A Media Repertoires Approach to Selective Exposure: News Consumption and Political Polarization in Eastern Europe

In recent years, links between selective news exposure and political polarisation have attracted considerable attention among communication scholars. However, while the existence of selective exposure has been documented in both offline and online environments, the evidence of its extent and its impact on political polarisation is far from unanimous. To address these questions, and also to bridge methodological and geographical gaps in existing research, this paper adopts a media repertoires approach to investigate selective news exposure and polarisation in four Eastern European countries – the Czech Republic, Hungary, Poland and Serbia. Using a combination of population surveys, expert surveys and qualitative interviews, the data for the study were collected between November 2019 and May 2020. We identify five types of news repertoires based on their relative openness to counter-attitudinal sources, and show that selective news repertoires are present in 29% of the entire sample. Our findings also reveal significant cross-country differences, with the more selective news repertoires more prominent in countries characterised by higher levels of polarisation. Furthermore, while the selection of news sources is in line with people's electoral (and to a lesser extent ideological) preferences, our findings show that exposure to counter-attitudinal sources can also be strongly correlated with political and ideological leanings. Our qualitative data suggest that this is because exposure to counter-attitudinal sources can reinforce attitudes, and potentially contribute to polarisation. Qualitative data also highlight the influence of environmental factors (e.g., family), and suggest that selective news consumption is associated with normatively different conceptions of media trust.

people's electoral (and to a lesser extent ideological) preferences, our findings show that exposure to counter-attitudinal sources can also be strongly correlated with political and ideological leanings.Our qualitative data suggest that this is because exposure to counter-attitudinal sources can reinforce attitudes, and potentially contribute to polarisation.Qualitative data also highlight the influence of environmental factors (e.g., family), and suggest that selective news consumption is associated with normatively different conceptions of media trust.

Keywords media repertoires, selective exposure, polarisation, news consumption, Eastern Europe
A growing number of democratic countries are experiencing increasing political polarisation (Carothers and O'Donohue 2019;McCoy et al. 2018).Often associated with deep social cleavages, growing divisions between opposing political camps, as well as with the disappearance of the political centre and shared political ground (Carothers and O'Donohue 2019: 1), polarisation has commonly been considered a serious threat to democracy (Arbatli and Rosenberg 2021;Mickey et al. 2017).Many political scientists and students of democracy have linked polarisation with the rise of populism, which exploits and further deepens political divides (Arbatli and Rosenberg 2021;McCoy et al. 2018).At the same time, communication scholars have examined parallels between political polarisation and the increasingly fragmented and personalised news media ecosystem, characterised by the proliferation of more partisan-oriented media and by a more prominent tendency towards selective news exposure on the side of the audience (e.g., Bennett and Iyengar 2008;Stroud 2010).
Selective news exposure has been studied by communication scholars for more than half a century, with research demonstrating the existence of a relationship between people's political beliefs and news exposure across different media types, from newspapers to television to the Internet (e.g., Klapper 1960;Lazarsfeld et al. 1948;Stroud 2017).With the proliferation of partisan media in recent decades, particularly following the arrival of online news outlets, opportunities for selective exposure seem to have expanded significantly (Garrett et al. 2014), reigniting debates about their impact of selective exposure on political attitudes and polarisation (Iyengar et al. 2019).
However, existing search on selective exposure has several gaps and open questions.First, the evidence of the continued existence of a link between selective exposure and political polarisation in the contemporary, high-choice information environment is inconsistent.Some of the work examining patterns of selective exposure in an online environment found no link between voter polarisation and exposure to partisan news (e.g., Wojcieszak et al. 2021) or even showed a link between increased exposure to ideologically diverse views and greater ideological distance (e.g., Flaxman et al. 2016).Second, the extent of selective exposure is uncertain as well, and is often exaggerated rather than investigated empirically (cf Garrett 2013).Third, virtually all research on selective news exposure and polarisation has been restricted to a handful Western countries (and the U.S. in particular), and largely based on single-country studies, thereby reducing our ability to generalise from the observed patterns onto news consumers in different types of media and political systems.Fourth, existing research on selective exposure tends to examine the impact of individual media outlets on polarisation separately from each other.However, in the contemporary, high-choice media environment, people's news media diets have arguably become much more complex, often consisting of multiple sources which are consumed simultaneously, and which might not necessarily be all in line with their consumers' preexisting political or ideological views.And finally, most of existing research utilises quantitative methods only, thereby limiting our ability to develop a more rounded understanding of the factors that shape people's news diets in everyday life.
In this paper, we seek to advance knowledge on the extent of selective news exposure and its links with political and ideological attitudes in three ways.First, we focus on Eastern Europe, a region that has been largely absent in the scholarship on selective exposure and related phenomena, and that has been experiencing significant democratic backsliding, polarisation and a shift towards authoritarian populism (Cianetti et al. 2018;Vachudova 2019).We also adopt a comparative perspective, examining four countries characterised by different levels of polarisation and democratic backsliding, and therefore potentially also by different levels of selective exposure: Czech Republic, Poland, Hungary, and Serbia.Second, we adopt a 'media repertoires' (Hasebrink and Domeyer 2012) approach to selective exposure, which is better equipped to capture the complexity of contemporary news consumption patterns, because it identifies typical combinations of news sources across different news brands and forms (from newspapers to news websites).This approach also allows us to distinguish between different types of media repertoires based on the extent of selectivity, thereby showing how widespread selective exposure is, and how this differs across countries.Third, we use a mixed methods approach, combining population surveys, an expert survey, and qualitative interviews.This allows us to examine how different (more or less selective) media repertoires relate to the ways in which people make decisions over which news sources to use in their daily lives, and to the relative role political attitudes have in this context.

