Influenced by Peers: Facebook as an Information Source for Young People

Introduction

There is good reason to suppose that the patterns of political information consumption may exert an influence on citizens’ political behavior. For example, the revolt of the 1968 generation, the first “television generation” that grew up in front of the screen, is often connected to changing media consumption habits. The reality presented by television, however, differed from that of previous generations, which resulted in generation gaps (Bodroghkozy, 2001). Social and technological changes are persistently shaping the political information environment, and this should be reflected by political science (Bennett & Iyengar, 2008).

Over the last few years, the political information environment has been significantly transformed. The greatest change is related to the rise of social media. Social media is increasingly pervading our everyday life, which is further strengthened by the development of mobile technologies enabling constant presence. However, qualitatively this change in the political information environment due to the rise of social media differs from earlier transformations, which always remained within the field of professional journalism. While changing formats used to be operated by journalists and media experts, today’s information universe of social media is largely formed by ordinary people and their communication.

The purpose of this study is to contribute to the understanding of the outcomes of this changing information environment by uncovering the distinct effects Facebook has on young people’s political behavior compared to more traditional, professional-operated media platforms. The study suggests that the growing importance of social media, especially of Facebook, in political information consumption results in the revitalization of the two-step flow model of communication (Katz & Lazarsfeld, 1955; Lazarsfeld, Berelson, & Gaudet, 1948), proposing that political behavior is strongly affected by social influences. The general hypothesis regarding the presence of social influences on Facebook is demonstrated indirectly by testing two sub-hypotheses. These sub-hypotheses are connected by a variable measuring satisfaction with the way democracy works, which is deduced from the thesis of the dominance of negativity in Facebook political communication. To test these hypotheses, a secondary analysis of a survey among Hungarian university students was carried out. The result of the first test shows that dissatisfaction with democracy is a strong predictor of the political sharing and posting behavior, and the second test demonstrates that being regularly informed about politics through Facebook leads to more negative perceptions among students about the way democracy works. The two findings are connected by the empirically not tested assumption that the second result is caused by the first, that is, the dissatisfaction of peers who share or post about politics on Facebook is reflected in the dissatisfaction of those who obtain information from it. An important contribution of this study is that influences that result from using Facebook are not investigated in themselves, but are embedded into the modern information environment where several information sources, including television, radio, online news portals, and daily newspapers, are used simultaneously. It is important to see that we cannot capture the effects of single platforms without taking into account the whole information universe in which our respondents moves.

The first part of the study discusses the growing importance of social influences on political behavior owing to Facebook as its special and peculiar feature compared to other information sources. Next, I argue for the dominance of negativity on Facebook political communication, which justifies the use of the satisfaction with democracy variable. The second part of the study presents the methods and variables, followed by the results and conclusions.

Facebook and a Return to Social Influences

Technological and social changes may shape what kinds of influences are exerted on citizens’ political behavior. ¹ Different effects could emerge from or be amplified by the different patterns of information acquisition about politics. Therefore, as Facebook is becoming crucial in many people’s political orientation, this may have consequences in terms of their political behavior. It seems that the stronger role of Facebook is associated with increasing social influences on political behavior. This implies that citizens’ political communication is becoming more important because the political behavior of the majority that gets informed through social media is shaped by the minority that communicates about politics there.

