People, Money, Ireland: An Analysis of British and Irish Online Media in the Year Following Brexit
Eoin Hennessy 11105151
eoin.hennessy@student.uva.nl
Master’s Thesis
Graduate School of Communication
Research Master’s Programme Communication Science
Damian Trilling February 2nd, 2018
Abstract
The study of intermedia agenda setting effects has been an extensively explored topic in communication science for a number of years. Despite this, there has been limited research conducted on intermedia agenda setting effects between countries, particularly in a time of political crisis. Therefore, it was the aim of the present study to examine the online news coverage of Brexit in Ireland and Britain in the year following the referendum. The impact of Brexit, along with Britain and Ireland’s tumultuous history, provided the ideal setting under which to explore international intermedia agenda setting effects in relation to a dramatic political event of relevance for both countries. Using a novel scraping protocol of Google, the current research was able to obtain over 45,000 articles in the year following the Brexit referendum. This data was then analysed using topic modelling and time series analysis to identify patterns in the online coverage of Brexit. The results of the
analysis indicate that British media were the dominant agenda setters on topics relating to Brexit. Additionally, the research also found that tabloid sources set the agenda for most Brexit issues in both countries. The findings of the study demonstrate that intermedia agenda setting occurs in online media between countries; an issue of particular interest in a post-colonial context.
Introduction
On the 23rd June 2016, Britain voted in favour of leaving the European Union (EU). The result, with a 51.9% “Leave” vote, sent shockwaves throughout Europe (Goodwin & Heath, 2016). To some the result merely confirmed a trend in
Euroscepticism on the continent, while others remained dumbfounded by the outcome (Goodwin & Heath, 2016). Over a year since it was announced, the implications of the referendum are still not fully realised. These implications are of even greater significance to Britain’s only remaining land border to the EU, Ireland. Aside from Britain; Northern Ireland and the Irish Republic are at the forefront of the Brexit aftershocks, whether they like it or not. Brexit has reignited conversations of Irish unity, as well as fears of a hard border and possible conflict resurgence (Stevenson, 2017). Further concerns have also been raised in relation to the economic and customs border that might be implemented post-referendum. However, are these concerns echoed in the media of Britain, as well as Ireland? The role of the media during the referendum was seen as significant, in particular from anti-EU, tabloid newspapers such as The Sun and The Daily Mail (Seaton, 2016). The role of tabloids like these in determining the public agenda cannot be overstated, yet there is little scientific
research observing how these sources interact between Britain and Ireland. (Vonbun, Kleinen-von Königslöw, & Schoenbach, 2016). As a result, it is the aim of this study to explore media coverage of Brexit from both an Irish and British online media perspective in the year following the Brexit referendum.
There are several reasons for focusing on online media post-Brexit
referendum. The ease and accessibility of online media has taken power from single mainstream outlets in determining the public agenda (Vonbun et al., 2016). Instead, recent research has found online news to be the main agenda-setter in today’s media
environment (Cushion, Kilby, Thomas, Morani, & Sambrook, 2018). No fixed deadlines or publishing schedules enable online media to be at the forefront of news (Vonbun et al., 2016). As a result, the media can react to news events much quicker than ever before. The sheer scope and scale of online news has resulted in a shift in agenda setting research to focus on intermedia agenda setting, i.e. since the media sets the public agenda, “how is the media’s agenda set?” (Rogers, Dearing, & Bregman, 1993, p.79). Findings presented by Vonbun et al. (2016) suggest that media outlets change roles between being followers or leaders depending on the issue. So who sets the agenda in relation to the issue of Brexit? Recent research carried out by Cushion et al. (2018) found that right-wing papers (mainly tabloids in this case) set the agenda in relation to stories about the 2015 UK General Election. Since there is limited research exploring the transfer of agendas between Britain and Ireland, the present study seeks to investigate the intermedia agenda setting effect between the two nations in the year following Brexit. As such, an overarching research question was formulated: Can an intermedia agenda setting influence be identified between British online media and Irish online media in the year following the Brexit referendum and if so, what does it look like?
Examining a potential intermedia agenda setting effect in online media the year after the Brexit referendum is important for a number of reasons. Firstly, there is limited analysis exploring the interaction between Ireland and the United Kingdom’s (UK) media institutions, with almost no empirical findings related to Brexit.
Secondly, if there is a significant difference between Ireland and Britain’s coverage of the Irish border question, it may point toward a further division between Britain and its gateway to Europe. Thirdly, the year following the Brexit referendum offers an opportune timeline to observe intermedia agenda setting effects, due to a number of
key events and topics, such as the Conservative-DUP agreement. Finally, online news remains understudied as it is often not systematically archived. The current research offers a novel method of big data collection in which Google’s site-specific search function is utilised. I will demonstrate how this new method of data collection can facilitate a comprehensive analysis of specific online-topics and timelines.
Theoretical Framework Online Media and Euroscepticism
As the Internet becomes more prevalent in society, so too does the presence of online and mobile news. The Internet has facilitated the rise of fast paced, largely free-to-read journalism. The widespread availability of the Internet has transformed the modern media environment. The Pew Research Centre now suggest that 54% of people in the UK get their news online, with that number increasing to 74% amongst young people (Pew Research Centre, 2018). Not only has the format of the news changed as a result of the Internet, but also the way the public consumes news media. From 2012 on, the uptake of mobile news and cross-media consumption reached new records, in contrast to print media’s continuing decline (Westlund & Färdigh, 2012; Roche, Pickett, & Gertz, 2016). These developments highlight the importance of studying online news sources in today’s diverse media landscape.
The Internet and online news are distinct from other forms of media for a number of reasons. To begin with, aside from television, the Internet is the only media system that people start using from a very young age. As a result, it means the Internet plays an important role in shaping the public’s worldview (Rideout, Foehr, & Roberts, 2016). Furthermore, the Internet and mobile technologies have enabled people to access news media at any time and any place (Roche et al., 2016). For young people,
the Internet has become a particularly important source for obtaining information on international, as well as local issues (Rosenstiel, Mitchell, Purcell, & Rainie, 2011). This fact is particularly relevant to the present study, as young people turned out in record numbers to vote in the Brexit referendum (Goodwin & Heath, 2016).