Selective Exposure and Political Polarization
The idea that people, if given a choice, prefer to be exposed to messages that are matching their existing beliefsrather than those that stand in opposition to themhas been advanced and tested since the early decades of modern-day communication research (Klapper 1960;Lazarsfeld et al. 1948).Although it has fallen out of fashion in the 1970s and 1980s, the selective exposure theory experienced a comeback with the arrival of cable TV and especially the Internet.As information sources multiplied and more partisan media started to emerge, people have been increasingly able to tailor their news media choices according to their personal preferences, including political attitudes (Bennett and Iyengar 2008).
Inspired by changes in the information ecosystem as well as in politics, several researchers examined the impact of selective news exposure on political polarisation.Earlier studies carried out largely in the offline environment have found compelling evidence that exposure to partisan news reinforces both attitudinal polarisation, i.e. attitudes towards social and political issues (Knobloch-Westerwick 2012; Levendusky 2013a; Stroud 2010), as well as affective polarisation, i.e. hostility towards the political opposition (Garrett et al. 2014;Iyengar et al. 2012;Levendusky 2013b).Subsequent scholarship exploring selective news exposure in the online domain has however brought mixed outcomes.While a three-wave panel study focused on U.S. presidential elections concluded that increased exposure to conservative online media was associated with increased affective polarisation among Republican voters (Garrett et al. 2019), a recent study using online data-tracking found no effect of exposure to partisan websites on U.S. voters' polarisation (Wojcieszak et al. 2021).
Research focusing on selective exposure and polarisation in the context of social media usetypically discussed with reference to the "echo chamber" thesis (Sunstein 2001)also brought conflicting results.On the one hand, research on user interaction with Brexit-related news posts on Facebook (Del Vicario et al. 2017) showed the existence of two separate communities with little overlap in terms of user activity and found that the two communities perceived the same topics differently.On the other hand, a study of browsing histories of U.S. users of online news (Flaxman et al. 2016) showed that the use of social networks and search engines was associated with increased exposure to ideologically diverse views, while alsocounterintuitivelylinked with greater ideological distance.An experimental study of Twitter users in the USA, which examined the ideological impact of exposure to cross-cutting views (Bail et al. 2018) reached a similar conclusion, but also pointed to the impact of political orientation.While republicans who were exposed to crosscutting views developed significantly more conservative views, democrats became slightly more liberal over time.To complicate the argument further, Yarchi et al. (2021) found evidence of homophilic interactions and polarisation on Twitter, but opposite patterns on WhatsApp and Facebook.Taken together, this body of research suggests that in a high-choice information environment, the link between selective exposure and polarisation is context-dependent; and that polarisation could sometimes be driven by exposure to opposing views, rather than by exposure to content one agrees with.
Claims about the extent of selective exposure have been questioned as well, with some authors pointing out that even in a polarised environment such as the US, a majority of citizens rely primarily on less partisan, mainstream news, and that their news choices are not determined solely by their political orientations (Garrett 2013).In a similar vein, alarmist claims about the relative prominence of "echo chambers" online have been questioned, too.Dubois and Blank (2018) showed that they are constrained to small segments of population and moderated by political interest and a diverse media diet, while a study based on data from nine Western democracies (Vaccari and Valeriani 2021) found that it is far more common for people to encounter disagreeable content online than in face-to-face conversations, and that ordinary social media users tend to encounter similar quantities of agreeable and disagreeable political messages online.
In sum, while selective exposure to like-minded outlets has been documented in both offline and online environments, the evidence of its extent is rather patchy and far from unanimous, as is the empirical support for its impact on political polarisation.In this study we seek to advance knowledge on the extent of selective exposure and its links with political polarisation by focusing on Eastern Europe, a region that has been neglected in existing research on the topic and that has been experiencing significant polarisation in recent years.In line with this, our first two research questions ask:

RQ1:
To what extent do citizens selectively consume news from ideologically and politically homogeneous sources?RQ2: How closely are people's electoral behaviour and ideological selfidentification linked with their (selective) news consumption patterns?

