The State of Europeanisation: between Clash and Convergence. A comparison of the media coverage of the 2019 European Elections in seven countries

Resumen The issue of the Europeanisation of national public spheres is a question as to how a discursive media space can be created within the EU. There are forces of convergence at work, such as networking within the borderless digital space. At the same time, there are counterforces: increasing nationalism and populists who identify ‘Brussels’ as a target for their criticism of elites. The vision of a European public sphere appears to share the same fate as the European project as such; as a result of years of crisis, optimism has given way to disillusion. Using coverage of the 2019 EU elections in seven European countries (a total of 57,943 articles from Germany, Hungary, Italy, Poland, Portugal, the Czech Republic, and the UK), we draw a picture of a heterogeneous EU public. What is particularly La “europeización” de las esferas públicas nacionales es una cuestion de crear un espacio mediático discursivo dentro de la UE. Hay fuerzas de convergencia, como redes en el espacio digital. Al mismo tiempo, hay fuerzas contrarias, un creciente nacionalismo y populistas que critican los élites a “Bruselas”. Aparentemente, la visión de una esfera pública europea comparte su suerte con el proyecto Europeo: después de años de crisis, el optimismo se convirtió en desilusión. Utilizando la cobertura mediática de las elecciones de la UE de 2019 en siete países europeos (57.943 artículos de Alemania, Hungría, Italia, Polonia, Portugal, la República Checa y el Reino Unido), dibujamos cuadro de un público heterogéneo de la UE. Resulta claramente que la


Introduction
The European elections of 2019 took place in the wake of the Brexit referendum and the success of Eurosceptic right-wing populist and far-right parties in national elections in countries such as Hungary, Poland or Italy, so that anti-EU sentiments were dominating the narrative even by then. The 2019 European elections appeared to be indicative of a legitimacy crisis within the European Union. At the same time, never before have so many issues affected the EU as a whole. During the euro crisis or the so-called 'refugee crisis', it was already obvious that polarising issues can cause the interests of national media audiences to converge (Benert & Pfetsch, 2021). This does, however, raise the question of whether different countries' media reporting on what is happening in other countries is what Habermas (2009) had in mind when he referred to the normative goal of a common public sphere, especially when such mutual thematisation appears to be couched primarily in terms of blame narratives (Müller et al., 2018).
We need to be clear about the difference between a toxic polarising clash and normatively desirable convergence to appreciate the current situation of high interest in the EU occurring simultaneously with rampant EU scepticism. It is not enough to measure the relative prominence of European issues, institutions, and developments in other countries in the coverage of national media-these must be placed in a cultural, social and political context. In order to show connections and interdependencies, we need to adopt a comparative approach. However, the individual publics are so heterogeneous that we need also to be aware of the risks of generalising from the particular.
All this presents researchers with serious challenges in terms of methodology. They need to collect representative data in different media systems, analyse content across language boundaries (which often cannot be adequately assessed without the necessary cultural expertise), and at the same time devise an analytical approach that allows for comparison.
We have addressed these challenges in two ways. On the one hand, this analysis was developed as an international research collaboration with communication scientists participating as experts in the media of their respective countries. On the other hand, the evaluations were based on computational content analysis, which allowed us to analyse 57,943 articles from seven countries. The articles were grouped in their original languages into country corpora and analysed with the help of unsupervised topic modelling (Blei et al., 2003), which reveals the latent thematic structures of a corpus. This made it possible to establish how much coverage had been devoted to which topics and, to a certain extent, how certain topic areas are covered in each country. Using this multi-method approach, we can combine the subjective insights of national experts with the objectifying approach of computational content analysis.
The overarching topic of our investigation is the 2019 election to the European Parliament. As the only election that calls all EU citizens to the ballots, this makes a perfect point of reference for analysing the degree of fragmentation or convergence of coverage in different national public spheres as the national media report on both on their EU neighbours and on issues of common concern. clear is that the phenomena of horizontal and vertical Europeanisation require more nuanced interpretations. While a high degree of horizontal Europeanisation indicates convergent and pro-European media coverage (as in the cases of Germany and Portugal), a high degree of vertical Europeanisation may indicate polarised publics or an unfree media landscape (as in the UK and Hungary). From a methodological point of view, the study shows that a combination of computational content analysis and international cooperation between scientists can advance research into the European public. europeización horizontal y vertical requieren interpretaciones diferenciadas. Un alto grado de europeización horizontal indica una cobertura mediática convergente y pro-europea (

