Eurosceptics into the Limelight? Eurosceptic Parliamentary Actors and Media Bias in EU Affairs

In the spotlight for some time now, the potential of national parliaments to legitimise EU politics has become even more salient given the growing politicisation and public contestation of EU issues. Their ability to realise this potential depends, however, vitally on citizens being actually aware of parliamentary involvement in EU affairs. Academic as well as political attention has therefore more recently turned to the communication function of parliaments, and here the media play a crucial role. Important is not only whether EU parliamentary affairs are covered in the media, but also who within parliament gets the opportunity to raise European issues in the media. In the context of this thematic issue, the question of media visibility is of particular interest with regard to Eurosceptic parliamentary party groups and their members. Do Eurosceptics in parliament get to dominate parliamentary EU news in the media and thus to take ownership of EU issues—or do the media freeze parliamentary Eurosceptics out of the coverage? Both would seriously undermine the legitimising potential of national parliaments. The article therefore analyses to what extent we can find a visibility bias in the print media coverage of Eurosceptic parliamentary actors and explores the factors that contribute to such bias. For the analysis, it draws on a quantitative dataset of all newspaper articles covering parliamentary EU affairs in six member states (Austria, Finland, France, Germany, Poland and the UK) over a period of four years (2010 to 2013).


Introduction
The 2016 referendum in the UK on leaving the EU and its aftermath not only sent shock waves through the Union, it has also brought home two very uncomfortable truths.First, it made, again, powerfully clear how deep the rejection of, but also the disconnect between, at least parts of the citizens of the EU and the Union is and, second, how little many citizens actually know about the EU.
Both insights are not exactly new, and have been discussed for some time now under the broad umbrella of the EU's infamous democratic deficit.Described as a move away from the 'permissive consensus' (Lindbergh & Scheingold, 1970), that is the friendly ignorance of citizens towards the EU, towards a more 'constraining dis-sensus' (Hooghe & Marks, 2009) or even 'destructive dissent' (Daddow, 2012), this development has, inter alia, also led to a greater interest in the role of national parliaments in EU affairs, and in their communication function in particular (Auel, Eisele, & Kinski, 2016;Auel & Raunio, 2014;Rauh, 2015;Rauh & De Wilde, 2018;Wendler, 2016;Winzen, De Ruiter, & Rocabert, 2018).By communicating EU affairs to their citizens, the argument goes, parliaments can not only legitimise national politics in EU affairs, but also overcome the disconnect between citizens and the EU: The communicative performance of national parliaments in EU affairs is directly related to the often discussed democratic deficits of supranational gover-nance: if MPs raise European issues, they offer a remedy to the otherwise opaque procedures, the overwhelming complexity, and the difficult attribution of political responsibility in decision-making beyond the nation state.(Rauh, 2015, p. 118) The ability of national parliaments to connect their citizens to the EU and EU politics does, however, depend crucially on whether they are actually able to reach a wider audience.The internet has clearly facilitated parliamentary communication via social media, websites, web streams or parliamentary TV.Yet although we lack systematic data, it remains questionable whether citizens actually make broad use of these opportunities.Here, the mass media still play an important role as a relais between politics and the citizens.Traditional media, such as newspapers and TV are certainly no longer 'the only contact many [citizens] have with politics' (McCombs & Shaw, 1972, p. 176), but surveys consistently show them still to be among the important sources of information about political issues (e.g., Reuters Institute, 2018).
In contrast to social media or websites, which establish an unfiltered, un'mediated' line of communication between MPs and citizen, traditional media are not simply an arena that can be strategically occupied by political actors, but autonomous players that engage 'in making European news' (Trenz, Conrad, & Rosén, 2009, p. 343, emphasis added) and thus contribute to the shaping of public discourse about the EU.Various studies have linked the coverage of EU news in the media to public perceptions of the EU (Schuck & De Vreese, 2006;Vliegenthart, Schuck, Boomgaarden, & De Vreese, 2008; for an excellent discussion see Galpin & Trenz, 2017) as well as turnout and vote choice in the European Parliament (EP) elections (Van Spanje & De Vreese, 2014) or in EU-related domestic referendums (Elenbaas & De Vreese, 2008).This raises not only the question to what extent the media cover parliamentary engagement in EU affairs more generally, but also who within parliament gets the opportunity to raise European issues in the media, to explain procedures and to attribute responsibility.
With regard to the former, studies have shown that parliaments generally tend to play a minor role in national EU news compared to domestic executives or EU actors (De Wilde, 2014;Koopmans & Statham, 2010).Still, Auel, Eisele and Kinski (2018) find that parliaments are still routinely covered in the media and, moreover, that the more active parliaments are in EU affairs, the greater the attention they gain in the press.Yet so far, we know little about the latter, namely about the visibility of different parliamentary actors in the media when it comes to the coverage of parliamentary EU affairs.A number of studies have investigated the media visibility of parliamentarians (see Vos, 2014, for an overview), but with very few exceptions (e.g., Gattermann & Vasilopoulou, 2015), these studies focus on domestic rather than EU politics.
In the context of this thematic issue, and especially given the recent surge in public Euroscepticism throughout the EU as well as the successes of Eurosceptic parties, the question of media visibility is of particular interest with regard to Eurosceptic parliamentary party groups (PPGs) and MPs.Research has shown that the relative visibility of parties and candidates can also have a strong impact on vote choice (Eberl, Boomgaarden, & Wagner, 2017;Takens, Kleinnijenhuis, van Hoof, & van Atteveldt, 2015).At least as important is the question who within parliament gets to take public ownership of EU issues.Do Eurosceptics in parliament get to dominate parliamentary EU news in the media-or do the media freeze parliamentary Eurosceptics out of the coverage?Both would indicate a bias in the media and seriously undermine the capacity of national parliaments to fulfil their communication function, which depends crucially on whether they succeed in making 'the choices and political alternatives involved in European integration visible to the wider public they mean to represent' (Rauh, 2015, p. 117, emphasis added).
Against this background, the article analyses the print media coverage of Eurosceptic compared to non-Eurosceptic parliamentary actors and explores the factors that contribute to a more or less balanced coverage.Drawing on the concept of media bias, the aim is to make a contribution at three levels: first, empirically, by providing the first investigation into the question of the relative visibility of Eurosceptic parliamentary actors in the domestic media coverage of parliamentary EU news, thus contributing to the literature on parliamentary communication in EU affairs, but also the media's role in shaping public discourse on the EU; second, methodologically, by developing two different types of visibility bias that take the relevance of Eurosceptic parliamentary actors in terms of their seat share and their level of activity in EU politics into account; and third, conceptually, by treating media bias not as an independent but as a dependent variable and exploring factors that might contribute to a bias in the first place.The article is structured as follows: the next section develops the two different types of visibility bias and discusses potential factors contributing to bias, distinguishing between partisan and structural bias and drawing on the notion of 'newsworthiness'.Section three presents the data and operationalisation, followed by the empirical analysis in section four.The final section discusses the findings and concludes.