There is little doubt that media matter when it comes to the democratic process and the public’s political attitudes (de Vreese, 2007). Of these attitudes and important for the present study, is how the media have been accused of promoting
Euroscepticism (de Vreese, 2007; Michailidou, 2015). Euroscepticism is the view that the EU and other European institutions do not sufficiently provide for individual EU nations and may hinder the development of national prosperity. Euroscepticism is on the rise and, for the UK, has culminated in a Leave vote winning the referendum (Scharkow & Vogelgesang, 2009). Torreblanca & Leonard (2013) have even gone so far as to liken Euroscepticism to a virus that has spread across the continent. The connection between Euroscepticism and the media has been highlighted in the past. De Vreese (2007) concluded that, “Euroscepticism is, at least partially, a function of the diet of information that citizens consume about European affairs” (p. 280). Of this information, online media appear to be pivotal in the amplification of EU discontent. In her content analysis of mainstream online news media in several EU member states, Michailidou (2015) concluded that online media contribute to the deepening of existing conflicts between member states rather than contributing to a common European understanding. By consistently discussing EU politics from a national perspective, online media contribute to an on-going EU divide (Michailidou, 2015). This is a particularly important finding in the case of Brexit.
The current literature on online media and Euroscepticism points to a continued need to analyse these concepts, particularly in relation to one another. Of
even greater significance, is the need to identify the media agenda setters, as they will be the ones who determine the public salience of these issues and their attributes. Since Brexit was largely seen as a culmination of discontent with the EU, the post referendum environment offers a distinct opportunity to examine one of the 21st Century’s most important political events (Stevenson, 2017).
Intermedia Agenda Setting
Paramount to studies on Euroscepticism and the media is agenda setting theory. In its most basic form, agenda setting theory explores the transfer of salience of objects and their attributes from the media agenda to the public agenda (McCombs & Valenzuela, 2017). Initial research by McCombs and Shaw (1972) found there to be a high correlation between issues on the media agenda and the public agenda. This correlation echoes the impact of news patterns on public perceptions of issues, for example, how Eurosceptic views presented in the media may foster an “us (the nation) and them (the EU)” perspective on the public agenda.
Furthermore, the transfer of salience of objects and their attributes between agendas has resulted in changes in individuals’ attitudes, as well behaviour
(McCombs & Valenzuela, 2017). As such, agenda setting has been a significant theory worthy of consistent exploration. The way the media impact public perceptions is a vital area of study, particularly if it can impact governmental policy (Eissler et al., 2014). The impact of the media in the political process has been far reaching, but ambiguous (Boydstun, 2013). In her comprehensive overview of the status of the media within the political process, Boydstun (2013) argues that the media’s effect is not constant, but rather intermittent. She argues that the media generally ignore most topics, but pay particular attention to selective subjects. This selectivity could have far
reaching consequences, particularly in the case of Brexit negotiations. In their
analysis of agenda setting research, Eissler et al. (2014) argued that the “media create and frame public perceptions regarding social problems, and the political agenda space is built on those realities, which are again perpetuated by the media agenda” (p. S78). If the media’s impact extends to the Brexit negotiations it could affect its outcomes.
Furthermore, even politicians themselves consider the impact of the media to be far reaching (Van Aelst & Walgrave, 2011; Walgrave, Soroka, & Nuytemans, 2008). In their survey of members of the Belgian parliament, Walgrave et al. (2008) found that politicians perceive the media’s agenda setting impact to be very high. Moreover, the politicians considered the media to be one of the primary political agenda setters, second only to the Prime Minister. Similarly, in a study of political systems in Denmark, Sweden, Belgium and the Netherlands, Val Aelst and Walgrave (2011) found that members of parliament consider the media to be of great
importance, and in some cases, even the most important political agenda setter. If politicians consider the media to have this great effect on the political agenda, with the possibility of policy also affected, it is imperative that media patterns are observed, particularly in relation to the Brexit negotiations. As previously
mentioned, one important facet of agenda setting research is finding out which media content influences and relates to other media content (Rogstad, 2016). Intermedia agenda setting theory is based around the idea that journalists and news contributors have a tendency to be influenced by their peers at other news outlets (Breed, 1955). As a result of this, the media tend to echo the views of particular types of sources, whether intentionally or not. One of the earliest studies in intermedia agenda setting found elite media, such as The New York Times, had an effect on the smaller news
outlets (Reese & Danielian, 1989). McCombs (2014) postulated that this was due to journalists validating their work by looking to elite media organisations.
The importance of intermedia agenda setting is further exacerbated with the prevalence of online news. News websites have dramatically changed the way people interact with news (Harder, Sevenans, & Van Aelst, 2017). The so-called “hybrid” news environment cultivated by the Internet has enabled news organisations to diversify in their presentation of the news (Harder et al., 2017). For example, social networking sites like Twitter and Facebook have enabled journalists to gather information directly from the source, e.g. in the form of civilians live Tweeting events. Thus, the contemporary news landscape may have shifted the power away from elite media organisations to something else.
The online news setting requires no fixed schedules and as a result enables journalists the opportunity to provide a constant flow of information (Russel-Neuman, Guggenehim, Mo Jang, & Young Bae, 2014). Furthermore, the flexibility of the Internet means that journalists no longer have to submit a completely “finished” product. Instead, journalists can edit as they please and adjust as the stories unfold. Since online articles can be updated so easily, it means that accuracy has taken second seat to publication time. Using a time series analysis of traditional media (radio, television and print), online media and Twitter, Harder et al. (2017) found news websites to have a significant agenda setting effect on the other media. The speed with which news can be disseminated online further emphasizes the importance of identifying the media agenda setter. For example, in reporting on a news story first online, other media might be inclined to validate (or at least compare) their reporting with the initial story. This could mean that relatively one-sided sources are setting the
agenda for other online news sources to follow. In the case of Brexit, this could mean that nationalistic, Eurosceptic papers are setting the agenda for the UK and Ireland.
As previously mentioned, both the time constraints and editing environment have changed with the advent of online publishing. Not only has publication time affected the way news is reported on, but also the availability of different sources of information from around the globe. The widespread availability of the Internet has enabled the public to not only rely on local news sources, but rather on a plethora of different worldwide publications. Unfortunately, the globalisation of the news environment could also mean its homogenisation. While intermedia agenda setting research has primarily focused on the agenda setting effects between media (e.g. traditional vs. online), there has been limited research examining its effects between two different countries. Guo et al. (2015) looked at intermedia agenda setting effects between the USA, China, Poland and Taiwan on coverage of the Iraq war. They observed heterogeneity in the reporting between the nations. While Guo et al.’s (2015) results are revealing about international patterns of media coverage, they fail to say anything about online media. Furthermore, since only the USA and Poland were involved in the Iraq War, their results reveal little about media coverage between effected countries.