como Alemania o Portugal), un alto grado vertical puede indicar un público polarizado o un panorama mediático poco libre (como el Reino Unido o Hungría).
Our study shows that the degree of Europeanisation varies from country to country. In countries where the EU is mainly regarded in a positive light and which enjoy a high level of press freedom, we observe a high degree of convergence in the coverage of EU-related topics, for example in Portugal and Germany. In other countries, the press is more nationally oriented. However, some countries where there is a low level of press freedom (Hungary, for example), EU-related topics occupy a surprisingly high amount of space in the media.

Dimensions of Europeanisation
Ever since the founding of the European Union, the question of the legitimacy of European governance and thus the importance of a European public has been a pressing issue. These questions have been made even more urgent in recent years by events such as the European financial crisis, Brexit and the success of Eurosceptic political parties (e.g. Risse The media are regarded as playing an important role in the emergence of a common European public sphere; however, the emergence of such a sphere does not require the same transnational media services to be consumed throughout Europe. In fact, even if European topics are discussed in national public spheres, a transnational common understanding of those topics is possible. Thus a European public can be seen as arising from the Europeanisation of national publics (Gerhards, 1993(Gerhards, , 2000. The coverage of European affairs by national media influences the degree of Europeanisation of national public spheres (Brüggemann & Kleinen-von Königslöw, 2009) and thus citizens' views on 'Europe' (Boomgarden & de Vreese, 2016). Several studies by Boomgaarden, de Vreese, and colleagues (Boomgaarden et al., 2010(Boomgaarden et al., , 2013Schuck & de Vreese, 2011) demonstrate that the visibility of the EU in media coverage potentially plays a key role in providing citizens with "information about the issues at stake and cues them about the importance of EP [European Parliament] elections" (Boomgaarden et al., 2013: 611).
Various indicators are used to measure the degree of Europeanisation (see Table 1). Some researchers argue that the intersection of national media debates (Wessler et al., 2008) is the most important condition for Europeanisation, which can be measured according to whether the same topics are discussed in the national media at the same time and within the same frame of reference (Kantner, 2003: 226) and how they are interpreted. An increasingly convergent approach to interpreting the issues can be regarded as the strongest indicator for Europeanisation, "as it can be interpreted as the result of an increasing openness and exchange between the public debates of the European countries" (Kaiser & Kleinen-von Königslöw, 2017: 800). Koopmans and Erbe (2004: 101) defined Europeanisation as a vertical or horizontal process: Vertical Europeanisation "consists of communicative linkages between the national and the European level"-for example when national media report on supranational issues or supranational actors such as representatives of the European Commission. Horizontal Europeanisation "consists of communicative linkages between different member states"-when different EU member states report about each other. When media report on national actors in other member states instead of just reporting on events, this is associated with a higher degree of Europeanisation.

Dimension of Europeanisation Mode
Vertical communicative linkages between the national and the European level Horizontal communicative linkages between different member states Intersectional same topics are discussed at the same time and within the same frame of reference Of course, the above definitions are somewhat simplistic; in practice, the differences between the dimensions tend to be rather blurred. Thus, mainly vertical topics (e.g. discussions of the decisions of central institutions) always have horizontal components, which are reflected in the news coverage. At the same time, vertical and horizontal topics are often intersectional. For example, exit negotiations between the UK and the EU (horizontal or vertical, depending on the perspective) are followed with keen interest in many EU member states, i.e., they generate intersectional attention.

