Skip to main content

What mattered most to you when deciding how to vote in the EU referendum? By Chris Prosser, Jon Mellon, and Jane Green

The British Election Study Team
11/07/2016

We expect election campaigns to follow the major concerns of voters. We also expect campaigns to focus on those issues considered strongest for each respective side. But what did voters think were the most important things when deciding how to vote in the EU referendum? The stark differences in concerns we show here are guides to the divided nature of the British electorate in terms of what they thought mattered in the referendum, and how they subsequently voted.

Wave 7 of British Election Study Internet panel (conducted prior to the formal campaign period) lets us look into the reasons for people voting in the EU referendum in more detail. We asked an open-ended question to let respondents tell us what mattered to them: “What matters most to you when deciding how to vote in the EU referendum?”. Respondents gave a wide variety of responses. Some typical ones included:

  • Ability to manage our own affairs
  • Britain influence and trade in the world
  • Economic stability
  • Keeping Europe together and no wars

Open ended data is always tricky to work with because respondents give so many different responses. In wave 7, respondents gave a total of 15,070 unique answers to the question. We show a few different ways of analyzing these unique data in this post, and the insights into the British public that arise.

Word clouds

An easy way to get an initial idea of what respondents are saying is to use word clouds – a visual display of all the text in our respondents answers. These scale the size of the text according to how frequently a word was used. Here we break down the responses into people who, between 14th April and 4th May 2016 (fieldwork dates) said they had decided they would vote leave, vote remain and those who had not yet made up their minds.

The first word cloud shows very clearly that a key concern for leave voters was immigration. The issue dominates the leave voter word cloud and other words like ‘borders’ also feature. Other words that feature prominently concern issues surrounding sovereignty, including the word ‘sovereignty’ itself and ‘control’, and ‘laws’.  

 

Leave voters word cloud

wordcloud_leave

The remain word cloud illustrates clearly that the primary concern of remain voters was the economy, which again dominates the word cloud along with related words like ‘trade’ and ‘jobs’. Other words that feature prominently for remainers are ‘rights’, ‘security’, ‘stability’ and ‘future’.

 

Remain voters word cloud

wordcloud_remain

These very stark differences display how voters of the respective sides were not just divided in terms of geography and demographic characteristics. They were fundamentally divided in what they considered important and relevant to how they were intending to cast their vote.

In some ways undecided voters were a mix of leave and remain voters – ‘economy’ and ‘immigration’ both feature prominently in the undecided word cloud, along with other words that pop up in the other clouds like ‘future’, ‘security’, ‘rights, and ‘control’. Neither side had an obvious advantage among these respondents if we examine only those issues respondents thought would be important in deciding how to vote. However, something that characterises the responses of the undecided voters is the number of words relating to uncertainty about the effect of Brexit, like ‘facts’, ‘information’ ‘affect’, and ‘impact’.

 

Undecided voters word cloud

wordcloud_dont_know

Future analysis will allow us to determine the relevance of these issues in deciding how undecided voters decided how to vote, and indeed whether to vote at all.

 

Coded responses

We also code the open-ended responses into a series of 54 categories covering the key topics that respondents mentioned. We did this coding using the codething platform that combines machine learning with manual coding. Any coding scheme will not be able to cover every possible answer that respondents can think up. Some of the more difficult responses to code included:

  • Keeping friends close and enemies even closer
  • if businessmen say it’s good to stay in it must be bad for everyone else

These unusual examples aside, most responses were able to be coded into a clear category. The following plot shows the main reason given by respondents in their answer, again categorised by their wave 7 pre-referendum vote intention:
reasons_categories

These results show that while the single largest word that leavers say is “immigration”, they were actually more likely to mention sovereignty related issues overall. The clear picture we get from this analysis is that leavers are concerned primarily about sovereignty and immigration. In fact reading responses shows that many respondents mention both sovereignty and immigration together, showing that these two issues were closely linked in the minds of British voters.

On the remain side, economic reasons are by far the largest single category, with other respondents split fairly evenly across other categories. These included people who felt European and didn’t want to be “little Englanders”, and people who worried that Britain’s influence in the world would decline.

Overall, our results suggest that the referendum campaign was not a fight about which side had the best argument on the issues: very few people voted leave to improve the economy and very few voted remain to reduce immigration. Instead, the fight was about which of these issues was more important.

 

Using the EU Most important Issue data

All the data used in this post are available in the wave 7 data release. You can look at the raw text data by using the variable MIIEU_text.  You can look at the 15 category coding using euMIISmall or the larger 54 category version using EUMIICategory.

 

  • John Barker

    It seems even at this stage there is no attempt to differentiate between immigration and migration ? I see this failure to be a massive flaw in the data and the presentation.

    • Justin Benn

      No attempt on behalf of whom and/or for whom? If this data represents what people actually say, rather than what you want them to say, how can it be a “failure”? At first glance, the premise of your criticism seems faulty.

  • Justin Benn

    Would be really useful to have these images available as versions that can be scaled up. Some of the words are hard to read!