The current research aims to utilise coverage of the post-Brexit aftermath as a means of comparing online media between nations in a time of political crisis. This should shed light on the intermedia agenda setting relationships that occur between nations involved in a significant political shift. The importance of this kind of analysis is amplified by Britain and Ireland’s long, tumultuous relationship. Following from this, two further research questions were formulated:
RQ1.1: To what extent do sources from Britain set the online news agenda for Ireland?
RQ1.2: To what extent is Euroscepticism promoted in the online news media of Ireland and Britain, and who (if any) is the dominant intermedia agenda setter of Euroscepticism?
Furthermore, based on the findings by Cushion et al. (2018):
RQ2: To what extent do tabloid sources set the online news agenda for other types of sources?
Brexit and The Irish Question
The Irish Question, or the Irish Problem as it was sometimes called, was a term predominantly used by British nationals to discuss questions of Irish nationalism and independence (Pašeta, 2003). These discussions normally centred on issues of British involvement in Ireland, as well as border controls and the conflict between Nationalists and Unionists in Northern Ireland. The Irish Question is seen as one of the most significant political problems to face Britain during the 19th and 20th Century (Mansergh, 1968), and with the result of the Brexit referendum, these issues may once again be at the forefront of political discourse.
Michel Barnier, the lead negotiator for the EU following the Brexit referendum, cited the Irish border issue as being one of the three priorities of the negotiations along with citizens’ rights and financial commitments. Or as one Irish journalist put it, “People, money, Ireland” (O’Toole, 2017, para. 1). Indeed, the significance of a border in Ireland is not to be understated, particularly with its long, bloody history. From the late 1960s up until the end of the millennium, Northern Ireland, as well as parts of the Republic and Britain, experienced what were known as
the Troubles. During this time, 3,600 people were killed, with 90% of these
causalities at the hands of illegal paramilitaries (Mulholland, 2002). In total, 40,000 Irish people were injured during the Troubles, equalling almost 3% of the population (Mulholland, 2002). An extremely complex history between Irish Unionists and Irish Nationalists in the North was at the heart of the Troubles, which saw pacification in 1998 with the Good Friday Agreement (Mulholland, 2002). Although Ireland has since experienced years of peace, Brexit seems set to dredge back up unpleasant memories, especially with respect to a hard border on the island of Ireland. Irish people are now left wondering, how does one figure out the logistics of an EU frontier in Ireland without unravelling the complex diplomacy that ended a 30 year conflict?
Even as recently as 2013, relations in Northern Ireland showed signs of strain. A dispute over the removal of a Union Jack in Belfast City Hall required a special envoy from the US to chair negotiations (Stevenson, 2017). After a year of debate, the negotiations broke down and lead special envoy, Richard Haass, to remark that
political violence in Northern Ireland “could very well re-emerge as a characteristic of daily life” (BBC, 2014, para. 10). Stevenson (2017), in his analysis of Brexit and the Troubles, also commented on how a hard border with stationed British troops might be the final provocation for the Irish Republican Army (IRA) to return to arms.
While these issues may loom large over the minds of the Irish public, it is unclear whether they are as much of a priority for Britain. The result of the referendum has opened up a logistical can of worms for the UK, in which every aspect of life is likely to be affected. As a result, it might be the case that Britons think they have “bigger fish to fry”. This should be of great concern to Northern Ireland which stands to lose the most as a result of Brexit, since their EU subsidy is higher per capita than any other UK constituent (Stevenson, 2017). Even despite Barnier’s
priorities, the BBC failed to mention Ireland even once in their list of key issues for the Brexit negotiations (Ahmed et al., 2017). With this in mind, a subsequent research question was formulated:
RQ3: Can an intermedia agenda setting influence be identified on the Irish border issue between British online media and Irish online media?
Since the border remains such a key issue to the people of Ireland, it is important to determine whether these feelings are paralleled in the media of Britain. Furthermore, it is important to see who is setting the agenda on the issue of the border, since the coverage might be different in each country. For example, British sources might focus on the impact to trade post-border, while Irish sources might focus on conflict. As such, it is imperative to see who is setting the agenda in relation to the border issue.
The aim of the present study is to provide a detailed comparison of Irish and British media institutions in the post-Brexit-referendum aftermath. Such an analysis should be able to assess whether the Irish Question is addressed by all nations involved, or whether it has taken a backseat in the British media. Furthermore, it should provide a wider context for testing intermedia agenda setting effects between nations in a time of political uncertainty.
Methods
The present study utilised computer-assisted content analysis to gain information on media patterns related to Brexit in Irish and British online news
sources. Time series analysis and natural language processing were then performed on the data. Additionally, the research employed a unique method of data collection,
exploiting Google’s site-specific search function. Since this method of data collection is new to the field, it will be outlined in greater detail below.
Case Selection
To begin with, an exhaustive list of online news sources from the UK and Ireland were compiled. Sources were seen as eligible as long as they were from any of the 5 nations involved in Brexit (i.e. Ireland, Northern Ireland, Scotland, Wales and England) and had an online presence. During the selection of sources, an effort was made to try balance the number of sources from the UK and Ireland, as well as both tabloid and broadsheet. The intention was to ensure that the selected sources would act as an accurate representation of online news sources available to the public. Although over 20 sources were selected to begin with, this number was reduced to 16 once pay-walls and the number of articles were accounted for. For example, although a popular local newspaper, Dublin’s Herald was excluded from the final list of sources since they did not publish enough online articles in the year following the Brexit referendum. The final list of sources included 8 from Britain (The Daily Mail, The Daily Express, The Daily Star, The Daily Mirror, The Sun, The Guardian and The Independent) and 8 from Northern Ireland and the Republic (The Irish Examiner, The Irish Independent, Buzz, The Irish Daily Mirror, The Belfast Telegraph, The Journal and The Sun). The national media of Ireland (Raidió Teilifís Éireann, RTE) and Britain (British Broadcasting Corporation, BBC) were also included in the analysis. An overview of the sources is presented in Table 1.
Although The Irish Independent and The Independent are completely separate entities, the same cannot be said of the Irish and British Sun and Mirror. The Sun, despite having the same appearance and owners (News UK) in both the UK and Ireland, publishes different content in both nations. For example, in quite a funny gaffe in 2006, The (British) Sun’s review of the film, The Wind That Shakes the
Barley, stated that it was “designed to drag the reputation of [Britain] through the mud”, while The Irish Sun hailed the film as giving “the Brits a tanning” (Greenslade, 2006, para. 1). Similarly, The Daily Mirror publishes different content in both nations
Table 1. Number of Articles in the final sample
Source Nation Type Nº of Articles
BBC British Non-Tabloid 2704
Belfast Telegraph Irish Non-Tabloid 1493
Buzz Northern Irish Tabloid 36
The Daily Express British Tabloid 7143
The Daily Mail British Tabloid 4790
The Daily Star British Tabloid 784
The Guardian British Non-Tabloid 10410
The Independent British Non-Tabloid 6926
The Irish Examiner Irish Non-Tabloid 855 The Irish Independent Irish Non-Tabloid 1806
The Irish Mirror Irish Tabloid 272
The Journal Irish Non-Tabloid 492
The Mirror British Tabloid 2442
The Sun British Tabloid 3953
The Sun Irish Tabloid 529
RTE Irish Non-Tabloid 950
but maintains the same owners and style. This makes both papers ideal sources to compare and examine in the aftermath of Brexit.
