Of Echo Chambers and Contrarian Clubs: Exposure to Political Disagreement Among German and Italian Users of Twitter

Abstract

Scholars have debated whether social media platforms, by allowing users to select the information to which they are exposed, may lead people to isolate themselves from viewpoints with which they disagree, thereby serving as political “echo chambers.” We investigate hypotheses concerning the circumstances under which Twitter users who communicate about elections would engage with (a) supportive, (b) oppositional, and (c) mixed political networks. Based on online surveys of representative samples of Italian and German individuals who posted at least one Twitter message about elections in 2013, we find substantial differences in the extent to which social media facilitates exposure to similar versus dissimilar political views. Our results suggest that exposure to supportive, oppositional, or mixed political networks on social media can be explained by broader patterns of political conversation (i.e., structure of offline networks) and specific habits in the political use of social media (i.e., the intensity of political discussion). These findings suggest that disagreement persists on social media even when ideological homophily is the modal outcome, and that scholars should pay more attention to specific situational and dispositional factors when evaluating the implications of social media for political communication.

Keywords

social media political discussion political homophily political disagreement political networks

Introduction

Democracy is founded on freedom of public opinion (Manin, 1997), and for opinions to be freely formed, citizens need diverse but reliable sources of information (Dahl, 1998). At the same time, the likelihood that individuals encounter diverse and reliable viewpoints depends on their informational environments (Prior, 2007). Whereas research suggests that the mass media are more likely to expose individuals to diverse information in comparison with face-to-face discussions, scholars have debated whether digital media, because of their choice-enhancing affordances, are more conducive to self-segregation. In this article, we investigate the role of social media in exposing individuals to different viewpoints on the basis of unique representative online surveys of Twitter users who posted campaign-related messages during the general elections of 2013 in Germany and Italy.

We demonstrate that the Twitter users we sampled were more likely to employ social media to engage with networks that supported rather than challenged their views, but that disagreement persisted on social media even when homophily was the modal outcome. The more individuals exchanged election-related messages on social media, the more likely they were to be part of networks that supported their views. At the same time, the ideological composition of respondents’ online networks closely reflected their face-to-face networks, so individuals encountering oppositional and mixed political viewpoints offline tended to have similar experiences on social media. We also show that, when online and offline experiences with agreement or disagreement do not perfectly overlap, political networks on social media are more likely to add to, rather than detract from, the overall diversity of political viewpoints that individuals encounter.

Taken together, our findings suggest that political homophily on social media is not a universal outcome that all users experience to an equivalent degree. On the contrary, it is a condition that citizens experience with different degrees of intensity depending on their broader patterns of political conversation and their specific habits in the political use of social media. In particular, individuals experiencing homophily in their offline discussion networks, and those who are more engaged in the exchange of political messages on social media, are more likely than others to encounter echo chambers on these platforms.

Antecedents to the Experience of Political Homophily on Social Media

In this study, we focus on two main themes: the extent to which social media foster exposure to political agreement or disagreement and the extent to which, and reasons why, individuals vary in their experience of political agreement and disagreement on social media. In particular, we assess whether social media users are prone to engage with three different types of political networks, as defined by Nir (2011): “supportive” networks, which primarily expose individuals to viewpoints with which they agree; “oppositional” networks that primarily confront members with viewpoints with which they disagree; and “mixed” networks that feature both congruent and divergent positions.¹

Political Agreement and Disagreement on Social Media

In contemporary media environments, citizens acquire and integrate political information received through interpersonal communication, the mass media, and digital media. Prior research suggests that these channels are not equally likely to expose individuals to diverse viewpoints. Studies of interpersonal communication show that individuals gravitate toward others who share their viewpoints (Huckfeldt & Sprague, 1995). This pattern has been understood within the rational choice paradigm as a strategy to acquire low-cost information from reliable sources (Downs, 1957) and within experimental social psychology as the result of individuals’ desire for belief confirmation (Festinger, 1957; Sears & Freedman, 1967). However, Huckfeldt, Johnson, and Sprague (2004) demonstrate that disagreement still persists, especially in low-density networks in which individuals interact with “weak ties,” that is, relatively distant acquaintances who are more likely to differ from them, in comparison with close interaction partners. Huckfeldt et al. (2004, pp. 21-23) also note that individuals choose with whom they talk not only on the basis of political homogeneity but also in terms of commonalities in lifestyles, hobbies, and family life.

The literature on selective exposure highlights that individuals, if given the opportunity, tend to choose media content that matches their political preferences (Garrett, 2009; Lazarsfeld, Berelson, & Gaudet, 1944) and that the mass media provide more exposure to contradictory viewpoints when compared with interpersonal conversations (Mutz & Martin, 2001). However, Iyengar and Hahn (2009) have shown that, in a high-choice media environment, citizens can craft personalized news diets that are consistent with their political views.

With respect to digital media, three different lines of research can be identified within the literature: (a) studies showing how the affordances of the Internet, by enhancing opportunities for the selection of sources, facilitate ideological self-segregation; (b) studies showing this type of selectivity, while resulting in increased exposure to consonant contents, does not necessarily lead to avoidance of dissonant ones; and (c) studies contending that inadvertent exposure to political content on the web might act as a counter-balancing mechanism increasing exposure to political disagreement.

