Political institutions and the gendered use of social media among political candidates: evidence from Tunisia

ABSTRACT Despite a recent boom in information and communication technologies (ICTs) in many developing countries, there is limited research on how social media is being used in political campaigns in emerging democracies. In these contexts, the introduction of electoral gender quotas often forms part of the constitutional reform. This paper investigates the gendered use of social media among political candidates during the 2019 parliamentary election campaign in Tunisia, which recently adopted an ambitious gender quota system. We have mapped the use of Facebook among a random sample of 400 men and women candidates and carried out interviews with 19 men and women candidates who ran in these elections. The results demonstrate how patriarchal structures interact with political-institutional features – the electoral gender quota system, the electoral system (a closed-list proportional-representation system), and the fragmented party system – in shaping the gendered use of social media among political candidates in Tunisia. We show that men are more likely to use Facebook for campaign purposes since they are more often placed at the top of the candidate lists. Hence, rather than having an equalizing effect, the study suggests that social media become a new tool for already privileged candidates in Tunisia.


Introduction
Social media platforms are becoming increasingly important campaign tools for politicians worldwide.As a consequence, research on the use of social media by political candidates has become a fastgrowing research field (see e.g., Dolezal, 2015;Filimonov, Russmann, & Svensson, 2016;Larsson & Kalsnes, 2014;Suiter, 2015).Within this literature, it has been debated how social media shape power dynamics among political candidates.Some scholars suggest that social media platforms, due to their low costs, accessibility, and wide reach, has the potential to equalize the playing field among political candidates, by presenting a new way for marginalized political actors with limited resources to bypass gatekeepers in traditional media and reach the electorate directly (see Gibson & Mcallister, 2015;Vergeer & Hermans, 2013).However, others have argued that the effective use of social media platforms demands considerable know-how and access to resources and that social media therefore risk to mainly strengthen already established and resource-strong actors (see e.g., Dolezal, 2015;Jacobs & Spierings, 2016;Quinlan, Gummer, Roßmann, & Wolf, 2018).
This article draws on, and contributes to, this body of literature, by zooming in on the gender dynamics of social media usage among political candidates.Throughout history, women have been a marginalized group in politics.Despite a large influx of women parliamentarians during the past two decades, only one of the four parliamentarians across the globe are women.However, while gender has been demonstrated as a relevant factor in relation to social media uptake among political candidates, there are few studies that explicitly focus on whether social media level out or reinforce inequalities between men and women candidates (see, however, Sandberg & Öhberg, 2017;Yarchi & Samuel-Azran, 2018).Among the analyses that have included gender as a factor, most have studied old democracies in Europe or North America, where women have been present in politics and parliament for a long time (see e.g., Dolezal, 2015;Evans, Cordova, & Sipole, 2014;Jackson & Lilleker, 2011;Sandberg & Öhberg, 2017;Vergeer & Hermans, 2013).
Therefore, this article redirects the focus to emerging democracies. 1 One central feature of most democratization processes since the early 1990s, has been the introduction of electoral gender quotas.This means that women's presence in politics often has increased rapidly after democratic transitions (see e.g.Dahlerup, 2006).Yet, despite these advances, women politicians face various constraints in countries undergoing democratic transitions.For instance, as relative newcomers in politics, they have more limited access to patronage and other financial resources, which among other things makes it more difficult for them to run their election campaigns (Clayton & Zetterberg, 2021;Murray, Muriaas, & Wang, 2021).In this article, we address the question as to whether social media can function as an "equalizer" between men and women candidates in election campaigns in emerging democracies where electoral gender quotas have been adopted as part of a constitutional reform -or if social media rather contribute to the reproduction of gender inequalities in politics.Despite a recent boom in information and communication technologies (ICTs) in many developing countries, we know little about political candidates' use of social media in these contexts -and even less so when we apply a gender lens on these issues.
To address the question, we analyze gender differences in the use of social media among political candidates during the campaign for the Tunisian general elections in 2019.Tunisia is a suitable case for addressing the question, because the country had at the time of the 2019 elections some typical features of an emerging democracy.For instance, the recently developed party system was still very fragmented and the political parties varied greatly in their organizational strength (Belschner, 2021;Hamid, 2014).Similarly to most other democratic transitions, an electoral gender quota system was installed at the national level in 2014.The Tunsisian electoral gender quota systema parity law (see Piscopo 2016) -is ambitious in the sense that it stipulates that all parties, coalitions, and independents must alternate between men and women candidates on their electoral lists for the lists to be accepted by the electoral authority (i.e.every second candidate must be a woman candidate) (IPU, 2021).However, in national elections, this requirement does not apply to the first position across their lists. 2The results of our study of Tunisia can potentially contribute to our understanding of other contexts at similar levels of development where electoral gender quotas have been adopted.In addition, there is a pragmatic reason for examining Tunisia: its high Internet and social media penetration.In contrast to many other countries at similar levels of development, a relatively large proportion of the Tunisian population (67%) had access to the Internet in 2019, and most of the Internet users in Tunisia were also active Facebook users at the time of the study (63%) (Datareportal, 2019).As in many other developing contexts, Facebook is by far the most used social media platform in the country, and it was also the dominating campaign platform among the candidates during the 2019 elections (Elswah & Howard, 2020;IRI/NDI, 2019).The combination of widespread Facebook usage and a fairly large pool of women political candidates provide a possibility to investigate the gender dynamics in social media use during the 2019 election campaign.