Data were collected for a 13-month timeframe from June 1st 2016 to June 30th 2017, thus mostly after the day of the referendum on June 23rd 2016. This enables the analysis to focus on the consequences of Brexit, and to disregard the scant discussions proceeding the referendum, when a large number of sources had not even anticipated the Leave vote succeeding. The 22 days prior to the referendum serve as a sanity check, as one would expect a bump in articles on referendum day. Once the United Kingdom’s exit from the European Union was confirmed, news sources started worrying about its impact for Ireland, including the possible hard border between Northern Ireland the Republic. Furthermore, the year following the referendum saw Ireland even more involved in Britain’s political discourse, as Theresa May’s Conservative party struck a deal with Northern Ireland’s right-wing Democratic Unionist Party (DUP).
Data Collection
One of the issues involved in getting online news data is that it is difficult to obtain retrospectively. Although web feeds, such as RSS, have made it easy to gather large amounts of news data, they cannot be used to gather old articles. Additionally, search functions of online news websites and databases (e.g. LexisNexis) are often messy, making it difficult to obtain backdated articles. As such, Google presented the most efficient way of retrieving a large number of articles within the study timeframe. The data was collected using the programming language Python in a virtual Linux environment. For each of the news websites, an individual parser was written to extract only the relevant information, such as title, text, teaser and date of publication.
The parsing was tested and carried out using XPaths, a computer language used for addressing parts of XML documents (Barton et al., 2003).
As well as parsing the news websites themselves, it was also necessary to parse the source of all the data, Google. This particular method of data collection made use of Google’s advanced search facilities, which enables a user to search specific sites for specific topics (also at specific times). This enabled the research to obtain specified information on the world’s dominant search-engine. For each news source, a special Google uniform resource locator (URL) was generated to search a particular term, on a particular source, at a particular date. For example, a URL would be generated to search the term “Brexit” on the BBC website
“http://www.bbc.com/news” for June 1st
2016. A special loop was then created in order to obtain the search results for each day of the chosen 13-month period. A scraper was designed to store the URLs, as well as their names. Furthermore, the web-scraper was designed to obtain every possible article on a particular date by scrolling through the pages that were available after the search.
The process of scraping from Google turned out to be much more challenging than it seemed initially. The search-engine is quick to block any scraping requests and will even implement a temporary ban on those who disobey. It is worth noting here that Google itself also scrapes content from other search-orientated sites (i.e. Yelp, TripAdvisor, Amazon) to use for its own benefit (Nsehe, 2012, para. 2; Garside, 2015, para. 8). Companies that complained about the scraping to Google were also
threatened with removal from the search engine rankings (Garside, 2015, para. 8). We therefore further investigated the possibility of scraping Google search pages for scientific use, and were able to do so using Tor. Tor is a free software and open network which enables its users to defend against traffic analysis through the use of
anonymous Internet Protocol (IP) addresses. In order to get access to the Tor network, we set up a Tor gateway and a http proxy using the packages Privoxy and PyTorCtl in a virtual Linux environment.
The data collection procedure was as follows: To begin with, an anonymous proxy and gateway to the Tor network was used to view Google search results’ pages in the form of “Brexit site:(online news source URL)” on a specific date. These pages were scraped for a URL and a title. If there was more than one page (as was often the case), a loop built into the Python script would retrieve a fresh IP address using Tor and then scrape the subsequent page until all pages were complete. In addition to changing the IP address to avoid Google blocks, a ‘sleep’ function was also used to delay requests made to the Google page (usually set between 20- and 40-second intervals). Once all of the pages for a certain date were complete, another loop in the script created a fresh search result with the next date (again, the IP address was
changed in conjunction with the sleep timer). These Python scripts were then run until all the URLs had been obtained for every relevant source over the desired time period. The URLs were stored in an ElasticSearch database. Once all of the URLs had been obtained, each corresponding webpage was downloaded and added to the
ElasticSearch database. In total, 279,174 news webpages were downloaded. Once all of the webpages had been downloaded, they were parsed using the scripts made earlier. These were then filtered, so that only articles containing Brexit in the article body (and not in ‘suggested articles’, ‘most read’, and similar sections) were
maintained and transferred to a comma-separated values (csv) file. This file contained 46,158 articles, which was later reduced to 45,585 once articles with dates outside the specified range (e.g. one article had a date from 2001) and articles with empty dates were removed.
Variables
In order to observe patterns of media attention, the first variable created was visibility. Assigning each article in the data set a value of 1 and then indexing and aggregating the data by the date created the visibility variable. In effect, the visibility variable counted the number of articles published each day on the topic of Brexit. An overall visibility variable was created, as well as individual variables for the Irish and British visibility. In addition, a visibility variable was also created for tabloids and non-tabloid publications. The latter variable, ‘non-tabloid’, not only constituted broadsheet sources but also national and web-based news agencies, none of which subscribed to typical tabloid reporting (i.e. sensational coverage, focus on gossip etc.). A new data-frame was created for each analysis (e.g. Visibility British vs. Visibility Irish, Visibility Tabloid vs. Visibility Non-Tabloid). Additional variables were also created using a bottom-up approach in the form of topic models. They are discussed in greater detail below.
Data Analysis
Topic modelling.
Using parsed and filtered texts; a topic model was constructed with the aid of the Python package, Gensim. For this, Latent Dirichlet Allocation (LDA) topic
modelling was implemented. LDA is an unsupervised machine learning technique that automatically creates topics based on patterns of (co-)occurrence of words in a
particular document (Jacobi, van Atteveldt, & Welbers, 2016). The results of the model are then interpreted manually, as each cluster of words is assigned a topic (Jacobi et al., 2016).