Examining Selective News Consumption Through a Media Repertoires Approach
Apart from shifting the empirical focus beyond the West, our study also adopts a different methodological approach to selective exposure, based on the concept of media repertoires (Hasebrink and Domeyer 2012).In contrast to the usual practice followed in selective exposure research, which tends to focus either on a single type of medium, or on a specific digital platform (typically, Facebook or Twitter), a media repertoires approach advocates for analysing news exposure as a combination of different outlets.This approach is also conducive to the adoption of mixed methods, which has been lacking in existing research on selective exposure.
According to Hasebrink and Domeyer (2012), a repertoire-focused approach is characterised by three fundamental principles: first, it is a user-centred perspective; second, it considers the entirety of the media that a particular person regularly consumes; and third, there is a focus on the relationality or connected functions of a repertoire's components, emphasising that a media repertoire is not just the basic sum of its components, but that there is meaningful coherence to its inner structure.This last point is particularly significant, because it implies that members of a particular repertoire not only share their preferences for certain media types over others, but also, to varying degrees, preferences over brands, purpose of use, demographic characteristics and so on (Peters and Schrøder 2018).Consequently, in addition to uncovering patterns of media use in the countries we study, a media repertoires approach will allow us to examine the attitudes of the members of each repertoire group, and its consequences for political polarisation in those countries.
Past research has examined a relatively limited set of repertoire structure components.Whilst many studies focus on medium-level questions of preferences for certain media platforms over others (Schmidt et al. 2019), others examine repertoires in terms of their content: for example, by genres or topics (Taneja et al. 2012), or specific outlets or shows (Mangold and Bachl 2018;Mourão et al. 2018;Strömbäck et al. 2018).More recently, the debate has moved beyond descriptive analyses, concentrating on behavioural and attitudinal variables associated with these profiles, particularly political and ideological ones, such as political participation (Strömbäck et al. 2018), political interest and knowledge (Kim 2016), partisan preferences (Edgerly 2015) or support for political figures such as Donald Trump (Mourão et al. 2018).
However, unlike previous research, we move beyond simple component structures that take media items as their components (mediums, brands, shows, etc.), focusing instead on the underlying political and ideological position of the media brands that respondents consume.In this way, the resulting groups of repertoires show the extent to which audiences are exposed to politically and ideologically diverse views, based on their relative closeness or relative openness to information and opinions from "the other side".
Another advantage of a media repertoires approach is its methodological diversity.Unlike selective exposure research, media repertoires literature uses both quantitative, survey-based approaches (e.g., Taneja et al. 2012), and qualitative, interview-based approaches (e.g., Je ̨drzejewski 2017; Kõuts-Klemm and Brites 2017).Our study adopts a mixed methods approach, combining surveys with interviews.This enables us to conduct a large-scale cross-country study whilst facilitating an in-depth examination of our respondents' reasonings behind their news choices.Accordingly, our third research question asks: RQ3: How do people justify their choice of news sources, and how can that help explain the links between their news media repertoires and their political and ideological attitudes?

Study Context: Polarization and Democratic Backsliding in Eastern Europe
In recent years, several countries in Eastern Europe have been affected by simultaneous trends of deepening polarisation and erosion of democratic standards.This has been particularly clear in Hungary (see e.g.Enyedi 2016) and Poland (Domalewska 2021, Tworzecki 2019), whose governments have been repeatedly criticised for disrespecting the rule of law and undermining democratic institutions, especially the judiciary and independent media (cf.Bayer et al. 2019;Przybylski 2018;Surowiec and Štětka, 2020;Vachudova 2020).In both countries, party polarisation has been rising since 2010, most notably on the social-cultural dimension (Vachudova 2019).
However, it is important to note that Eastern Europe is a diverse region, marked by different degrees of polarisation and democratic deconsolidation, and different levels of media freedom.The contrast between Hungary and Poland on the one hand, and the Czech Republic on the other hand, is instructive.Data from the Chapel Hill Survey indicate that party polarisation along the identity axis is happening also in the Czech Republic (Vachudova 2019), with increasing gaps between the liberal/ pro-European and the illiberal/anti-Western camps.However, unlike Hungary and Poland, where the quality of democracy and of media freedom have plummeted since the right-wing populist parties took power, the Czech Republic still maintains a comparatively higher position in most global rankings; according to the Freedom House's 2020 Nations in Transit report, the Czech Republic is a "consolidated democracy", while Poland has been classified as "semi-consolidated democracy", and Hungary has been downgraded to "transitional or hybrid regime" (Freedom House 2020).This is mirrored by the evaluation of media freedom in those countries, where the Czech Republic stands apart from the other three countries (RSF 2020).Serbia, on the other hand, shares some traits with Hungarymost notably, it had been classified as a "transitional or hybrid regime" in the Nations in Transit report (Freedom House 2020).However, unlike Hungary, Poland and the Czech Republic, Serbia had a different political trajectory, experienced delayed democratisation (Bieber 2003), has not joined the EU, and never reached the standards of democracy reached (at least temporarily) by the other three countries.
Similar patterns of differences are displayed across the media systems in these countries regarding the extent of their politicisation and divisions along the pro-versus antigovernment axis.In both Hungary and Poland, the media scene is deeply divided between the opposing political camps, with the Hungarian government already exerting control over a substantial proportion of the media market (Bátorfy and Urbán 2020), and the Polish government attempting to muzzle oppositional media by legislative means or economic pressures, especially via distribution of state advertising (Klimkiewicz 2019).In Serbia, the media landscape is relatively more fragmented, but the majority of media are under heavy political influence, largely from owners siding with the government (Milutinovic ́2021).In all three countries, the space for not just independent, but also for politically impartial media has been shrinking in the last years.In contrast, the Czech media system, despite some of the largest news media brands belonging to the Prime Minister Andrej Babiš (Kotisova and Císařová 2021;Vojtěchovská 2017), and a number of media actively mobilising the voters against the governing coalition, is characterised by the presence of a relatively higher number of non-partisan news outlets, including public service media (PSM), which belong to the most trusted brands in the country (Newman et al. 2021).

