Community Research and Development Information Service - CORDIS

Twitter analysis shows immigration drove UK Leave voters

Six months on from the UK’s decision to leave the European Union, researchers are still trying to ascertain how and why the Leave campaign was ultimately victorious. In new research partly supported by the EU-funded PHEME and SOBIGDATA projects, 3 million tweets were analysed over a 6-month period that showed immigration was by far the most important issue driving Leave voters.
Twitter analysis shows immigration drove UK Leave voters
Researchers based at the UK’s University of Sheffield identified 41 443 Leave supporters and 41 445 Remain supporters based on their pattern of using campaign hashtags on Twitter in the run-up to the referendum date. They then analysed the issues those users tweeted about in relation to Brexit every day from June to November (e.g. #VoteToLeave, #SaferIn).

However, in order to remain as accurate as possible, for tweets containing Leave/Remain hashtags, only the last hashtag used was considered by the researchers as the one indicating the tweet’s intended stance (pro-Leave or pro-Remain). They also filtered out all users with fewer than three stance-expressing tweets. This allowed them to arrive at a more reliable sample of Twitter Leave/Remain supporters.

In order to break down tweets into topics, the research used the UK Government’s stated high-level ‘policy areas’, such as public health, immigration and law. For each of these topics, sets of keywords were linked to them (e.g. for public health: NHS, nursing, doctors), which were taken automatically from party manifestos and UK election tweets, then post-edited and extended manually. This was also complemented by a more traditional search-based approach in order to capture statistics on popular Brexit-specific topics (such as ‘Article 50’).

Lessons from Twitter

The analysis showed that in the run-up to the referendum, Leave voters sent twice as many tweets about borders and immigration as they did about sovereignty, employment, the justice system or the National Health Service (NHS). The issue of law and justice, though far behind immigration, was still a major issue for Leave voters, who sent 4 times as many tweets on this topic as Remain supporters.

Another one of the most interesting findings from the research was that although Twitter is often seen more as a liberal echo chamber, in the month running up to the referendum it was Leave supporters who were tweeting most actively, with immigration and the economy being the most active topics. After the referendum, Remain voters tweeted more actively across all topics, closing the gap with Leave voters. However, Twitter acted more as an echo chamber overall, with only 7 % of tweets being replies and over 58 % being retweets, although it was noted that Leave voters made more effort to spread their views.

The gap between the number of tweets from Leave and Remain supporters about borders and immigration became most apparent in the final week of the referendum campaign, with a large surge in tweets from Leave supporters. This coincided with the release of a now infamous poster by Leave.eu (which was not the official Leave campaign) showing a queue of refugees. The poster gave the impression that the individuals depicted were heading specifically to the UK, whilst the photo was actually taken on the Croatian-Slovenian border.

‘The key challenge and benefit of this work is in getting the insights in real time,’ commented senior study researcher Kalina Bontcheva. ‘This means always having up-to-the-minute knowledge of the topics, participants, and opinions, and being able to study changes in these over time.’

Whilst the researchers are careful to insist that Twitter and other social media channels cannot by themselves explain the complex reasons behind some of the most unexpected political events of 2016, such as the UK public’s decision to go for Brexit, they do highlight that social media allows for the study of the general public’s immediate reaction to ongoing events in almost real time. This thus complements traditional polling methods that have had less than a stellar record in accurately making predictions in what has been a politically tumultuous year.

For more information, please see:
PHEME project website
SOBIGDATA project website

Source: Based on media reports

Related information

Record Number: 126608 / Last updated on: 2016-12-16
Category: Policy making and guidelines
Provider: EC