A Certain Kind of Flavour? Measuring and Explaining Visibility Bias
Bias can be defined most basically as 'any tendency in a news report to deviate from an accurate, neutral, balanced and impartial representation of "reality" of events and social world' (McQuail, 2010, p. 549).Related to media coverage more generally, bias means that a specific actor or group of actors gets a consistently different coverage than others according to a predefined of bench-mark for balance or neutrality (Hopmann, Van Alst, & Legnante, 2012; see also Eberl et al., 2017).The problem is, of course, to define what should be taken as the benchmark for a balanced coverage of political actors or groups.

Measuring Media Bias
A first option is treating all (groups of) actors equally, i.e., providing them with the equal amount of coverage.This is often the benchmark used in two-party systems and especially for the news coverage of presidential elections in the US (D'Alessio & Allen, 2000).For European multiparty systems, however, equal shares of the news coverage ignore the relative size and thus relevance of the PPGs.Indeed, in many European countries the relative amount of parties' coverage during election campaign is regulated for public broadcasting to achieve balanced coverage through allocated shares (see Hopmann et al., 2012, for examples).In line with much of the literature, a first measure of bias will therefore be based on the definition of balanced coverage as a share of the coverage that is equal to the relative size of the Eurosceptic group(s) in parliament, and thus to their share of the seats.
Yet measuring bias solely in the basis of the seat share ignores what groups actually do in parliament.It is entirely possible, for example, for smaller groups to be very active in parliament, especially in policy areas of importance to them.The second measure of visibility bias therefore takes the activities within parliament into account.This measures how much Eurosceptics are engaged in EU affairs within parliament compared to non-Eurosceptics, and thus their comparative news supply.In addition, given that PPGs and MPs have to make a conscious choice to invest scarce resources in EU affairs-rather than in domestic affairs-it captures how salient EU affairs are for Eurosceptics in parliament compared to non-Eurosceptics.Visibility bias here is thus defined as a share of the coverage that does not reflect the Eurosceptic parliamentarians' share of activities in EU affairs.