Obviously, in political communication research, it is not a new idea that social influences effect political behavior. Concerning the role social communication plays in social processes, as early as the end of 19th century, Gabriel Tarde (1898/1989) asserted that “if people did not talk, it would be futile to publish newspapers . . . they would exercise no durable or profound influence; they would be like a vibrating string without a sounding board” (cited by Katz, 2006, p. 265). A few years later, American war propagandists put Tarde’s thought into practice as reflected in the idea of having “four-minute men” during World War I (WWI). To create a war-supporting public, the US Committee on Public Information recruited local opinion leaders and provided them with scripts and instructions for everyday conversations with neighbors on a weekly basis (Ewen, 1996, pp. 117–121). However, following WWI, the direct media effects approach, hallmarked by Harrold Lasswell (1930, 1935), became prevalent. Emerging totalitarian political systems in Europe and the expansion of mass media were a fertile ground for the image of the citizen as a passive recipient. However, in the middle of the 20th century, this image was refuted by the Columbia School’s election studies, and social influences came to the fore again. Paul Lazarsfeld et al. (1948) found that in their voting choices, voters are more reliant on opinion leaders in their personal contexts than on the mass media. This result led the authors to form the two-step flow model of communication (Katz & Lazarsfeld, 1955), arguing that the political behavior of the wider, politically uninterested public is shaped by a small number of highly interested peers. The latter group, however, are active mass media consumers, indicating that through the mediation of these opinion leaders, the mass media could exert an influence on the larger public. Despite its subsequent critiques for its theoretical roots (social pluralism) and poor empirical groundedness (Gitlin, 1978), the theory has become a prominent theoretical approach in communication research. However, many modifications have been proposed to weaken its focus on opinion leaders and connect it to the idea of the more horizontal personal influences within information network flow (e.g., Robinson, 1976).

Despite its popularity in communication research, over the last decades, less attention has been given to the two-step flow model in political communication research. Frequently cited trends like weakening traditional social ties, growing individual control over information (Bennett & Iyengar, 2008), as well as the taboo nature of politics in social conversation (Eliasoph, 1998) turned scholars’ attention to other influences on political behavior. Empirical investigations show that people rarely talk about politics (Conover, Searing, & Crewe, 2002), and even when they do, these conversations take place in tight circles with close ties (Huckfeldt & Sprague, 1995; Wyatt, Katz, & Kim, 2000) and like-minded people (Smith, 2015). As a result, changes in attitude or political behavior are hardly expected to occur due to everyday political conversations. Furthermore, Bennett and Manheim (2006) proposed that the increased possibility of direct targeted strategic communication and narrowcasting techniques also weaken the empirical validity of the two-step flow model, arguing that a one-step flow of communication model may be more applicable in the current social and technological context.

However, the rise of the Internet, and especially social media, has reopened the door to bringing the idea of social influences back into political communication research. Already in the Web 1.0 period, Norris and Curtice (2008) proposed that the actual relevance of political websites is to directly reach opinion leaders rather than voters in general. Their results demonstrated that those small segments of voters who follow such websites are more likely to discuss politics with their families and friends. Later, Web 2.0 in general, and Facebook in particular, actually changed the relevance of citizens’ political communication. Facebook has become an important source of information about politics for a number of voters, which information in turn is largely created and disseminated by their peers. The reach of ordinary citizens’ communication on Facebook has grown to an extent never experienced before. Potentially, any communicative act performed in the Facebook public can be seen by nearly all offline acquaintances of the performer, including family members, friends, and colleagues. As a result, although the past research showed that only a minority of users talk about politics offline, on Facebook the voice of even a few can be heard far away.

Despite these facts, few studies have investigated how citizens’ political communication affects political behavior and attitudes in this new communication context. The strongest evidence about the influence of Facebook friends’ information on political behavior is offered by an experiment conducted at the University of California. The experiment involved 61 million Facebook users; on the 2010 US Congressional Election Day, at the top of their news feeds, treatment groups got a call-for-voting statement. The “Social message” group was shown the statement displaying how many Facebook friends, with some of their profile pictures included, had already voted. The members of the “informational message” group could see the statement without showing the above information. The results are that users who saw their voting friends were more likely to vote than those who did not see such information (Bond et al., 2012). In a laboratory experiment, Turcotte, York, Irving, Scholl, and Pingree (2015) examined the effects exerted by social mediation on the evaluation of the source of the message. The results showed that news sources are trusted more when their content is mediated by Facebook friends rather than when they are seen in themselves. In addition, the perceived credibility of the news source increased when the mediating Facebook friend was thought to be politically informed and honest about politics. The importance of social cues appearing in messages was supported by Messing and Westwood’s (2014) laboratory experiment. The study found that users suspend even their partisan bias in their news selection if social cues turn up in the news. Moreover, in this research, social cues were impersonal, as only the number of users recommending the articles was shown to the treatment group.

