The results of the 2017 General Election have come as a shock to the UK; Labour has defied expectations and pre-election opinion polls, leaving the UK with a hung parliament. The huge conversation on social media gives us an insight into opinion, patterns and voter behaviour on a large scale.
We tracked social dialogues and sentiments during the election to get insights in the driving forces behind the changes in the political landscape in the UK, and to see what hidden patterns we can identify within the enormous volumes of related chatter.
For this report, we limited ourselves to Twitter as an information source and all data was collected via our flagship STORY platform on a non-tweaked standard set-up. To capture all relevant Tweets –independent of volume and speed– we used Twitter’s/Gnip’s Firehose for harvesting.
When preparing this we obviously didn’t know what was going to happen or what volumes and patterns to expect, so we started by simply tracking a few typical hashtags in the days leading up to the event and kept doing this until the whole circus was over.
The figure below shows the evolution of the usage of hashtag #GE2017 on Twitter from 2 to 8 June with the time in CEST (i.e., UK time + 1:00).
From this figure alone we can already identify some interesting patterns:
- Tensions were slowly rising the days before the election. The peak on the night before the election was at least twice as high as a few days earlier and the daily volume had already tripled.
- From this picture, it seems as if the London terror attacks on the night from 3-4 June did not affect the elections at all. No special pattern is visible on or immediately after the events.
- In the UK, no numbers/predictions are published in the media until the doors of the polling stations have closed, so on election day itself, Twitter volumes were higher than usual, but still very much under control, averaging around 10 Tweets/sec.
- This changed dramatically the moment the doors had closed and the first exit polls were published, predicting a disastrous outcome for the Conservative party. Within seconds, Twitter volumes Skyrocketed and when scrutinising the feeds, we found individual seconds with over 200 election-related Tweets. The average volume over the first full hour after closing time was 3 times that of the hour before.
- The impact of the first predictions was so large that Twitter even forgot to go to bed: the usual day-night pattern was completely distorted and it took another 24 hour for a normal UK night to set in again.
We found it hard to believe that the London terror attacks had no effect on people’s sentiment around the elections, so we dug into this a little deeper. We correlated #GE2017 occurrences with various other keywords and found the following interesting pattern when looking for “enough” or “resign” among the #GE2017 Tweets:
Clearly, sentiments trended on at least 3 occasions, not seen in the global patterns:
- On Sunday 4 June “enough” occurrences had a sharp peak and after little investigation, we found that this was the direct result of just 1 Tweet going viral, namely this one:
Remarkably and unexpectedly, this Tweet was Retweeted faster and more often than people quoted Theresa May’s own “Enough is enough” statement.
- On Monday 5 June “resign” suddenly trended, and this turned out to be the direct result of Corbyn’s statement where he asked for May’s resignation and accused her of cutting police budgets while she was home secretary.
- On Friday 9 June “resign” trended again for almost the complete day: immediately after the first exit polls were made public, around her speech for winning her district, the moment the first morning news shows were broadcast on radio/TV and finally around lunch when she announced not to step down and to form a coalition.
It is remarkable that individual Tweets with such an enormous impact can start to trend so suddenly and be shared on a large scale so quickly (we’re talking minutes here). There lies enormous value in having the right toolset so that key events such as these can be identified at an early stage, while still having sufficient time for analysis and/or appropriate actions. Obviously ‘appropriate actions’ may mean something completely different for a journalist, a TV producer or a political campaign leader…
Patterns in Party Loyalty
We also tracked social volumes attributed to explicit support of one of the participating parties. For each political party, we monitored occurrences of #Vote<xxx>, where <xxx> is either the party name/abbreviation (e.g. #VoteLabour) or the party’s main candidate (e.g. #VoteMay), and of course various variations and alternative spellings. For a culture that’s known and often praised for its reserved nature, people in the UK turned out to surprisingly forthcoming when it came to sharing their political support online. This led to a steady flux throughout the day of 10 Tweets/sec in which people called out their support.
The pattern found while tracking specific party loyalty this was extremely one-sided, see the figures below:
Three conclusions are obvious:
- Labour (pink) was leading the social arena, with no other party coming even close. In fact, each other party collected less than 10% of Labour’s volume and added together the others didn’t even reach 25% of Labour.