The first strand of research emphasizes that the Internet’s selective nature—the fact that it allows, and to some degree compels, users to make frequent choices among a variety of sources and content (Bimber & Davis, 2003)—leads most individuals to engage primarily with views similar to their own. The general argument is that the Internet functions as an “echo chamber” in which individuals are exposed more or less exclusively to consonant views, and it is supported by some empirical research. Gaines and Mondak (2009) found that students in a large American university clustered on Facebook according to their ideological proclivities. Bakshy, Messing, and Adamic (2015) studied 10 million Facebook users in the United States who declared their ideological preferences and showed that—even taking into account the fact that Facebook’s algorithmic ranking of news according to users’ preferences limits the diversity of content they are exposed to—individuals are more likely to engage with stories that are consistent with their viewpoints. Studies analyzing large quantities of behavioral data likewise suggest that political networks on Twitter exhibit high levels of homophily (Barberá, 2015; Colleoni, Rozza, & Arvidsson, 2014; Conover et al., 2011). At the same time, Barberá, Jost, Nagler, Tucker, and Bonneau (2015) observed that the degree of homophily in information sharing on Twitter varied to a considerable extent according to users’ ideology, context, and issue type.

The second group of studies suggests that the choice affordances of digital media are less likely to produce self-segregation, such as the filtering out of dissonant viewpoints, in comparison with traditional forms of media. Based on extensive research, Holbert, Garrett, and Gleason concluded that the extent to which people deliberately avoid attitude-discrepant information online had been exaggerated by previous research. To the contrary, they argue, “individuals exhibit a stronger bias toward attitude-consistent information than against attitude-discrepant information” (Holbert, Garrett, & Gleason, 2010, p. 19). In other words, political interactions on digital media may entail increased exposure to viewpoints users agree with, but not necessarily an equivalent avoidance of contrary ideas, and thus do not necessarily result in self-segregation.

Finally, according to the third line of research, even if individuals select online content and sources based on their political inclinations, digital media also facilitate exposure to different viewpoints, perhaps inadvertently. Given that individuals are more likely to come across attitude-discordant political content (often unintentionally) in non-political online environments than in overtly political spaces (Wojcieszak & Mutz, 2009), social media—insofar as it is often used for non-political reasons—may enable serendipitous exposure to views that are quite different from one’s own. Gil de Zúñiga and Valenzuela (2011) show that social media facilitates fortuitous contact with weak ties, which is likely to expose individuals to different political views. With respect to Twitter, Colleoni et al. (2014) suggest that the connections users build between each other exhibit high degrees of homophily, but information can also circulate across different networks and, in the process, expose individuals to dissonant viewpoints (see also Barberá et al., 2015).

Given that most of the literature suggests that individuals are more likely to seek agreement than disagreement on social media, we expect political homophily to be prevalent on social media. Thus, we hypothesize that in general individuals are more likely to engage with supportive than oppositional or mixed networks on social media (H1).

Nevertheless, the literature also provides some evidence that, under certain circumstances and for certain kinds of users, political networks on social media may differ from the modal outcome of homogeneity. However, research on what these factors are and under what conditions they operate has been very limited so far. Although most studies assume that conversations on social media are characterized by some degree of user control, scholars have generally failed to study how different users employ control affordances, and what implications such choices may have. This is because most scholars have treated important aspects of social media usage as constants rather than variables. To address this gap in the literature, our next hypotheses move beyond a general assessment to address how different individual-level characteristics may explain variation in exposure to supportive versus contrarian viewpoints on social media.

The Nexus Between Online and Offline Patterns of Political Discussion

We proceed from the premise that the role of social media in fostering or forestalling exposure to a diversity of political opinions cannot be understood without considering the specific goals and circumstances of different individuals’ social media activity as well as their broader patterns of political conversation. This is because (a) social media are fundamentally intertwined with offline dynamics and (b) social media are high-choice environments in which individuals to some extent choose their levels of engagement with politics and the kinds of contents and sources they encounter.

Political networks on social media do not exist in a vacuum; they are part of a broader ecosystem of information flows in which individuals play different roles and exercise different degrees of agency. According to Chadwick (2013), contemporary political communication systems must be understood as “hybrid,” that is, combining different logics from older and newer media as well as integrating face-to-face and digital modes of engagement. This model implies that we cannot consider social media as separate from, or independent of, face-to-face conversational contexts—as most prior research has done—and that if we are to understand the role of social media in facilitating encounters with viewpoints that are dissonant versus consonant, we must take into account individuals’ offline discussion networks as well. Accordingly, research shows that individuals often use social media to connect (and reconnect) with members of their extended offline social networks (Ellison, Steinfield, & Lampe, 2007; Subrahmanyam, Reich, Waechter, & Espinoza, 2008). It is thus reasonable to expect that individuals’ patterns of offline conversation may be largely reproduced on social media, especially insofar as online interactions involve the same partners as face-to-face encounters.

Moreover, investigating the nexus between online and offline networks of political discussion helps to illuminate the various ways in which individuals approach political discussions in general. The “uses and gratifications” theory contends that individuals take advantage of the affordances of any medium of communication to fulfill their needs and preferences (Campbell & Kwak, 2010; Cho, Gil de Zúñiga, Rojas, & Shah, 2003). Taking offline networks into account when assessing online networks not only allows researchers to compare communication experiences in two different domains and platforms, it also facilitates an understanding of how given individuals differ in their overall approaches to political discussion writ large.