State of Europeanisation
Empirical studies about the convergence of national media debates reach very different conclusions.
While Díez Medrano and Gray (2010) found a similar framing for reporting on European integration in seven countries and Wessler et al. (2008) found there to be convergence in the reporting on genetically modified food in five countries, they did not find any for the coverage of military interventions. Kaiser and Kleinenvon Königslöw (2017: 800) conclude that the degree of Europeanisation of national public spheres is highly topic-specific.
The result of Kaiser and Kleinen-von Königslöw's quantitative content analysis of the coverage of the euro crisis in two German and two Spanish online newspapers between 2010 and 2014 supports Risse's theory (2014) that the euro crisis contributed to the Europeanisation of national public spheres. The researchers noted a convergence of national media discourses, which has continued to increase. While there were still marked differences between the German and Spanish media analysed between 2010 and 2012, these differences decreased over the next few years so that by 2013-14 the media discourses on the topic were much more similar -"[i]t seems that the Europe-wide negotiation process on how to understand the crisis needed time for a common perception to develop (Kaiser & Kleinen-von Königslöw, 2017: 811)".
These differences were also identified by Picard (2015), who investigated the media coverage on the European debt crisis between 2010 and 2012 in ten European countries (Belgium, Germany, Finland, France, Great Britain, Greece, Italy, the Netherlands, Poland, Spain). His analysis shows that the media coverage followed a national rather than European approach and that the respective national narratives on the crisis and Europe were based primarily on national concerns and the individual country's relationship with the European Union. In contrast, Risse (2014: 1208) argues that the euro crisis promoted the Europeanisation of national public spheres and increased the degree of identification with the EU, as it resulted in national media paying more attention to the importance of Europe's political institutions. It also led to a greater sense of solidarity among many EU member states.
A study by Borchardt et al. (2018) Risse (2014: 11) is necessary for the emergence of a European public sphere.
Studies have also shown that the Europeanisation of national public spheres is more advanced in some countries than in others. One reason for this could be the duration of EU membership-the longer a country has been a member of the EU, the more the national media report on Europe and the EU ( As Demeter and Goyanes (2020) indicate, the Europeanisation process can also be influenced by subjective features such as nostalgia. Drawing upon cultural backlash theory, they investigated how a general feeling of nostalgia and a rejection of neoliberal values such as social and cultural diversity have an impact on Eurosceptic attitudes in Mediterranean countries that were badly affected by the financial crisis -especially Spain and Greece.
In most recent studies, the degree of convergence is generally established by comparing national media coverage of topics of a supranational nature-for example, the financial crisis, migration, or, most recently, Brexit. A characteristic feature of such issues is that they are relevant to the European Union as a whole, as well as to individual member states. These topics tend to combine the vertical, horizontal and intersectional dimensions, and it is difficult to disentangle these aspects when analysing such topics. Accordingly, these spotlight studies make it difficult to determine whether the Europeanisation of a country is primarily vertical, horizontal or intersectional. These distinctions are still relevant, however, because the presence of different dimensions of Europeanisation can indicate different dynamics: they can point to a European public that is growing together, but also to a populist and anti-elite framing of the EU. A very pronounced vertical Europeanisation, for instance, may indicate that national reporting is focused on 'the EU' as an outside unit, and thus also indicate a tendency towards Euroscepticism and populism. In countries where this is the case, we may expect to encounter a more confrontational tone in media coverage of the EU. Horizontal Europeanisation, on the other hand, may indicate an interest in the policies of neighbouring countries. This phenomenon is to be expected in smaller countries that are economically dependent on larger economies. Intersectional topics, on the other hand, can appear in countries characterised by an interdependence between their media public and the publics of other countries, for example, due to their historical-cultural or linguistic similarities.
Problems arise when the methodological perspective does not allow us to make distinctions between different dimensions of Europeanisation. The contradictory (and in recent years conspicuously stagnant) state of research suggests that some comparative studies have failed to take into account certain decisive characteristics, and that this is hampering research into the current state of the common European public sphere. We therefore propose an approach focused on distinguishing between the different dimensions of Europeanisation (see method, below).

Research questions
Based on the current state of research and taking into account the considerations described above, we formulated the following questions with the aim of differentiating between intersectional, horizontal and vertical Europeanisation: Which intersectional issues have shaped election coverage in all the countries we studied?

RQ2
Viewed from the horizontal dimension, what degree of Europeanisation can be discerned in the public spheres?

RQ3
Viewed from the vertical dimension, what degree of Europeanisation can be discerned in the public spheres we studied?
In addition, we place Euroscepticism within the context of the various dimensions of Europeanisation:

RQ4
What connections can we identify between Euroscepticism and vertical / horizontal Europeanisation?

