Media in Context and The 2015 General Election: How Traditional and Social Media Shape Elections and Governing

Lead Research Organisation: University of Exeter
Department Name: Politics


The 2015 British election media study we propose is timely and important in an era of declining support for the major parties. In the last two elections the leading parties have formed governments with roughly 35 percent of the vote, with 2015 currently promising a similar scenario and possibly another coalition. The election also promises debates--including more leaders' debates--on major constitutional issues pertaining to national and regional devolution and their consequences for Westminster, as well as on the possibility of a referendum on continuing EU membership, all in a context of continuing austerity. Media reporting and framing of such subjects is likely to be critical to the dynamics of the election, to public opinion, and to the election's aftermath.

Yet although their effects on voters have consumed research on electoral politics in Britain, the US and other democracies since the 1940s, the question of media effects remains unsettled. Moreover, the issue of the difference that social media has made, supplementing or replacing the information provided by traditional media--television, newspapers and radio--further muddies the waters. In this research we both seek to address several pressing questions pertaining to media effects on governance and elections and also to gather timely and high quality data on media coverage during and after the 2015 British general election to share with the user community.

The substantive questions we will examine within our study are:
1) The flow of campaign information. Traditional academic models depict campaign information flows as linear, from elites to opinion leaders to masses, but this may no longer be accurate in a world in which social media can provide a platform for opinion leaders (and masses) to produce information. While some think that social media have made opinion leaders even more important, others argue that it has cut them out of the picture, with information flowing directly to the consumer.
2) The changing media landscape matters in a second way--not in terms of the flow of information but, more straightforwardly, for where if at all people obtain political information in a world of declining newspaper readership and trust in media. Moreover, the traditional media no longer play the same gatekeeping role, potentially diluting their influence on the issue agenda. For example, traditional campaigns in the UK followed a pattern in which parties held morning press conferences that launched the "theme of the day." While the media may not have always framed the theme in the way parties would have wished, the press conference set the issue agenda for 24 hours. That no longer seems to be the case.
3) The role of the media, both social and traditional, in the post-election period. Interpretations of election results may be important in two respects: in conferring legitimacy upon the outcome and thus fostering what is sometimes known as "losers' consent," and in providing a narrative about the mandate the incoming government enjoys.

Our study will also address four deficiencies in existing studies of British media election coverage: that they tend to focus on election coverage, ignoring non-election coverage and thus not permitting analysis of the overall news context or the prominence of the election as an issue; that the data are either not made publicly available or only made available years after the election; that recent British election studies have permitted little understanding of media effects due to very few questions about media habits; and that British media studies tend to rely exclusively on survey data, ignoring the benefits for establishing causation and effect sizes offered by field experiments.

The proposed research brings together investigators with a unique combination of expertise in human and automated traditional and social media content analysis and statistical modelling skills.

Planned Impact

By examining flows of campaign information, as well as by linking them to individuals, we are able to understand both where social and traditional media fit in the contemporary information environment, the implications for governance, and media's effects on individual attitudes and behaviours. These research foci allow us to address ESRC strategic priorities such as influencing behaviour and informing interventions. In addition, based on the implications of the research outputs our project will inform debates about how to enhance the quality of political life in Britain by illustrating where media appear to have a positive influence and where its influence is less positive or detrimental. In light of ongoing discussions regarding media (e.g., Leveson inquiry, Political and Constitutional Reform Committee's inquiry on voter engagement in the UK), the data will allow us to identify and deliver policy recommendations by providing a more in-depth understanding of the nature of media coverage of politics and its effects on governance and the British public.