Quantitative Methods: Data and Measurement
The quantitative part of our study is based on data from two surveys.The first is a population survey (N = 4,092), collected online (CAWI, 75%) and via telephone interviews (CAPI, 25%) in the Czech Republic, Hungary, Poland and Serbia.Respondents were selected by quota sampling, designed to be representative of the general population for key socio-demographic quotas: age, gender, education, region, size of municipality and the frequency of Internet usage.Pilot testing started in lat|e November 2019 and the main fieldwork was conducted in December 2019 and January 2020.
We also used a second online expert survey (N = 60), which was conducted in the same four countries between February and May 2020.Participants were selected through a combination of purposive and snowball sampling, following our project partners' suggestions and the experts' further recommendations.We aimed to include a proportionate number of experts with professional, academic, and civil society backgrounds.The resulting dataset provides data on media partisanship, independence and disinformation in each country, based on experts' assessments of a countryspecific set of media brands (matching those used in the population survey).

Quantitative Methods: Variables and Analysis
Our main dataset consists of a four-country population survey, which identifies respondents' "regular" media consumption; 1 specifically, those media brands (14-17 countryspecific selection, listed in Appendix A of the Supplementary Information File) that users indicated they consume at least once a week.Based on this data, we created two sets of variables: the first three identify the proportion of anti-, pro-and neutral government brands that each respondent consumes regularly, when taking all media brands they declared to use at least once a week as the whole.Similarly, the other three variables identify the proportion of liberal, conservative and neutral brands that each respondent consumes. 2The brands' political and ideological position was determined based on the results of our expert survey (for details see Appendix A).
In the next step, we used these six variables in a latent profile analysis (LPA) to explore the media repertoires that emerge in our four countries.The analysis is weighted by demographic distribution, and we used the Akaike's information criterion (AIC) and Schwarz's Bayesian information criterion (BIC) for determining best fit in terms of parsimony, and the entropy criterion value to select the model providing the greatest evidence for partitioning data in terms of well separated groups or clusters (Celeux and Soromenho 1996).Subsequently, we performed ordinary least squares linear regressions to examine the relationship between membership in our repertoire groups and the ideological and political preferences of our respondents in each country separately.Accordingly, our dependent variables were the probability of belonging to each repertoire group (a continuous variable calculated by LPA), and the two independent variables measured individual-level ideological and political preferences.The first places respondents' own viewpoints on a 1 (liberal) to 7 (conservative) scale.The second is a binary variable measuring whether a respondent voted for the governing party or not. 3 We also included control variables for gender, age, education, domicile size and religiosity, and all regressions were weighted by demographic distribution.

Qualitative Methods
Qualitative analysis is based on semi-structured qualitative interviews with 30 participants from each country, totalling 120, conducted in February and March 2020.
Participants were recruited from quantitative surveys and from personal connections.The sample was limited to participants who are somewhat engaged with news consumption and who are minimally politically engaged, meaning they follow politics on a regular basis (minimum weekly).Quota sampling was used to ensure the purposive sample was sufficiently diverse on several key demographic dimensions known to shape both media use and political behaviour, namely age, gender, domicile size and political preference.The aim was to divide each country's sample equally between three age groups (18-34, 35-59, 60 + ), gender (male, female), domicile type (urban, rural), and political preference (conservative, liberal, and, to a lesser extent, centre).
The interviews lasted an hour on average, and covered everyday routines, political engagement, the media environment, and news consumption routines and preferences.Particular attention was paid to news preferences, and to facilitate this, a card exercise was developed for each of the countries, modelled on the exercise used in a study conducted by Swart et al. (2017).Participants were provided with a pile of cards that contained names of 14-17 country-specific news brands (used also in the survey).They were asked to consider how important these sources are in their everyday life, rank them in order of importance, while also explaining the reasons for their choices.All interviews were transcribed from local languages to English and material was analysed using thematic analysis (Braun and Clarke 2006) with support of NVivo.The analysis focused on participant's justifications for the choice of specific news brands.To ensure consistency and comparability, and resolve potential translation problems, all interviewers used the same interview protocol (translated from English) and attended a joint workshop prior to fieldwork, and several online meetings during fieldwork and data coding.