Explaining Media Bias
To explore factors that may explain bias in the news coverage, I draw on the distinction between partisan and structural bias (McQuail, 2010;Van Dalen, 2012): partisan bias is the result of journalistic decisions driven by an ideological rationale leading to systematically greater attention to parties on a specific side of the political spectrum.Historically the press in Europe was indeed rather closely linked to specific political parties (Hopmann et al., 2012;Seymour-Ure, 1974).Reporting in a politically unbiased manner has become much more of a journalistic norm since, yet newspapers still feature editorials and opinion pieces that more or less explicitly promote certain political views, actors, candidates or parties, while criticising others (Takens, Ruigrok, & van Hoof, 2010).Moore and Ramsay (2017), for example, have found the media coverage of the UK referendum campaign to be highly partisan.Studies have generally shown that partisan views expressed in editorials tend to seep into the general news coverage (Brandenburg, 2006;Kahn & Kenney, 2002).Thus, partisan bias, or at least biases towards a specific political position, can still be expected to be present in the media (Hopmann et al., 2012).A first expectation is therefore: H1: The relative visibility of Eurosceptic PPG and MPs is positively affected by a matching editorial line of the newspapers.
Structural bias, in turn, is the result of journalistic routines and the judgement and selection of events for coverage based on their newsworthiness (Van Dalen, 2012, p. 34).Since the seminal study of Galtung and Ruge (1965), news value research focuses on the criteria that guide journalistic judgments regarding the selection of specific events or actors for coverage.'This news judgment is the ability to evaluate stories based on agreedon news values, which provide yardsticks of newsworthiness and constitute an audience-oriented routine' (Shoemaker & Reese, 1996, p. 106).Here, studies have shown that some news values are rather consistently applied to political events across a range of news organizations: in general, events that involve powerful actors or institutions (power/influence), have an entertaining or dramatic (negativism/conflict) element, or are perceived as relevant to a significantly large audience (relevance), are more likely to be selected than those not featuring one of these factors (for a comprehensive overview see O 'Neill & Harcup, 2009;also Brighton & Foy, 2007;Eilders, 2006).
The basic guiding assumption for structural bias is therefore that visibility bias is the result of newspapers according Eurosceptic parliamentary actors greater or lesser newsworthiness than non-Eurosceptics in EU affairs.Here, I distinguish between two sets of news factors or values that potentially impact media bias: 1) news values related directly to the object of coverage, i.e., Eurosceptic parliamentary actors; and 2) news values related to expectations of the addressees of the coverage, i.e., the readership.

News Factors Related to Eurosceptic Parliamentary Actors
First, I assume the position of Eurosceptics towards the EU to have an impact.Here, Taggart and Szczerbiak (2002, p. 7) famously distinguish between 'hard' and 'soft' Euroscepticism.The former refers to principled opposition to the EU that might be associated with demands to leave the EU or halt further integration, the latter to a more qualified opposition to specific EU policies or institutional choices, and Eurosceptic positions can be located on a continuum between the two, an approach also followed by the Chapel Hill Expert Survey (CHES; e.g., Bakker et al., 2015; for a discussion of conceptualisations and measurements of party Euroscepticism see Vasilopoulou, 2017).Given the importance of news values such as conflict and negativity, I expect: H2: Eurosceptics with a more hard-line stance on the EU have greater relative newsworthiness and thus relative visibility than those with more moderate views.
The underlying assumption is that journalists expect readers to be engaged by political conflict, but bored by political consensus (Vliegenthart, 2012).
Political conflict as a news value, however, also indicates the importance of taking the position of the other parliamentary actors and the overall level of conflict over EU issues within parliament into account as well (Van der Pas & Vliegenthart, 2016).As Auel et al. (2018) show, for example, a greater conflict potential regarding EU issues within the governing coalition has a clear positive effect on parliamentary news coverage.This suggests that where the potential for conflict within parliament is overall higher regarding the EU, media attention for parliamentary EU affairs will generally increase, thus reducing the relative attention paid to Eurosceptics.Accordingly, I expect: H3: Eurosceptics have greater relative newsworthiness and thus relative visibility, the lower the overall level of conflict within parliament.