To conduct the analysis, we have firstly manually mapped the use of Facebook among a random sample of 400 men and women candidates from all the parties elected (those having 2 or more seats) to the Parliament in the 2019 elections.This enabled us to examine to what extent social media were used among men and women candidates and thus to identify any gender differences within the sample of candidates.Secondly, we have carried out semistructured interviews with 19 men and women candidates for the 2019 parliamentary elections.The interview material enabled an analysis of the motivations behind the use (or nonuse) of social media during their election campaigns -and thereby an increased understanding of possible gendered patterns demonstrated by the statistical analysis.
The analysis shows that men candidates more frequently use social media for campaign purposes than women candidates running for the same political party in the same district.To understand this finding, we suggest that patriarchal structures interact with political-institutional features -the electoral gender quota system, the electoral system (a closed-list proportional-representation system), and the fragmented party system -in shaping the gendered use of social media among political candidates in Tunisia.Men are more often placed as top candidate of the lists, and important financial and human resources for using social media are often reserved for these candidates.Moreover, the interviews show how (women) candidates placed lower on the lists perceive it as useless to campaign on social media, since they are unlikely to be elected.Rather than having an equalizing effect, the study hence suggests that social media in Tunisia become a new tool for already privileged candidates.

Gendered dynamics in social media use among political candidates
The theoretically relevant question we address in this article is whether social media platforms contribute to making the political playing field more gender equal, or whether they reproduce (or even reinforce) gender inequalities.There is a considerable literature that has analyzed the gender bias in traditional media's covering of political elections.Studies consistently show that women candidates are less visible and portrayed in a different manner than their male counterparts (Hooghe, Jacobs, & Claes, 2015;Kahn, 1996).Drawing on equalization theory, some scholars therefore suggest that social media offer a possibility for women candidates to overcome the gender biases produced by traditional media (Jacobs & Spierings, 2016;Sandberg & Öhberg, 2017;Yarchi & Samuel-Azran, 2018).The theory suggests that social media platforms, because of their wide reach, low cost, accessibility, and simplicity, are particularly attractive to politicians from historically marginalized groups (such as women) that have fewer resources.Thereby, they present a way for candidates to reach out to the electorate, transmit political messages, and mobilize voters without having to deal with gatekeepers within traditional media (see Dolezal, 2015;Gibson & Mcallister, 2015;Sandberg & Öhberg, 2017).According to this argument, we would expect more women candidates than men to use social media as a political tool.
However, others have argued that social media rather provide already privileged and resourceful political actors with new benefits.The normalization theory suggests that those with more resources are better equipped to effectively use social media platforms as political tools, which reproduce or reinforce existing power structures.For instance, incumbent candidates make great investments to maintain their positions.Such candidates are therefore more likely to adopt social media and to devote great resources in order to maintain their power and status (see Jackson & Lilleker, 2011;Quinlan et al., 2018;Samuel-Azran & Yarchi, 2020;Sandberg & Öhberg, 2017). 3According to this argument, we would expect men candidates to use social media more than women do.Moreover, it has been argued that the widespread diffusion and professionalization of social media use over time will make it less of an alternative channel for marginalized political actors, including women, which increases the expectation that normalization will take place (Gibson & Mcallister, 2015;Jacobs & Spierings, 2016).Given these developments, more fine-grained analyses with regard to the equalization/normalization theory has been called for, which also include different electoral contexts (Jacobs & Spierings, 2016;Sandberg & Öhberg, 2017).
There are, to the best of our knowledge, only a few empirical studies that have explicitly focused on the gendered dynamics of social media use among political candidates (see Sandberg & Öhberg, 2017;Yarchi & Samuel-Azran, 2018 for notable exceptions).In most studies, mainly on the US or on European countries, gender has been included as only a control variable.Based on existing research, there is no unified answer as to whether gender affects the use of social media by political candidates.Some studies of Western contexts suggest that women are more likely to use social media (and especially Twitter) for political campaigns (Evans et al., 2014;Jackson & Lilleker, 2011).Others have found that men candidates are more likely to use social media for campaign purposes (Dolezal, 2015) or that there are no great gender differences when it comes to candidates' adoption of social media tools during election campaigns (Sandberg & Öhberg, 2017;Vergeer & Hermans, 2013).
