Election Unspun FAQs
What is Election Unspun?
Election Unspun is a weekly analysis of UK online national news coverage of politics in the run-up to the 2015 UK General Election. It looks at the parties, people and issues being covered across national news sources. Every news article from selected news websites is gathered and tagged according to their text content and metadata, from Monday to Sunday each week. We then analyse the data and, every Thursday morning from 26th February to 7th May, publish statistics on the previous week’s political news.
Over the course of the campaign we will also be publishing special reports, exploring various aspects of campaign news in more depth, such as coverage of party leaders, share of voice on specific policy areas, constituency coverage, and policy coverage in UK political magazines and blogs.
Election Unspun is driven by the Media Standards Trust’s new digital analysis tool Steno. Steno can be targeted at selected news websites, and gathers the URLs they publish, alongside certain information, including date of publication, headline, byline, etc. Algorithms can then be developed to tag and delete articles based on their content or metadata.
Developed in order to analyse media content about specific policy areas, the 2015 election is the first proper road test for this new tool and approach to news analysis. Following the election we aim to refine the research process to perform smaller-scale targeted analyses of online news coverage.
What does Election Unspun cover?
At the moment, Election Unspun analyses news articles published on the websites of 16 UK national news providers:
- BBC (bbc.co.uk/news)
- ITV (itv.com/news)
- Channel 4 (channel4.com/news)
- Sky News (news.sky.com)
- The Daily Mail and Mail on Sunday (dailymail.co.uk)
- The Sun (thesun.co.uk)
- The Daily Express and Sunday Express (express.co.uk)
- The Daily Mirror, Sunday Mirror and Sunday People (mirror.co.uk)
- The Daily Star and Daily Star Sunday (dailystar.co.uk)
- The Daily Telegraph and Sunday Telegraph (telegraph.co.uk)
- The Times (thetimes.co.uk) and Sunday Times (thesundaytimes.co.uk)
- The Financial Times (ft.com)
- The Independent and Independent on Sunday (independent.co.uk)
- The Guardian and Observer (theguardian.com/uk)
- The Huffington Post UK (huffingtonpost.co.uk)
- Buzzfeed UK (buzzfeed.com/uk)
We are also gathering data from the sites of political magazines (including The New Statesman and The Spectator), a selection of political blogs from across the political spectrum, and a pilot-run of some local and regional newspapers. We hope over the course of the project to bring these into the analysis, but to begin with we are concentrating on UK national news sites publishing daily news articles.
How does the analysis work?
Each week, Steno gathers approximately 30,000 articles from the various sites that it targets. This number is then whittled down to around 7,000 as all articles with no relevance to UK politics or public policy are removed. This is done firstly through automatic removal of all articles with URLs denoting sports or showbiz articles, reviews, recipes, crosswords, and so on. Second, areas where there may be some policy significance (i.e. technology, science, lifestyle, etc.) are automatically scanned according to their text content and removed or retained accordingly. All articles published in general news sections are retained for analysis.
This sample of around 7,000 articles is then automatically tagged on the basis of whether or not they contain – or are highly likely to contain – references to:
- The main UK political parties
- The leaders of the main parties
- The various policy issues contained in our adapted version of the Issues Index used by Ipsos-MORI (see below)
Finally, all articles in which there is a degree of doubt about whether they have been correctly tagged are manually scanned and re-tagged, untagged or deleted if necessary.
How does issue tagging work?
The Election Unspun analysis uses as its reference the Ipsos-MORI Issues Index. This consists of a list of 36 issue areas used to gauge public opinion on the most important issues facing the country. For our analysis, we have condensed these into 14 categories. Eight remain the same:
- Transport/Public transport
- Local government/Council tax
- Devolution/Constitutional reform
- Fuel and energy
Certain other categories were combined to create the following categories broadly corresponding to government departments:
- Crime/Law & order/Justice
- Defence & Foreign Policy
Currently, the Ipsos-MORI catch-all ‘Other’ category is not being used in the Election Unspun analysis.
Each issue is tagged according to a series of search commands. For example. All articles containing the phrase “jobseeker’s allowance” would be tagged “welfare”, as would the phrase “DWP”. These tagging rules are continually being revised to ensure that articles are not mis-tagged, or that relevant articles are not being missed. Where there is ambiguity around a search term (e.g. “schools” or “Europe”) these articles are flagged to be manually checked.
These categories are not mutually-exclusive. An article may contain one, two, or many references to different policy areas. There are also a few areas of overlap. For example, fracking may – depending on the context – relate to both ‘Environment’ and ‘Fuel and Energy’. These articles are manually tagged where they occur.
What’s being measured?
In each weekly analysis, we’ll be repeating a number of measurements. Below is an explanation of what they are, and how the numbers are arrived at:
This is a comparison of the most-repeated phrases in campaign news articles. It lists the number of articles in which the phrase was mentioned, based on a text search of the weekly dataset. For example, between 16th and 22nd February, covered in our first newsletter published on 26th February, the phrase “tax avoidance” appeared in 126 different articles.
Issues of the week
This is a comparison of how many different news articles each issue appeared in. These numbers are derived from the dataset once it has been fully tagged and manually checked. More than one issue may feature in any article, but each issue is only counted once per article.
Each week, we choose an issue (usually the most featured issue that week) and investigate how it breaks down, i.e. which major stories dominated coverage of that issue.
Party leaders in the news
This lists the party leaders in order of how many articles they appeared in. Simply, it is calculated by tags which are based on whether or not the leader’s name appeared in the headline or text of the article. The tagging criteria ensure that we get Ed Miliband and not David Miliband, and David Cameron, not Cameron Diaz.
This graph also shows how the party leader appearances compare with the previous week, so it’s possible to see who had more or less exposure that week.
Prominent this week
Each week we analyse which figures (political and non-political) feature often in political news articles. As with party leaders, this is calculated on the basis of how many articles they appeared in.
Coverage of the parties
Like party leaders, this shows how many articles contained references to the main political parties. As with both issues and party leaders, articles can refer to more than one party. In order to ensure accuracy, phrases containing party names (i.e. “labour supply”, “religious conservatives”) are flagged and – if necessary – untagged or removed.
Cumulative party coverage
This shows the aggregate number of articles containing references to each party from the beginning of week 1 of the study (Monday 5th January) to the end of the relevant week.
Who’s covering what?
This graph breaks down the total number of articles referencing each issue by each website, to show the balance of issue coverage on the different news outlets.
Day by day coverage
This gives a day-by-day account of the number of articles published during the week of analysis containing references to one or more of the policy issues in the adapted Issues Index.
Campaign vs. Kardashian
In order to show the campaign coverage in context, each week we select five campaign-related stories, and compare them with other prominent stories that week. These stories (campaign and non-campaign) are selected from the full weekly sample of all c.30,000 news articles, rather than the modified sample of policy-related articles.
All leader articles published by the different newspaper sites are tagged for further analysis. We plan to do a number of different analyses on the content of leader articles, though in Week 1 (26th Feb) we focus on how parties are depicted by the different news sources. All articles were coded according to whether they contained positive or negative statements (or both) about each political party.
So, in Week 1, 74 leader columns were gathered. Of these 74, 20 contained one or more evaluative statement about political parties (28 in total).
(NB: of all the newspaper sites analysed in Election Unspun, two (mirror.co.uk anddailystar.co.uk) do not publish leader articles online)