News Values Based on Readership Expectations
Second, newsworthiness as an audience-oriented concept also depends on public opinion, and thus on the relevance of EU affairs for the potential readership.In line with the argument about the importance of negativity as a news factor, it can be assumed that stronger public Euroscepticism is more likely to increase the relative newsworthiness and visibility of Eurosceptic parliamentary actors.'People naturally pay attention to things that are [or that they perceive to be, the authors] dangerous or threatening' (Shoemaker, 2006, p. 107), and Eurosceptics voicing such concerns can be expected to have greater relative news value in member states, where public opinion is more critical towards the EU.In addition, it has be argued that what matters for media coverage of EU affairs is not so much the general public attitude towards the EU (Brüggemann & Kleinenvon Königslöw, 2009), but whether EU issues are actually salient in public opinion, i.e., relevance as a news factor.Where citizens do not really care about EU politics, journalists have little audience related incentive to focus on EU coverage beyond mere routine reporting, in general, and thus on the views or activities of Eurosceptic PPG or MPs, in particular.
Yet the relationship between public opinion and media coverage is not a one-way street.As mentioned above, and as the literature on media effects more generally shows (see Schuck, 2017, for an overview), the media are not only 'mirrors', but also 'shapers' of public opinion.Indeed, De Vreese (2007, p. 280) found public Euroscepticism to be, 'at least partially, a function of the diet of information that citizens consume about European affairs'.Similarly, the extent to which the media cover EU affairs may also have an impact on how salient the public regards EU politics.Thus, assuming a more reciprocal rather than straightforward onedirectional relationship, I expect: H4: Parliamentary Eurosceptics' relative visibility is positively associated with the level of public Euroscepticism.
H5: Parliamentary Eurosceptics' relative visibility is positively associated with the salience of EU issues in public opinion.

Data and Research Design
The empirical analysis focuses on Austria, Finland, France, Germany, Poland and the UK.These six member states provide a representative subgroup in terms of size, length of membership, geographical location as well as public opinion on EU integration.Importantly, their parliaments differ in terms of the size of Eurosceptic PPGs, the degree of their Euroscepticism, as well as their position on the left-right political scale.
The basis for the analysis is a dataset including all articles covering parliamentary involvement in EU affairs in three newspapers per member state over four years (2010 to 2013, N = 5589).For each member state, the two largest quality broadsheets (one conservative, one liberal) and the largest tabloid were selected (see Table A1 and further information on the coding process in the Appendix).Each article was coded according to whether it mentioned Eurosceptic or non-Eurosceptic actors their own, together or not at all.The distinction between Eurosceptic and non-Eurosceptic parliamentary actors was made based on the CHES 2010 (Bakker et al., 2015), and from 2012 onwards, the CHES 2014 (Polk et al., 2017).PPGs, and accordingly their MPs, were considered as Eurosceptic if they had a score of 3.5 or below1 .Measuring Euroscepticism for MPs using party data is, of course, somewhat fuzzy, as parties can be internally split.There is, however, no reliable comparative data accessible on the position towards the EU for individual MPs.
Figure 1 gives an overview over the distribution of articles featuring either Eurosceptics or non-Eurosceptics alone, together or not at all, across newspapers.As the figure shows, in all newspapers we find a sizable share of articles that focus on parliament as an institution and do not mention PPGs or MPs at all.In turn, across almost all newspapers the proportion of articles that feature both types of PPGs or MPs is astonishingly low.The Finnish newspapers are the only ones where the share is over 20 per cent.France, Germany and the UK, in turn, are at the very lower end of the scale.Thus, readers in the six countries rarely seem to get both parliamentary sides of the story within the same article.

Dependent Variables: Two Measures of Visibility Bias
For the two measures of visibility bias, I selected all articles where either Eurosceptic or non-Eurosceptic actors were mentioned on their own (N = 2.535), as articles that mention both types of parliamentary actors automatically display a weaker bias, although it may still exist.Mentions included direct or indirect quotes by parliamentary actors as well as references to parliamentary activities by or positions of parliamentary actors within an article.Given that MPs can also have (had) other functions, such as party chair (as opposed to chair of the PPG) or member of the government, a strict rule was applied: articles were only selected if the current role of MP/the status as parliamentary party group was explicitly mentioned or clear from the context, current government members were excluded.I then calculated the share of articles mentioning only Eurosceptic actors (in per cent) out of all articles that mentioned either type of actor on their own, aggregated at the monthly level.Where newspapers featured no articles mentioning either Eurosceptic or Europhiles on their own during a spe-cific month, the bias was coded as 0 for both measures for that month.
For the first measure of bias, the visibility bias based on the seat share (from here on simply termed seat share bias), I used the combined seat share of all Eurosceptic party groups for each parliament (recalculated after elections) as the benchmark to determine the share of articles they ought to have received to guarantee balanced coverage.I then subtracted their share of seats from their share of articles for each month.If party A had 30.5 per cent of the seats, but was only mentioned in 25 per cent of the articles in a given month, for example, the resulting bias was −5.5 percentage points.The visibility bias thus measures whether Eurosceptics receive more or less coverage in any given month than would have been balanced according to their size within parliament.
For the second visibility bias based on parliamentary activity (from here on simply termed activity bias), the analysis draws on the PACE parliamentary activity dataset, which covers all parliamentary activities in EU affairs within the six member states' parliaments over the same period (Auel et al., 2016).The share of activities by Eurosceptic actors was again calculated at the monthly level by selecting all activities, where the identification of the PPG or MP is straightforward, i.e., oral and written parliamentary questions as well as motions and censure motions introduced in parliament.Clearly, these four activities do not fully reflect the overall parliamentary work of MPs or PPGs.All other parliamentary activities, however, such as debates, votes, hearings etc., automatically involve both types of PPGs and MPs.The measure also accounts only for the sheer volume, but not the content of the activities, which may be more or less relevant for journalists to report on.Yet parliamentary involvement via questions or motions on EU affairs reflects at least to some extent the overall activity in EU affairs and thus provides an imperfect, but workable proxy.To calculate the bias, I subtracted their share of activities from their share of articles for each month.Thus, the visibility bias measures whether Eurosceptics receive more or less coverage (in percentage points) than would have been 'their due' according to their level of activity in EU affairs within parliament.