Hitherto, many studies on social media use among political candidates have focused exclusively on Twitter (see e.g., Jungherr, 2016;Sandberg & Öhberg, 2017;Vergeer, 2015;Vergeer & Hermans, 2013).However, when moving beyond a Western focus that has dominated social media analyses of political candidates (see, however, Ndlela & Mano, 2020;Prihatini, 2020), there are good reasons to switch social media platform.In developing countries, Twitter is not as widely used as in Western democracies.For instance, in Tunisia only 6% of Internet users use Twitter, while 85% of Tunisian Internet users use Facebook (Statista, 2021).Therefore, there is reason to pay attention to other platforms such as Facebook, which has become synonymous with the Internet in many developing countries (Nothias, 2020).Do men and women candidates in such a context differ in the extent to which they use Facebook as a political campaign tool?
Emerging democracies also differ from Western democracies in other respects that may impact the gender dynamics in political candidates' use of social media in perhaps contrasting ways.First, patriarchal structures in politics are often strong.This means that women are more confined to the private sphere than men, which can make political campaigning difficult.In these types of contexts, social media platforms can present new possibilities for women and other marginalized groups to participate politically, which not least could be seen during the Arab uprisings (Howard & Hussain, 2011).Relatedly, research suggests that in repressive regimes in the Middle East, Internet use can open up new possibilities for politically marginalized groups in general to gain political knowledge and to participate politically (Wagner & Gainous, 2013).It has also been demonstrated that among Arab women in particular, the mere access to the Internet is positively correlated with viewing women as capable as men when it comes to political leadership (Shamaileh, 2016).Second, in contrast to Western democracies, emerging democracies have more often seen a "fast track" to women's descriptive representation through the adoption of electoral gender quotas (Dahlerup & Freidenvall, 2005).This means that most women can be described as relatively newcomers to politics.Therefore, we can expect that they have fewer resources to invest in political campaigning, which may affect their use of social media platforms.Finally, emerging democracies are typically characterized by fragmented party systems and by parties with weak organizational structures.These features can make the lack of resources an even greater problem for less privileged candidates (Belschner, 2021;Hamid, 2014).In sum, we believe a priori that gender differences in social media use in the contexts of democratic transition could go in either direction.

The Tunisian case
While Tunisia presently seems to be moving in a more autocratic direction, at the time of the 2019 elections, Tunisia was considered as one of the most democratic countries in the Middle East and North Africa (MENA) region (Freedom House, 2020).Yet, Tunisia's transition to democracy is very recent.Following the ousting of President Ben Ali in 2010-2011, elections were held to a Constituent Assembly in 2011, which devised a new electoral code for national elections in 2014.Thereafter, two elections to the national assembly have been held, in 2014 and 2019, respectively.Tunisia has a proportional representation (PR) system with closed electoral lists.There are 33 electoral districts, of which 27 are in Tunisia (electing 199 members) and six abroad (electing 18 members).To run as a candidate you must belong to an electoral list, and voters can only cast their vote for one candidate list in their constituency.In the 2019 general elections, nearly 15,000 candidates from over 1,500 party, coalition, and independent lists competed for the 217 parliamentary seats (IRI/NDI, 2019).
Since the 2011 revolution, Tunisia's party system can be described as typical for an emerging democracy.Partly, it is characterized by a large number of small parties; in the 2019 elections the parliamentarians elected to parliament represented 30 different electoral lists and most parties have only 1-4 seats.Thus, Tunisia has a fragmented party system.The resulting low party magnitude is accompanied by a moderate district magnitude: the number of seats for each district ranges between four and ten (Carey & Hix, 2011).Moreover, the Tunisian party system is also very volatile.New parties constantly emerge and parties that are seemingly established and resource strong can quickly fall apart.This, for example, happened with the former governing party, Nidâa Tounés.Before the parliamentary elections in 2019, Nidâa Tounés split into several parties due to internal contention, and the party only received 1.5% of the vote in the elections (compared to 37.6% in the 2014 elections) (Martin & Carey, 2022).Except for the Islamist party Ennahdha, which has a large membership and great organizational strength, most Tunisian parties are hence newer parties with less access to resources and weaker organizational structures (e.g.Front Populaire, Afek Tounés, and Courant Démocrate) (Belschner, 2021;Hamid, 2014).