Closed, Open and Balanced Consumption: Media Repertoires in Eastern Europe
To answer RQ1 we first had to identify prevailing types of media repertoires.To this end, we performed LPA analysis using our 6 indicators for the political and ideological content of the brands that our respondents consume.We focused only on those participants who regularly consume news media (minimum once per week), so that the resulting media repertoires exclude news-avoiders or disconnected citizens.The LPA model was estimated using gaussian regression for continuous variables and was performed iteratively five times with increasing number of groups to find the model with the optimal model fit.The lowest AIC and BIC scores (the most parsimonious model), and the highest entropy value (the model with the best ability to provide well separated clusters) indicated that the optimal number is five clusters or profiles, i.e. five different types of media repertoires. 4 Figure 1 shows these five media repertoires and the average proportion of the political and ideological positions of the media brands that people belonging in each repertoire regularly consume.Two of these repertoires (the closed liberal Those belonging in the closed liberal anti-government repertoire (11% of respondents) selectively consume media that are almost exclusively liberal and mostly antigovernmentalthough there is some exposure to brands that are considered neutral in in their political leaning.A good illustration of a participant in this group is our Hungarian participant Hun-12 (female, 59, rural), who is university-educated, liberal, and openly and emphatically anti-government.Her most trusted news sources are anti-government online sources such as Index and 444.hu, and press such as Magyar Narancs.She actively avoids PSM (other than Bartok Rádió) and progovernment news sources: "I am too old to read through an Origo article and then vomit.I feel that honourable journalism has disappeared."She is also highly suspicious of news sites that are not independent or, as she describes them, "are made by government's orders" (Hun-12).
Respondents in our second group, the open liberal anti-government repertoire (33%), also prefer liberal and anti-governmental media.However, as opposed to the closed repertoire, they display more "open" consumption patterns as they are open to media that show neutral or even opposite viewpointsmost notably, their consumption of both anti-and pro-government media shows remarkable balance, with average scores of 0.51 and 0.41, respectively.A telling example is our Polish participant Pol-04 (female, 62), who is university-educated, liberal, anti-government, and lives in a rural area.Although she is clearly opposed to the government, she occasionally consumes pro-government/right-wing media, but is also very aware of their political bias.In contrast to participant Hun-12, who has a closed liberal anti-government repertoire and refuses to engage with pro-government sources, participant Pol-04 is more open, learning about different perspectives, though she is clear on her political standing: For me, it's about finding the middle ground.I'm aware that each source of information has its own point of view, its agenda.I like to learn about different viewpoints.I stand by my convictions but, at least, I know what my opponents think…For some, a given thing will be good, for others it will be bad.I sometimes find it difficult to see why that is and so I'm trying to understand it.(Pol-04) The third group, which we describe as the balanced repertoire, comprises the smallest proportion of our sample with only 8%.On average, these individuals' media consumption is varied and mostly non-discriminant, showing a slight preference for politically and ideologically neutral media, but still exposed to plenty of liberal and conservative media and, to a lesser extent, to pro-government brands.The only exception is anti-government media, which are mostly absent.In the qualitative sample, only Czech participants were found in this category, for example participant Cze-26 (male, 26) who is university-educated, liberal, and lives in a rural area.He follows the Czech public service television (C ̌T24) and radio (C ̌eský rozhlas), both of which are politically neutral and ideologically liberal, but also occasionally switches to the more progovernment and ideologically conservative commercial TV channel TV Nova.At the same time, he is highly distrustful of sources that are ideologically more extreme, like the commercial channel TV Barrandov: "Since C ̌eská televize is a public service medium, I feel they try to present information objectively.I would not trust private media like I trust these.Specifically, not TV Barrandov." Respondents in the open conservative pro-government repertoire (30%) clearly show a preference for both pro-government and conservative media brands, although their open diet is reflected in their consumption of some neutral and anti-government/ liberal media.A typical example is Hungarian participant Hun-23 (female, 62), who is university-educated, conservative, steadfastly progovernment, and lives in an urban area.She regularly watches the news on the pro-government channels HírTV and M1, but also occasionally checks anti-government sources like 444.hu, 24.hu, and ATV to better understand the opposition, and reads the anti-government news website Index as they are "the fastest at reporting breaking news."She finds neutrality and diversity to be very important in news sources, though she maintains her support for the government.
Finally, 18% of the respondents from our sample belong to a closed conservative pro-government repertoire.Respondents in this group rely almost exclusively on pro-government and conservative news sources, with very little exposure to other viewpoints.A good example is Serbian participant Srb-11 (female, 46) who is university-educated, liberal, pro-government, and lives in a rural area.She is very distrustful of media in general but does consume pro-government and pro-Russian news websites like Sputnjik and Srbija Danas, some pro-government TV channels like RTS and Prva, and pro-government press (Vecernje novosti, Blic, Alo).She is clear in her right-wing views, considering herself a 'nationalist'.She does not trust foreign-owned commercial TV station N1, nor the radio station Slobodna Evropa, and sees them as "a foreign body", while at the same time having not issue with the pro-Russian Sputnjik, which suggests that by foreign sources she means primarily Western sources.
These five repertoires can be ordered on a political/ideological spectrum according to relative levels of selectivity (closeness) or diversity (openness); in this way, we can identify media consumption repertoires from an exclusively liberal/anti-government extreme to an opposite, exclusively conservative/pro-government extreme.In between, we see groups that have a less selective diet, preferring one end of the spectrum but open to content from more neutral or indeed counter-attitudinal brands.Lastly, in the middle of the spectrum are those whose media consumption does not lean to either side ideologically, and displays a preference for politically and ideologically neutral brands.
Having outlined the different types of media repertoires, we can now tackle RQ1 more directly.In line with studies that question the relative extent of selective news consumption (e.g., Garrett, 2013), most individuals in our study belong to the more diverse or 'open' groups, each comprising around one third of the respondents from our full dataset (33% and 30%).In contrast, a "closed" consumption is much less prominent (11% and 18% of the whole four-country sample).To put it differently, participants who selectively consume news from ideologically and politically homogeneous sources are in a minority.That said, the distribution of these repertoires differs considerably from country to country (Figure 2).
Taking together the closed and open groups at either side of the political and ideological spectrum, we can get a clearer picture of the extent of selective consumption and consequently polarisation of media consumption in these countries.First, it is apparent that in Hungary and Poland most respondents consume liberal and anti-governmental brands, whilst in the Czech Republic and Serbia it is the conservative-pro-government media viewers who are in a majority.However, the much larger closed conservative pro-government repertoire in Serbia also indicates that selective consumption is much more prevalent here than in the Czech Republic.Moreover, it is in Serbia that the two opposing closed groups make up the largest proportion of the sample (nearly 50% of respondents), indicating a society much less willing to consider competing political and ideological positions.In contrast, the Czech Republic shows the most "balanced" society, not only because of the lowest percentage of membership in the two closed groups (17%), but also because they are virtually the only country with a balanced repertoire: in Poland this makes up less than 1% of the respondents, and none of the respondents in Hungary or Serbia are classified to this group.On the other hand, Hungary and Poland show more similar patterns of repertoire groups distribution, though selective consumption is notably more common in the former, with 34% of the respondent belonging to the two closed groups, as opposed to only 20% in Poland.