Finally, the extent to which individuals discuss politics—and the networks they develop—may be a function of their psychological characteristics, among other things. For instance, it is well known that individuals differ considerably in the extent to which they value openness to new experiences—as opposed to the preservation of what is familiar and traditional—and the extent to which they exhibit open-mindedness in the context of opinion exchange (e.g., Kruglanski, Webster, & Klem, 1993; McCrae, 1996; Rokeach, 1960; Schwartz, 2012). There is some evidence that these individual differences manifest themselves in terms of online behavior: people who score higher on openness tend to have larger and more diverse social media contacts and networks (Bachrach, Kosinski, Graepel, Kohli, & Stillwell, 2012; Gosling, Augustine, Vazire, Holtzman, & Gaddis, 2011).

The hybridization of political communication, the uses and gratifications theory, and research on personality and social psychology all suggest that some degree of concordance is to be expected when it comes to the degree of ideological homophily in individuals’ face-to-face and social media networks. Thus, we hypothesize that the political composition of discussion networks that individuals engage with on social media is similar to the composition of their offline networks (H2). Understanding the connections between face-to-face and online environments is also critical to assessing whether and under what circumstances the use of social media increases or decreases the overall diversity of political information to which one is exposed.

Political Habits on Social Media and Experience With Agreement and Disagreement

Once these individual circumstances and dispositions have been taken into account, we also need to consider how specific habits in the political use of social media may affect the likelihood that users encounter either agreement or disagreement on these platforms. In a high-choice media environment, people are likely to be selective when it comes to political content they are exposed to and with whom they discuss such content, and this selectivity is likely to be greater for those who are high (vs. low) in the intensity with which they engage with politics on social media. This is because the sheer abundance of potential content and interaction partners makes it not only possible, but also necessary, for individuals who are highly engaged in politics to filter out truly divergent perspectives. The more an individual is exposed to political content and engages in political discussion, the more rational it is for him or her to filter messages in order to maximize utility while minimizing effort. Selecting predominantly like-minded sources and conversational partners is clearly one way of attaining efficiency in this regard. Moreover, the intensity of political discussion is likely to be a reliable indicator of the strength of political identification and discursive involvement.

The theory of motivated reasoning suggests that people tend to search for information that reinforces their preexisting opinions and to avoid information that challenges them (Lodge & Taber, 2000; Stroud, 2008). The more involved in politics a given individual is, the more likely it is that he or she wishes to be part of an ideological community and, thus, to exercise selectivity in the context of social networks. Lawrence, Sides, and Farrell (2010) noted that most U.S. political blog readers—who are generally quite politically engaged—gravitate toward blogs that reinforce their own ideological inclinations, whereas very few read blogs across the entire ideological spectrum. Those who are highly involved in political discussions on social media may also be more likely to engage with like-minded others. Therefore, we hypothesize that the quantity of political messages that individuals exchange on social media is positively associated with the likelihood of engaging with supportive political networks on these platforms (H3).

Case Selection

Most studies of online discussion networks have focused on the United States. This means that we cannot be sure whether findings from these studies can be generalized to other Western democracies. We address this omission by focusing on Germany and Italy—two large, relatively affluent European democracies that held general elections in 2013 and possess similar levels of Twitter diffusion (9% in Italy, 7% in Germany).² Unlike the United States, both Italy and Germany are parliamentary multiparty systems with mixed, but predominantly proportional, electoral laws. Although they differ from each other in some key respects—such as mean levels of political trust and the relative stability of party systems and governments—we do not expect these differences to bear on the structure and function of online political networks. It should be noted, given the goal of this research, that in multiparty systems it is often the case that parties pursue niche, bonding strategies aimed at mobilizing relatively narrow segments of the population (Norris, 2004, pp. 100-101). As a result, European citizens’ ideological preferences may be more fragmented and multidimensional in comparison with citizens of majoritarian systems such as the United States.

In testing our hypotheses in Germany and Italy, we want to assess the robustness of our findings across different national systems and to expand existing knowledge beyond singular case studies of the United States.

Data

Investigating political discussions on social media requires that we focus on those individuals who take part in such discussions rather than on the entire voting-age population or even social media users in general. We focus on Twitter because it is one of the most popular social media platforms worldwide and its structure makes it especially germane to our hypotheses, insofar as it facilitates serendipitous encounters with unanticipated information and is highly accessible to study, because most of the interactions can be retrieved and archived. Testing our hypotheses requires valid and reliable measures of the political activities performed by social media users, what motivates them, and what kinds of information they encounter. We, therefore, devised unique surveys of representative samples of individuals who engaged in election-related conversations on Twitter in Germany and Italy. Compared with analyses of behavioral data of individuals’ interactions on social media, surveying representative samples of these users allows us to measure constructs, such as characteristics of face-to-face interactions, which may not be observable on the basis of social media activities alone. The potential downside is that our results could be an artifact of the extent that respondents systematically misreport characteristics of online and offline networks.³ Moreover, as with all cross-sectional surveys, our data are subject to some degree of endogeneity and, therefore, do not support causal interpretations of claims regarding the associations we observe.