Prior to creating the model, a number of topics must be specified, as well as the number of passes over the data. To identify the ideal number of topics, several tests were run with different numbers of topics specified in the model (between 20 and 200 in steps of 20). From this, it emerged that 40 topics with 5 passes yielded a sufficiently low perplexity score, indicating that fewer topics were better at predicting the held-out data (Jacobi et al., 2016). Several pre-processing steps were also
implemented into the model. Firstly, the text data was reduced so that it only contained nouns and adjectives. Secondly, unigrams and bigrams were used in the model to attain more precise topic prediction (e.g. rather than have two words, “per” and “cent”, we would have “per_cent”). Finally, a term frequency weighed by the inverse document frequency (TF-IDF) approach was implemented, while extreme words were filtered out (no words that occurred in fewer than 5 documents or more than 50% of all documents). To check whether the topics were semantically coherent with one another, Umass values were also obtained (Mimno, Wallach, Talley,
Leenders and McCallum, 2011). Most importantly, however, the topics were able to contribute to answering our proposed research questions.
Time series analysis.
Vector autoregression models (VAR) were used for the analysis, which allow for multi-directional relationships to be observed (Strauß, Vliegenthard & Verhoeven, 2016; Strycharz, Strauss & Trilling, 2018; Vliegenthart, 2014). This is a useful
analysis for the present study since it takes the interdependence of both variables into account (Vliegenthart, 2014).
The procedure for VAR recommended by Vliegenthart (2014) was adopted. In order to perform VAR analysis the data has to be stationary, otherwise it must be
differentiated. The stationarity was tested using Dickey-Fuller (DF) tests. Once significant DF tests were achieved, an appropriate number of lags were selected. The number of lags (one lag = one day) was based on research observing the “shelf-life” of online articles (Castillo, El-Haddad, Pfeffer, & Stempeck, 2014). Shelf-life is the time elapsed between the first visit an article receives and the time at which it has received 90% of the visits it will ever receive (usually 3 or 4 days) (Castillo et al., 2014). As such the maximum number of lags was set to 3 days, with fit statistics also consulted to help find the best model fit. The selection-order statistics (Akaike’s Information Criterion) then determined the final number of lags to be used.
Since ordinary least squares regression (OLS) can cause issues with multicollinearity in VAR analysis, Granger-causality tests were preformed instead (Vliegenthart & Mena Montes, 2014). Granger-causality takes into account a variable’s (x) own past, as well as the past of another variable (y) in question
(Vliegenthart & Mena Montes, 2014). Granger-causality indicates whether x exerts a significant influence on y, but does not indicate the size or direction of the influence (Vliegenthart & Mena Montes, 2014). As such, we also consulted the cumulative impulse response function (CIRF) and the decomposition of the error forecast variance (FEVD). Additionally, Portmonteau (Q) tests were performed to check residuals and the squared residuals. This was done to check whether a series
adequately reflected its past. Finally, a VAR stability test was performed on the data to make sure that the CIRF and FEVD values could be interpreted. All the VAR stability tests proved to be stable.
Results Topics
To begin with, the results of the topic model will be discussed, since this was needed for the subsequent VAR analysis. A large proportion of the topics in the model were substantively interpretable. The number of topics was reduced from 40 to 7 based on interpretability and relevance. Topics were removed if they were
uninterpretable or if they were not relevant to the present research (e.g. there was a topic related to Scotland). The final topics were border, economy, public_services, euroscepticism, Irish_government, fishing and farming. The economy, public_services and euroscepticism variables were aggregated in an approach similar to Tsur, Calacci and Lazer (2015). The economy variable comprised three topics, while the
public_services and euroscepticism variables comprised two topics each. Here it is necessary to mention the difficulty in acquiring a Eurosceptic topic. Since LDA does not incorporate any sort of sentiment analysis it is hard to tell, simply from a group of words alone, whether a topic is Eurosceptic or not. In order to get around this
problem, a combination of two topics was used to specify euroscepticism: the use of a positive word (e.g. “glee”) in combination with any word loosely associated with the pro-Brexit movement (e.g. “purge” and “british_waters”). Nevertheless, the results from this topic should be interpreted with caution. The topics are presented in the Appendix, along with some example words from the model.
Agenda Relationships Ireland vs. Britain.
The overarching research question in the present study is concerned with whether there was an intermedia agenda setting effect between Irish and British online
Table 2. Results of comparison between British and Irish sources with Granger score, CIRF and FEVD
British → Irish Irish → British
visibility Granger 7.43* 2.11 CIRF (L) .022 (3) .071 (3) FEVD (L) 1.6% (3) 45.6% (3) border Granger 6.04* 1.33 CIRF .148 (3) -.603 (3) FEVD (L) 23.6% (3) 4.4% (3) economy Granger 2.12 1.64 CIRF .296 (3) .038 (3) FEVD (L) 48.1% (3) 1.2% (3) public_service Granger 2.60 0.86 CIRF -.347 (3) .077 (3) FEVD (L) 32.6 (3) 1.9% (3) eurosepticism Granger 4.95* 4.69* CIRF .064 (3) .846 (3) FEVD (L) 26.8% (3) 3.9% (3) irish_government Granger 2.17 2.86* CIRF .147 (3) -.258 (3) FEVD (L) 15.6% (3) 0.5% (3) fishing Granger 4.05* 1.81 CIRF .054 (3) -.075 (3) FEVD (L) 42.3% (3) 2.6% (3) farming Granger 6.56* 1.16 CIRF .016 (3) -.227 (3) FEVD (L) 44.9% (3) 3.9% (3)
media in the year following the Brexit Referendum. Furthermore, I also wanted to see to what extent sources from Britain set the online news agenda for Ireland [RQ1.1]. Results from the initial VAR analysis indicate that an intermedia agenda setting influence can be identified between British online media and Irish online media in the year following the Brexit referendum. The initial VAR analysis examined Brexit topics between Irish and British sources. Granger-causality tests are presented in Table 2 along with the CIRF and FEVD scores. These findings indicate that the visibility of Brexit topics in British online media significantly Granger-caused the visibility of Brexit topics in Irish online media, but not the other way around.
All tests between Britain and Ireland were conducted with a 3-day-lag based on AIC scores. RQ3 was concerned with coverage of the Irish border issue in British and Irish online sources. The coverage of topics relating to the border is presented in
Figure 1. Surprisingly for the border topic, British sources were found to granger-cause coverage of the issue in Irish sources. When it came to the economy and public_services, there was no significant effect in the coverage between the two nations. RQ1.2 was concerned with the promotion of Euroscepticism in online news media. For the euroscepticism topic, a significant Granger-relationship was found between British sources and Irish sources, as well as vice-versa. This could mean that both Ireland and Britain influence each other’s media on Eurosceptic topics. As previously mentioned, however, this result should be interpreted with caution. This is discussed in greater detail in the Conclusion section.
