#GE2017 – Tories vs. Labour: Party Resonance & Personality Analysis [Not a Poll!]

If you are interested in getting an update on the latest UK election polls, this is not what you are looking for. This article is clearly not intended to predict the outcome of the elections in any way. So if it is polls you are looking for, you might want to have a look at polls from the BBC, the Independent, the FT or the Telegraph – of course considering them with the necessary caution as none of the pollsters have predicted the UK 2015 election results, the Brexit vote or the US electing Donald Trump as their new president.

With that in mind, I am not aiming at predicting the election outcome but rather want to shed some light on the communication performance of the two main parties in this snap election – specifically as predictions based on non-representative (i.e. social media) data sets are generally vulnerable, nicely illustrated by widely contradicting predictions based on exactly these datasets, see here and here.

Also, having already established the unrepresentative character of the dataset itself, there is obviously no point in counting quantitative reach and engagement differences between the parties – meaning we can’t conclude anything from the over 8 times higher amount of #VoteLabour tweets (16,974) over the amount of #VoteConservative tweets (2,048). That said, what we should be actually looking at is the content resonance of the parties’ messages with their respective followership.

To that end, we have extracted all @Conservatives and @UKLabour tweets and have been comparing their content linguistically with the topics that the respective party followers have been talking about. In order to visualise this process, we have created a direct word cluster comparison for each party. For the Tories, this comparison looks like this:

What we can see here is that the Conservatives have focused their Twitter campaign very much around Theresa May, her “strong and stable” leadership and to some degree around the “weaknesses” of her opponent Jeremy Corbyn. Looking for the resonance of this message in the topics of the self-expressed Tory voters (on the right hand side of the visual above), we find all of these topics represented, although in the case of Theresa May and Brexit to a much lesser degree. This is how the same comparison looks like for the Labour side:

Again we can see how this time the Labour party has tried to put their leader Jeremy Corbyn very much in the middle of their campaign. Different to the pattern in the Tory voters, the self-expressed Labour voters showed quite strong support for their front man Corbyn. Also, we can not only see more of the topics of the Labour party’s political agenda being picked up by their supporters but also topic clusters emerge independently from the main social media agenda of the party, for example the BBC debate, the declared goal to push the “tories out” as well as Corbyn’s message of “hope”. Another difference to the Conservatives’ resonance pattern is that the Labour voters seem to not only have picked up but strongly resonated with the political message “for the many” and have again independently from topical agenda of the party engaged around the “labour manifesto”.

Overall, the Labour voters seem to show a stronger resonance with the message of their party based on this topical analysis. Also, for both parties, Brexit itself seems to be of much less importance for the supporters than the actual political key messages and positions of both parties suggest – which indicates that the EU Referendum might actually have been more of a “symbolic vote”, with which the electorate wanted to give the political elite a wake up call [A/N: interpretation of the author].

However, we did not stop at this level of resonance analysis but also calculated the language style matching score for both parties, using 500 random tweets from both parties and followers. The language style matching score (LSMS) determines the degree to which any two contents of language are similar in their language styles. Scores range from about 50 to 100. The closer the score is to 100, the more in sync two language styles are. Most LSM scores for email, IM and transcribed conversations range between 75 and 95. The more that the two people are paying attention to each other in their interaction, the higher the LSM. Research has shown that LSMS are associated with the quality of interpersonal relationships, and how long relationship last. Without further ado, here are the results of our language style matching analysis for both parties and their supporters:

This quantitatively confirms what we have analysed in a qualitative way before and supports our first impression that the Labour party seems – in this snap election – been better able to communicate with their supporters in a way that resonates with their own key interests and way of communicating.

Now, whether this will have an effect on the outcome of the election, we don’t know and can’t predict from this data due to the general characteristics of this dataset. It however can very well explain why Labour has been able to show such a good performance in the polls over the last week, gaining massive points whilst the Tories were basically stagnating or gradually losing. But again, polls are only polls.

If you haven’t voted yet, please do so. Here is a short side to side personality profile comparison between Conservatives and Tories based on their Twitter posts – hopefully this will help some of you make your final decision. #govote