Sampling Political Users of Twitter

Because no comprehensive list of Twitter users—let alone, Twitter users who discuss elections—is publicly available, we devised a strategy to construct sampling frames that are as inclusive as possible with respect to our populations. Because most of the sources and messages posted by Twitter users are publicly accessible,⁴ we rely on the contents of these messages to identify our populations, which we define as those Twitter users who posted messages concerning the German and Italian elections of 2013. We pinpoint these users on the basis of election-related keywords—the names of the main parties, their leaders, and the topical hashtags for the elections⁵—contained in the messages they posted. We used these keywords to query Twitter’s Streaming API during each country’s extended campaign season⁶ and retrieved about 5.8 million tweets from over 151,000 unique users in Germany and 3 million tweets from over 275,000 unique users in Italy.⁷

Fielding a Survey of Twitter Users

From these lists of users, we randomly selected 43,000 users in Germany and 35,000 in Italy and contacted them via Twitter through an automated script that delivered a personalized message as follows: “@[username] University research on social media use: Would you like to participate? [link to the survey].”⁸ Because our invitations were delivered in such a way that recipients were asked to share their opinions with strangers via social media, respondents to our surveys may differ from those who refused to answer in terms of their willingness to express their opinions to strangers. People who are more open to engaging with strangers on social media might also be more likely to encounter disagreeing opinions on these platforms. As a result, it is possible that our sample may overestimate exposure to political disagreement online. Although we acknowledge this potential bias, we emphasize that it works against our first and third hypotheses (see above), while it does not affect our ability to validly test our second hypothesis. In sum, we have no reason to believe that our findings are an artifact of the method we chose to contact our respondents.

A total of 1,143 (Germany) and 1,493 (Italy) individuals answered at least half of the questions, which corresponds to response rates of 4%.⁹ Because these are by no means high figures—although they are not much lower than the single-digit response rates that are common in telephone surveys¹⁰—in Appendices B and C¹¹ we illustrate evidence suggesting that our respondents may be considered representative of Germans and Italians who discussed the 2013 elections on Twitter, and that the differences that could be measured between these two groups were taken into account in our analysis.¹²

Political Users of Twitter Versus General Population Samples

Understanding the behaviors of representative samples of Twitter users who commented on their countries’ elections enables us to understand political communication on social media platforms and the factors that shape it (see Bekafigo & McBride, 2013 and Bode & Dalrymple, 2014, for other representative surveys of Twitter users). At the same time, focusing on the specific population of individuals who posted at least one election-related tweet does not allow us to generalize to other populations, such as citizens who read (but do not post) political messages on Twitter. To the extent that our survey respondents were more engaged in politics than the general population, we would expect—on the basis of prior research (e.g., Mutz, 2006)—that they would be less eager to engage with contrary views; we take this possibility into account by controlling for relevant political attitudes in our analyses.¹³

Variables

Although our respondents were recruited via Twitter, our primary independent and dependent measures focus on general social media use because individuals’ online interactions, political and otherwise, are not limited to one platform, but often integrate many of them: For instance, a person can use Twitter to share a YouTube video, a Facebook status update, or an Instagram picture. When we asked respondents to indicate social networking platforms on which they had a profile, the median respondent had profiles in 4 of 10 platforms we asked about in Italy and 5 of 10 in Germany.

Our dependent variables measure the types of political networks that individuals engage with on social media, in response to two questions: “How often do you [agree/disagree] with the political opinions and contents that other people post on social media?” Respondents could answer with one of four categories:

Always or nearly always (4% of German and 0.6% of Italian respondents for agreement; 1% of Germans and 0.8% of Italians for disagreement);

Often (43.1% of German and 42.4% of Italian respondents for agreement; 18.1% of Germans and 21.8% of Italians for disagreement);

Only sometimes (50.3% of German and 55.9% of Italian respondents for agreement; 77.8% of Germans and 76.3% of Italians for disagreement);

Never (2.6% of German and 1.1% of Italian respondents for agreement; 3.1% of Germans and 1.1% of Italians for disagreement).

As Nir (2011) showed, mixed political networks have important implications for political engagement, and should thus not be treated simply as an intermediate category between supportive and oppositional ones. This is why we use as dependent variables dichotomous measures representing each of these types of networks rather than a combined ordinal measure of the continuum of supportive, mixed, and oppositional networks. Because only about 5% of respondents used the most extreme categories, we combined the first two and the last two response categories and then constructed a composite measure corresponding to four types of network structures:

Respondents who claimed to always or often encounter agreement and rarely or never disagreement were classified as participating in supportive networks;

Respondents who claimed to always or often encounter disagreement and rarely or never agreement were classified as participating in oppositional networks;

Respondents who claimed to always or often encounter both agreement and disagreement were classified as participating in mixed networks,

Respondents who claimed to rarely or never encounter agreement or disagreement were classified as participating in neutral networks.¹⁴

The independent variables required to test our hypotheses involve the characteristics of respondents’ offline political networks (H2) and the proportion of political messages they exchange on social media (H3).

We test H2 by focusing on responses to the questions “How often do you usually [agree/disagree] with the opinions of people with whom you talk about politics?” These questions immediately followed a specific question about how frequently respondents discussed politics in face-to-face contacts with friends and family. The response modes were the same as for the questions about social media, and we derived combined measures of supportive (36.5% of German and 34.8% of Italian respondents), oppositional (16.6% of German and 25.2% of Italian respondents), mixed (8.6% of German and 8.2% of Italian respondents), and neutral (38.3% of German and 31.8% of Italian respondents) offline networks in the same ways as we did for social media networks.

We test H3 by averaging answers to two separate questions, one for posting and one for reading political messages. The questions were: “Thinking about everything you have recently [posted/read from people you follow or are in contact with] on social media, such as status updates, comments, or links to news stories—about how much is related to politics, political issues or the 2013 elections?”