Selective News Consumption, Voting and Ideology
In response to RQ2 we first examined the extent to which voting patterns (voting for or against the governmental party in the last national elections) correlates with probability of membership in each repertoire group (Figure 3).First, it is clear that the correlations are strongest for members of the two groups with most selective or 'closed' diets; members of the closed liberal anti-government repertoire in all countries are more likely not to have voted for the government party, while those whose media consumption consists of exclusively conservatives pro-government sources more likely to vote for the government.However, significant correlations also exist for some of the more 'open' (i.e., less selective) repertoires in some countries: in Serbia, those with an open anti-government repertoire are also significantly more likely not to have voted for the governmental party, while in Poland and Hungary, those with an open conservative pro-government diet are more likely to have voted for the government.This suggests that voters on both sides of the political spectrum are not bound to consuming only the news that align with their political views, but are being regularly exposed to counterattitudinal sources.
Figure 4 explores the relationships between membership of media repertoires and ideological self-placement.In this case, links with selective news consumption are   even less straightforward.While statistically significant correlations do exist for the closed liberal anti-government repertoire across all four countriesi.e., those with such a repertoire are more likely to place themselves on the liberal end of the ideological scalepatterns at the opposite end of the ideological spectrum are far less clear.Here, a statistically significant link between a closed conservative pro-government repertoire and self-placement as conservative exists only in Hungary.Furthermore, and counterintuitively, it is the members of the open (i.e., less selective) conservative progovernment repertoire that are more likely to present themselves as conservative; a significant correlation exists in three of the four countries (Czech Republic, Hungary, and Poland).Yet again, we find evidence that suggests that exposure to counter-attitudinal sources can sometimes be as closely linked with attitudes as selective exposure to attitude-consistent sources. 5