1. Potential Policy and Societal Impact
- The user community outside academe is large and diverse. It consists of all those institutions and individuals who have a professional interest in media, media regulation, elections and electoral processes. One segment of this extra-academic user group is centered around political parties: elected office holders, party officials, campaigners, and those working in research institutes and thinktanks connected to political parties. A slightly different group consists of those representing social groups and organised interests and who equally have a stake in the outcome of elections, and sometimes in providing their members with relevant information and advice (labour unions, employers organisations, churches, sundry cause groups, formal lobbyists, etc.). Another component consists of media organisations and journalists who provide audiences (the mass audience as well as more specialised and targeted audiences) with information on elections. Finally there is a plethora of firms (mainly, but not exclusively, SME's) that cater to the rest of the extra-academic user community (market research companies, media and campaign specialists, consultancy firms, etc.).
- The research findings will also contribute towards evidence-based policy making in the area of media and media regulation. The findings should give clear indications of the impact of social media competition on the production of electorally relevant information, and of variation by source and mode (e.g., TV vs. newspapers). Public broadcasters in particular are interested in fulfilling the remit of informing citizens and can base policy recommendations on the indications from our research about the influences on, and of, their election coverage. The workshops at the 2015 EPOP conference and in London in September will advance these areas of impact. Furthermore, the briefing paper we will write after the stakeholder workshop, and the discussion paper with IDEA, will have implications for policy recommendations regarding engaging citizens in representative democracy that may be of interest to bodies such as the Leveson Inquiry and the House of Commons Political and Constitutional Reform Committee in the UK (to which Banducci and Stevens gave evidence as party of its inquiry into voter engagement on July 10, 2014).
2. Research Area Impact
Another benefit of the research is to the UK Research Area: The visibility of the UK as a venue for research on matters pertaining to electoral behaviour, electoral representation, the role of the media, and democracy will be promoted. The innovation of automated media coding also places the UK at the forefront of research using content analysis, while the methods we use to discern media effects using the survey data will have applicability to a rich set of causal questions in the social sciences in general.
Description The research project had four objectives. Our key findings thus far with respect to each are:

1. To develop a model of information flows in the campaign and post-election periods
Also known as 'intermedia agenda setting' we have explored these with respect to media coverage of issues and of party leaders and their effects on public opinion:
• We see some evidence of partisan slant in terms of the amount of coverage of issues such as the NHS and the EU in Labour and Conservative newspapers.
• However, in the case of the NHS this did not seem to translate into any influence on public opinion on the NHS from newspapers, and on the issue of the EU, television appeared to hold more sway.
• Overall, the prevalence of the EU in media appeared to be more driven by the Conservative party agenda than the NHS was driven by Labour efforts. This is consistent with anecdotal evidence that the media were more responsive to Conservative messages during the 2015 campaign.

With respect to party leaders:
• The relationships between media for coverage of leaders are conditional and the patterns of flows between different media outlets are complex and contingent. For example, we observe extensive intermedia agenda setting, i.e., influence of different media on each other, for the volume and tone of the relatively unknown leader Nicola Sturgeon, but much less pervasive or clear effects for Cameron and Miliband.
• This suggests that intermedia agenda setting, and the potential for the media to "presidentialize" the campaign, is conditional on the party leader and perhaps on the extent to which coverage of them has already settled into an established set of norms.
• Intermedia agenda setting for leaders may differ from that for issues. While television may take its cues about issue salience from newspapers because of their larger newshole and the depth of coverage they can provide, this does not appear to apply to coverage of party leaders, about whom there are no manifestos to launch and fewer "themes of the day" that could sustain multiple spokespeople on the campaign trail.
• While our results show an interplay amongst media-with some outlets more influential for some leaders-there is little evidence that all media converge on the salience and tone of leaders. Therefore, the ability of media to act as an independent force in personalizing the campaign is limited.

2. To examine media effects in the campaign and post-election periods

During the campaign:
• We found stronger effects of media in constituencies that are competitive and where the outcome is therefore uncertain.
• In an experiment using articles from The Guardian and The Daily Telegraph we uncovered relatively large and consistent effects of the Guardian treatment in particular in competitive constituencies.
• Given that this experiment took place at the end of an election campaign during which media coverage of the issues we asked about had been extensive, this demonstrates a greater capacity for British media to affect perceptions than has been acknowledged heretofore. It implies that the suggestion that media influence in Britain can only unfold over the long-term is overblown.
• This impact appears to be the result of source rather than content effects. We found no evidence that media effects are limited to, or strongest upon, the specific issues mentioned in stories. Rather, source effects appear to pervade views on a range of issues.
• The effects we found were most consistent on respondent's perceptions of their own positions on issues; we saw little impact on perceptions of the parties on issues or on perceptions of the parties and their leaders. This suggests that concentrating on the impact of the media attitudes toward partisan positions or leaders, while understandable, also misses a more important part of the picture.
• In general the research illustrates the difficulty in finding media effects in observational research, such as in separating the influence of media from broader campaign effects that we could see in our control groups but that would be harder to discern in a non-experimental setting.
• The impact of the Guardian treatment in particular, but also of the Telegraph treatment when it was statistically significant, tended to be in the opposite direction to the partisan orientation of the newspaper, i.e., The Guardian moved respondents' perceptions of issues to the right and The Telegraph moved respondents to the left. This indicates motivated resistance and counter argument when respondents are exposed to media sources and content they would not usually select to read.
• Our research suggests that endorsement effects stem from the overall reputation of the newspaper and the broad tenor of its election coverage rather than from its specific content.