Editorial Line
To measure the editorial line, two dummy variables were included: the first distinguishes between clearly Eurosceptic (= 1) and other newspapers.As the EU editorial line of a newspaper is often difficult to measure, only those with a rather clear Eurosceptic stance were coded as such.These include the Kronenzeitung (Austria), the BILD Zeitung (Germany), The Times and The Sun (UK), but neither a Finnish, French or Polish newspaper (for France, Germany, Spain and the UK, see the country reports of WP3 in EuroPub, 2004; for the UK, see also Daddow, 2012; for Poland see Filas & Płaneta, 2009).The second dummy variable indicates whether the editorial line and the position of the party match broadly with regard to their position on the left-right scale (1 = match).The basis are the CHES 2010 and, from January 2012 onwards, CHES 2014 mean scores for left-right party position (scores range from 0 to 10, left to right).Parties were very broadly categorised into left (< 5) and right wing (> 5) and then matched with the respective political orientation of the newspapers.

Party Euroscepticism
Data draws on the CHES 2010 and, from January 2012 onwards, CHES 2014 mean scores for 'EU position' for the Eurosceptic actors.

Conflict Potential over EU (Dispersion)
Drawing on Gattermann and Hefftler (2015, p. 314) parliamentary conflict potential is operationalised as the parliamentary party system dispersion: where SS jk denotes the seat share, P jk the position of party j in country k towards EU integration (using CHES data), and P k the weighted mean of all party positions in country k.

Public Euroscepticism
Eurobarometer data on the percentage of citizens stating that they 'tend not to trust the EU' (European Commission, 2010, 2011, 2012, 2013).To test the robustness of the measure, the percentage of citizens stating that they 'have a negative image of the EU' was used as an alternative, but the results remained the same.

Salience
To measure salience, I developed an index measuring both stated and actual public interest in EU politics.The index is based on the factor scores for the following variables obtained in a principal component analysis: share of respondents (European Commission, 2010, 2011, 2012, 2013) stating that they had an interest in EU politics, share that had an opinion on their image of the EU (both negative and positive answers), EP election turnout 2009 and EP election trend in turnout 2009 to 2014.
Finally, I added a number of control variables:

European/National Event
The two variables indicate whether important European or national events took place during individual months (event = 1).European events include European Council meetings and Eurozone summits, national events include the parliamentary ratification of important EU Treaty changes or EU related international agreements.

Election
Assuming that national parliamentary elections will impact media coverage as well as parliamentary activities generally, a dummy control variable was added, with a value of 1 indicating elections taking place in that month.
A detailed overview over all dependent and independent variables can be found in Table A2 in the Appendix.