In terms of gender equality, women's rights were often used instrumentally during the dictatorship to gain international recognition and to consolidate power.This meant that while Tunisia often was perceived internationally as a "women-friendly backwater surrounded by patriarchal states," the Tunisian women's movement was coopted by the regimes and politically dissent women were harshly suppressed internally (Marks, 2013, p. 233).During the drafting of the new constitution, however, Tunisia's women's movement as well as international gender advisers gained traction and the design of an electoral quota system became an important part.Due to the electoral gender quota, women's participation in Tunisian politics has increased rapidly.In the 2014 general elections, when the quota was applied for the first time in, 31% of the elected representatives were women (IPU, 2021).However, while electoral lists must alternate between men and women candidates (i.e.'vertical parity'), alternating the sex of the top candidate (i.e.'horizontal parity) is not required (see Piscopo, 2016). 4This means that political parties, coalitions, and independents can continue to put men candidates as number one on as many lists as they like.Hence, in the 2019 parliamentary elections only 14% of the lists were headed by women (i.e.only 219 out of 1,506 registered lists) (IRI/NDI, 2019, p. 6).Moreover, many gender-related problems remain in Tunisia, which can have consequences for women's political participation and representation.There are still major equality problems in particular within the "private sphere," which remain highly unequal regarding inheritance, the relationship between spouses, and child custody.Women are also increasingly discriminated against in the labor market, and harassment of women remains a major problem (Marks, 2013).
With respect to the use of social media in the 2019 elections, Facebook was extensively used by individual candidates, parties, and coalitions to communicate with potential voters during the campaign period (Elswah & Howard, 2020; IRI/ NDI, 2019).While social media played an important role for the candidates themselves, it was also reported that numerous Facebook pages without clear party or list affiliations emerged that targeted candidates or lists, and that these pages frequently contained hate speech and disinformation (IRI/ NDI, 2019).

Data and methods
We employed a mixed-method approach to conduct the empirical analysis.First, we ran regression analyses on a random sample of 400 men and women candidates (for detailed information about how this dataset was compiled, see the Supplementary Information file) to examine whether there are gender differences in the extent to which the candidates used Facebook during the 2019 electoral campaign.Second, to analyze the (gendered) motives behind the candidates' use (as well as nonuse) of social media during the campaign we conducted semi-structured interviews with 19 candidates (both men and women, and non-elected as well as elected candidates).
To conduct the statistical analysis on the random sample of 400 candidates, we measure the dependent variable by distinguishing between candidates who have a Facebook public figure page (coded 1) from those who have no such profile (coded 0), in an attempt to understand who uses Facebook in their public role as politicians.These Facebook pages are specifically designed for the purposes of public figures, and there is a particular type of page for political candidates.Hence, public figure pages have many features that can help political candidates to promote themselves vis-à-vis potential voters and to build a network of supporters.For instance, they have an unlimited number of fans (personal profiles in contrast does only allow a maximum of 5,000 friends).In addition, they have the possibility to verify the profile, filter derogatory comments, customize messages, and rank comments and commenters (Facebook, 2021).Given these features specifically tailored for political work, having a public figure page has been argued to be a good indicator of professionalization in relation to politicians ' Facebook use.Therefore, we build on previous mappings of politicians' social media use, who only have included public figure pages and excluded private profiles (see Larsson & Kalsnes, 2014). 4Previous research has moreover shown that public figure pages are the most used type of profile by political candidates during election campaigns (Enli & Skogerbø, 2013;Lin, 2017).It has also been shown that having a public figure page can increase the vote share received by the candidate in the election (Lin, 2017).
Our main independent variable is gender, for which we code men as 1 and women as 0. We include fixed effects for party and district in order to control for the possibility that there are variations in the use of Facebook across political parties and districts (e.g., Gibson & Mcallister, 2015).Therefore, by using these fixed effects, we in reality compare men and women on the same candidate lists.In line with previous work (e.g., Dolezal, 2015;Sandberg & Öhberg, 2017), we also control for incumbency (coded 1; non-incumbents are coded 0).It is possible that those who are members of parliament communicate with constituents through social media already before the start of the election campaign.Thus, any intra-party gender differences in the usage of Facebook could be a result of the fact that a majority of the incumbents are men.Ideally, we would also have liked to control for age, as research on Western democracies has shown that young candidates are more likely to use social media than older candidates (e.g., Dolezal, 2015;Larsson & Kalsnes, 2014;Sandberg & Öhberg, 2017).However, unfortunately, the candidate lists in Tunisia do not include such information.
As additional control variables, we include candidates' positions on the list (coded as their number on the list) as well as whether they were elected (coded 1 for elected and 0 for not elected).The reason for including these variables is that the usage of Facebook as a campaign tool could be a strategy that only those who stand a fair chance to be elected use.As Tunisia has a closed-list proportional system, and thus the rank order of candidates cannot change once the lists have been presented, candidates themselves and their parties know fairly well which candidates have a chance to be elected.This fact may increase or decrease candidates' propensity to use Facebook in their campaign.In addition, while the electoral gender quota system stipulates that parties need to alternate between men and women on their lists, they are, to reiterate, not obliged to alternate between men and women among the head of the candidate lists in elections at the national level.Leaving this decision to the parties themselves has hence implied that most lists still have men at the very top (i.e. as number one of the lists).As the vast majority of the lists only get one candidate elected, it is likely that mainly men candidates at the top of the lists believe that they have a good chance to be elected.
