E-expression in a comparative perspective: contextual drivers and constraints of online political expression

ABSTRACT As the opportunities for online political expression grow exponentially, aggregate levels of e-expression vary strongly across countries worldwide. The paper explores contextual factors enabling or restraining e-expression, particularly media dependence, democratic experience and civil society robustness and combines them with micro-level demographics, capacities, and motivations. Based on multilevel logistic modelling of 2014 ISSP 'Citizenship II' data [ISSP Research Group. 2016. “International Social Survey Programme: Citizenship II – ISSP 2014 (Version 2.0.0) [Data file].” GESIS Data Archive.], it shows that e-expression is not dependent on a robust civil society, but on the years spent under democratic rule and the level of media dependence. The latter mediates the predictive effect of political trust, which is negative but ceases in countries with dependent and unfree media. The findings challenge assumptions on the mobilizing potential of digital tools in less free countries, particularly for critical citizens who wish to express grievances outside the circuit of official but closed or monitored channels. In contrast, a reinforcement effect is not only found on the individual-level but also in terms of a democratic digital divide between free and consolidated as well as dependent and young/no democratic regimes. Thereby, the paper contributes to our theoretical understanding of the institutional factors shaping e-expression.


Introduction
Due to the ever-rising number of digital tools, opportunities for citizens to express themselves politically have increased exponentially (Gibson and Cantijoch 2013;Theocharis and van Deth 2018). Yet, aggregate levels of e-expression 1 vary significantly across countries worldwide. According to an ISSP survey conducted between 2013 and 2016, 11.9% of citizens in 34 countries have expressed themselves politically online. Evidently, countries do not cluster according to geographic regions as the numbers are ranging from 24.7% in India, 22.2% in the U.S., and 20.1% in Denmark to as low as 1.6% in Japan, 5.2% in Poland and 5.5% in South Korea (ISSP Research Group 2016). This empirical and changing opportunities for political expression as ever new digital tools and social media emerge and the affordances of existing platforms are constantly updated. Among others, existing research on e-expression has focused on content-expressive behaviour, e.g. creating original political and news media content (Barnidge, Ardèvol-Abreu, and Gil de Zúñiga 2018), embedding or sharing content about politics with others, for example on social media profiles (Gibson and Cantijoch 2013), commenting on and discussing existing media content (Vaccari et al. 2015) or modifying profile pictures to raise awareness for political and social causes (Weeks, Ardèvol-Abreu, and Gil de Zúñiga 2015).
Arguably, there are offline counterparts to some of these forms of e-expression. While these activities may not appear revolutionary at first, their underlying digital technology is indeed of disruptive character (van Dijk and Hacker 2018). Due to the logic of social media programmability, popularity, connectivity and datafication (van Dijck and Poell 2013)information monopolies vanish and communication structures become increasingly horizontal, which deprives traditional authorities and gatekeepers of their powers. Users turn into content producers and consumers alike, which may be shared with a geographically dispersed audience at significantly lower costs in terms of time, money and even cognitive resources (Castells 2007).
Studies that have looked at what motivates or restrains e-expression at the individuallevel underline its mobilizing potential in terms of age, not least because young cohorts are more tech-savvy and familiar with digital tools (Bennett 2008;Literat and Kligler-Vilenchik 2019). Among young adults, political interest and conflict avoidance are the most important predictors in a study of e-expression on Facebook (Vraga et al. 2015). Vraga et al. (2015) argue that the degree of network heterogeneity and, thus, the amount but also the tone of political disagreement may hamper or encourage eexpression. Politically interested individuals who are less conflict-averse are more likely to express themselves despite political disagreement, while disagreement may encourage 'provocateurs' to post about politics even though they are less politically interested. Barnidge, Ardèvol-Abreu, and Gil de Zúñiga (2018), who do not distinguish between age cohorts, find a positive association between social media news use and content-expressive behaviour, which grows stronger in more heterogeneous social networks. The authors also address the potential of the Internet to perpetuate ideological extremism, i.e. due to filter bubbles and echo chambers (Bail et al. 2018;Ferrucci, Hopp, and Vargo 2020;Sunstein 2017). Strikingly, they find that content-expressive behaviour is not significantly related to ideological extremity without accounting for the moderating role of emotional intelligence. Hence, respondents with high emotional understandings result in more moderate ideologies and vice versa.
The literature also underlines the link between e-expression and political participation. Analogues to offline political talk, e-expression may put citizens 'on a pathway to political participation' (Gil de Zúñiga, Molyneux, and Zheng 2014, 613). Shah (2016) argues that all kinds of formal and informal political conversations online offer expressive and informational opportunities that foster political participation. Hence, research found a consistent connection between e-expression and online political participation (Ferrucci, Hopp, and Vargo 2020), political actions during electoral campaigns (Vaccari et al. 2015), boycotting (Becker and Copeland 2016), protest events (Macafee and De Simone 2012), and partyrelated engagement (Vaccari and Valeriani 2016).
Regardless of the substantive amount of literature addressing e-expression, the majority consists of single-case studies focusing on individual-level variables. Hence, there is a striking scarcity of research engaging in comparative analyses of the contextual drivers and constraints, especially beyond Western democracies. Among the few exceptions are meta-analyses by Boulianne (2019) and Skoric, Zhu, and Pang (2016) who find that press freedom and regime type play an important role in the relationship between online political expression and political participation. Barnidge, Huber, Gil de Zúñiga and Liu (2018) show that a positive relationship between network heterogeneity and political expression in 20 countries is stronger if freedom of expression is low. Vaccari and Valeriani (2018) spot that the association between political talk on social media and on and offline political participation is stronger in four established than three third-wave democracies. Except for Barnidge, Huber, Gil de Zúñiga, and Liu (2018), however, these studies mainly focus on e-expression as means to political participation. Thereby, they do not provide enough information on the specific contextual factors that are associated with, and presumably stimulate online political expression as the main dependent variable and across a large variety of countries. In the following, I will elaborate on potential macro-level explanatory factors for e-expression to be tested in this paper.