Respondents could answer both questions by indicating a number between 0 (none) and 10 (all).¹⁵ The resulting variable averaged 4.46 (SD = 2.58) for the German and 4.46 (SD = 2.27) for the Italian sample.

Findings

Bearing on H1, Figure 1 shows the percentages of German and Italian respondents who engage with supportive, oppositional, mixed, and neutral political networks on social media. Respondents are substantially more likely to engage with supportive rather than oppositional networks (40.6% vs. 12.5% in Germany and 35.8% vs. 15.3% in Italy), and in both cases, the differences are statistically significant.¹⁶ Participation in networks that exhibit disagreement with respondents’ opinions is not, however, infrequent: In both countries, oppositional and mixed networks (combined) affect one in five respondents—one in three if we exclude those in “neutral” networks that are disengaged from politics. These findings support H1, insofar as respondents on average encounter more agreement than disagreement, but they also suggest that for some people, social media platforms are not “echo chambers” of univocal agreement, but “contrarian clubs” where political disagreement is common. Another interesting finding is that about two in five of our respondents participate in networks in which no particular political opinions emerge, and this group is the modal one in the Italian sample. Thus, even among those social media users who communicated at least once about the election, exposure to very few political opinions is approximately as likely as exposure to attitude-congruent opinions.

Figure 1.

Engagement with different types of political networks on social media (percentages among country respondents).

We address our remaining hypotheses by conducting three logistic regressions (summarized in Table 1)¹⁷ that, in each country, predict whether respondents report being part of supportive, oppositional, or mixed political networks on social media as a function of the characteristics of their offline political networks (H2), and the proportion of political messages they exchange on social media (H3).

Table 1.

Dependent Variable(s): Types of Political Networks Respondents Engage With on Social Media.

	Germany			Italy
	Supportive	Oppositional	Mixed	Supportive	Oppositional	Mixed
Face-to-face political networks (ref. neutral)
Supportive	0.818**	−0.201	0.832	0.998***	0.183	0.580
Oppositional	−0.381	0.899*	1.200*	0.040	1.756***	0.228
Mixed	−0.365	0.489	1.696**	0.032	1.082**	2.072***
Ratio of political messages on the total exchanged on social media	0.153***	−0.041	0.149	0.120**	−0.058	0.126
Preferred use of social media	1.240**	−1.032	0.224	1.714***	−1.056*	−0.675
Source of political information
Internet	0.078	1.023	−0.386	0.171	−0.488	−0.983
Newspapers	−0.460	0.360	−0.167	−0.690*	0.060	−0.433
Radio	−1.046***	0.612	0.562	−0.508*	0.005	0.815
Television	−0.406	−0.264	−0.008	0.078	−0.068	0.911
Political efficacy (disagreement with following sentences)
“People don’t have any say”	0.108	−0.201	−0.581	0.042	0.469	−0.048
“Public officials don’t care”	0.167	−0.375	0.849	0.338	−0.239	−0.886
“Politics is too complicated”	−0.660	−0.039	0.799	−0.098	0.000	0.311
Interest in politics	1.601**	−0.923	0.846	−0.141	0.805	1.969
Trust in political parties	−0.276	−0.641	0.597	0.379	−0.386	−0.768
Offline political discussion (frequency)	0.079	0.661	−0.118	0.257	−1.014	−0.587
Gender (male)	−0.128	−0.227	0.367	−0.211	0.041	0.212
Age	−0.712	−1.051	−0.401	−0.561	−0.347	−0.798
Education	−0.431	1.025	−1.042	0.361	0.290	0.614
Income	−0.582	1.421	0.812	−0.094	0.227	−0.528
Constant	−0.667	−3.087**	−5.569***	−1.348**	−1.893**	−4.822***
N	727	727	727	1,167	1,167	1,167
Percentage correctly predicted	69.9	86.5	93.4	68.7	85.1	93.4
Nagelkerke R²	.241	.125	.141	.164	.157	.143
Log-likelihood	834.045	518.951	312.860	1,391.319	889.100	503.732

Note. Cell entries are estimated logit coefficients where the dependent variable is 1 for the reported network type, and all other network types are coded as 0. See Appendices E to J for complete results with standard errors. Dummy variable identifying missing observations for income omitted from table, see note 18. All variables range from 0 to 1 apart from political messages exchanged (0-10) and preferred use of social media (−1 to 1, with respondents who claimed that agreement was more important than disagreement having positive values, whereas those who stated that disagreement was more important than agreement having negative values, and those who attributed equal importance to agreement and disagreement having a score of zero).

***

p ⩽ .001. **p ⩽ .01. *p ⩽ .05.