The 2019 European Parliament Elections
Traditionally, EP elections are seen as "second-order elections" (Reif & Schmitt, 1980;Schmitt, 2005); in other words, they are viewed as less important by voters, parties and the media.  Reif and Schmitt (1980: 9), many citizens do not vote at all in 'second-order elections', and if they do, they use them as an opportunity to express their dissatisfaction with their national governments and vote for smaller parties-nowadays including "those that are critical of the EU" (Curtice, 2019: 9).
Despite this, Eurosceptic parties failed to become the dominant force that many observers had predicted. However, the elections "delivered a blow to the two major party groups" (Hobolt, 2019: 16). For the first time in four decades the centre-right European People's Party (EPP) and the centre-left Progressive Alliance of Socialists and Democrats lost their overall majority. Citizens voted instead for the smaller Eurosceptic, liberal and green parties. It is noteworthy that while public interest in the European Union has never been so high, the elections produced the most fragmented European Parliament so far.

Comparative Approach
The present study seeks to provide a comparative perspective on coverage of the European Parliament elections in countries with different political systems and journalism cultures ( Model, which are based on four variables, namely the structure of the media market, political parallelism, journalistic professionalisation and the role of the state (Hallin & Mancini, 2004: 296).
According to Hallin and Mancini (2004: 11) the Liberal Model, "characterised by a relative dominance of market mechanisms and commercial media" can be attributed to the United Kingdom [1] , the Democratic Corporatist Model, categorised by a "historical coexistence of commercial media and media tied to organized social and political groups, and by a relatively active but legally limited role of the state" to Germany, and the Polarised Pluralist Model, which shows an "integration of the media into party politics, weaker historical development of commercial media, and a strong role of the state" to Italy and Portugal [2] . The grouping of Poland in the Hybrid Liberal Model has long since ceased to be valid. After the conservative populist Law and Justice party (PiS) came to power in 2015, extreme political pressure was brought to bear on public service media (Kuś 2019)-and increasingly also on private media. The Polish government regularly maintains that foreign companies-especially German ones-own too high a stake in the Polish media and claims that this is an undesirable state of affairs; on several occasions, it has announced its intention to 're-polonise' the private media sector (Szynol 2020). Poland's ranking in the Reporters without borders world press freedom index fell from 18 in 2015 to 62 in 2020 (Reporters without borders, 2015, 2020a). Press freedom is also declining In the Czech Republic. According to Reporters without borders (2020b), "Czech journalists face major challenges, including threats to public broadcasting from governmentcontrolled supervisory bodies and online smear campaigns, with the result that there are fewer and fewer critical journalists". The Czech Republic's ranking in the world press freedom index fell from 13 in 2015 to 40 in 2020 (Reporters without borders, 2015, 2020a).
Hungary is grouped together with Bulgaria, Croatia, Romania, and Serbia in the Politicized Media model. These countries are in an "early consolidation phase of democracy" and in the "secondary transition stage of the media reform", classified as 'partly free' according to the 2015 press freedom index, and rank between 50th and 57th place in the 2014 democracy index (Dobek-Ostrowska, 2015: 19, 28). Recent developments confirm this trend: Hungary's leading opposition newspaper Népszabadság was shut down in 2016 (Polyák, 2016); and since its establishment in 2018, the pro-government Central European Press and Media Foundation has dominated the media landscape (e.g. Kennedy 2019; Reporters without borders, 2020c). Hungary's ranking in the world press freedom index fell from 65 in 2015 to 89 in 2020 (Reporters without borders, 2015, 2020a). In July 2020, pressure on the leading independent news site Index.hu-the last of Hungary's main independent media outlets-brought matters to a crisis point: after governmentcritical editor in chief Szabolcs Dull was laid off, almost the entire staff of Index resigned in protest, and a new editorial team was assembled by the government-friendly owner. (Polyák, 2020).

Trade-offs of representativeness
The analysis of different public spheres presents a number of challenges. Two scientific quality criteria are key: representativeness and comparability. The issue of representativeness (see Figure 1) arises when it comes to selecting the countries to be studied (macro level)-from a comparative perspective, it is desirable to include different countries and for there to be at least one example of each of the different types of systems under consideration. At the next level, we need to select representative lead media according to similar criteria (meso-level) to allow for a comparison between the selected countries based on their systemic differences (see section on data gathering). This step reveals the conflicting demands of representativeness at the different levels. And finally, it is important to select a representative topic for investigation (micro level), which again presents conflicting demands that have to be weighed up carefully. As already noted, topics of supranational scope, such as Brexit or the financial crisis, are too specific to allow us to explore the different dimensions of Europeanisation, as these only become apparent when the full range of the national topic agenda is taken into consideration. We therefore need to select an overarching topic that is relevant to all countries, but at the same time allows us to observe national sub-themes in each country.