Unsurprisingly, for the irish_government topic, coverage in Irish sources was found to significantly Granger-cause coverage in British sources. Conversely, British sources were found to significantly Granger-cause coverage of farming and fishing topics in Irish sources. It is worth noting that the variables in the visibility and farming models were differentiated since the Irish sources were not stationary.
Tabloid vs. non-tabloid.
The second research question was concerned with the agenda-setting
relationship between tabloids and other forms of publications in the year following the Brexit referendum. In answering this research question, the number of lags varied per-topic based on the recommendation from the AIC scores. The visibility of Brexit topics in tabloids was found to significantly Granger-cause the visibility of Brexit topics in other news sources at a lag of one day. The visibility of tabloids and non-tabloids is illustrated in Figure 2. Although the effect size is relatively small, it means that the amount of attention given to Brexit topics in the tabloids predicted how much
attention they would receive in non-tabloid publications. The results of the Granger-causality tests, as well as the CIRF and FEVD scores are presented in Table 3.
On the topic of the border, significant Granger-causality was found in both directions between tabloid and non-tabloid online news sources at a lag of two days.
Tabloids were found to significantly Granger-cause coverage of topics relating to the economy at a lag of one day. No significant relationship was found between the type of publication on the topic of public_services at a lag of three days. Interestingly, however, a significant bidirectional relationship was found between tabloids and other publications for the euroscepticism topic at a lag of two days. Once again, however, this result must be interpreted with caution. Tabloids were found to significantly Granger-cause coverage of topics related to the irish_government and fishing at a lag of three days and two days respectively. Finally, tabloids were found to significantly Figure 2. Visibility of Brexit topics in tabloid and non-tabloid online sources
Granger-cause coverage of topics related to farming in other publications at a lag of one day.
Table 3. Results of comparison between Tabloid and Non-Tabloid sources with Granger score, CIRF and FEVD
Tabloid → Non-Tabloid Non-Tabloid → Tabloid
visibility Granger 10.76* 0.01 CIRF (L) .368 (1) .009 (1) FEVD (L) 77.8% (1) 0.0% (1) border Granger 3.79* 8.87* CIRF .159 (2) .545 (2) FEVD (L) 35.9% (2) 0.0% (2) economy Granger 17.94* 0.02 CIRF (L) .441 (1) -.012 (1) FEVD (L) 73.7% (1) 0.0% (1) public_services Granger 0.17 2.18 CIRF (L) .054 (3) .442 (3) FEVD (L) 52.0% (3) 0.0% (3) eurosepticism Granger 31.75* 10.16* CIRF (L) .430 (2) .416 (2) FEVD (L) 48.2% (2) 0.0% (2) irish_government Granger 2.87* 0.74 CIRF (L) .625 (3) .661 (3) FEVD (L) 53.3% (3) 1.2% (3) fishing Granger 15.86* 0.09 CIRF (L) .852 (2) -.253 (2) FEVD (L) 70.6.3% (2) 2.8% (2) farming Granger 9.97* 1.31 CIRF (L) .319 (1) .088 (1) FEVD (L) 70.5% (1) 0.0% (1) Note: *=p < 0.05
Discussion and Conclusion
This study examined the intermedia agenda-setting relationship between Britain and Ireland on the subject of Brexit in both the month prior to, and the year following the Brexit referendum. In the process, the research has also proposed and tested a novel way of collecting large amounts of media data retrospectively from Google search pages. For the sake of clarity, I will first consider the use of the topic models and subsequently consider the VAR analysis that was used to test the various research questions.
The results of the topic modelling provided sufficient topics to use within the VAR analysis, however, the interpretation of the subjects were subjective and varied depending on the topic. Although most of the topics were very easy to interpret based on the words generated, the variable euroscepticism proved to be more difficult. There is a great challenge in determining whether a topic is Eurosceptic or not solely based on a group of words with no meaning attached. This difficulty in generating the topic can be interpreted in two ways. It might be that it is not possible to discern a
Eurosceptic topic without sentiment or more context, or it could be the case that neither online sources from Ireland or Britain endorse much of a Eurosceptic view in the first place. If the latter option is the case, it goes some way to providing an answer to RQ1.2. Indeed, the results could echo the findings of Michailidou (2015), in that national topics remain the primary concern of the media, rather than EU-topics. Future research examining the relationship between Euroscepticism and the media should look at ways of incorporating framing research, similar to de Vreese (2007), into a time series analysis.
Despite the difficultly in answering RQ1.2, the results of the topic model are interesting in that they point to very clear reference patterns in the media of both
Ireland and Britain. For example, the border, public_services, economy, farm and fishing variables were abundantly clear and obviously point to topics of concern for all nations involved. Since my analysis was concerned with articles published after the referendum, it means that these matters were the primary focus of the media following the vote.
The analysis was also able to answer the overarching research question and show that there is an intermedia agenda setting influence between British and Irish online media in the year following the Brexit referendum. Unsurprisingly I found that British media attention on the topic of Brexit was able to predict media attention on the same topic in Ireland. Furthermore, Britain set the agenda when it came to topics on the Irish border, farming and fishing. Irish online media only appeared to have an intermedia agenda setting effect when it came to topics on the Irish government. My research question about the Irish border revealed a surprising finding in that Britain seem to be the dominant intermedia agenda setter on the topic of the Irish border. Luckily for the Irish, the issue of the border seems to be a shared burden. From Figure 1, it is clear to see that the patterns of attention given to the border issue are similar in both nations’ news sources; however, one interesting observation is the spike in border coverage in Irish sources on January 27th. It is difficult to determine exactly what caused this, but it could be in part due to the news that pharmaceutical firm Almac announced that they would expand their Northern Irish firm into the Irish Republic on that date in order to gain access to the single market. It is also interesting to note that the border topic did not contain any reference to the Troubles or the IRA. Instead the border topic consisted of terms like “united_ireland”, “frictionless” and “special_status”. Although this could be a case of “not mentioning the elephant in the
room”, it might be that the media are weary of stoking an issue such as Brexit with fears of terrorism.