Although the dependent variables are coded as binary (Facebook profile or no Facebook profile), for which logistic regression is more appropriate, we use ordinary least squares (OLS) regression to conduct the statistical analysis.We do this for ease of interpretation, as the OLS regressions show the difference in the propensity to have a Facebook public figure page (see Hellevik, 2009).However, the robustness checks conducted with logistic regression (presented in Supplementary Information file, Table A2) showed no significantly different findings from those presented here.
In our 19 interviews with candidates, we aimed for a variation regarding political parties, districts, and list positions (see Supplementary Information file, Table A1, for detailed descriptive statistics on the interviewees).In line with the approved application to the Swedish Ethical Review Authority (Dnr 2019-06159), detailed information about the study's potential risks and consequences as well as a consent form was provided to the interviewees before each interview.The interviewees are all anonymized in the presentation of the results.The interviews were carried out online between November 2020 to June 2021, mostly via messaging services like WhatsApp and Facebook Messenger.We asked the candidates questions about their rationales for using, or not using, social media during the electoral campaign, why they had chosen a certain type of Facebook profile as their main online campaign platform, and if and how using social media as a campaign tool was related to their list positions.Most of the interviews were conducted in Arabic (one was conducted in English), and they were then translated to the authors' mother tongue upon transcription.All transcripts were coded in NVivo.

Gendered patterns in Facebook use during the 2019 election campaign
Before presenting the results of the regression analyses, let us first see a few initial pieces of evidence by looking at the descriptive statistics.Table 1 shows that men candidates are significantly more likely than women candidates to have a Facebook public figure page: almost 23% of the Tunisian men candidates have such a profile, whereas about 12% of the women candidates have one.These descriptive statistics suggest that men more often than women use this social media platform for campaign-related purposes.To understand these differences, one potential factor could be our choice of social media platform.It is possible that we see these gender differences because we look specifically at Facebook: Tunisian women may prefer other platforms (such as Instagram).Although we cannot test this proposition directly, we include private Facebook profiles in the descriptive statistics to examine whether there are gender differences also for this profile.As shown in Table 1, there are no such differences: approximately 17-18% of both the men and women candidates have a private Facebook profile.Thus, the differences lie only in the type of profile that is used primarily as a political tool, such as during an electoral campaign.
With respect to the control variables, incumbency has no gender differences: approximately 5% of both men and women candidates in our sample are incumbents.However, we do identify large and significant differences regarding where on the candidate lists women and men candidates are placed as well as regarding men's and women's likelihood to be elected.Men are placed higher on the candidate lists (on average slightly lower than as number two, whereas women are on average placed almost as number three) and they are (consequently) more likely to be elected (almost 20% of the men were elected, whereas among women only 8% were elected).
Moving to the regression analysis, a first analysis addresses the question as to whether the observed gender differences in the use of a Facebook public figure page (shown in Table 1) have to do with the political parties men and women candidates represent and/or with the electoral districts in which they run.There is a possibility that these pages are more commonly used in some political parties than in others and that men candidates in our sample are more likely to represent these parties.If that is the case, we would observe no differences when we control for political party.Similarly, Facebook public figure pages may be more common in some parts of the country, such as in urban electoral districts.If men candidates in our sample are concentrated in these districts, we would observe no gender differences when we control for electoral districts.
To statistically address these possibilities, we use fixed effects for political party and electoral district.To reiterate, this means that we compare men and women on the same candidate lists.Model 1 in Table 2 shows that the observed relationship in Table 1 persists: approximately 10 percentage points more men candidates have a Facebook public figure page than women candidates who are placed on the same candidate lists (significant at p < .01). 5  In a second test, presented in Model 2 in Table 2, we examine whether gender differences in political experiences -captured by incumbency -account for the intra-party variation.Yet, as expected based on the descriptive statistics (see Table 1), that is not the case: the gender differences in having a Facebook public figure page remain -both in size (approximately 10 percentage points) and significance (p < .01)-when we control for those who were already serving as MPs by the time of the 2019 elections.