The models include control variables for socio-demographic characteristics,¹⁸ political attitudes (political efficacy, interest in politics, and trust in political parties), frequency of offline political discussion, and frequency of use of different media to get political information. We also control for respondents’ preferred use of social media, distinguishing between those who claimed to consider social media to be more important for finding people with similar (vs. different) viewpoints, in comparison with their own.¹⁹ We introduced this variable in accordance with the “uses and gratification” theory, which contends that individuals’ preferences may shape the type of conversational experiences they have on social media. Controlling for respondents’ preferred use of social media provides a more precise assessment of the specific role played by broader patterns of political conversation as well as habits in the political use of social media when it comes to the development of citizens’ online networks.²⁰

As can be seen from the first block of coefficients in Table 1, H2 is supported in both countries, insofar as we see positive and significant associations between all relevant pairs of independent and dependent variables. As an example of the strength of these associations, if we construct a hypothetical Italian respondent that has values equal to the median (for interval and ordinal variables) or modal (for categorical variables) values in the sample, then the probability that this hypothetical (male) respondent would engage with supportive political networks on social media is 48% if he also engages with supportive offline networks, but only 26% if he engages with either mixed, oppositional, or neutral offline networks. The results also highlight some interesting differences between the two countries. While in Germany respondents who engage with oppositional networks offline are significantly more likely to encounter either oppositional or mixed networks online, in Italy engagement with offline oppositional networks is solely associated with oppositional networks on social media. By contrast, whereas in Germany those who engage with mixed offline networks are more likely to engage with the same types of networks on social media, in Italy those who are part of mixed offline networks are significantly more likely to interact with either oppositional or mixed networks on social media. Individuals’ online and offline experiences tend to overlap, but to the extent that they do not overlap perfectly, social media functions as an echo chamber only for those individuals who also possess homogeneous offline networks. Another way to interpret these associations is that political networks on social media are more likely to add to than detract from the overall diversity of political viewpoints to which individuals are exposed.²¹

The data also support H3, namely the expectation of a positive association between the intensity of online political involvement and the probability of participating in supportive networks.²² In both countries, the more individuals post and read political messages on social media, the more likely they are to engage with supportive networks. As an example, if we set all variables to their median or modal levels in the German sample, a hypothetical respondent has a 39% probability of engaging with a supportive network. If, however, the intensity of his or her activity is increased to one standard deviation above the median, the probability of engaging with supportive networks increases to 49%. Conversely, if the intensity of his or her activity is set one standard deviation below the median, the probability of engaging with supportive networks decreases to 30%.

Finally, the associations pertaining to the control variable measuring preferences for political agreement versus disagreement on social media deserve a brief comment. In both countries, respondents who attribute greater importance to social media for encountering agreement as opposed to disagreement (positive values of the variable) are significantly more likely to engage with supportive networks on these platforms. To the contrary, the association between such variable (where respondents valuing social media as more important for encountering disagreeing than agreeing others have negative values) and engagement with oppositional views is negative in both countries, but it is significant only in Italy. This pattern may suggest that, for individuals who approach social media deliberately in search of agreement (vs. disagreement), it may be relatively easy to obtain such agreement. Instead, for those preferring to seek out disagreement, it may be more difficult to locate and participate in “contrarian clubs.”

Conclusion

We have shown that German and Italian Twitter users who communicate about elections are more likely to do so in networks that support rather than challenge their views, consistent with the notion that social media facilitates the emergence of echo chambers. At the same time, contrarian clubs, which involve frequent encounters with dissonant opinions—whether in oppositional or mixed networks—are less exceptional than expected. We may have come across an important parallelism between studies of political communication offline and online: As noted by Huckfeldt et al. (2004) with respect to offline networks, heterogeneity persists on social media even though homogeneity is the modal outcome.

We have approached citizens’ experiences with political agreement and disagreement on social media through the theoretical lenses of hybridization in political communication and of “uses and gratifications” theory, while at the same time taking into account the importance of individual attitudes in high-choice media environments. This, in turn, led us to focus on aspects that are likely to differentiate individuals, as opposed to treating everyone as guided by technological affordances in the same way. Thus, we have been able to demonstrate that the extent to which social media functions as an echo chamber (as opposed to a contrarian club) varies across individuals. This, in turn, suggests that understanding political dynamics in choice-enhancing platforms may be better served by an appreciation that different users have different traits, preferences, and social networks that affect their behaviors and experiences rather than an assumption that most or all users employ the selective features of social media to pursue the same goals, thus leading to fairly predictable and monolithic outcomes.

More specifically, we hope to have shed light on broader dynamics of political communication as well as specific habits pertaining to the political use of social media that help to explain how various platforms give rise to different types of political networks. Our observation that online and offline networks tend to resemble one another suggests that understanding the dynamics of political communication requires a holistic approach that encompasses both contexts. As increasing numbers of citizens rely on social media for political information, which they often encounter by discussing public affairs with others, the overall diversity of viewpoints in contemporary democracies is not likely to be dramatically reduced when compared with face-to-face discussions; in some cases, it may even be broadened. However, the use of social media seems to diminish political diversity for those who participate in more or less entirely supportive offline networks and who prefer engaging with people with whom they tend to agree.

At the level of individual behavior online, our finding that the more people post and read political messages on social media, the more likely they are to encounter supportive networks indicates that, all else being equal, the greatest proportion of social media messages exchanged involve interactions among individuals who agree with one another. This highlights a crucial methodological issue in the study of online political communication. To the extent that the quantity of messages and interactions that can be observed on social media is associated with the levels of agreement among individuals who take part in these exchanges, scholars interested in the implications of social media for political diversity should be aware that taking messages or connections as their unit of analysis may overestimate the pervasiveness of homogeneity as actually experienced by individuals.

Finally, our study has confirmed the centrality of hybridity in contemporary environments of political discussion. Building on Chadwick’s (2013) theorizing, we have investigated the relationship between newer (i.e., social media) and older (i.e., face-to-face) networks of political discussion, and have observed that these two types of environments—and their underlying logics—are integrated rather than separated. Our findings also have important implications for power—a crucial component of Chadwick’s theory—insofar as they suggest that politically active citizens will be unlikely to find much challenging content on social media, but they may be able to reach less engaged users—who according to our findings are less likely to be part of exclusively homophilic networks—with oppositional points of view. Under certain conditions, these interactions could create opportunities for political persuasion and, thus, the possibility for some to exercise influence over others.