The significance of social influences on political behavior through social media was supported by some survey-based research. During the 2008 US presidential election, Vitak et al. (2011) examined the effect of Facebook usage on students’ political participation at a large Midwestern university. They found that being exposed to Facebook friends’ political activities is positively associated with political participation. Diehl, Weeks, and de Zúñiga (2016) also showed that the more intensive social media use was, the more frequently users changed their political opinions based upon what someone they regarded as influential posted on social media.

According to the two-step flow model of communication, social influence occurs because citizens communicating about politics are opinion leaders in terms of their personality traits, opinion formation intention, or self-definition. A few empirical results suggest that the personal traits of users who communicate about politics on Facebook correspond to some features usually ascribed to opinion leaders. Examining Facebook communication about climate change among republican voters, Vraga, Anderson, Kotcher, and Maibach (2015) found that the more a user perceives himself or herself as an opinion leader, the more he or she is apt to engage in political communication on Facebook, especially about climate change. Another recent study has also shown that so-called “prosumers,” who intensively communicate about politics on social media, are more likely to consider themselves opinion leaders and try to persuade others of their own opinion (Weeks, Ardèvol-Abreu, & de Zúñiga, 2017).

Overall, there are signs that the relevance of social influences on political behavior has increased owing to social media in general, and Facebook in particular. This may mean that citizens or opinion leaders actively communicating about politics play a vital role in shaping political behavior and public opinion. However, the literature discussed above does not address the question of how these social influences relate to other media effects. For example, it is an important question whether, in terms of their form or size, social influences on Facebook are different from the effects of other media that citizens consume. This is important because most people are not only informed through Facebook but also, accidentally or deliberately, exposed to political content from other media. The effect of one information source cannot be grasped without considering the entire information universe wherein users live. Another deficiency of the few studies examining social influences on social media is that they were seeking short-time effects on the activity level, especially on political participation and did not focus on how peers shape political attitudes in the longer run. We may assume that social influences are significant in the long-term political opinion formation process, shaping political experiences and perceptions for a longtime—this could be especially relevant for young people with flexible political preferences. As most political information is acquired in indirect rather than direct ways, the perception of political situations is strongly shaped by indirect experiences. Facebook has increasingly become an integral part of people’s everyday lives; therefore, experiences stemming from here may strongly impact how they perceive politics and reality.

This overview leads to the following general hypothesis:

The general hypothesis is tested indirectly rather than directly by breaking it down into two sub-hypotheses. The two sub-hypotheses are connected by a complementary variable based on the idea that citizens’ political communication on Facebook is characterized by intense negativity.

Negativity in Facebook Political Communication

Over the past decades, citizens’ negative attitudes, growing cynicism, and distrust toward politics have been well-discussed popular topics in the political science literature (see Cappella & Jamieson, 1996; Norris, 1999 etc.). There is reason to believe that this negativity applies to citizens’ political communication as well, and is especially dominant in Facebook political communication.

Political communication on Facebook is mostly driven by expressive motives (see Marichal, 2013; Svensson, 2011). This means that the acts publicly performed here essentially aim at the actor’s self-presentation and make an impression on his or her social context. Even when there are other motives behind the act (e.g., the instrumental motive of persuading others or the wish to reach a mutual understanding based on communicative rationality), users must always be aware that their communication can be seen by the whole circle of their acquaintances—any utterances are part of their publicly presented selves. As a result, supposedly, those people communicate about politics on Facebook, whose publicly presented self-image includes some political aspects.