As a third and final test, we analyze whether it is the prospects for being elected that reveal a gender bias in the use of a Facebook public figure page.Men candidates may simply have greater chances for being elected and therefore use Facebook for campaign-related purposes.We address this question by using two items: in addition to controlling for whether the candidate was elected, we also look at the candidate's position on the candidate list.As Tunisia has implemented a parity law in its closedlist PR system with low party magnitude, it is possible that political parties -because of their patriarchal structures -have mainly chosen to place a man candidate as head of the list and that it is mainly this candidate who stand a fair chance of being elected.If that is the case, we should observe no gender differences in the use of a Facebook public figure page when we control for list position as well as for whether the candidate was elected or not.We see a first piece of evidence in Table A4 in the Supplementary Information file, which shows that men candidates in our sample were much more likely to be top candidates.If we focus only on the number of observations (n) in the table (thus, leaving information about Facebook public figure pages aside), we see that more than half of the men (100 of 198 men candidates) in our sample were listed as top candidates (i.e.List position 1), whereas only 10% of the women candidates (21 of 201) were list heads.
Moving then to the regression analysis, Model 3 in Table 2 indeed shows that there is no longer any differences in the extent to which men and women candidates have a Facebook public figure page when we control for list position and whether the candidate was elected.The coefficient is close to zero and not statistically significant.On the other hand, all the control variables are statistically significant in the expected direction.For instance, the negative sign of list position suggests that for every additional place down the candidate list we go (from position 1 to 2, etc.), the fewer candidates are expected to have a Facebook public figure page.
These findings are important, because they show that the gender differences are not a result of women's limited interest in using a social media platform for campaign purposes.In fact, when we look at Table A4 in the Supplementary Information file again, but concentrate our attention this time to the proportion with Facebook public figure pages by gender and list position, we observe that women who are placed as the top candidate are even more likely to have such a page than men candidates: 62% of the women list heads had a public profile during the 2019 elections whereas 37% of the men top candidates had one.Rather, it is the political parties -and the party gatekeepers -who contribute to creating a gender bias by prioritizing men candidates when implementing the parity law.In other words, the observed gender differences in Table 1 were mainly a reflection of where on the candidate lists parties placed men and women.Men candidates were more likely to be placed by their parties on the list positions that maximized their chances to be elected, and it was this type of candidates that were most likely to use a Facebook public figure pages as part of a campaign strategy.In that sense, rather than having an equalizing effect, social media in the 2019 parliamentary elections were mainly used as a political tool by those whose prospects for electoral success were already high -and they were mostly men.

Qualitative analysis: list position and gender differences in resources and motivation
Our interviews provide additional nuance to the findings of the statistical analysis.They suggest that we can understand the (gendered) patterns in two main ways.First, using a Facebook public figure page during the electoral campaign is perceived as strongly related to financial and human resources that are primarily available to the (men) candidates at the top of the lists.Second, it was perceived as useless for (women) candidates placed lower on the lists to campaign on social media, as these candidates did not have a chance to win a seat in the elections.

List position and access to (party) resources
One important factor behind the use of social media during the campaign concerned the extent to which the candidates had access to resources, such as assistants or social media advisers.
Interviewees commonly stated that parties invested more resources in the candidates that were at the top of the lists, as these candidates were seen as the main representatives of their parties: "If I was to compare, the head of the list receives more resources locally [from the party].(. ..)The head usually has an assistant, and get more attention and support." 6This meant that candidates placed lower on the list did not have the same amount of resources to invest in their use of social media during the campaign, which could make them refrain from using a more resource-demanding social media strategy.In particular, the (lack) of access to crucial resources was by the interviewees related to which type of Facebook profile they had used during the campaign.
In line with the observations made by civil society groups and organizations (see Elswah & Howard, 2020; IRI/NDI, 2019), all interviewees claimed that Facebook was the most important social media platform during the 2019 general elections.However, while Facebook was perceived as the most widespread and thereby accessible platform to both voters and politicians, previous research on political campaigning has characterized Facebook as a particularly resource-demanding platform (see e.g., Jacobs & Spierings, 2016;Quinlan et al., 2018).Resources were also brought up as one of the main reasons for which type of Facebook profile the candidates had used during the campaign. 7Among the interviewed candidates, the public figure page was commonly perceived as the most professional one and therefore most suitable for campaign purposes, in particular due to its wide reach.However, running a public profile was also perceived as more resource demanding than running a private one.

As one woman candidate explained:
To create a public page you need an assistant that takes care of your profile.(. ..) Nw I have been a parliamentarian for one and a half year, so maybe I will soon be able to become more professional and create a [public] profile for the parliamentary activities.8 Thus, it was perceived as necessary to have a certain amount of financial and human resources (such as at least one assistant that could help you with running the profile) to use a public figure page during the campaign.The relationship between access to resources and the use of public profiles clearly had gendered consequences in Tunisia.In the 2019 parliamentary elections, 86% of the candidate lists were headed by men candidates (IRI/NDI, 2019).This means that men candidates -as heads of the list -more often had access to the resources that are perceived as necessary for running a public figure page.As a woman candidate from Ennahdah expressed it: Yes there is a difference [between men and women candidates].We have men who are leaders within the party and who have groups of young volunteers working for them and taking care of their [Facebook-] profiles, videos, pictures and so on.(. ..) I is a patriarchal mentality.9 Thus, the relation between list position, gender and access to crucial resources was one important factor that helped explain how social media became a tool for already privileged (men) candidates in the Tunisian context.However, list position was also perceived as strongly related to another important factor behind the gendered patterns of social media use, namely the motivation to campaign on social media or not.