Empirical Analysis
Figure 2 provides an overview over the means of both types of visibility bias for the 17 newspapers.A first immediate result is that the seat share bias is overall far less pronounced than the activity bias.In addition, variation between newspapers and countries is also greater regarding the activity bias.Second, there is also a positive bias in some other papers, especially the French Le Monde, but the British papers The Times and The Sun are the only ones with a strong positive bias across both measures.
Turning to the factors impacting bias, I calculated both types of bias at the monthly level by newspaper and fit an ordinary least squares regression with standard er- rors clustered by newspaper for each.Since the UK is something of an outlier in the sample with the largest Eurosceptic party, which was also in government over most of the period under investigation, I also fit the regression omitting the data for the UK newspapers.The results are presented in Figure 3 (see also Table A3 in the Appendix).
As Figure 3 shows, there are stable patterns across all models, but omitting the UK data also leads to some important differences.Interpreting the coefficients, however, is somewhat challenging as they only signal the direction and strength of the impact of the variable.Thus, it remains unclear whether a positive effect is, for example, a decrease in the negative bias leading to more bal- anced coverage, or an increase in the positive bias leading to greater overrepresentation.The following figures therefore present the predicted values for the two types of bias for all significant variables (all other covariates held at their means) as the basis for the presentation of the results.The analysis draws on the full dataset unless stated otherwise.
Turning to measures of partisan bias first, Figure 4 shows the impact of the editorial stance of the newspaper on the degree of bias.Here, the impact of a Eurosceptic editorial line is positive and quite substantial, although it mainly reduces a negative bias.When the UK data is omitted, however, EU editorial line still has a positive effect, but it is no longer significant at the 95 per cent level (p = 0.089) for bias based on seat share.Thus, for seat share bias this result is driven to some extent by the UK.The effect of a match regarding the left-right position is positive as well, but not significant.There is overall only some support for H1 on partisan bias.
Turning to measures of structural bias, I assumed that more moderate Eurosceptics will generally have less newsworthiness compared to non-Eurosceptics than hard-line Eurosceptics and thus be less well represented in the media (H2).The effect is indeed significant for both bias measures (Figure 5), with one additional percentage point increase in the position towards the EU (e.g., from 2 = opposed to 3 = somewhat opposed) resulting in an increase in the negative seat share bias by about 5.5 percentage points and 7.5 percentage points for activity share bias.Both effects become even slightly stronger if the UK data is omitted, H2 is thus confirmed.
Yet although more hard-line Eurosceptics benefit from greater relative visibility-i.e., less negative bias-Eurosceptics overall do not benefit from a greater conflict potential regarding EU integration within parliament (Figure 6).Indeed, conflict potential has a fairly strong negative impact, both with and without the UK data.An increase by 1 leads to a decrease of the seat share bias by ca.0.6 percentage points and of around 2 percentage points for the activity bias.Although these values may seem small, the effect can be fairly sizeable given the range of the conflict potential from around 6 to a little over 19.Thus, the expectation that where the conflict potential within parliament is overall higher, the relative newsworthiness of Eurosceptics actually decreases, is confirmed (H3).
To explore the relationship between the party position and the overall conflict potential within parliament further, I fit a regression including the interaction between the two (see Table A4 in the Appendix for the full results).The marginal effect is negative in both cases, but only significant for activity bias (Figure 7a) both with and without the UK data.Thus, as the conflict potential within parliament increases, newspapers pay relatively   less attention to the parliamentary activities of more moderate Eurosceptics.To illustrate the effect, Figure 7b provides the predicted values for party Euroscepticism at 2 = opposed and 3 = somewhat opposed, showing that the difference between more hard-line and more moderate Eurosceptics is rather substantial.
Turning to the impact of public opinion, the results are surprising.Figure 8 confirms H4 on the positive relationship between public Euroscepticism and visibility in the full sample, but the effect becomes not only insignificant if the UK data is omitted, the coefficient for activity bias actually becomes negative.Thus, the effect is to a large extent driven by the UK, which is also the country with the strongest public Euroscepticism in the sample.
The expectation regarding the salience of EU affairs in public opinion (H6), in turn, can not be confirmed, quite the opposite is the case (Figure 9): salience has a significant negative impact on both types of bias, and these results remain the same when omitting the UK data.Thus, where the public regards EU affairs generally as more salient, Eurosceptic parliamentary actors have less news value compared to non-Eurosceptic actors.
I also included an interaction between public Euroscepticism and salience, but the effect is not significant in the full sample.It is, however, positive and significant for activity bias if the UK data is omitted (see Figure 10a, for the full regression results see Table A4 in the Appendix).Figure 10b visualises the ef-fect for public Euroscepticism at the minimum (32 per cent) and the maximum (60 per cent) level in the subsample.It shows that, outside of the UK, a low level of public Euroscepticism has a more positive impact where the public does not care very much about EU issues.The more public salience increases, however, the difference between low and high levels of public Euroscepticism diminishes.Where salience is fairly high, newspapers pay relatively more attention to the activities of Eurosceptics, or at least ignore them relatively less, the more Euroscepticism resonates with the public.