Explaining e-expression through its context
The paper argues that the dependence and, thus, closure of the media (system) restrains e-expression and mediates the predictive effect of individual-level political distrust on online political expression. Furthermore, it hypothesizes that democratic experience and the robustness of civil society offer theoretical explanations for cross-country differences in the propensity of citizens to express their views about political issues online. Each of the mechanisms will be explained in more detail below.

Media dependence and political distrust
As e-expression is taking place with (online social) media as its stage, I argue that the independence and freedom of the media should encourage citizens to express their views openly while dependence and state control would constrain citizens' action repertoires. A tight control over the media is usually undertaken to manipulate rational citizens' behaviour and avoid electoral punishment (Larreguy and Marshall 2020), for example, by dictating media content that distorts actual government performance (Geddes and Zaller 1989) or concealing cases of corruption (Besley and Prat 2005). A state may also want to demonstrate strength (Simpser 2013) and censor and suppress relevant information that could mobilize individuals to take political actions against it.
Under such circumstances, a rational individualweighing the costs and benefits of her actions (Olson 2003)would engage in so-called 'preference falsification' (Kuran 1997). Even in mildly authoritarian countries, individuals will weigh the external costs of revealing any preferences running counter to the regimein the worst case being prosecuted and risking one's lifeas well as the internal costs of passively acquiescing to a system that does not conform to their privately held beliefs. According to the collective action literature, external costs are reduced as soon as there is a critical mass of individuals who go first and motivate others to follow (Lohmann 2020). In the absence of a free and independent media, however, individuals will find it difficult to learn about the true preferences and actions of fellow citizens and, thus, be discouraged to make a first move. Moreover, leaving traces of one's opinion may be particularly risky in the public and open environment abounding online (Cantijoch Cunill and Gibson 2019). Hence, citizens remain silent. In contrast, if information is freely available in systems with independent media, citizens can access this information which may incentivize them to speak out as well.
However, state intervention in the mediaeven in a strong manneris not only limited to autocracies. In contrast, as shown by Hallin and Mancini (2004) it takes places in Western democracies, too. Take the case of Hungary where the constitutional court in May 2014 ruled that Internet providers can be held liable for user-generated content. This was criticized for encouraging websites to disable comment features. Meanwhile, employees of private and public Hungarian media outlets admitted to growing selfcensorship due to possible fines or dismissals (Freedom House 2015). Beyond Hungary, it is argued that law enforcement agencies in various democratic states are increasingly monitoring online and social media activities (Shaked 2017). Hence, the widespread belief that one's communication is under surveillance could put individuals to silence. It is for this reasoning that I expect countries with a relatively high dependency of the media on the statewhich is interfering and dictating the rules of communication and expressionto discourage citizens from freely expressing themselves politically online (Farrell 2012).
However, a qualification must be made here: e-expression does also take place on different platforms. Hence, beyond popular social media, such as Facebook and Twitter, which are pushing for persistent identities, there are also fully and quasi-anonymous fora, such as 4Chan (Knuttila 2011), which invite respondents to express their views in an incognito fashion. Thereby, the Internet is also characterized by a certain degree of autonomy and anonymity, which makes universal control impossible (Röllgen and Bug 2012). This may be particularly attractive for citizens living in countries with high media dependence and displaying low levels of political trust, who are found to be more drawn to participate in non-institutionalized channels anyways (Hooghe and Marien 2013;Kaase 1999). Hence, detractors of the political systemsuch as those distrusting governmental actors and political institutionswill be pushed to withdraw to safe spaces outside the government's control if official participatory channels are limited and monitored. Boulianne (2019), for example, postulates a relationship between press freedom and social media use arguing that the latter take on an especially important role for political activities in countries without a free and independent press as 'policing and influencing social media communications is more challenging because intervention must occur at many different nodes' (Larreguy and Marshall 2020, 601). Hence, I do not merely assume contextual factors, such as media dependence, to enable or constrain individual-level behaviour, but also to change the meaning and mediate the strength of micro-level variables, such as political trust (Goertz 1994).
Based on this line of thinking, the predictive effect of political trust on e-expression should be conditioned through media dependence. I expect that it is generally negative but stronger in countries with higher media dependence as politically vocal opponents who are distrusting government officials are more likely to switch to online platforms to express their grievances in a less controlled and monitored environment: H1: The higher the media dependence, the stronger the negative effect of political trust on eexpression.