The notion of expressive motive-driven political communication fits well with Lance Bennett’s personalized political identity approach. Bennett (2012) suggests that growing fragmentation, individualization, and consumer culture result in personalized political identities. It is mainly the investigation of new and successful social movements that leads Bennett to the conclusion that in the modern, digitalized political sphere, broad and inclusive action frames are gaining relevance. These are such action-stimulating contents that do not require strong identification. Their central features are that people can connect to them for different reasons and with different narratives; they can be easily personalized, and through them different identities can be performed and experienced. As a result, they can easily and quickly spread through digital networks. Thus, a new kind of political action has emerged, that Bennett and Segerberg (2012) call “connective action.” However, this connectivity can be extended beyond the movement sphere and used for understanding changing approaches toward politics: political content, issues, and events are appreciated and become subjects of citizens’ political communication if they can be applied to performing, experiencing and shaping individual, personalized identities and involve less of an identification burden, rather than content that aims to transform these identities and fit them into a homogenized collective identity-block. It appears that connective action can be more easily carried out on negative than on positive ground. A positive connection to any political subject entails a greater identification burden than any negative connection. Performing personalized identity is easier by refusing something rather than by identifying with something. Moreover, the “collapsed contexts” (Marwick & boyd, 2011) and the dominance of weak ties (Vraga, Thorson, Kligler-Vilenchik, & Gee, 2015) on Facebook also urge users not to undertake too much of an identification burden. They are forced to present a self-image in their public communication that they could undertake in front of their distinct, heterogeneous (Diehl et al., 2016), and offline separated but on Facebook “collapsed” social contexts.

Some empirical findings stemming from investigations of citizens’ reactivity on politicians’ Facebook pages support the claim regarding the dominance of negativity on citizens’ Facebook political communication. Larsson (2015) found that critical Facebook posts were mostly commented and the second most liked on Norwegian party leaders’ Facebook pages during the 2013 parliamentary elections. In the United States, Xenos, Macafee, and Pole (2017) also found that opponent-attacking posts were liked and commented more than self-promoting posts on congressional and gubernatorial candidates’ pages during the 2010 elections. Similar results were obtained in a study about the reactivity of Hungarian parliamentarian candidates’ Facebook posts during the latest election in 2014. Among the post elements examined, negativity was the most frequently commented and shared by the candidates’ followers. In contrast, the presence of positive content elements within the posts did not significantly affect the number of likes, comments, or shares (Bene, 2017). On the whole, these findings indicate that followers of politicians’ Facebook pages are more likely to connect to and use negative content for performing political communication activities in front of members of their networks.

Of the database used, the variable measuring satisfaction with the way democracy works relates mostly to political dissatisfaction and negativity. There is a remarkable debate in the literature about what this variable actually measures. Some scholars caution against using this variable, since they argue that this is not an indicator of supporting the principle of democracy as it is intended to be. It is more about the performance of a political system, the satisfaction with incumbents, or even a summary indicator of satisfaction with politics in general (see Canache, Mondak, & Seligson, 2001; Linde & Ekman, 2003). Paradoxically, this is a benefit for this study, rather than a limitation because the research is not at all interested in the legitimation of democracy. It intends to capture a general dissatisfaction with or negativity toward politics. There are several interpretations of what this variable exactly measures but a common feature of the explanations is a negative attitude toward politics. This is the attitude the study intends to capture, and from this perspective, it is irrelevant whether this dissatisfaction is coming from a negative attitude toward the principles of democracy, or from evaluations of political elites, incumbents, or people’s personal economic situation.

A crucial benefit of using this variable is that it is sufficiently general and broad to capture lasting political experiences. Asked for a general evaluation of the political situation, respondents are forced to draw on their impressions. These impressions may come from different sources. Kovács (2014) found that the degree of satisfaction with the way democracy works is shaped by subjective financial well-being, degree of uncertainty about one’s future, and party affiliation. Beyond the fact that these points offer additional evidence to show that the variable is more about a general negativity toward politics rather than an attitude toward the principles of democracy and its state, this result indicates that personal experiences and party identity contribute relevant impressions to the evaluation of the way democracy works. However, crucial impressions may also stem from indirect information about politics. It might be relevant what kind of political reality is unfolded through the respondents’ mediated experiences.

As argued above, I expect that negativity is attractive in Facebook political communication because negative political content is more readily applied for connective political actions than positive content, which entails more of an identification burden. Based on this assumption, we have the following hypotheses:

If hypothesis H1.1 is supported, we can reasonably assume that Facebook political communication is dominated by a dissatisfied minority’s political critiques. It follows that because spreading political content on Facebook may provide meaningful impressions for the evaluation respondents make. The latter question has already been investigated in other contexts. In a secondary analysis of Eurobarometer data of 27 European countries, Ceron and Memoli (2016) found that social media news consumption has a significantly negative effect on satisfaction with the way democracy work, and that effect was mediated by a rough proxy variable of disagreement on social media. ² This study aims to connect this result to another possible explanation: to the negativity of citizens’ political communication on Facebook. If both hypotheses are supported, the effect on political behavior of obtaining information through Facebook can be connected with the communication the minority of young people performs on Facebook which is the main idea of the general hypothesis. However, this connection is not directly tested; instead, it is only a plausible explanation for the observed effect of being informed through Facebook.