Discussion and Conclusion
The article investigated potential visibility biases in the print media regarding Eurosceptic parliamentary actors.As the article shows, the benchmark used to calculate visibility bias matters.Overall the coverage in most newspapers reflects the seat share of the groups much more closely than the share of activity.In other words, relevance and electoral support of Eurosceptic actors matter more for their relative media visibility than what they actually do in parliament.Still, in some newspapers, most notably the Austrian and Polish broadsheets as well as the German newspapers, the seat share bias is still fairly negative, reaching between 8 and 11 per cent underrepresentation on average.While the lack of a clear Eurosceptic stance of any of the broadsheets in these   three countries may explain their more negative bias, the same explanation does not hold for the Tabloids.Both types of bias are indeed less negative in the Austrian Neue Kronenzeitung, but the negative bias in the German Bild Zeitung is rather close to the German average.Here, at least, the question whether the ideological position of the Tabloids on the left-right scale does (Kronenzeitung) or does not (BILD) match that of the Eurosceptic PPG does seem to make a difference.The same is true for the real outlier in the study, the UK.While both types of bias are also negative in The Guardian, a consistent positive bias regarding the Conservatives was found in The Times and The Sun.
The results regarding structural bias show that hardline Eurosceptics are indeed relatively more visible in the media (i.e., suffer less from negative bias) than more moderate Eurosceptics.This suggests that news values such as 'conflict' or 'negativity' do indeed also matter with regard to bias and not just with regard to absolute visibility.Due to the same news values, by contrast, Eurosceptics are relatively less visible in the media, the greater the overall conflict potential regarding EU affairs within parliament.Although this affects moderates more than hard-liners, Eurosceptics, and their activities in particular, are less newsworthy where positions differ more strongly between the non-Eurosceptic groups in parliament as well.This mirrors the findings by Auel et al. (2018), namely that the conflict potential over EU politics within the governing coalition has a strong positive effect on the general coverage of parliamentary activities in EU affairs.To put it bluntly, where European politics within parliament are overall consensual and thus boring, Eurosceptics are more interesting for the media.Where, in turn, the conflict potential is overall higher, Eurosceptics are relatively less interesting on their own.
Turning to audience related news factors, the results for public opinion, in turn, were surprising.Stronger public Euroscepticism is indeed positively associated with relative visibility of Eurosceptic parliamentary actors, but the result is driven by the UK data and does not hold for the subset without the UK.This supports the argument by Brüggemann and Kleinen-von Königslöw (2009) that public attitudes towards Europe matter less in terms of newsworthiness than the public salience of EU issues, but in an unexpected way: the more salient EU politics are for citizens, the lower the relative visibility of Eurosceptic parliamentary actors in the media.Thus, although Auel et al. (2018) have found public salience to increase the coverage of parliamentary EU news more generally, Eurosceptic actors do not seem to benefit from that trend, at least not in relation to non-Eurosceptic parliamentary actors.This emphasises that if parliamentary EU politics have greater newswor-thiness in general, due to the overall conflict potential within parliament (see above) and/or due to the public salience of EU issues, the relative newsworthiness of Eurosceptics decreases.Outside of the UK, salience does, however, have an interesting effect when interacted with public Euroscepticism.Overall, the activity bias regarding Eurosceptics decreases the more the public actually cares about EU issues, but the effect also depends on the level of public contestation of the EU.Eurosceptics are more visible (or at least less underrepresented) where salience and contestation go hand in hand, i.e., where both are at a lower or at a higher level.
Before concluding, a number of limitations of the study need to be addressed: First, given the overall research interest related to the communication function of national parliaments, the data only captures the relative media visibility of domestic parliamentary Eurosceptic actors.It therefore cannot provide a full picture of visibility bias regarding Eurosceptic actors in general.Clearly, Eurosceptic parties such as UKIP, for example, are not only rather visible in the media despite having no parliamentary representation, but they also have a major impact on public discourses on EU affairs.
Second, the impact of governing status could not be tested in the present study given that only the British Conservatives were in government during the period of observation.As Vos (2014Vos ( , p. 2249) ) shows in her meta analysis of studies on politicians' visibility in the news, however, it not governing status by itself that hances news coverage, but 'political standing': 'Cabinet members, party leaders, and committee chairs have a higher political standing and therefore receive additional coverage' (Vos, 2014(Vos, , p. 2448)).Accordingly, the bonus attached to governing status in terms of media visibility mainly applies to politicians with a high political office, 'and not to ordinary politicians, such as members of parliament' (Vos, 2014(Vos, , p. 2448)).Green-Pedersen, Mortensen and Thesen (2017), for example, find governing status to be decisive for media visibility, but of the government actors appearing in the media, 79 per cent are either 'the government' in general or ministers (Green-Pedersen et al., 2017, p. 137).Yet the study here focuses exclusively on MPs, excluding Cabinet members from the analysis.With regard to the visibility of PPGs, in turn, the division between Eurosceptics and non-Eurosceptics does not neatly follow the government/opposition distinction in the present sample: in most of the member states in the study, a sizeable part of the opposition consists of non-Eurosceptic PPGs, while the governing British Conservatives, in turn, were in a coalition with the most pro-European party in the UK, the Liberal Democrats.The greater media visibility of the Conservatives may therefore have had more to do with the fact that the Conservatives had a number of intense internal conflicts over EU affairs during the time-most importantly over holding the referendum on the UK's membership in the EU and the related renegotiations of the UK's status in the EU-and thus only indirectly with their governing status.Still, given that the effect could not be tested, it can also not be ruled out.
Third, the time period under observation in this study (2010 to 2013) covers the most turbulent period of the eurozone crisis.This may not only have affected the coverage of parliamentary activity in general (see Auel et al., 2018), but also the relative visibility of Eurosceptic parliamentary actors in the media, calling for caution regarding generalisations beyond the observed period.As also emphasised by Auel et al. (2018, p. 641), however, the EU has hardly seen calmer times since 2013 due to the refugee crisis or Brexit.It would therefore be interesting to analyse whether the increasing politicisation of the EU impacted not only the electoral fortunes, but also the relative media visibility of Eurosceptics.
To come back to the question posed at the outset of the article, do newspapers support national parliaments in their communication function by providing fairly balanced coverage of different political perspectives on EU issues represented in parliament?As the analysis has shown, an answer is not easy as the assessment depends very much on what a bias is measured against.Overall, the coverage is more balanced when measured against the relative size of the party groups, although it is still negative in a number of newspapers.This is both good and bad news.While the analysis shows that Eurosceptics by no means generally dominate the media coverage of parliamentary EU affairs, few newspapers provide a truly balanced coverage, especially when using parliamentary activity as the benchmark.Thus, which benchmark to use depends on the specific subject of the study, but it is, in the end, also a normative question.Should Eurosceptic and Non-Eurosceptic actors be represented in the media according to their relative strength within parliament (and thus their electoral support), or should it matter what parliamentarians actually do in parliament?
What clearly seems problematic, however, is the fact that newspapers devote a large share of the parliamentary EU coverage (on average between ca. 25 per cent in Germany and just over 50 per cent in Poland) to parliaments (or parliamentary bodies) as institutions in EU affairs-and not the actors and groups working within them.While such articles may provide citizens with information on the parliaments' involvement in EU affairs, they offer little in terms of the different political positions on specific EU issues within parliament.Indeed, the share of articles mentioning both types of actors was surprisingly low in all countries-reaching above 20 per cent only in Finland.The media thus rarely present their readers with both Eurosceptic and more Europhile political views on EU issues simultaneously, which is precisely one of the main advantages of parliamentary involvement in EU affairs.Yet the possibly most important question remains so far unanswered, namely how this media coverage affects public perception of, and trust in, parliamentary representation in EU affairs-and whether it does so at all.