Trade-offs of standardisation
In a comparative approach, in addition to selecting an object of investigation at each individual level, it is crucial to develop a sufficiently standardised investigation instrument to ensure comparability. Here, the challenge is to achieve the right balance between the standardisation of the method and ensuring that researchers are free to use their background knowledge and cultural expertise in order to discover unexpected connections.
The method design-data selection and the investigation instrument-must find a middle ground between these conflicting objectives.

Data Gathering
The study relies on seven large-scale corpora from countries from different media systems (

Levels of representativeness and operationalisation at country, media and topic level
The media outlet selection was guided by media reach, diversity (political direction, distinction between quality newspaper and tabloid) and availability in databases. [3] Our European research consortium, consisting of journalism and communication science scholars from renowned research facilities in the Czech Republic, Germany [4] , Poland, Portugal, Italian-speaking Switzerland and the United Kingdom, was tasked with selecting print media agenda-setters which have a high impact on the national news agenda and on public debates in the countries under consideration. In most countries a liberal, a conservative and a tabloid were chosen; in Portugal, a political profile distinction was not feasible as Portuguese newspapers do not have a specific political orientation.  [5] , Origo.hu (4,550) An overview of the countries, the media outlets, and the national article corpora For the United Kingdom, the liberal "Guardian", the conservative "Daily Telegraph", and the tabloid "Daily Mail" were selected, for Germany the liberal "Süddeutsche Zeitung", the conservative "Welt", and the tabloid "Bild", for Portugal the broadsheets "Jornal de Notícias" and "Público", and the tabloid "Correio da Manhã", and for Italy the centre-left "Repubblica", the centrist "Corriere della Sera" and the rightist "Il Giornale". In Italy, there is no clear demarcation between quality newspapers and tabloids and the distinctions are rather based on political leaning. For the Czech Republic, the left-wing "Právo", "Hospodařské noviny", a conservative daily focused on business and economics, and the tabloid "Blesk" were chosen, for Poland the liberal "Gazeta Wyborcza", the conservative "Rzeczpospolita", and the tabloid "Fakt", and for Hungary the pro-government newspaper "Magyar Idők", the pro-government online news portal "Origo.hu", and the liberal news site "Index.hu" [6] . For Hungary, two online news sites were chosen as selecting a conservative, a liberal and a tabloid newspaper would not have represented the reality of the Hungarian media landscape. The only newspaper which is critical of the government is the leftist newspaper Népszava, but it has a very small circulation and limited reach. The liberal online portal index.hu was until July 2020 one of the most popular news sites in Hungary.
All articles were published between 1 May 2018 and 31 May 2019, and include the words "European parliament elections" or "European elections" or "EU parliament elections" or "EU elections" or "European parliament" or "EU parliament", or at least twice the words "European Union" or "EU", or "European Union" and "EU" [7] [8] . For the gathering of articles, the databases Factiva, Nexis and Newton Media Search (the latter for the Czech newspapers "Blesk" and "Právo") were used. Hungarian articles were scraped directly from the news media's homepage. In total, our corpora consists of 57,943 articles (see Table 2).