Both political communication on Facebook and satisfaction with the way democracy works may be shaped by several other variables, which are certainly to be controlled for. The theory of the two-step flow model of communication also states that citizens communicating about politics (a) are opinion leaders and (b) consume mass media content to a larger degree than the general population. These factors should be considered when examining political communication on Facebook. Therefore, the first model pair contained a proxy variable of being an opinion leader and the frequency of professional mass media consumption. Several studies have demonstrated that political communication on Facebook is strongly affected by political interest (e.g., Vitak et al., 2011; Vromen, Loader, Xenos, & Bailo, 2016). Moreover, since Facebook communication is described above as guided by expressive motives and connected to identity, the effects of the intensity of political identity on performing political communication on Facebook can be assumed.

The effect of Facebook as an information source on the attitude examined can be settled as long as the potential effects of other information sources are considered as well. Political impressions may stem from other media consumed by respondents. Hence, the effect of being informed about politics through Facebook on the satisfaction with democracy is investigated within the whole information universe in which our respondents move. This enables us to directly compare the effect sizes and forms of different information sources.

Kovács’ (2014) study shows that satisfaction with the way democracy works is shaped by subjective financial well-being, degree of uncertainty about the future, and party affiliation. Also, political interest may influence the dependent variable because it affects the cognitive processing of information (see Kazee, 1981; MacKuen, 1984).

Young People and Satisfaction With Democracy in a Post-Socialist Context

Although young people differ from the general population in their political attitudes and participation, there are minimal cohort differences in political trust, cynicism, or dissatisfaction (see Siemienska, 2002; Zukin, Keeter, Andolina, Jenkins, & Carpini, 2006, pp. 89–119). Studies examining satisfaction with the way democracy works have also found that age is not a significant predictor of dissatisfaction (Cordero & Simón, 2016; Magalhães, 2016). However, research also shows that dissatisfaction with democracy is generally higher in the post-socialist countries than in older democracies (Kestilä-Kekkonen & Söderlund, 2015; Magalhães, 2016). Furthermore, analyzing survey data from 1990 to 2003, Neundorf has found that in these countries there was a slight generational variation regarding the satisfaction with the way democracy works because people being socialized before the regime change were more dissatisfied with democracy than younger generations. However, the evaluation of economic performance played a more important role in this respect, having an intergenerational equalizer effect on satisfaction with democracy (Neundorf, 2010). Unfortunately, there is no similar research from later periods, but the remarkable dissatisfaction observed among Hungarian university students may suggest that intergenerational differences have further declined. Even though the degree of this dissatisfaction has slightly decreased since 2011, even in 2015 two-thirds of students were dissatisfied with the way democracy works (see Appendix).

Methods

To test these hypotheses, a secondary analysis was carried out. The database used in this study comes from a survey among Hungarian university students conducted by the Active Youth in Hungary Research Group. ³ The population examined is rather specific, whose results cannot be generalized to the entire society. However, this sample is suitable for testing a newly formed hypothesis: this is the segment of society where the effect of two-step information flow is more likely to occur, as young people may be more reliant on interpersonal communication than their older peers (see Yang & Stone, 2003). The sample consists of 800 students recruited from 35 universities and interviewed face-to-face. Interviewees based on random selection were asked at their home universities. ⁴ The data collection took place between 1 and 25 April 2015. The sample is representative of the university and college student population in Hungary.