List position and (lack of) motivation to campaign on social media
While most candidates on viable list positions claimed that their use of Facebook played an important role in their campaign, it was generally perceived as a less important campaign tool for candidates placed lower on the lists, as these did not have a chance to win a seat in their districts. 10Since the Tunisian party system is so fragmented and characterized by a low party magnitude, only very few lists elect more than one candidate.In the 2019 elections, only the two largest parties, Ennahda and Qalb Tounes, managed to elect more than one candidate on their lists, and this occurred only in a few districts.Therefore, it is in general only the first candidate on each list who perceives that they have a chance to be elected and thus that it is worth running a campaign.As one man candidate put it: In Tunisia, when it comes to the lists, it is always the head of the list that has a greater chance [to be elected].In best-case scenario, also the one placed second on the list can enter [the Parliament], if they belong to a large party, such as Ennahda or Qalb Tounes.Therefore you often see that they [those placed under the head of the list] are not as active [on social media] as the list head.It is because they do not have any hope that they will be elected. 11  It was also commonly stated that candidates who were placed lower on the list primarily campaigned for the top candidates on their own Facebook profile, if they had one. 12To understand the lack of motivation to campaign for oneself on social media that followed from one's list position, we need to consider institutional factors.Evidently, Tunisia's PR system with closed lists decreases the motivation for those who perceive that they have a limited chance to be elected.In addition, the fragmented party system with low party magnitude accentuate these tendencies, as also those who are placed among the top candidates (as number two or three) may perceive that their chances to be elected are slim.
While (lack of) access to resources and motivation were dominating as explanations for the choice of Facebook profile among the candidates, an additional argument for refraining from using a public figure page brought up by a woman candidate was that personal profiles were less exposed to negative reactions, including harassment and defamation. 13 These accounts may provide an increased understanding of the gendered motivations behind the use of specific type of social media profile that future studies could further investigate.

Conclusion
This article has analyzed the gendered dynamics in Facebook use among political candidates during the 2019 Tunisian parliamentary election campaign.We limited the analysis to a single social media platform because Facebook is largely dominating the Tunisian social media landscape, although other platforms such as Instagram are increasingly used (16% of Tunisians were active Instagram users in 2019 (Datareportal, 2019)).Moreover, other platforms were scarcely used during the election campaign (Elswah & Howard, 2020;IRI/NDI, 2019;Statista, 2021).
The statistical analysis showed some interesting links between candidate gender and Facebook use.Specifically, we observed in our bivariate analysis that men candidates were significantly more likely to have a Facebook public figure page than women candidates were.However, this direct relationship disappeared when we controlled for list position and whether the candidates were elected.This means that it was the prospects for being elected that revealed a gender bias in the use of a Facebook public figure page: men were more likely to have such a page because party gatekeepers had prioritized to put men as top candidates when implementing the parity law -and it was mainly these candidates who had a Facebook public figure page.Our interviews revealed that Facebook public figure pages are viewed as demanding more resources, which were perceived as primarily accessible to the very top candidates.Therefore, the motivation for lower-ranked candidates to use a public figure page was relatively low.
Our results are in line with previous research that has demonstrated the importance of access to resources as well as incumbency and list position to explain individual candidates' use of social media in Western democracies (e.g., Dolezal, 2015;Quinlan et al., 2018;Sandberg & Öhberg, 2017).However, studies of Western contexts have rarely linked the importance of access to resources and list position to the gendered use of social media in election campaigns.By contrast, this link is clearly visible in the Tunisian context where district magnitude is moderate, party magnitude is low, party resources are scarcer, and patriarchal structures are still strong.This means that very few candidates per party stand a chance to be elected to parliament, and the large majority of these are men.Thus, women candidates were less likely to have access to crucial resources to run a public profile and/or did not perceive it as useful to campaign in social media due to their small chances of being elected.In this way, our findings support the normalization theory: Facebook public figure pages became a tool of already privileged candidates in the Tunisian context.If we want to level out gender inequalities in politics, and we believe that using social media is an increasingly important campaign tool, these findings are problematic.
The analysis contributes to the literature on the (gendered) use of social media by political candidates in at least two ways: First, we show that there are clear gender differences in social media use among candidates in an emerging democracy, which have seen a recent influx of women in politics due to the adoption of electoral gender quotas.Second, we show how these differences are a result of party gatekeepers' preference for men candidates as list heads as well as of institutional factors such as a fragmented party system and low party magnitude, which result in that only the very top candidates stand a chance to be elected.