After the campaign, we examined the impact of different interpretations of the outcome and of the consequences of breaking manifesto promises in further survey experiments:
• Receiving the information that the Conservatives had a 'decisive' parliamentary majority made respondents in marginal seats more likely to believe in the government's ability to deliver on campaign promises, while receiving a 'narrow/weak' majority interpretation made them more likely to consider having a single-party government in office important for the government's accountability in elections.
• In both case, the treatments affected the expectations of voters in marginal constituencies regarding the government's likely performance and the electorate's ability to hold the government accountable, thus, indicating the moderating effect of seat competitiveness on framing effects.
• Had Cameron backed down from his campaign pledge to carry out the EU referendum, the damage might have been manageable. Only a quarter of participants in our study were 'strict punishers' for whom such inconsistency would have translated into disapproval of the decision to the tune of a loss of approval scores of 18% to 27% (relative to having carried out the referendum), and who would have viewed Cameron as significantly less competent for reneging on his promise. His perceived competence following inconsistent behaviour remained untarnished for the remaining three quarters of respondents in the study.
• Furthermore, a third of our participants did not punish Cameron at all for not implementing the referendum when told he backed down due to internal or external political pressures. This last group included both Labour and UKIP supporters.

3. To assess the contrasting influence of national, regional and local media on specific constituencies and subsets of the electorate
• The data we gathered on local media show that the election receives minimal coverage on local television news or in local newspapers. Regional coverage in Scotland is more extensive, providing a Scottish perspective on the election. This is less pronounced in Wales.
• Experimental evidence suggests very little influence of local media on the public. Stories from local newspapers did not have the same influence as national newspapers like The Guardian and did not even prompt respondents to weigh local issues more heavily in their evaluations.

4. To understand the role of social media in the electoral process
The research has looked at both the supply of information via social media and their impact on the public. Key findings include:
• The two major party leaders, Cameron and Miliband, dominated Twitter traffic by a ratio of more than 3:1 compared to Farage and Clegg. Sturgeon's status as a possible coalition partner with Labour, perhaps along with her performances in debates, also brought her a lot of attention.
• The leaders' debates and interviews led to a surge in mentions of most of the leaders, most clearly Sturgeon after the 7-way and challengers' debates but also Miliband after the consecutive Question Time interviews of him, and Cameron.
• There are somewhat stronger effects of traditional media on Twitter in terms of what is discussed and its tone than vice versa.
• Conservative constituency candidates' Twitter activity pertaining to the NHS follows increases and decreases in both public opinion and other media rather than leading it. This is also true of Labour candidates' activity.
• On average, candidates in competitive constituencies Tweet more.
• The tone of Tweets for candidates in all parties but UKIP was more often positive than negative.
• Sharing urls on Twitter is associated with a reduction in vote shares, while sharing photos is associated with an increase.
• Both the number of followers and the number of people followed are associated with increases in vote shares.
• Social media could provide an inexpensive way to mobilize voters in safe constituency when party mobilization efforts are targeted in marginals.
Exploitation Route Our findings have several implications for our overall objective of informing and shaping policy debates about the role of media in contemporary elections.
• The evidence on intermedia agenda setting lends empirical support to some of the anecdotal evidence about a tendency for public broadcasters such as the BBC and ITV to have followed the agendas of the Conservative press in 2015. Given our focus on a single election we cannot say whether it was because the press was predominantly supportive of the Conservative Party, or because of something distinctive in the nature of its coverage, but we observed a greater influence of Conservative-supporting outlets like the Daily Mail and Telegraph on the BBC than vice versa, on issues like the NHS and on the coverage of leaders such as Nicola Sturgeon. This seems to be in part due to an integral aspect of daily election news now being a description of press headlines. Our analysis suggests that these kinds of tendencies compromise public broadcasters' neutrality and might therefore be reconsidered.
• We show some of the promise of social media as an additional tool to mobilize voters. Previous studies have demonstrated the efficacy of Facebook but our study suggests other social media such as Twitter could also provide an inexpensive means of mobilization.
• Many of our findings about context illustrate differences between individuals living in marginal as opposed to safe constituencies. Individuals in safe consitutencies appear far more responsive to flows of information during and after campaigns, i.e., more engaged. Given that one of the intentions of campaigns is to inform and educate prospective voters, this suggests that the trend towards fewer marginal constituencies in the UK is detrimental to democracy.
Sectors Digital/Communication/Information Technologies (including Software),Government, Democracy and Justice