Both sub-hypotheses were tested using binary logistic regression models. However, respondents having ever posted or shared political content on social media sites were removed from the models of the second sub-hypothesis. This is reasonable since according to the first sub-hypothesis, they are more dissatisfied with democracy and the causal mechanism behind the casual relationships of the two sub-hypotheses assumed by the general hypothesis is that their dissatisfaction is what infects those who are informed about politics through Facebook. Including these young people within the models of the second sub-hypothesis would distort the validity of this assumed mechanism.

Variables

Results

First, we should take a look at the descriptive statistics of political information sources and at posting and sharing activity. As Table 1 shows, Facebook has become the most important political information source for university students. Half of the sample regularly, and a further third sometimes acquire political information through Facebook. Another crucial information source is online news portals, while television and radio seem to be relevant in casual exposure to political information. The popularity of news portals, however, may not be independent from Facebook, as it is known that the highest share of news portal visitors come from Facebook. ⁷ On the other hand, only a third of young people actually create political content on social media sites (see Table 2). This means that while a considerable percentage of university students are informed about politics through Facebook, this information is mostly created and distributed by a small active segment. This finding raises the question of what influence they have on their more passive peers.

OPEN IN VIEWER

Table 1.

How Often Are You Informed About Politics From the Sources Listed Below?

	Regularly (%)	Sometimes (%)	Never (%)	N
Online news portal	43	44	13	798
Television	26.4	41	32.6	797
Radio	18.5	38.8	42.7	799
Daily newspaper	7.6	27.6	64.8	799
Facebook	51.6	33.9	14.5	799

OPEN IN VIEWER

Table 2.

Have You Ever Been Engaged in the Activities Listed Below?

	No (%)	Yes (%)	N
Sharing	68.3	31.7	798
Posting	76.7	23.3	797

The first step is to examine people who create content on social media. What factors account for sharing and posting behavior? Table 3 shows the logistic regression models for sharing and posting on social media. In accordance with previous findings, political interest is a strong predictor of sharing and posting behavior. Surprisingly, the results indicate that the intensity of political identity is irrelevant when it comes to Facebook political communication. This result may suggest that personalized political identities cannot be grasped by the old opposites. However, hypothesis H1.1 also gains strong support, since the results show that sharing and posting are triggered by dissatisfaction with the way democracy works. The more dissatisfied the students are with democracy, the more likely they are to express themselves politically on social media sites. Also, in accordance with the two-step flow theory, the strongest predictor in both models is the proxy variable of opinion-forming intentions, the “encourage to vote” variable. This may indicate that people who create content on social media are generally more inclined to attempt to form their peers’ political attitudes. However, those who communicate about politics on social media are not characterized by stronger professional mass media consumption.

OPEN IN VIEWER

Table 3.

Binary Logistic Regression Models of Sharing and Posting About Politics on Social Media.

	Sharing		Posting
	B (SE)	Odds ratio	B (SE)	Odds ratio
Satisfaction with democracy	−.884 (.189) ***	.413	−.619 (.207)**	.538
Urge to vote	.857 (.228)***	2.357	1.049 (.234)***	2.854
Intensity of left–right identity	.001 (.095)	1.001	−.079 (.105)	.924
Intensity of liberal–conservative id	.158 (.088)	1.171	.154 (.096)	1.166
Professional mass media consumption
Regularly	.915 (.649)	2.497	1.502 (1.046)	4.492
Sometimes	.873 (.653)	2.394	1.596 (1.049)	4.934
Political interest	.382 (.074)***	1.466	.494 (.0083)***	1.639
Constant	−2.033 (.701)**	.131	−3.676 (1.087)**	.025
χ2	91.831***		94.188***
R2 (Nagelkerke)	.158		.174
N	765		764

SE: standard error.

The reference category is the “never” answer in case of the professional mass media consumption variable.

*

p < .05; **p < .01; and ***p < .001.