Instrument of analysis: Standardised exploratory procedure with computational methods
The basis of the investigation is the topic modelling method LDA (Blei et al., 2003). This text-mining approach makes it possible to find latent thematic connections in large text corpora, independent of the language of the text. The starting point for this automatic pattern recognition is the correlation between linguistic units of similar meaning (words, phrases) and a similar distribution over certain linguistic contexts. In LDA, these relationships are modelled as topics and as probability distributions over all words. Each document in the corpus can also be represented as a probability distribution (over the topics). The LDA topics can reflect general topic structures as well as more subtle contexts-their granularity depends on certain variables: the nature of the corpus, the length and number of documents, and their thematic mix and composition; it also depends on the settings chosen by the researcher. Even unsupervised procedures require calibrationin the case of LDA, for example, the variable K is of great importance (Jacobi et al., 2016: 100). It indicates the number of topic clusters to be formed and is therefore decisive for the 'resolution' of the corpus: a larger K value means finer granularity.
Unsupervised approaches like LDA allow for an explorative-quantitative approach and are therefore well suited to procedures that require both standardisation and the openness of qualitative approaches. In the following, we describe our multi-stage comparative procedure based on LDA models: 1.) The multilingual data sets described above were fed into the system by means of different R packages ('LexisNexisTools', Gruber, 2020; 'tm.plugin.factiva'; Bouchet-Valat et al., 2019) and converted into a uniform data format (as 'textmeta' objects, see R-package 'tosca', Koppers et al., 2020). The pre-processing steps were identical for each corpus: For example, languagespecific "stop words" (generic words such as pronouns, numerals, conjunctions, etc.) were removed, as were punctuation marks. Upper case letters were converted to lower case. Only words that occur more than five times were transferred to the corpus, etc. 2.) LDA models were estimated on the basis of the corpus prepared in this way. The most important output is word lists (called the 'top words'), which are representative for a topic and also indicate its share in the overall corpus. The number of topic clusters to be generated (the variable K) was chosen depending on the consistency and interpretability of the word lists. To this end, we sent various models to our research partners-after coordinating this process, we selected a version with 30 topics. For the sake of comparability, we chose not to select different K values per corpus. We thought about taking different corpus sizes into account-however, higher K values did not yield any decisive gains in insights, even for larger corpora. 3.) The models we selected in this way were each labelled, i.e. the word lists were provided with headings. This enabled us to identify intersectional topics and their shares in the respective corpus. At the same time, we searched the first five top words for indicator terms for horizontal or vertical Europeanisation. Horizontal indicators were, for example, mentions of other EU countries or foreign politicians; vertical indicators were European institutions, representatives, mentions of the European capital Brussels, etc. 4.) Parallel to this quantitative evaluation, our international research partners qualitatively interpreted the LDA results in their capacity as country experts, drawing out inferences that would not have emerged from a more mechanical interpretation of the word lists. The aim was to explain the individual topics and the main focal points of coverage that are evident from the word lists, and to interpret them in the context of political and cultural factors specific to each country.

RQ 1
Which intersectional issues have shaped election coverage in all the countries we studied?
We identified three intersectional topics in the different countries' news coverage: Brexit, Donald Trump, and migration (see Figure 2). The most important pan-European topic for all the countries under consideration is Brexit. More than a fifth of the British corpus is directly associated with Brexit issues, while other topics deal with its indirect effects, for example on Ireland and Scotland. Scotland has long been more pro-European than England or Wales, and Scotland's backing of the 'Remain' campaign in the 2016 Brexit vote (with 62% voting to stay in the EU) reignited the long-simmering debate on Scottish independence. UK media coverage of the sub-theme Brexit/Ireland in the run-up to the EU election focused mainly on the potential impact of Britain's withdrawal from the EU on relations between Northern Ireland and the Republic of Ireland and the status of the border between the two countries.
The dramatic events surrounding the UK's exit referendum did not only dominate British reporting: The German corpus also reveals three topic clusters on Brexit -in addition to a general topic on the withdrawal of the United Kingdom, there is a topic primarily concerned with the role of Theresa May, plus a third one that mainly reflects the perspective of the EU. Apart from the German corpus, this vertical perspective is only found in the Czech media coverage. In Italy, almost 10 percent of the corpus is devoted to the topic of Brexit; news coverage even raised the hypothetical option of an 'Italexit'. In Poland, the issue is particularly relevant because many Poles are resident in the UK as guest workers. While all the countries report similarly on Brexit, the issue of migration is contextualised differently (see Table  3). In the Czech Republic, for example, migration is a symbol of a troubled Europe. The two topics related to migration have clear horizontal references to Italy (top words: Italy, Salvini) and Germany (top words: Merkel, CDU, CSU, Chancellor): Czech journalists report on the rise of Italy's far-right leader Salvini and political agitation in Germany-particularly in Chancellor Merkel's party-attributing both to immigration (comparative relationships can be also found in the Hungarian corpus). Similar horizontal references are found in the media of individual countries: while Germany focuses on the situation in Italy (especially the disputed legal status of those giving aid to refugees on the Mediterranean)-Italy, on the other hand, reports a lot on German political attitudes, with a special focus on Chancellor Angela Merkel. In both countries, the topic is prominent in news reporting. Horizontal references are also made in Poland, including the treatment of refugees in Hungary, which the anti-refugee Polish government regards as a kind of role model. In Hungary itself, the main narrative relating to the so-called 'refugee crisis' is the one created by the government: Hungarian Prime Minister Victor Orbán presents himself as the defender of a Christian Europe against the 'leaders of the EU'.
As with Brexit, 'Trump' is not really perceived as a pan-European challenge, i.e. vertical references are also missing here. Instead, Trump is often presented in connection with other geopolitical actors, such as Russia (e.g. in Italy) or China (in the UK). In Poland, discussions about the United States of America are closely linked to the US military presence in the country (top word: Nato).
Other EU-related topics also fail to make it onto the shared agenda of national publics. Even current global issues of huge importance, such as the climate crisis or the regulation of US-American platform companies, are not covered everywhere. In general, there is a lack of vertical, common themes that could have a Europeanising effect-for example, articles on joint projects that could highlight the advantages of joint action within the European Union. Very rarely are reports in individual countries presented from an EU perspective.