Explaining the Nature of News Repertoires and Links with Voting and Ideology
To gain a better understanding of the patterns of correlations found in our quantitative data, and specifically the findings about exposure to counter-attitudinal sources, we examined our qualitative data, focusing on different reasons our participants mentioned when discussing their preferred news sources.We can divide these reasons into three broad groups: those that are more common among participants with more diverse media repertoires, and which help explain why they consume sources that are not aligned with their own views; those that are common among participants in the two closed repertoires, whose news consumption is highly selective, and which help explain why they are reluctant to engage with counter-attitudinal sources; and those that can be found across all types of repertoires and have to do with systemic or environmental factors rather than specific political or ideological orientations.
For participants with more diverse repertoires (i.e. the open liberal anti-government repertoire, and the open conservative pro-government repertoire) there were two key reasons for consuming media sources from the 'other side'.First, there were those participants who were curious to hear views different from one's own and/or felt that it is important to understand what the 'enemy' thinks.For instance, Polish participant Pol-08 (female, 36, urban) described checking both pro-government and antigovernment news: "I look at what they say on TVN and Polsat and on TVP1, but these are three extremely different viewpoints on the same events.I just watch all three just to see how they present these different outlooks."In Serbia, the antigovernment participant Srb-28 (male, 60, urban) used similar arguments, and explained she watches pro-government TV channels "[T]o compare them to the ones I usually watch and to see what their copycat TV shows look like" (Srb-28).
It is important to note that in these cases, openness to counter-attitudinal sources did not lead participants to question their own views.If anything, exposure to 'the other side' seemed to reaffirm one's own political stance, or even provoke palpable frustration and anger.One Hungarian participant (Hun-11, female, 61, urban) explained that she listens to the PSM radio channel Kossuth Rádió "because it is the government's propaganda, mostly…"; it is how she "can identify which party they would work together with if they didn't have to hide it."Serbian participant Srb-19 (female, 62, urban) described sometimes consuming pro-government news to "listen to what Vučic ́[President of Serbia] has to say"; however, she tends to have a negative reaction to this type of news as she "keep[s] swearing all around the house, as if I'm arguing with my TV."These examples help explain why more diverse media repertoires could be equally, if not even more closely correlated with voting preferences and ideological attitudes than the more selective news repertoires.
In contrast to examples examined so far, some participants explained that they consume both pro-and anti-government sources because they feel neither provide a full account, and therefore one has to listen to both to be able to establish what is actually happening.Unlike participants who were simply interested in hearing 'the other side', these participants believed that the 'truth' lies somewhere in-between and requires them to make their own independent judgment.For example, one Polish participant explained: TVP, pro-governmental propaganda broadcaster where every piece of news is a lie.I like to watch it because then I know that the truth is the complete opposite of what they say.Then, I also tune to another broadcaster, TVN for example.They show a different side of things.I combine these portrayals together, which gives me the real picture.male,34,urban) One Hungarian participant (Hun-30, male, 38, rural) mentioned watching both proand anti-government TV channels, observing "how they communicate given facts" in order to come up with his own conclusions; as he states, "one says black and the other says white."In Serbia, participant Srb-09 (male, 48, rural) often follows the same news story across several contradictory news sources: "I check it in a few different places, then check if it fits, whether it's the truth or not.I make my own judgement." Looking at participants in the two more selective, closed repertoires, a different set of factors emerges.Here, one's preferred news sources, which are aligned with one's political preferences, are seen as more trustworthy.However, the basis for establishing trustworthiness varies.On the one hand, some participants associated trustworthiness with perceived objectivity and/or independence of news sourcesand, by contrast, described sources they do not use as untrustworthy because they are lacking in objectivity and/or independence.For the Serbian anti-government participant 62,urban) it is important that the anti-government TV channel N1 tends to be "concise, objective, realistic, critical of both the ruling and the opposition structures," which they describe as being "music to [their] ears."Another Serbian antigovernment participant, who likewise follows only anti-government sources (Srb-27, female, 60 + , urban) explained his reluctance to follow pro-government sources with reference to their lack of autonomy: "I mostly just watch news on N1 and Nova S. I don't watch other news because they are at the hands of the regime of Aleksandar Vučic ́." On the other hand, some participants explained their loyalty to specific news brands by emphasising the alignment with their own views.For some, this meant that they consumed specific sources precisely because they were biased, and because they lacked autonomy, rather than this serving as a reason for distrust.Even objectivity was occasionally explained in these terms and ascribed to news sources that are in tune with one's own opinions, or to sources that are subjected to control of the governing party.For instance, Hungarian pro-government participant Hun-14 (female, 56, urban) explained that she consumes the pro-government channel HírTV because the channel's bias aligns with her own convictions, and that she sees the governmentcontrolled PSM channel M1 as 'objective' because she believes it does what a public service channel is supposedly meant tonamely, 'report on the government's work': I think M1 is objective, even though they say it is biased, I think…public service media should report on the government's work, and this is what they do…Of course, they say that HírTV is biased, it is true that it is a right-wing channel but since my way of thinking is also similar, so… [laughs] This is what I want to listen to.Such reasoning was present among participants belonging to closed repertoires on both sides of the political spectrum.For instance, one of our Czech participants in the closed anti-government liberal repertoire explained her preference for anti-government portals by linking them to objectivity, but then went to argue that these portals also shared her own point of view: Because I suppose I find them somehow objective or more objective than some others.Perhaps, I would say that those moderators or journalists have a pretty similar point of view as I do, so I sympathise with it.female,26,urban) Arguably, this kind of media trustbased on alignment with one's views rather than perceived objectivity and/or independencefits particularly well with the notion of selective exposure being driven by people's desire to avoid cognitive dissonance.
The shared trait of all reasons explored so far is that they are associated with individuals' political or ideological convictions, or with their perceptions of the relative objectivity or autonomy of specific news brands.However, our analysis also revealed important environmental factors involved in shaping news consumption.First, some participants consumed certain sources simply because they were the only ones available.For example, Hungarian anti-government participant Hun-12 (female, 59, rural), while being steadfastly anti-government, consumes the pro-government station Bartok Rádió as it is all that is available in her rural area.
Second, several participants explained that they are exposed to certain brands because they are consumed by members of their household, typically partners or parents.For example, pro-government Czech participant Cze-03 (male, 65, rural) watches TV Nova when his wife is at home and turns it on, while Polish participant Pol-25 (female, 79, urban) mentioned listening to radio due to her husband (Pol-25, female, 79, urban), and the Hungarian participant Hun-20 (male, 18, urban) explained that he watches TV because of his father (Hun-20, male, 18, urban).These findings are likewise important for our understanding of selective exposure, because they remind us that people's news choices may be determined by factors beyond their control, independent of their personal attitudes and preferences.