Democratic experience
The second argument concerns the level of democratization (Norris 2002) and the recent democratic history of a country more specifically. As shown by Teorell, Torcal, and Ramon Montero (2007), the democratic legacy, i.e. years spent under democratic rule, does well in predicting institutionalized forms of political participation. Thus, Western, mature democracies usually display higher levels of participation than their counterparts in other regions of the world. In contrast, the suppression of autonomous and independent associations in non-democratic regimes deprives citizens of important 'schools of democracy' that are teaching the civic skills fundamental for political engagement (Hyden 1997;Tocqueville, Mansfield, and Winthrop 2002). In Central and Eastern Europe, for example, the Soviet legacy has impacted negatively upon the latter. Due to the extraordinary penetration of the party system into all kinds of public and private spheres, mistrust in post-communist institutions was fostered leading to deeply ingrained political apathy (M. M. Howard 2003; Pop-Eleches and Tucker 2013). I assume that this mechanism holds for all kinds of political activities with citizens living in countries with comparatively shorter or no democratic history displaying lower levels of e-expression: H2: The shorter a country's democratic history, the lower aggregate levels of e-expression.

Civil society robustness
My last argument concerns the relationships between e-expression and civil society robustness. As shown by Putnam, Leonardi, and Nanetti (1994), social capital is key to the maintenance of democracy. In particular, the authors reveal that a vibrant associational life and civil society encourage democratic practices. Beyond formal associations, civil society also consists of a variety of less institutionalized (new) social movements, groups and even individuals creating solidarities to advance their interest (Linz and Stepan 1996). For the organization and political communication of civil society, digital tools have become indispensable (Castells 2015). According to Obar, Zube, and Lampe (2012), social media in particular do help all kinds of civil society organizations (CSOs) in extending their public outreach, raising awareness about their activities, facilitating mobilization and creating efficient feedback loops. The online sphere is not only a place where a multitude of locally dispersed individuals can interact, pool resources and form new groups to advance their interests, but also a means through which already existing advocacy groups can get their message across more effectively and reach a potentially unlimited number of people. Not least because of this, Castells argues that ' [F]or new social movements, the Internet provides the essential debate, their means of acting on people's mind, and ultimately serves as their most potent political weapon.' (Castells 2007, 250). One can thus expect CSOs to take advantage of digital tools to express political grievances.
In this paper, a robust civil society is defined as being relatively autonomous from the state and in which citizens pursue their collective interests actively, i.e., there are many diverse CSOs with a high degree of citizen involvement (Bernhard et al. 2015). To this end, I assume that a country hosting such a robust civil society taking advantage of the technological advancements online to display higher aggregate levels of e-expression: H3: The more robust the civil society of a country, the higher aggregate levels of e-expression.