Figure 2: Intersectional topics
Share of topics identified as intersectional in the national corpora.

RQ2
What is the degree of Europeanisation in the public spheres we studied in the horizontal dimension?
On average, the corpora show a horizontal Europeanisation degree of 12.5 percent, i.e. on average, 12.5 percent of a corpus can be assigned to topics that are strongly associated with another EU member state. Germany stands out with the highest share of horizontal topics (18.2 percent), United Kingdom with the lowest share (7.7 percent). Six EU countries play a central role in German reporting on the EU elections (see Table 4): Italy (migration), Greece (Chinese investments in South-East Europe), France ('Franco-German partnership'), United Kingdom (Brexit) as well as a joint topic: Hungary and Poland (freedom of justice and rule of law in Central Eastern Europe).
In addition to Brexit, as mentioned above, the issue of migration is often the reason for horizontal references in different constellations, especially to Italy and Germany. In general, we observed that small countries rarely get any coverage-except for UK reports on Ireland (border problems due to Brexit) and the coverage of Spanish topics in Portuguese media (especially the election in Catalonia).  Surprisingly, the horizontal and vertical dimensions vary greatly between countries. Most topics with vertical references are found in the corpora from Hungary and the UK (see Table 5). When we add up the share of topics with horizontal and vertical dimensions, it becomes clear that Hungary has the highest general degree of Europeanisation of all the countries we studied. While the media in other countries used the occasion of the EU elections to report on national political issues (especially in Portugal and Italy, where the EU election was considered a test run for upcoming national elections), Hungary and UK focused on a range of EU topics.

RQ4
What connections can we identify between Euroscepticism and vertical / horizontal Europeanisation?
As we have seen, analyses of horizontal and vertical Europeanisation produce a disparate picture. This shows that horizontal and vertical dimensions of Europeanisation have different functions within the different public spheres. A comparison of our results with demographic data on the approval ratings accorded to EU membership within individual countries (Schulmeister et al., 2019) provides some hints as to how to account for them.
It is striking that the countries with the highest approval ratings, Portugal (69 percent) and Germany (76 percent) are also the countries with the highest proportion of horizontal reporting. A positive attitude towards the EU seems to correlate with a country's interest in its European neighbours. It is in Portugal and Germany that we find the highest degree of convergence of public opinion with regard to the EU.
In contrast to this, the countries with the lowest approval ratings regarding EU membership are the Czech Republic (33 percent) and Italy (36 percent). These countries are also the ones with the lowest degree of vertical references. The EU itself, its institutions and representatives, played only a marginal role in the election coverage of these countries-overall, national issues were the main focus of attention. In the Czech Republic, for example, many articles deal with the resignation of ministers, or insurance and healthcare, stressing the lack of medical doctors. In Italy, the agenda features mostly national issues, such as the TAV project (high-speed train between Turin and Lyon) or the budget law. In general, the Czech Republic and Italy are thus representative of a group of public spheres with a low degree of internationalisation, which can be described as national in the traditional sense.
The third group is the one with average approval ratings (UK: 43, Hungary: 61, Poland: 68, EU average: 61). It is particularly striking that these countries show the highest difference between vertical and horizontal Europeanisation values. More often than on average, the EU is presented as a separate entity, especially in the context of the conflicts between the respective member state and the European Union. Other member states receive comparatively little attention. This characteristic is most pronounced in Hungary, where the press mainly reflects the government's EU-critical view. Conflict topics can also be found in Poland: the topic of turbulent relations (in the wake of the parliamentary elections of 2015) between the PiS government and EU institutions formed an important part of the pre-electoral media agenda, highlighting problems such as European Commission demands on the Polish government, rule of law, and changes in the Polish judiciary system. In this sense, a high divergence between the dimensions points to a confrontational attitude towards the EU-a form of populist Europeanisation.