Footnotes

Appendix A

Appendix B

Appendix C

Appendix D

Appendix E

Appendix F

Appendix G

Appendix H

Appendix I

Appendix J

Appendix K

Appendix L

Appendix M

Appendix N

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: We acknowledge the support of the Italian Ministry of Education “Future in Research 2012” initiative (project code RBFR12BKZH) and the INSPIRE program of the National Science Foundation (Awards # SES-1248077 and # SES-1248077-001) as well as New York University’s Global Institute for Advanced Study (GIAS) and Dean Thomas Carew’s Research Investment Fund (RIF). Pablo Barberá gratefully acknowledges financial support from the Gordon and Betty Moore Foundation and the Alfred P. Sloan Foundation. We thank Duncan Penfold-Brown, Jonathan Ronen, and Yvan Scher for programming assistance. This article is the product of a collaboration between the research projects “Building Inclusive Societies and a Global Europe Online”(http://www.webpoleu.net) at the University of Bologna and “Social Media and Political Participation” () at New York University. In accordance with Italian academic conventions, we specify that Cristian Vaccari wrote the paragraphs titled “Antecedents to the Experience of Political Homophily on Social Media” and “Findings”; Augusto Valeriani wrote the paragraphs titled “Case Selection” and “Variables” and the Appendixes; all the authors collaborated in the design of the study, editing the text of the article, and in writing the paragraphs titled “Introduction,” “Data,” and “Conclusion”.

Notes

Author Biographies

Cristian Vaccari (PhD, IULM University in Milan) is a Reader in Politics at Royal Holloway, University of London, and an Associate Professor in Political Science at the University of Bologna. His research interests include political communication and political engagement in comparative perspective.

Augusto Valeriani (PhD, University of Siena) is an Assistant Professor in Media Sociology at the University of Bologna. His research interests include digital media and politics, both in national and international contexts.

Pablo Barberá (PhD, New York University) is an Assistant Professor of International Relations at the University of Southern California. His research interests include social media and politics, quantitative political methodology, and electoral behavior and political representation.

John T. Jost (PhD, Yale University) is a Professor of Psychology and Politics at New York University. His research interests include stereotyping, prejudice, political ideology, social justice, and the theory of system justification.

Jonathan Nagler (Phd, California Institute of Technology) is a Professor of Politics at New York University. His research interests include political behavior, economics and elections, political methodology, and social media.

Joshua A. Tucker (PhD, Harvard University) is a Professor of Politics and an affiliated Professor of Russian and Slavic Studies and Data Science at New York University. His research interests include elections and voting, partisanship, public opinion formation, protest, post-communist politics, and social media and politics.

References

Bachrach

Kosinski

Graepel

Kohli

Stillwell

(2012). Personality and patterns of facebook usage. Retrieved from http://research.microsoft.com/pubs/163535/FacebookPersonality_michal_29_04_12.pdf

Bakshy

Messing

Adamic

(2015). Exposure to ideologically diverse news and opinion on Facebook. Science. Advance online publication. doi:10.1126/science.aaa1160

Barberá

(2015). Birds of the same feather tweet together: Bayesian ideal point estimation using Twitter data. Political Analysis, 23, 76–91.

Barberá

Jost

J. T.

Nagler

Tucker

Bonneau

(2015). Tweeting from left to right: Is online political communication more than an echo chamber? Psychological Science, 26, 1531–1542.

Bekafigo

M. A.

McBride

(2013). Who tweets about politics? Political participation of Twitter users during the 2011 gubernatorial elections. Social Science Computer Review, 31, 625–643.

Bimber

B. A.

Davis

(2003). Campaigning online: The Internet in US elections. Oxford, UK: Oxford University Press.

Bird

Klein

Loper

(2009). Natural language processing with Python. Sebastopol, CA: O’Reilly.

Bode

Dalrymple

(2014). Politics in 140 characters or less: Campaign communication, network interaction, and political participation on Twitter. Journal of Political Marketing. Advance online publication. doi:10.1080/15377857.2014.959686

Campbell

S. W.

Kwak

(2010). Mobile communication and civic life: Linking patterns of use to civic and political engagement. Journal of Communication, 60, 536–555.

10.

Chadwick

(2013). The hybrid media system: Politics and power. New York, NY: Oxford University Press.

11.

Cho

Gil

Zúñiga

Rojas

Shah

D. V.

(2003). Beyond access: The digital divide and Internet uses and gratifications. IT & Society, 1, 46–72.

12.

Colleoni

Rozza

Arvidsson

(2014). Echo chamber or public sphere? Predicting political orientation and measuring political homophily in twitter using big data. Journal of Communication, 64, 317–332.

13.

Conover

Ratkiewicz

Francisco

Gonçalves

Menczer

Flammini

(2011). Political polarization on Twitter. Proceedings of the ICWSM, Barcelona, Spain, 17–21 July.

14.

Dahl

(1998). On democracy. New Haven, CT: Yale University Press.

15.

Downs

(1957). An economic theory of democracy. New York, NY: HarperCollins.

16.

Ellison

N. B.

Steinfield

Lampe

(2007). The benefits of Facebook “friends:” Social capital and college students’ use of online social network sites. Journal of Computer-Mediated Communication, 12, 1143–1168.

