most politicians in the sample from both genders consciously or unconsciously present a more relatable, humanized version of themselves instead of using Instagram to show their authority or establish legitimacy. The use of these cues, however, led to a decrease in likes and comments or showed no significant effects on engagement. Instagram users seemed to prefer posts in which politicians radiate more authority: posts in which a politician was not interacting with others, photographed alone, and from further away (full-body or long shot) received more likes and/or comments. An interaction effect with gender was found for camera perspective (full-body or long shot) and distance to others. Thus, multiple forms of visual self-presentation that may help politicians to radiate more authority and establish legitimacy (see Verser & Wicks, 2006), have an even stronger effect on audience engagement for female politicians. This strengthens Yarchi and Samuel-Azran’s (2018) claim that social media might provide greater opportunities for female politicians in comparison to their male colleagues ‘to promote themselves and improve their status in the political power play’ (p. 978). Secondly, this study sheds light on which political issues are brought forward by female politicians compared to male politicians on Instagram. We found that a majority of Instagram posts contain references to one or more political issues, confirming that politicians predominantly use Instagram to share their political stance, despite the platform’s visual focus. Our findings also show that the use of political issues depends on gender: some topics are more often referred to by female politicians than by male politicians. However, contrary to earlier findings, this study found that while female politicians indeed refer to feminine issues more often than to masculine issues, male politicians do the same (cf. Huddy & Terkildsen, 1993a; Lee & Lim, 2016; Meeks & Domke, 2016). Since previous studies on gendered issue-use found that masculine issues generally receive more coverage by traditional news media (e.g., Zulli, 2019), this is an interesting result. Whether this stems from societal trends or from sample characteristics is not clear and should be further investigated. Additionally, the use of gender-congruent issues did not lead to more audience engagement, as argued in the literature (Herrnson et al., 2003; Lee & Lim, 2016). Moreover, the use of one or more feminine issues leads to significantly less likes and comments for male and female politicians, although no interaction-effect was found for gender. This seems congruent with earlier findings that masculine issues are associated more with leadership by the traditional media and are considered ‘more important to political success’ (Zulli, 2019, p. 19), and implies that these preconceptions continue to exist on Instagram.

Not one individual issue leads to a significant increase in likes or comments. Instead, a small number of issues, such as environment, lead to less audience engagement. This might imply that Instagram users are not necessarily interested in the issues politicians bring forward in their posts; since several forms of visual self-presentation do lead to more audience engagement, it might be the case that Instagram users predominantly focus on the way a politician looks on Instagram, instead of focusing on what they say. The fact that the use of not one of the issues leads to significantly more audience engagement, indicates that politicians from both genders may benefit from shifting their focus more to the visual component of Instagram. This confirms the extant literature on visuals and political communication stating that images can be more powerful than text (see, e.g., Lilleker et al., 2019; Muñoz & Towner, 2017, Russmann et al., 2019; Schill, 2012). Only one issue showed a significant interaction effect on gender and audience engagement: female politicians receive more comments compared to male politicians when referring to the feminine issue of family.

However, some issues did lead to more discussion among Instagram users, considering the amount of comments that were placed below other comments. Interestingly, the use of masculine issues overall led to more discussion when posted by female politicians compared to posts by male politicians. Such an effect was not found for feminine issues and male politicians. The results imply that Instagram can serve as a public space that can foster democratic deliberation. Given the findings that certain issues, of which some can be currently considered ‘hot’ topics (such as the environment and human rights), lead to more discussion among Instagram users, the platform seems to offer voters a space where they can discuss political issues, as has been shown to be the case for other social media platforms, such as Facebook and YouTube (Halpern & Gibbs, 2013).