The second step is to investigate the effect of being informed through Facebook on satisfaction with way democracy works. Note that respondents having ever posted or shared political content on social media sites were removed from these models. The first logistic regression model (see Table 4) indicates that the perception of the working of democracy is strongly shaped by party affiliation and perception of personal uncertainty. All non-Fidesz voters, including those who do not have clear party affiliations, regard the functioning of democracy as worse than voters of the government party, Fidesz. In addition, students, perceiving their future as insecure, are more dissatisfied with the political system. Interestingly neither political interest nor subjective financial well-being affects satisfaction with democracy. Among the political information sources, only being regularly informed through Facebook and television influence the dependent variables. However, their effects are in opposite directions: while being regularly informed from television positively influences satisfaction with democracy, Facebook as a regular information source decreases the level of satisfaction. Although being sometimes informed through Facebook also influences the dependent variables negatively, this deviation from the reference category (i.e., never being informed through Facebook) is not significant. This may bolster the above discussed claim that Facebook has a long-term effect on political attitude by shaping young people’s political reality perceptions. This may occur by being for a longtime regularly rather than occasionally exposed to political content.

OPEN IN VIEWER

Table 4.

Logistic Regression Model of the Satisfaction With the Way Democracy Works.

	Model 1		Model 2
	b	Odds ratio	b	Odds ratio
Facebook ***
Regularly	−.963 (.308) **	.382	−.975 (.313)**	.377
Sometimes	−.398 (.294)	.672	−.421 (.298)	.656
Online news portal
Regularly	.371 (.341)	1.450
Sometimes	.370 (.318)	1.447
index.hu
Regularly			.141 (.373)	1.152
Sometimes			.312 (.314)	1.365
origo.hu
Regularly			.569 (.355)	1.767
Sometimes			−.162 (.262)	.850
444.hu
Regularly			−.671 (.423)	.511
Sometimes			−.236 (.251)	.790
kuruc.info
Regularly			.393 (.654)	1.347
Sometimes			.948 (.302)**	2.579
Television
Regularly	.605 (.303)*	1.832	.453 (.315)	1.573
Sometimes	.452 (.264)	1.572	.420 (.271)	1.522
Radio
Regularly	.320 (.317)	1.377	.362 (.329)	1.437
Sometimes	.128 (.234)	1.136	.145 (.240)	1.156
Daily newspaper
Regularly	.469 (.472)	1.598	.551 (.488)	1.734
Sometimes	.273 (.241)	1.314	.216 (.249)	1.241
Party***
Left-wing	−2.921 (.521)***	.054	−2.990 (.524)***	.050
Jobbik	−2.335 (.401)***	.097	−2.566 (.419)***	.077
LMP	−1.995 (.446)***	.136	−2.103 (.455)***	.122
Others	−3.261 (1.159)**	.038	−3.230 (1.161)**	.040
No party/no resp.	−1.822 (.360)***	.162	−1.899 (.364)***	.150
Personal uncertainty	.285 (.123)*	1.329	.262 (.126)*	1.300
Financial well-being	−.175 (.156)	.839	−.163 (.160)	.850
Political interest	.124 (.095)	1.132	.124 (.099)	1.133
Constant	.062 (.756)	1.064	.366 (.772)	1.442
χ2	98.265***			113.745***
R2 (Nagelkerke)	.236			.270
N	515			515

Dependent variable is dummy, 0 = “dissatisfaction” and 1 = “satisfaction” with the way democracy works. The reference category is the “never” answer in case of information sources, and Fidesz, the government party, in case of the party affiliation variable.

*

p < .05; **p < .01; and ***p < .001.

Since only media are included in this model, an alternative explanation may tell us why most of them are inefficient: different outlets stimulate distinct perceptions of the state of democracy, and within an aggregate variable they may cancel out each other’s effects. As the questionnaire contains some particular online news portal outlets, in their case, this alternative explanation can be tested. The second model includes four news portal outlets instead of the collapsed news portal variable: two openly government-opposition sites (index.hu, 444.hu), a more neutral site (origo.hu), and a radical right-leaning site (kuruc.info). The data show that only one variable influences significantly the dependent variables, but casual visiting of kuruc.info increases the satisfaction with the functioning of democracy rather than decreasing it. Following opposition news sites regularly or occasionally has no effect on the dissatisfaction with the functioning of democracy—unlike being regularly informed through Facebook. Furthermore, the significant effect of regularly consuming television disappeared from this second model.