Description Our findings led us to be invited to a Reporting Europe event in London on February 8th and to a debate to discuss what the major academic studies have to teach us about the 2015 General Election, sponsored by Ipsos-MORI on April 19th.
First Year Of Impact 2015
Sector Government, Democracy and Justice
Impact Types Societal

Title The 2015 UK General Election Coverage in Newspapers (machine-coded) 
Description These data consist of computer coding of election-related stories in national and some regional or local newspapers from February 1 2015 to May 28 2015 (just over 21,000 stories). Variables include story length, issue topic, and the proportion of words mentioning each party leader. 
Type Of Material Database/Collection of data 
Year Produced 2016 
Provided To Others? Yes  
Impact Findings from this data set were presented at a stakeholder workshop in London on January 15, 2016. 
Title The 2015 UK General Election Coverage in Traditional Media (human coded) 
Description These data are a sample (roughly 20% of days) of stories in national newspapers, national television news, some regional or local newspapers and regional or local television news from February 1st to May 28th coded by human coders. Variables include length, primary topic and tone (of election-related stories). 
Type Of Material Database/Collection of data 
Year Produced 2016 
Provided To Others? Yes  
Impact Findings from these data were presented at a stakeholder workshop in January 2016. 
Title The Twitter Activity of Prospective Parliamentary Candidates in the 2015 UK General Election 
Description These data summarize the Twitter activity of parliamentary candidates with a Twitter account from the top seven parties in Britain between March 30 and May 7, based on searches using 70 different search terms, e.g., "tuition fees", "terrorism". Variables include characteristics of the candidates, the number of Tweets they sent, and the number of Tweets they sent that included photos. 
Type Of Material Database/Collection of data 
Year Produced 2016 
Provided To Others? Yes  
Impact Findings from these data were presented at a stakeholder workshop in London on January 15, 2016. 
Title Traditional and Social Media data deposited with UKDS 
Description 1. British Election Study Wave 5 data with codes for media outlets added. This allows users to combine BES survey data with: 2. Data from a content analysis of 27 media outlets - 17 newspapers, 9 TV and 1 radio for the period between 1 February 2015 and 30 May 2015. The dataset contains 135 variables and 2826 observations. 3. Data from an analysis of social media, including tweets using #GE2015, major and minor party accounts, major and minor party Facebook pages, and major and minor party candidates. The dataset contains 75 variables and 850 observations. 
Type Of Material Database/Collection of data 
Year Produced 2017 
Provided To Others? Yes  
Impact The data were sent for deposit with UKDS on February 17, 2017. 
Description London stakeholder workshop 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Professional Practitioners
Results and Impact The project hosted a day-long workshop at the Nuffield Foundation in London for a broad range of practitioners from polling organizations, political parties and a vote advice organization. The purpose of the workshop was to:
a. Present key findings from the research
b. Get feedback from practitioners on the findings and on additional analysis that would interest them
c. Get feedback from practitioners on how to maximize the usefulness of the data for them
d. Discuss how we might conduct a similar EU Referendum project in ways that would be of maximum benefit
e. Announce the public release of data from the project
The workshop, which was divided into three topic-related sessions, prompted questions about the findings, discussion of their meaning, and suggestions for additional analysis.
There have been increased visits to the project in the wake of the workshop.
Year(s) Of Engagement Activity 2016
Description Pre-EPOP Workshop 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Postgraduate students
Results and Impact The workshop sparked discussion about:
1. Our data and how it was collected
2. Capturing social media data effectively
3. Automated content analysis
4. How postgraduate students might use our data or our methods
5. Issues of causation in media effects research

Participants in the workshop (14 graduate students and 2 full-time academics) filled out a feedback questionnaire afterwards. The workshop was rated 4 of out of 5 for 'understanding the data and its application' and 4.7 out of 5 overall.
As a follow-on to the workshop, one participant arranged a subsequent meeting with us in Exeter, in which we discussed how our data could help with her project on negative campaigning. We have since shared data with her, broadening the impact of the content analysis.
Year(s) Of Engagement Activity 2015