Discussion and Conclusions
This paper investigated selective exposure in news consumption in Eastern Europe and its links with political polarisation, through an exploration of audiences' news consumption preferences in four countries.Our analysis has identified five types of media repertoires, distinguished by the degree of selectivity, i.e., degree of openness/closeness towards politically and ideologically opposite news brands.In response to RQ1, we found that overall, selective news consumption is not particularly prominent; the "closed liberal anti-government repertoire" and the "closed conservative pro-government repertoire" compose only 11% and 18% of the entire sample, respectively.Most citizens fall within either the "open liberal antigovernment repertoire" (33%) or the "open conservative pro-government repertoire" (30%), characterised by politically and ideologically relatively heterogeneous diets.This result provides further support for arguments that suggest the extent of selective exposure is exaggerated (e,g., Garrett 2013), and suggests that avoidance of cognitive dissonance is not as strong a factor in news consumption as the original theories of selective exposure assume, at least not in the context of a contemporary, high choice news environment.
However, there are significant differences between the countries in our sample, illustrating internal heterogeneity political and media landscapes within the Eastern European region; while only 16% of Czech respondents fall within the "closed" repertoires, they comprise half of the population in Serbia, while in Hungary the closed liberal anti-government repertoire includes a quarter of all news consumers.Arguably, this pattern is linked to greater polarisation in Serbia and Hungary (Jiries et al. 2020;Vegetti 2019)also illustrated by the absence of significant politically neutral news brands in these two countries, as our expert survey documented. 6 Further underscoring the correspondence between the character of individual news exposure and the extent of country's polarisation, it is telling that the only country with a sizeable "balanced repertoire" is the Czech Republic, undoubtedly also because of the strong presence of politically neutral PSM (which do not exist anymore in the other three countries).These cross-country differences suggest that the extent of selective exposure is dependent on systemic factors and increases with higher political polarisation and lower media independence.
Exploring the correlations between individual news diets and their political and ideological preferences (RQ2), we found while news consumption does tend to be in line with people's electoral preferences, this does not necessarily mean that people are surrounding themselves only with likeminded news outlets.As we have demonstrated, electoral preferences are in some cases also correlated with more diverse media repertoiresi.e., those that lean towards one political/ideological end of the spectrum but also include counter-attitudinal sources.The argument that attitude-consistent news consumption is far less prevalent than what the traditional theories of selective exposure might imply finds an even stronger support when examining the links between people's repertoires and their ideological selfpositioning, where correlations were actually more common and stronger for more diverse media repertoires rather than for the more selective ones.This finding is aligned with previous studies that showed a link between increased exposure to ideologically diverse views and greater ideological distance (e.g., Flaxman et al. 2016).Finally, and consistent with some past research (Bail et al. 2018), we also found significant differences in correlations between political preferences and repertoires belonging to opposite sides of the political or ideological spectrum.This suggests that the likelihood of exposure to counter-attitudinal news sources may differ depending on one's political/ideological leanings, as well as on individual country context; for instance, in Hungary and Poland, exposure to counter-attitudinal sources appears to be linked to stronger pro-government attitudes, while the same type of exposure is more common for people who vote against the government in Serbia.
The qualitative investigation of reasoning behind news consumption habits helps explain these quantitative findings (RQ3).First, the results show that exposure to diverse (including counter-attitudinal) sources can end up reinforcing attitudes, and second, they draw attention to environmental factors that affect news consumption and that are beyond individual control (e.g., the influence of family and friends).Taken together, these results help explain why people's news choices are not always aligned with their political and ideological leanings, and why we can occasionally find significant correlations between more diverse news repertoires and political and/or ideological attitudes.Third, the results confirm the importance of studying selective exposure in conjunction with media trust, but also point to the importance of paying attention to normative differences in the basis of media trust, i.e., between trust based on perceived impartiality and independence, and trust based on attitudinal consistency.The latter type of media trust is arguably driven by a desire to avoid cognitive dissonance, and hence more consistent with the expectations of selective exposure theory.Future research could examine whether higher levels of selective exposure are linked with a normatively different conception of media trust.
The study is certainly not without limitations.First, it relies on self-reported news consumption, which is not necessarily a true reflection of exposure.Second, it includes only a selection of news brands from each country, leaving aside those that are less popular or politically significant, but might still be important for some audience members.This limitation extends to the focus on professional news outlets and the exclusion of alternative platforms, as well as social media.Lastly, this study provides only a snapshot in time, not considering the temporal dimension of media repertoires (Peters and Schrøder 2018).Further research could attempt to address these issues and utilise some of the methods that allow for more direct measurement of news consumption, such as digital data tracking.A longitudinal design could be adopted to explore the evolution of media repertoires over time.

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.
anti-government repertoire, and the closed conservative pro-government repertoire) indicate more selective news consumption, while the other three indicate a more diverse news consumption.In what follows, we summarise the key traits of each type of media repertoire and offer illustrative qualitative examples taken from our pool of interview participants.

Figure 2 .
Figure 2. Membership of repertoire groups by country.

Figure 3 .
Figure 3. Voting patterns and membership in media repertoires.

Figure 4 .
Figure 4. Ideological self-placement and membership in media repertoires.