A shortcoming of this study is that the nature of comments is not taken into account. We can therefore not say whether the comments were predominantly positive or negative. But the fact that the use of masculine issues only leads to more discussion for female politicians is interesting and warrants more research. Moreover, the fact that references to feminine issues lead to less likes and comments for politicians from both genders implicates that feminine issues may still be undervalued in politics. At the same time, masculine issues are still predominantly owned by male politicians on Instagram – only 18.1% of posts by female politicians contained a reference to one or more masculine issues. Male politicians, however, refer to both masculine and feminine issues, and might feel more free to do so. Positive interaction effects were, furthermore, found for gender and the issues of immigration, national defense and terrorism, and foreign affairs: they led to more discussion amongst Instagram users when referred to by female politicians. No positive effects were found for male politicians in comparison to female politicians. Whether more discussion below a post is beneficial for politicians is an important line for future research.

The results at large imply that politicians from different genders and different countries adopt similar social media strategies. Male and female politicians and politicians from the US and the Netherlands were very close in many aspects, such as their use of visual self-presentation styles. They are, however, not always evaluated equally by the audience. Perhaps the most pressing question following from the results is the extent to which female politicians can actually benefit more from Instagram than male politicians, especially since they are not treated equally by the traditional media. Is the optimism of Yarchi and Samuel-Azran’s (2018) justified, when they state that ‘social media not only creates an equal ground for female politicians, but also enables them to generate more engagement, compared to male politicians’ (p. 987)? In some respects, this seems to be the case: our study shows that female politicians can generate more discussion amongst users by mentioning certain issues and that they receive more likes when using certain forms of visual self-presentations, especially those focusing on authority. This raises the question of what it means that female politicians receive more audience engagement when they show more authority or represent public distance. Since these visual cues have been related to masculinity (see, e.g., Carpinella & Bauer, 2021; Chen et al., 2020), this implies that gender stereotypes may be maintained on Instagram. This would mean that, although Instagram offers female politicians a way to bypass the traditional media, they may not be able to bypass gender stereotypes. In other words online opportunities for female politicians may backfire due to gender stereotypes or due to more incivility on social media. Furthermore, our results point out that female politicians, compared to male politicians, received significantly more likes when present in a picture, which implies that Instagram users are more focused on the appearances of female politicians.

At the same time, Instagram offers female politicians a stage to write their own stories. Although scholars have argued that female politicians need to fight damaging stereotypes to not be perceived as weak or incompetent (Flicker, 2013; McGregor, 2018; Van Zoonen, 2006), the current study does not confirm that women actively do so on Instagram, even though this may lead to less engagement. Female politicians may feel less urgency to prove their competence just because they are female. At the same time, male politicians seem to shift more towards female politicians, partly by referring more to feminine issues. It should be noted that the effects of female politicians’ presence on Instagram in general can be of importance for society at large: prominent female politicians can ‘serve as role models for other women interested in political careers, causing an increase in female candidates’ (Ladam et al., 2018, p. 369). In this regard, young Democrat congresswomen like Alexandria Ocasio-Cortez are pioneering on Instagram by livestreaming their personal as well as their political life, bringing ‘followers into the political process and [convincing] them that they, too, could someday be in her shoes’ (Kozlowska, 2018; Minsberg, 2018). At a time where men still dominate the political system, this can be of vital importance for future generations.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Notes on contributors

Charlotte Brands

Charlotte Brands is a media analyst at the Dutch Ministry of Finance. She graduated cum laude from the University of Amsterdam in 2020 and holds a master's degree in Political Communication. She is interested in the way politicians interact with the electorate through (social) media. email: [email protected].

Sanne Kruikemeier

Sanne Kruikemeier is Associate Professor Political Communication and Journalism at the Department for Communication Science, University of Amsterdam. Her research focuses on the consequences and implications of online communication for individuals and society. e-mail: [email protected].

Damian Trilling

Damian Trilling is Associate Professor for Communication in the Digital Society at the Department for Communication Science, University of Amsterdam. He is especially interested in how the changing media environment changes how people engage with news and current affairs. e-mail: [email protected]

Notes

1 The European Parliament elections in the case of the Netherlands, which were held between 23 May and 26 May 2019, and the House of Representatives elections in the US, which were held on 6 November 2018.