Conclusion

The study argues that the growing importance of Facebook in the political information environment results in stronger social influences on political behavior. Unlike previous research, this article does not investigate the direct, short-time effects manifested in concrete action, but assumed that influences take place on the level of political attitudes, as well as in shaping longtime political perceptions. This assumption has been tested through the evaluation of the working of democracy based on the thesis that negativity is dominant in Facebook political communication. The results of the secondary analysis show that, for university students, Facebook is the primary source of political information. However, while many young people are informed through Facebook, only a minority of them share or post political content there. The study demonstrates that political posting and sharing are strongly influenced by dissatisfaction with democracy, and, accordingly, those who regularly obtain information about politics through Facebook also negatively perceive the working of democracy. Satisfaction with democracy is negatively affected by no other media than Facebook. This effect is likely to result from the fact that young people mostly see their dissatisfied peers’ negative political content and this experience shapes their perceptions of political reality; thereby, the results offer indirect evidence of the presence of social influences.

Overall, these findings point to the recurrent relevance of the two-step flow model of communication. This means that owing to the changing patterns of political information consumption, the political behavior of young people is significantly shaped by their peers communicating about politics. Moreover, the study detects an influence manifested in attitude levels through longtime shaping of political perceptions. This means that if we want to understand ordinary people’s political behavior, it is worth turning to politically engaged citizens’ communication. In this approach, the main question then is how and what these active people communicate and what are the factors that influence it.

The study certainly has several limitations. The most significant limitation emanates from the method of indirect demonstration. Although there are strong arguments that if the two sub-hypotheses are supported, it proves the general hypothesis, but this cannot be clearly stated because of the lack of direct evidence. Moreover, the indirect demonstration is based on an assumption which could not be empirically tested in this study. This assumption is that if dissatisfaction with the way democracy works significantly influences political posting and sharing, that also means these posts are generally negative or critical political contents. The secondary analysis did not allow applying direct demonstration, but the future research using first-hand data should attempt to prove directly the presence of social influences on political attitudes on social media. Another limitation is that the survey asked only whether students are informed through Facebook, but it does not specify which sources provide this political information. It may happen that political information causing dissatisfaction with democracy does not exclusively originate from citizens’ communication. However, the tested variable, satisfaction with the way democracy work, connects political information consumed by students to peers’ political communication, and the results offer support for the argument that being informed through Facebook is largely caused by peers’ communication.

The wording of some variables may also be problematic. The survey asked respondents about posting and sharing in social media in general, while the question regarding political information consumption pertains to Facebook in particular. While I indicated these differences in word use, the research design treated them as pertaining to the same thing. This decision may be justified by the fact that social media use of Hungarian young people is still dominated by Facebook, and all other platforms are rarely used (see Kitta, 2012). Another problem is that in case of most media the questionnaire asked about media type, while for online news portal specific outlets were asked about. Although it had advantages in the second model pair, this constraint should be considered in understanding the “online news portal” variable. Some respondents may have been placed into the “never” or “sometimes” category, while actually they actively follow online news portals different from the four listed. The future research using first-hand data should craft better formulated questions which are more suitable to investigate social influences on political attitudes on social media. Also, we do not know what role the specific context plays in these results. The relationship between being informed through Facebook and dissatisfaction with democracy has been settled in other contexts as well (see Ceron & Memoli, 2016), but dissatisfaction with democracy and Facebook communication has not been connected before. The future investigations in different contexts are needed to assess the context-boundedness of the results.

There are several possible future directions based on this study. First, it would be important to know what exactly it means that young people’s political behavior is shaped by peers’ political communication on Facebook. This study has highlighted only one aspect, namely, political negativity, but there may be many other characteristics of citizens’ political communication on Facebook that might affect young people’s political behavior. I hope that the results discussed above underline that examining the content of young people’s political communication is very much needed. Furthermore, we should raise the question how Facebook is used as a source of political information. It can be assumed that most political content is generally spotted by accident, usually in a context dominated by personal content. How is political information consumed in this context? This political content may exert an influence on users’ political behavior by not being extensively processed. Political information may be obtained from reading just titles, headlines or short notes, or from no more than a quick impression, and not only from reading a long text. That is why it should also be investigated how exactly Facebook political information is processed.