The analysis was able to show that tabloid sources still play a significant intermedia agenda-setting role in Britain and Ireland (at least on the subject of Brexit nonetheless). From Figure 2 it is quite clear that tabloids and other sources gave very similar levels of attention to Brexit related topics. However, our analysis shows that attention given to Brexit in the tabloids had an intermedia agenda setting effect on the attention it was given in other publications. Furthermore, online tabloid publications appear to be the dominant intermedia agenda-setters on subjects such as the economy, the Irish government, fishing and farming. The fact that sensationalised publications are setting the media agenda on this many topics is quite worrying, particularly in relation to a big political event such as Brexit. In my analysis, other publication types did not appear to have an influence on tabloids, with the exception of the reciprocal relationship on the subject of the Irish border. These findings echo those of Cushion et al. (2018), who found tabloids and right-leaning newspapers to set the topic agenda during the 2015 UK General Election. In their study, Cushion et al. (2018) interpreted their results as being a product of journalistic structural constraints and professional routines. In the current study, I would also like to suggest that these findings are endemic of the online news environment, which relies on hits for revenue. It could be the case that sensationalised tabloid reporting is better equipped to generate website hits and as such, other sources attempt to follow suit.
My research suggests that British online media set the agenda of Brexit related topics following the referendum. Since Britain is the source of Brexit, this finding should come as no surprise. However, the overwhelming influence of tabloids on other news sources should perhaps be a warning sign to media institutions. This
finding may not be new to communications research, but it does reinforce the importance of maintaining independent reliable news sources (especially online). Another notable finding of the research is on the coverage of border issues. Although border issues were obviously discussed in the online media of Britain and Ireland, it was mainly discussed in relation to trade. Indeed, as far as my analysis can determine, there was scant discussion of border conflict, an issue that is paramount to Irish relations with Britain.
Finally, one of the most important contributions of the present study is that it identified a way to use Google search information for academic purposes. Since Google is by the far the most dominant search-engine on the web, its data is a potential goldmine to communication scientists. Data scientists should consider adopting similar methodologies as the one presented here, as the current research was able to provide a thoroughly exhaustive amount of articles on a single topic.
Nevertheless, the present research is not without its limitations. Firstly, despite the success of the data collection, it was extremely slow and memory heavy.
Researchers looking to implement a similar methodology would want to be prepared. Secondly, although a sufficiently large number of articles were obtained for the analysis planned in the present study, there remains a difficulty in obtaining data from smaller tabloid sources. As mentioned previously, newspapers like The Evening Herald and Buzz published very little online. Future research should look at acquiring both online and print articles in order to remedy this. Additionally, I was unable to obtain articles from news sources with paywalls. Unfortunately, this meant that I had to leave out one of Ireland’s largest papers, The Irish Times, as well as one of the North’s, The Irish News. Finally, future research should look at better ways of
approach similar to Jacobi et al. (2016), but this left a fair deal of interpretability, which in turn may have lead to inaccuracies. As such, researchers should look for ways of measuring Euroscepticism, so that it can be implemented more easily into time series analysis.
The current study not only sheds light on the topic of Brexit and how it is covered in the media of Ireland and Britain, but it also provides a wider context for the intermedia agenda setting relationship between nations in a time of political uncertainty. It should be cause for concern for governments if tabloids are setting the online media agenda between countries. During times of great political shift (such as Brexit) it is important that the public are informed through balanced and fair reporting (the kind rarely found in tabloids). Additionally, the present study also warrants reflection on the relationship between media institutions post-colonialism. As was seen, Britain has the dominant say in the media agenda of Brexit, and even on topics highly significant to Ireland, such as the border. Patterns like these may highlight the limits of independence and warrant further empirical investigation.
References
Ahmed, K., Easton, M., Jack, S., Shaw, D., Pym, H., Jeffreys, B., & Shukman, D. (2017, March 29th 2017). What are the key issues for the Brexit
negotiations?. BBC. Retrieved from http://www.bbc.com/news/uk-39196315 Barton, C., Charles, P., Goyal, D., Raghavachari, M., Fontoura, M., & Josifovski, V. (2003). Streaming XPath processing with forward and backward axes. 19th International Conference on Data Engineering. IEE, Bangalore, India. 455-466. 10.1109/ICDE.2003.1260813
Boydstun, A. E. (2013). Making the news: Politics, the media, and agenda setting (1st ed.). Chicago, IL: University of Chicago Press.
Breed, W. (1955). Newspaper ‘opinion leaders’ and processes of
standardization. Journalism & Mass Communication Quarterly, 32(3), 277-328.
Castillo, C., El-Haddad, M., Pfeffer, J., & Stempeck, M. (2014). Characterizing the life cycle of online news stories using social media reactions. Proceedings of the 17th ACM Conference on Computer Supported Cooperative Work & Social Computing, Baltimore, Maryland. 211-223. 10.1145/2531602.2531623 de Vreese, C. (2007). A spiral of euroscepticism: The media's fault? Acta
Politica, 42(2-3), 271-286. 10.1057/palgrave.ap.5500186
Eissler, R., Russell, A., & Jones, B. D. (2014). New avenues for the study of agenda setting. Policy Studies Journal, 42(S1), S71–S86. 10.1111/psj.12048
Garside, J. (2015, March 20th 2015). Google 'illegally took content from amazon, yelp, TripAdvisor,' report finds. The Guardian. Retrieved from
https://www.theguardian.com/technology/2015/mar/20/google-illegally-took-content-from-amazon-yelp-tripadvisor-ftc-report
Goodwin, M. J., & Heath, O. (2016). The 2016 referendum, brexit and the left behind: An aggregate-level analysis of the result. The Political Quarterly, 87(3), 323-332. 10.1111/1467-923X.12285
Greenslade, R. (2006, 6th June 2006). A classic example of newspaper spin. The Guardian. Retrieved from
https://www.theguardian.com/media/greenslade/2006/jun/06/doublestandardso verloachmo
Harder, R. A., Sevenans, A., & Van Aelst, P. (2017). Intermedia agenda setting in the social media age: How traditional players dominate the news agenda in election times. The International Journal of Press/Politics, 22(3), 275–293. 10.1177/1940161217704969
Jacobi, C., van Atteveldt, W., & Welbers, K. (2016). Quantitative analysis of large amounts of journalistic texts using topic modelling. Digital Journalism, 4(1), 89-106. 10.1080/21670811.2015.1093271
Mansergh, N. (1968). In Allen G., Unwin Ltd. (Eds.), The irish question 1840-1921: A commentary on anglo-irish relations and on socal and politcal forces in ireland in the age of reform and revolution (2nd ed.). London: Uwin University Books.
McCombs, M., & Valenzuela, S. (2017). Agenda-setting theory: The frontier research questions. In K. Kenski, & K. Hall Jamieson (Eds.), The oxford handbook of political communication (). Oxford: Oxford
Handbooks.10.1093/oxfordhb/9780199793471.013.48
McCombs, M. E. (2014). New directions in agenda-setting theory and research. Mass Communication and Society, 17(6), 781-802. 10.1080/15205436.2014.964871
Michailidou, A. (2015). The role of the public in shaping EU contestation: Euroscepticism and online news media. International Political Science Review, 36(3), 324–336. 10.1177/0192512115577230
Mimno, D., Wallach, H. M., Talley, E., Leenders, M., & McCallum, A. (2011). Optimizing semantic coherence in topic models. Proceedings of the
Conference on Empirical Methods in Natural Language Processing, 262-272. Mulholland, M. (2002). The longest war: Northern ireland's troubled history (1st ed.).