Data and models
To test my hypotheses on the relation between contextual factors and e-expression, I combine data from the 2014 ISSP 'Citizenship II' round with several macro-level indicators and analyse them with multilevel logistic regressions (ISSP Research Group 2016). In total, my sample consists of 34 countries and 28,397 respondents. My dependent variable eexpression is operationalized using a survey item from a battery of 'some different forms of political and social action that people can take' (ISSP Research Group 2014).
Respondents were asked to indicate whether they have 'expressed political views on the internet' ranging from 1 ('have done it in the past year') to 4 ('have not done and would never do it'). I recoded it into a dummy whereas 1 indicates participation, i.e. 'have done it in the past year' and 'have done it in the more distant past', and 0 its absence, i.e. 'have not done it but might do' and 'have not done it and would never do it'. Data is mean ± standard error. On average, 11.9% (±11.8%) of respondents across all nations have expressed their political views online. Appendix 1 plots the mean and confidence interval for each country.
The macro-level independent variables are taken from a variety of sources and as of 2013-2016, depending on the time of the survey's fieldwork in each country. Media dependence is measured using Freedom House's Freedom of the Press Index ranking from 0 indicating free to 100 not free with a mean of 28.6 (±17.3) (Freedom House 2021). Moreover, I am looking at democratic experience with a self-generated variable indicating the number of years since the last transition to democracy with a mean of 61.8 years (±50.5 years). The robustness of civil society is indicated by V-Dem's core civil society index, which measures both the organizational environment of and level of citizen activism in civil society based on responses by country experts. It ranges from 0 least to 100 most robust with a mean of 83.4 (±16.6) (V-Dem 2021). Appendix 2 plots the mean for each variable per country. The means of the macro-level indicators across countries are summarized in Table 1.
To investigate the relationship of media dependence and political trust with eexpression, I include a cross-level interaction in the analysis. In the ISSP, political trust is measured on a 1-5 scale, whereas 1 indicates no trust and 5 much trust in 'the people in government to do what is right'.
On the macro-level, I include several covariates. First, I control for national elections coded as a dummy, whereas 1 indicates that a national election has taken place in the year(s) of fieldwork of the survey and two months after to account for e-expressive activities during an electoral campaign (ElectionGuide 2021). Electoral campaigns serve to illustrate a highly politicized period in the electoral cycle and the heyday of political expression, which could incentivize a broad mass all over a country to engage in eexpression (Cantijoch Cunill and Gibson 2019). As illustrated in Appendix 3, national elections happened in 14 countries, including Austria, Denmark, Switzerland, Estonia, Georgia, Croatia, Hungary, Japan, Poland, Sweden, Slovakia, Turkey, United States and Venezuela. The variable does not account for national elections that have taken place in the more distant past. For robustness purposes, I re-ran the analysis for e-expressive activities done in the past year only and included the results in Appendix 4. I also control for regime type to account for differences of liberal principles, particularly freedom of expression, that are traditionally (more) restricted in autocracies. Thereby, the number of channels and possibilities to engage in e-expressionwithout facing sanctions or being controlledare significantly lower than in less restrictive regimes (N. Howard 2010; Skoric, Zhu, and Pang 2016). I use V-Dem's Regimes of the World (RoW), which classifies countries as closed or electoral autocracies and electoral or liberal democracies ranging from 0 to 3 (Lührmann, Tannenberg, and Lindberg 2018). Moreover, eexpression is also fundamentally dependent on resources and technical accessibility. The ability to engage in self-actualization activities, such as e-expression, depends on the satisfaction of basic physical needs (Maslow 1943) as well as digital opportunities, such as access to and usage of the Internet (Norris 2001). Hence, I also control for GDP per capita and per cent of the population using the Internet as reported by The World Bank (2021). Following the reasoning of previous research, I also include a battery of micro-level indicators measuring resources and motivations to engage in e-expression. First, I include age measured in years, gender, and socio-economic status (SES) in the form of education and income. Education is a categorical variable ranging from 0 (no formal education) to 6 (university education). As it was not inquired about in Hungary, I re-ran the analysis also without education and included the results (N = 35) in Appendix 5. Income is not directly comparable in the ISSP, which is why I look at the relative position of income earners in a country and assign observations to the country-specific income deciles. In addition, I include political interest, which is measured on a 4-point-scale with 1 indicating not at all and 4 very interested. I also add a variable on ideological extremism measured on a L-R ideological scale that was folded so that moderate ideology equals 0 and extreme ideology 5. Finally, I also consider the relationship between e-expression and political participation as posited by the literature. First, I control whether a respondent has voted in the last election as a more institutionalized form of political participation. Second, I include a dummy measuring participation in at least one of the following informal activities included in the ISSP: petitions, boycotts, demonstrations, rallies, and donations.
Before running the analyses, I checked for multicollinearity. The VIFs of all variables are below 10 and acceptable (Hail et al. 1995). 2 To test my hypotheses, I run several multilevel specifications with random intercepts, adding all the macro-level arguments as well as the interaction one after the other and combining them together into a final model. All individual-level variables are unstandardized and unweighted. 3  Table 2 presents the results of the multilevel analyses. Models 1, 2, 3, 4 and 5 depict the impact of country-level variables and the interaction while controlling for national elections, regime type, GDP, and Internet usage. According to Model 1and irrespective of the level of political trustrespondents from countries with more dependent media which are subject to state intervention are less likely to express their views on political issues online (b = −.02, p < .10). In fact, the odds to engage in e-expression decrease by 2% with a one-unit increase on the 0-100 scale of press freedom. As illustrated in Model 2 and Figure 1, the interaction between media dependence and political trust is statistically significant (b = .003, p < .05). However, while the statistically significant effect of political trust is generally negative, its predictive power decreases in countries with higher media dependence. In other words, political trust has a stronger negative effect on e-expression in countries with relatively free and independent media and it becomes insignificant beyond a Freedom of the Press score of 60. As illustrated in Appendix 3, however, this may be explained by the fact that only three countries fall into that range, namely Turkey (65), Venezuela (81) and Russia (83). Nonetheless, I can reject H1: citizens from closed media systems who are displaying high levels of political distrust are not relatively more likely to retreat to potentially safe harbours online. On the contrary, e-expression as an outlet to voice grievances is relatively more frequented by politically distrustful individuals who are living in countries with relatively independent and free media. According to Model 3, I find evidence that supports H2: democratic experience is statistically significant (b = .01, p < .10) and the relationship is positive. Substantially, with each additional year spent under democratic rule, the odds for e-expression increase by 1%. The findings correspond to broader developments observed in consolidated democracies, where publics are becoming increasingly critical of authority and political authority in particular, which is why they are more engaged in elite-challenging activities outside conventional action repertoires and via alternative (online) channels (Dalton 2006;Dalton, Van Sickle, and Weldon 2010;Inglehart 1999;Marien, Hooghe, and Quintelier 2010;Norris 1999). In more restrictive and less free countries, the results also point to limitations to the extent that digital tools may enable 'risky' political expressions (Barnidge, Huber, et al. 2018) and accelerate democratization processes, if the risk is defined as being prosecuted for expressing preferences running counter to the regime. Collective action against the status quo is mobilized by a critical mass that dares to speak up and, thereby, reduces the external costs for others to follow (Lohmann 2020;Olson 2003). The results suggest, however, that even opponents are more likely to engage in preference falsification and passive acquisition instead of eexpression, which facilitates a continuation of current affairs (Kuran 1997). Hence, it appears that a democratic digital divide (Norris 2001) is reinforced by macro-level factors: control, surveillance and restrictions and low or no democratic experience prevent citizenseven the most critical onesfrom using digital tools for political expression. In contrast, e-expression serves to illustrate an additional, well frequented channel for citizens in consolidated democracies that already have large action repertoires at hands.