17.

Festinger

(1957). A theory of cognitive dissonance. Stanford, CA: Stanford University Press.

18.

Gaines

B. J.

Mondak

J. J.

(2009). Typing together? Clustering of ideological types in online social networks. Journal of Information Technology & Politics, 6, 216–231.

19.

Garrett

R. K.

(2009). Politically motivated reinforcement seeking: Reframing the selective exposure debate. Journal of Communication, 59, 676–699.

20.

Gellman

Hill

(2007). Data analysis using regression and multilevel/hierarchical models. New York, NY: Cambridge University Press.

21.

Gil

Zúñiga

Valenzuela

(2011). The mediating path to a stronger citizenship: Online and offline networks, weak ties and civic engagement. Communication Research, 38, 397–421.

22.

Gosling

S. D.

Augustine

A. A.

Vazire

Holtzman

Gaddis

(2011). Manifestations of personality in online social networks: Self-reported facebook-related behaviors and observable profile information. Cyberpsychology, Behavior, and Social Networking, 14, 483–488.

23.

Holbert

R. L.

Garrett

R. K.

Gleason

L. S.

(2010). A new era of minimal effects? A response to Bennett and Iyengar. Journal of Communication, 60, 15–34.

24.

Holmberg

(2007). Partisanship reconsidered. In Dalton

Klingemann

H. D.

(Eds.), The Oxford handbook of political behavior (pp. 557–570). Oxford, UK: Oxford University Press.

25.

Huckfeldt

Johnson

P. E.

Sprague

(2004). Political disagreement: The survival of diverse opinions within communication networks. New York, NY: Cambridge University Press.

26.

Huckfeldt

Sprague

(1995). Citizens, politics and social communication: Information and influence in an election campaign. New York, NY: Cambridge University Press.

27.

Iyengar

Hahn

K. S.

(2009). Red media, blue media: Evidence of ideological selectivity in media use. Journal of Communication, 59, 19–39.

28.

King

Honaker

Joseph

Scheve

(2001). Analyzing incomplete political science data: An alternative algorithm for multiple imputation. American Political Science Review, 95, 49–69.

29.

Kruglanski

A. W.

Webster

D. M.

Klem

(1993). Motivated resistance and openness to persuasion in the presence or absence of prior information. Journal of Personality and Social Psychology, 65, 861–876.

30.

Lawrence

Sides

Farrell

(2010). Self-segregation or deliberation? Blog readership, participation, and polarization in American politics. Perspectives on Politics, 8, 141–157.

31.

Lazarsfeld

P. F.

Berelson

Gaudet

(1944). The people’s choice. New York, NY: Columbia University Press.

32.

Liu

Kliman-Silver

Mislove

(2014). The tweets they are a-Changin’: Evolution of twitter users and behavior. Palo Alto, CA: AAAI.

33.

Lodge

Taber

(2000). Three steps toward a theory of motivated political reasoning. In Lupia

McCubbins

Popkin

(Eds.), Elements of reason: Cognition, choice, and the bounds of rationality (pp. 183–213). Cambridge, UK: Cambridge University Press.

34.

Manin

(1997). The principles of representative government. New York, NY: Cambridge University Press.

35.

McCrae

R. R.

(1996). Social consequences of experiential openness. Psychological Bulletin, 120, 323–337.

36.

Mutz

Martin

(2001). Facilitating communication across lines of political difference: The role of mass media. American Political Science Review, 95, 97–114.

37.

Mutz

D. C.

(2006). Hearing the other side: Deliberative vs. participatory democracy. New York, NY: Cambridge University Press.

38.

Nir

(2011). Disagreement and opposition in social networks: Does disagreement discourage turnout? Political Studies, 59, 674–692.

39.

Norris

(2004). Electoral engineering: Voting rules and political behavior. New York, NY: Cambridge University Press.

40.

Prior

(2007). Post-broadcast democracy: How media choice increases inequality in political involvement and polarizes elections. New York, NY: Cambridge University Press.

41.

Rokeach

(1960). The open and closed mind: Investigations into the nature of belief systems and personality systems. New York, NY: Basic Books.

42.

Schwartz

S. H.

(2012). An overview of the Schwartz theory of basic values. Online Readings in Psychology and Culture, 2, 1–11. Retrieved from http://scholarworks.gvsu.edu/cgi/viewcontent.cgi?article=1116&context=orpc

43.

Sears

D. O.

Freedman

J. L.

(1967). Selective exposure to information: A critical review. Public Opinion Quarterly, 31, 194–213.

44.

Stroud

N. J.

(2008). Media use and political predispositions: Revisiting the concept of selective exposure. Political Behavior, 30, 341–366.

45.

Subrahmanyam

Reich

S. M.

Waechter

Espinoza

(2008). Online and offline social networks: Use of social networking sites by emerging adults. Journal of Applied Developmental Psychology, 29, 420–433.

46.

Vaccari

Valeriani

Barberá

Bonneau

Jost

J. T.

Nagler

Tucker

J. A.

(2015). Political expression and action on social media: Exploring the relationship between lower-and higher-threshold political activities among twitter users in Italy. Journal of Computer-Mediated Communication, 20, 221–239.

47.

Wojcieszak

Mutz

(2009). Online groups and political discourse: Do online discussion spaces facilitate exposure to political disagreement? Journal of Communication, 59, 40–56.