Building on our research, we suggest that future research should continue to explore gender differences in social media use in contexts where electoral gender quotas have been recently adopted but patriarchal norms are entrenched.Thereby, we are able to better know how far the findings from Tunisia travel.In particular, our findings highlight the importance of conducting comparative work that looks specifically at certain institutional configurations -or at the interaction of institutional rules.For instance, given the demonstrated importance of how formal political institutions (e.g.electoral system) interact with the gender quota system, an important future extension of our analysis would be to address gendered dynamics in candidates' social media use in other non-Western democracies with different types of election systems and/or quota systems.One avenue for future research would hence be to address how these dynamics play out in similar contexts regarding the levels of democracy and development, but with more candidate-centered electoral systems, in which political parties exercise less of a gatekeeping role in election campaigns.To give an example, political parties have fewer incentives to allocate party resources to a few top-listed (men) candidates in an open-list PR system in which candidates are likely to often sell an individual position and thus pursue their own agenda (Zetterberg, 2013).Moreover, it would be relevant to compare our findings with a case that has a similar electoral system but where the gender quota system is designed differently.One such example is Costa Rica, which similarly to Tunisia has a PR system with closed lists, but which recently adopted a parity system that requires the parties to respect 'horizontal parity' between men and women at the top of the lists (IPU Parline, 2021).By conducting such analysis, we are able to gain additional insights into how gender and institutions interact to shape patterns of social media use among political candidates.
As a final avenue for future research, we build on the interviewee who suggested that some candidates refrained from using a public Facebook profile because these were associated with increased risks for being exposed to harassment and defamation.It is therefore important to continue exploring the consequences of gendered violence against politicians in online environments.While a few studies have investigated how exposure to violence affects political ambition (see e.g.Bjarnegård, Håkansson, & Zetterberg, 2020;Erikson, Håkansson, & Josefsson, 2021), studies on how online violence against politicians affects campaign styles and subsequently the possibility to be elected to public office are hitherto lacking.Taking the first steps on such a research endeavor is crucial if we are to understand how social media platforms shape gendered power relations in politics.

Notes
1.When using the terms "emerging democracy" and "transition context" in this text, we refer to contexts which are (or have been) undergoing democratic transitions from an authoritarian regime, but are not yet consolidated democracies, which increases the risk for democratic backsliding (see, eg., Kalyvas, 2000).However, the latest political developments in Tunisia have made the country's democratic path more uncertain.On July 25, 2021, President Kais Saied dissolved the parliament, fired the prime minister and consolidated his own powers.Saied has now appointed a new prime minister and cabinet, but the parliament is still suspended and public protests against Saied's concentration of power to himself have increased.In February 2022, Saied moreover dissolved the High Judicial Council, a constitutional institution set up to guarantee the independence and well functioning of the judiciary.See BIC-RHR report for an overview.https://www.bic-rhr.com/research/analyzing-latest-political-developments-tunisiaunclear-pathsstructural-change 2. In contrast, at the local level it is required that parties, coalitions and independents must alternate between men and women among their top candidates across their electoral lists.Moreover, in constituencies with four or more seats on national level, a candidate of 35 years of age or younger needs to be included among the top four candidates, or the list will lose half of the public funding available for electoral expenses.3.There is also another argument as to why we would expect more men to use social media.Women politicians are particularly exposed to online intimidation and threats (Bardall 2013) as well as to more sexualized and gendered harassment online (Erikson et al., 2021), which can impede on the use of social media among women political candidates.4.However, we include private Facebook profiles in the descriptive analysis, with the intention to compare gender differences across profiles.5.However, when we compared the number of posts and comments on the candidates' public figure pages during the three-week official campaign period (Sept 14-Oct 4) for nearly half of the candidates (men as well as women, and elected as well as non-elected), we did not find any gender differences.Thus, women who had a public page were as active as men who had such a Facebook page (see Supplementary Information file, Table A3).6. Interview W19, see also interviews W6, W10, W14. 7. E.g., Interviews M3, W10, W19. 8. Interview W10. 9. Interview W14. 10.E.g., Interviews M2, W6, W10, M13.11.Interview M2. 12. E.g., Interviews W10, M11, M12. 13.Interview W19.

Table 1 .
Descriptive statistics by candidate gender.

Table 2 .
Candidate sex and Facebook public figure page -OLS regression.OLS regression coefficients.The coefficient for Man reports the difference in the extent to which men candidates have a Facebook public figure page compared to women in percentage points.Fixed effects (FE) for 32 districts and 13 political parties.Standard errors in parentheses.Adjusted R-squares.*** p < 0.01, ** p < 0.05, * p < 0.1.