Results
According to Model 4, I can reject H3 on civil society robustness. Considering that eexpression is not related to the diversity of CSOs and the degree of citizen involvement, one could conclude that e-expression is more of an individual activity that is not stimulated due to its relevance for and usage by associations, movements, or collective interests. Model 5 combines all macro-level variables and the controls at both levels. The results for both, the interaction term of media dependence and political trust as well as democratic age remain robust.
Even though the focus of this paper is on the contextual drivers and constraints of eexpression, discussing the individual-level controls is substantially interesting and an important contribution to the ongoing scholarly debate on mobilization and reinforcement effects of digital tools. This section, therefore, briefly reflects on micro-level capacities and motivations of e-expressive citizens. As revealed in Model 5, male respondents are 25% more likely than females to express their political views on the Internet (b = .22, p < .001). In line with the literature, age has a negative effect with each additional year decreasing the odds by 4% (b = −.04, p < .001). Regarding capacities, any additional educational degree increases the likelihood that respondents engage in e-expression by 4% (b = .04, p < .001). Strikingly, it appears that education is also absorbing the effect of income, which is statistically insignificant in all models featuring education but becomes significant in Appendix 4 when education is excluded. Considering that the control of GDP on the macro-level, too, renders insignificant, one could argue that material resources do not enable e-expression.
In substantive terms, non-institutionalized political activities and e-expression have by far the strongest relationship: engaging in at least one petition, boycott, demonstration, rally or donating increases the chances of e-expression more than five-fold (b = 1.75, p < .001). Voting, too, is statistically significant, but the relationship is negative (b = −.11, p < .05). This underlines the informal character of e-expression. In terms of motivation, political interest, too, is highly significant: respondents positioning themselves one point higher on the scale are about twice as likely to engage in e-expression (b = .70, p < .001). In terms of ideological extremism, a one-unit increase on the 5-pointscale from moderate to high extremism increases the odds by 12% (b = .11, p < .001). This suggests that moderate citizens tend to remain silent online. One can therefore conclude that e-expression is done by predominantly young males with high education that are politically active and interested and tend towards ideological extremism. This supports previous findings on reinforcement effects of digital tools (Oser and Boulianne 2020).