Oxford: Oxford University Press.
Nsehe, M. (2012, January 31st 2012). Report: Google fires kenyan manager over data poaching scandal. Forbes. Retrieved from
https://www.forbes.com/sites/mfonobongnsehe/2012/01/31/report-google-fires-kenyan-manager-over-data-poaching-scandal/#72a3e5d26adc O'Toole, F. (2017, September 28th 2017). Brexit’s irish question. The New York
Review of Books. Retrieved from
http://www.nybooks.com/articles/2017/09/28/brexits-irish-question/
Pew Research Center. (2018). Publics Globally Want Unbiased News Coverage, but Are Divided on Whether Their News Media Deliver. Washington D.C., USA. Retrieved from http://www.pewglobal.org/2018/01/11/publics-globally-want- unbiased-news-coverage-but-are-divided-on-whether-their-news-media-deliver/
Reese, S. D., & Danielian, L. H. (1989). Intermedia influence and the drug issue: Converging on cocaine. In P. J. Shoemaker (Ed.), Communication campaigns about drugs: Government, media, and the public(pp. 29–46). New York: Lawrence Erlbaum Associates.
Richard haass warns NI violence could re-emerge without progress. (2014, 12th March 2014). Retrieved from http://www.bbc.com/news/uk-northern-ireland-26535987
Rideout, V. J., Foehr, U. G., & Roberts, D. F. (2010). Generation M2: Media in the lives of 8- to 18-year olds. (). Menlo Park, CA: Kaiser Family Foundation. Roche, S. P., Pickett, J. T., & Gertz, M. (2016). The scary world of online news?
internet news exposure and public attitudes toward crime and justice. Journal of Quantitative Criminology, 32(2), 215–236. 10.1007/s10940-015-9261-x Rogers, E. M., Dearing, J., & Bregman, D. (1993). The anatomy of agenda-setting
research. Journal of Communication, 43(2), 68-84. 10.1111/j.1460-2466.1993.tb01263.x
Rogstad, I. (2016). Is twitter just rehashing? intermedia agenda setting between twitter and mainstream media. Journal of Information Technology & Politics, 13(2), 142-158. 10.1080/19331681.2016.1160263
Rosenstiel, T., Mitchell, A., Rainie, L., & Purcell, K. (2011). Mobile news & paying online. (). Washington, DC: Pew Research Center’s Project for Excellence in Journalism.
Russell Neuman, W., Guggenheim, L., Mo Jang, S., & Young Bae, S. (2014). The dynamics of public attention: Agenda-setting theory meets big data. Journal of Communication, 64(2), 193–214. 10.1111/jcom.12088
Scharkow, M., & Vogelgesang, J. (2009). Effects of domestic media use on european integration. Communicaitons: The European Journal of Communication Research, 31(1), 73-91. 10.1515/comm.2010.004
Seaton, J. (2016). Brexit and the media. The Political Quarterly, 87(3), 333–337. 10.1111/1467-923X.12296
Stevenson, J. (2017). Does brexit threaten peace in northern ireland? Survival: Global Politics and Strategy, 59(3), 111-128. 10.1080/00396338.2017.1325606 Strauß, N., Vliegenthart, R., & Verhoeven, P. (2016). Lagging behind? emotions in
newspaper articles and stock market prices in the netherlands. Public Relations Review, 42(4), 548-555. 10.1016/j.pubrev.2016.03.010 Strycharz, J., Strauss, N., & Trilling, D. (2018). The role of media coverage in
explaining stock market fluctuations: Insights for strategic financial
communication. International Journal of Strategic Communication, 12(1), 67-85. 10.1080/1553118X.2017.1378220
Torreblanca, J. I., & Leonard, M. (2013). In Bechev D., . . . Wiersma J. M.(Eds.), The continent-wide rise of euroscepticism European Council on Foreign Relations. Tsur, O., Calacci, D., & Lazer, D. (2015). A Frame of Mind: Using Statistical Models
for Detection of Framing and Agenda Setting Campaigns. In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (pp. 1629–1638). ACL.
Van Aelst, P., & Walgrave, S. (20111). Minimal or massive? the political agenda– setting power of the mass media according to different methods. International Journal of Press/Politics, 16(3), 295–313. 10.1177/1940161211406727 Vliegenthart, R. (2014). Moving up. applying aggregate level time series analysis in
the study of media coverage. Quality & Quantity, 48(5), 2427–2445. 10.1007/s11135-013-9899-0
Vliegenthart, R., & Mena Montes, N. (2014). How political and media system characteristics moderate interactions between newspapers and parliaments:
Economic crisis attention in spain and the netherlands. The International Journal of Press/Politics, 19(3), 318–339. 10.1177/1940161214531335 Vonbun, R., Kleinen-von Königslöw, K., & Schoenbach, K. (2016). Intermedia
agenda-setting in a multimedia news environment. Journalism, 17(8), 1054 - 1073. 10.1177/1464884915595475
Walgrave, S., Soroka, S., & Nuytemans, M. (2008). The mass Media’s political agenda-setting power. A longitudinal analysis of media, parliament and government in belgium (1993-2000). Comparative Political Studies, 41(6), 814-836. 10.1177/0010414006299098
Westlund, O., & Färdigh, M. A. (2015). Accessing the news in an age of mobile media: Tracing displacing and complementary effects of mobile news on newspapers and online news. Mobile Media & Communication, 3(1), 53–74. 10.1177/2050157914549039
Appendix
Table 4. Topics in Brexit related articles and example words
Word (Rank) Topic
Topic 1 united_ireland (2); special_status(5); irish_border(6)
border
Topic 1 economy (4); bank (5); growth (6) economy Topic 2 sales (1); profits (3); retailers (5)
Topic 3 inflation (1); pound (7); interest (5)
Topic 1 tourism (1), pension (2); interest_rate (6) public_services Topic 2 national_insurance (1); state_pension (2); earners
(5)
Topic 1 obsolete (1); bye (5); british_waters (6) euroscepticism Topic 2 brc (1); glee (3); movement_end (7)
Topic 1 kenny (1); taoiseach (2); varadkar (3) irish_government Topic 1 fish (2), waters (4); port (7) fishing
