Pretty much everybody has heard by now that on June 23, citizens of the UK voted by a narrow margin to leave the European Union. This was one of the most hotly contested political issues in many years — not just for UK and EU residents, but around the world. The news sent shock waves through the global economy, and some are speculating what the move may say about the upcoming US Presidential election as well. The most common question people seem to be asking about the news is some version of “how did this happen?”.
CrowdFlower is in the business of rapidly collecting useful data from humans all over the world, and we viewed this as a unique opportunity to leverage our position to learn more about what people thought about the Brexit vote, what brought it about, and what its effects would be. We constructed a short poll asking our contributors for their thoughts on the issue, launched it over the weekend, and waited to see what kind of response we would get.
By Monday, we had received over 4,200 responses to our poll. This was without any special promotion of it on our platform, and despite the fact that it was run over the weekend (when data collection is usually slower). Moreover, as just a single page of questions, this was not the type of job to attract most contributors (who generally prefer committing to jobs offering large volumes of a given task). All this serves to underscore how impressive a resource for data collection the CrowdFlower platform is. Let’s start looking into the data we collected.
Overall, the people responding to our polls oppose Brexit by more than a 2:1 margin, but that margin narrows when we look just at responses coming from the UK. We also see how few in the UK responded with mixed or no opinions on the referendum, underscoring how important many viewed the issue to be.
At this point, one might ask “why do more people indicate they oppose Brexit, if 52% of voters supported Brexit in the referendum?”. There are several possible explanations. First and foremost, CrowdFlower is not a survey company, and we make no claims as to the representativity of this sample. It’s entirely possible that people using our platform tend to skew liberal, and nearly 75% of our respondents in this poll indicated they had at least a university degree. It is also quite likely that many of them are younger, and voters under 50 tended to oppose the referendum more than support it. This article has lots of great demographic visualizations about how the vote broke down, if you’re curious.
One possible additional factor, though, is regret especially given the post referendum political infighting. 2.6% of our responses from the UK explicitly indicated that they had voted for Brexit but now regret that decision, possibly as a result of the Leave campaign admitting one of their key campaign promises was misleading or maybe because Leave voters thought they were registering a protest vote without real consequences. That’s probably a conservative estimate — there are likely as many or more individuals who voted Leave, but now claim to have opposed the referendum all along, due to a well-known psychological effect known as social desirability bias. However, as noted by the Washington Post, until truly representative polling data is obtained, we will not begin to know how widespread this voter regret truly is. Finally, some voters may view the issue as ongoing, as a petition for a second vote has already received millions of signatures, while another petition is requesting a Parliamentary debate on whether or not to invoke Article 50.
Let’s dig a bit deeper into underlying rationales: what led people to support or oppose the referendum? Economic concerns dominate the discussion, along with the oft-cited issues of sovereignty and immigration particularly on the Leave side. Possibly most interesting is that opponents to Brexit outside of the UK view it as more of a global economic issue than a UK specific one.
In addition to polling questions such as these, we also asked participants to tell us, in their own words, what they thought about the referendum — specifically, what led it to become an issue and why they think the vote to leave ultimately succeeded. To identify common themes in these responses, we made use of topic modeling, a technique for automatically grouping together pieces of text based on the words they have in common. We took responses from within and outside of the UK and modeled them separately (using maul), and the differences are revealing.
When asked about the factors that led Brexit to become an issue, the role of immigration is much more salient for respondents in the UK, with broader European economic issues being much less emphasized. When looking at responses from outside the UK, exactly the other pattern is seen, with regional factors being overwhelmingly cited. Interestingly, UK voters are also keen to point out details of the political process leading up to the vote itself, such as perceived deceptive practices by the Leave campaign and the fact that Prime Minister Cameron allowed the vote to occur in the first place.
Similar themes can be seen in accounts for why the Leave campaign ultimately succeeded, with non-UK respondents tending to focus more heavily on the political issues, and UK voters focusing more on the (potentially dubious) claims of the Leave campaign. In the charts below, each bubble represents a recurring theme in the data, and the words inside the bubble are key terms that help to define that theme.
People both in the UK and beyond point to the issues of immigration and broader economic factors as driving the success of the Leave vote. However, the two groups differ sharply when it comes to how they discuss the ways that voters’ attitudes were shaped by the campaign. Nearly every topic identified in the UK responses contains one or more keywords like “lies”, “gullible”, “false”, “promises”, or “understand” — indicating that many voters feel that the driving issues were not portrayed accurately.
Finally, we wanted to compare what our poll respondents were saying to what could be learned from another worldwide information source: Twitter. We pulled random samples of 10,000 tweets from just before and just after the referendum vote, and identified recurring topics in both sets of data. Leading up to the vote, as one might expect, the conversation was very heavily dominated by discussions of highly publicized debates, and particularly individual performances by several key participants, along with a handful of highly publicized comments on the issue:
Following the closing of polls, however, the picture changed substantially. The conversation became very focused on monitoring results, particularly from regions expected to be very consequential. Others tweeted out messages as they went to bed hoping for a given outcome when they woke up in the morning. Following the announcement of the decision, discussions of the immediate financial impacts became the predominant topic:
Ultimately, what we have learned from this data serves to echo much of what has been said elsewhere in recent days — the success of the Leave vote comes as a surprise to a lot of people (and some are regretting the way they voted in the very close decision). Many issues have contributed to the way things turned out including economic instability, ongoing immigration and refugee crises, rising nationalist and populist sentiment, and questionable politicking. The economic effects of Brexit are foremost on most people’s minds right now. While these things have been said before, to the best of our knowledge no other inquiry into opinions on the Brexit vote has combined responses from so many different countries. That has opened the door to a comparison of internal and external perspectives, which is valuable in understanding such a complex and wide-scoping issue.