Figure 1 .
Figure1.Share of articles mentioning Eurosceptic and non-Eurosceptic parliamentary actors by newspaper (in percentage points).Note: Due to the extremely small number of articles (N = 9), the Polish Superexpress is omitted from the following analyses.

Figure 2 .
Figure 2. Types of visibility bias by newspaper (means).

Figure 3 .
Figure 3. Regression results.Note: The figure provides the OLS regression coefficients with 95 per cent CI (for the coefplot Stata package, see Jann, 2014).

Figure 4 .Figure 5 .
Figure 4. Adjusted predictions for seat share bias and activity bias dependent on EU editorial line with 95% CI.

Figure 6 .
Figure 6.Adjusted predictions for seat share bias and activity bias dependent on conflict potential with 95% CI.

Figure 7 .
Figure 7. (a) Marginal effects of party Euroscepticism on activity bias dependent on salience; (b) Adjusted predictions for activity bias dependent on party Euroscepticism interacted with salience, both with 95% CI.

Figure 8 .
Figure 8. Adjusted predictions for seat share bias and activity bias dependent on public Euroscepticism with 95% CI.

Figure 9 .
Figure 9. Adjusted predictions for seat share bias and activity bias dependent on salience with 95% CI.

Figure 10 .
Figure 10.(a) Marginal effects of public Euroscepticism on activity bias dependent on salience; (b) adjusted predictions for activity bias dependent on public Euroscepticism interacted with salience, both with

Table A4 .
Regression results with interactions.