- Although being the leading party when it comes to the popular vote, the Conservatives (green) are extremely marginal online and generate an online presence comparable to that of UKIP (blue), SNP (grey) or LibDem (orange).
- The sharp decline of the sharing party loyalty exactly at the time the voting stations closed, probably indicates that people share their political preferences only when it makes sense to do so, i.e., while there is still time left to persuade others to vote similarly.
However, we did (and still do) not claim that the volumes seen are direct indications of the election outcome. First of all, the UK system with electoral districts makes it hard to map nation-wide numbers onto parliament seats.
Secondly, and probably much more importantly, the pattern above is of course the result of a generational bias: younger generations use social media more often and more openly than older ones, and at the same time those same younger generations are expected to support a more progressive political agenda, i.e., are more likely to favour Labour over the Conservatives. Hence the pattern itself is perhaps not surprising, but the enormous size of it was certainly not what anyone had expected.
Finally, it is interesting to see what topics the supporters of the parties addressed as they publicly shared their preferences. To investigate this, we created a Word Cloud for the supporting Tweets for the two main players, Labour and the Conservatives:
Interesting common factors are that both talk about their own leader, the other party (which is surprising), ‘10PM’ (the moment the polling stations closed), ‘polls’ (anticipation?) and ‘Britain’ itself.
For the Conservatives ‘Brexit’ was a big topic, supported by related terms, such as ‘deal’, ‘strong’ and ‘economy’. And the Conservatives also wanted to give an outlook, with ‘future’, ‘tomorrow’ and ‘brighter’. Labour on the other hand, had a much more now and people-focused agenda with ‘NHS’, ‘share’, ‘people’, ‘#forthemany’. At the same time, terms like ‘please’ and ‘chance’ also express a certain level of uncertainty about the future among the Labour supporters.
Patterns in Candidate Mentioning
We closed our analysis by looking when the various candidates where mentioned throughout election day. Not only did we consider true Twitter @Mentions, but we added references to political candidates in Tweets with an election context:
Clearly, one can differentiate between the world before closing of the election and announcements of the first exit polls, and the world after that moment. Before the breaking point, the pattern is qualitatively similar to that of party loyalty: Labour/May (pink) is consistently higher than the Conservatives/Corbyn (green), albeit it with much smaller margins. The social behaviour is also quite smooth, without too many spikes, apart from one at the end of everyone’s work day.
That changes dramatically after the first results were made public. Theresa May becomes the most talked about candidate within seconds and throughout the night and the whole following day, here volumes are significant higher than would have been anticipated in advance. Both May’s and Corbyn’s chatter become much more erratic than before and this continues to be the case for at least 24 hours, completely ignoring the fact that it was night.
In those volumes one can see profound peaks the moments both candidates gave their victory speech in the middle of the night for winning their district. In addition, there are other peaks when those who went to sleep, got up again, and when May read her statement that she would not resign.
The most interesting thing to learn from this is that sentiments like chaos, disorder, disbelief and despair are so obvious from these graphs. It is almost as if after closing time, the UK Twitter universe entered some bizarre reactive modus, which caused the social chatter to explode and reach extreme volumes (certainly for the time of day), combined with equally high fluctuations.
Dogs at Polling Stations
If you read the above you might think that the UK has fallen back to an effective 2-party democracy as everything seems to be just about Labour vs. Conservatives. We did find an exception for this and it actually links back to a pattern we observed earlier during the Brexit referendum, namely trending of the hashtag #dogsatpollingstations.
When looking at how individual party candidates correlated with #dogsatpollingstations to measure the political preferences among dogs or their owners, we not only found that Corbyn (pink) and May (green) had a very comparable online presence, but also discovered that both are easily beaten by Sturgeon of the SNP (grey). One must conclude that Scottish canines take their civic duties much more serious than their friends elsewhere in the UK do.
By Taco Nieuwenhuis, Chief Technology Officer at never.no
Being the sole Dutch employee, Taco frequently smuggles cheese and stroopwafels from his home country into Oslo and constantly tries to understand Norwegian culture. In his spare time he’s obsessed with cats and penguins, theoretical physics and leads a parallel existence as garagerocker and concert-dj. Not always simultaneously though.
Related Content: Social Reaction to the EU Referendum