Discussion
This paper took as its starting point the realisation that in the past, the theoretically described dimensions of Europeanisation could not be brought to bear in a methodologically satisfactory way. A new comparative, computer-based approach was used to ascertain whether contradictions in the current state of research have anything to do with this lack of differentiation. In fact, we were able to show that horizontal and vertical Europeanisation do not correlate in any reliable way-the lack of correlation indicates that these are two different dynamics that must be interpreted differently. While a high degree of horizontal Europeanisation indicates a convergent and pro-European media coverage, a high degree of vertical Europeanisation (with low horizontal Europeanisation) may indicate a polarised or unfree media landscape.
The contradictory results of previous studies, for example on the euro crisis (Risse, 2014;Picard, 2015;Müller et al., 2018), can thus be reassessed: a focus on the European Union may in some countries be associated with confrontational, populist reporting, which does not necessarily go hand in hand with an increased interest in other member states.
In addition to confrontational and convergent countries, we were able to identify nation-state oriented countries with a low overall degree of Europeanisation. The developed classification is not directly related to the countries' media systems. However, it is striking that the countries in the sample with the lowest degree of freedom of the press (Hungary and Poland [10] ) can be assigned to the confrontational group, while the countries with the highest degrees of freedom of the press (Germany and Portugal) are representative of the convergent type. The concept of Europeanisation must therefore also be seen in terms of whether or not it helps to promote EU ideals-confrontational reporting could be viewed more as 'dark Europeanisation' in conflict with EU norms.
It is important to acknowledge the limitations of our study. Though newspaper readership continues to fall steadily in all the countries under consideration, our study focused on print media and took only a few online news sites into account. This is due to the study design and the limited free-of-charge access to articles in electronic archives. Ultimately, the databases Factiva and Nexis, which contain mainly articles from print media, suited our purposes. As the Hungarian media were not present in these databases, we had to scrape the data-a practice that comes with technical and legal uncertainties. Due to our selection it was not possible to represent each country's media landscape in its entirety. A common research database containing print and online articles from all European countries, from which it is possible to download data in a consistent format, would definitely facilitate the carrying out of such large-scale studies.

Conclusion
The issue of Europeanisation of national public spheres is also a question of how a discursive media space can be created within the EU. There are forces of convergence at work in this process, such as networking in the borderless digital space, cross-border journalism, and the increasing political and social interconnections within the European Union. At the same time, there are counterforces: rising nationalism, populists who identify 'Brussels' as a target for their criticism of elites, restricted freedom of the press, and dwindling journalistic resources due to the crisis of media business models. The vision of a European public sphere appears to share the same fate as the European project as such-as a result of years of crisis, optimism has given way to disillusionment. Development perspectives have been reassessed, and there are no longer any easy answers.
Against this background, the present paper is intended to help us to arrive at a more nuanced understanding of the complex phenomenon that is Europeanisation. To do this, we need to be aware of its vertical and horizontal dimensions, relate it to political and cultural events, and compare the dynamics of different countries. Computational methods now make it possible to compare representative media corpora across language barriers. However, without the cultural background knowledge of human experts, the results of such methods would remain one-dimensional.
Our study shows that the combination of computational content analysis and international cooperation between media scholars can advance research into the European public, allowing such research to move forward following a period when it became mired in contradictions. It offers some answers to researcheconomic dilemmas, allows for a standardised exploratory approach, and could provide a model for future studies.