Discussion
This paper proposes to approach e-expression differently than in previous studies: not solely as means to political participation or as a subject of single-case studies in advanced democracies, but as a standalone act with aggregate levels varying across countries that do not cluster according to geographic regions. Hence, the central question explored in this paper is the extent to which contextual drivers of media dependence, democratic age, and civil society robustness shape aggregate levels of engagement in e-expression. This is of particular relevance as the literature acknowledges that e-expression fulfills central functions for democracy and democratization and is highly context-dependent (Cantijoch Cunill and Gibson 2019;Chadwick 2013;Gil de Zúñiga, Molyneux, and Zheng 2014;Lane, Do, and Molina-Rogers 2021;Margetts et al. 2016;Shah 2016).
The results indicate that respondents from countries with higher media dependenceirrespective of their levels of political trustusually display lower levels of participation, so do those with shorter democratic histories. The media environment also mediates the effect of political trust on e-expression: it is generally negative but ceases the less free and independent media can operate in a state. This runs counter to my assumption that critical, distrusting citizens are more drawn to engage in e-expression in relatively anonymous and autonomous channels online if others are monitored. E-expression is also predominantly done in advanced, old democracies. These findings raise questions on the extent to which digital tools indeed posit an alternative route for citizens to express themselves if formal paths are closed. Instead, it appears to be an additional option for politically active and critical citizens who already have a wide range of possibilities to choose from. A democratic digital dividebetween those that use or do not use digital tools for political expressionseems to be reinforced by institutional factors: surveillance and control of the media and short or no democratic experiences serve to illustrate constraining factors for e-expression. The individual-level variables, too, suggest that instead of mobilizing traditionally disadvantaged or disengaged citizens, a reinforcement is taking place with politically active and interested citizens also being more drawn to express themselves online. Thereby, my findings fall into the hands of supporters and opponents of democratic consolidation alike: for political expression online to flourish, the media and Internet should be as open and free as possible, though patience may be needed in countries with late transitions to democracy. By implication, governing a tight regime over the media discourages parts of the population to participate in eexpression, as those with preferences against the status quo are less politically outspoken.
The conclusions articulated above are limited in two ways: first, the data was gathered in the mid-2010ths and the number of digital tools for political expression has increased exponentially in the meantime. Thereby, e-expression as a subject of study is constantly changing and the analysis at hand merely provides a snapshot in time for the period of fieldwork. Second, there are several outliers -India in terms of e-expression, Venezuela and Russia in terms of media dependence and the US in terms of democratic age. To overcome the (statistical) challenges posited by such outliers and increase the balance of samples, more countries should be included in future surveys featuring e-expression. Despite these limitations, the present study provides an important and unprecedented empirical exploration of the importance of media dependence and democratic age in stimulating e-expression and a starting point for future work to build on.

Notes
1. For the remainder of this paper, I will refer to online political expression as e-expression.

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