May 06 2015

General Election 2015: what Twitter tells us

Balbir Barn and Franco Raimondi are members of the Middlesex Applied Software Engineering research group, which aims to provide robust solutions for software and system development. With one week to go until the UK general election, in collaboration with Giuseppe Primiero from the Foundations of Computing Group, the group has analysed millions of tweets about the five main political parties in England to see what insights can be gained.

We started collecting tweets related to the UK general election exactly 100 days before polling day on the 7 May.

Our framing idea for the data collection was to evaluate the extent of public engagement in politics. So far we have collected approximately 2.3 million tweets that have mentioned either the leader or the political party for the Conservatives, Labour, Liberal Democrats, UKIP and Green Party. We chose to explicitly exclude tweets written by the candidates.

Proportional representation

Thus far, our findings of the analysis of these tweets provide some interesting observations.

Unsurprisingly, Cameron (or The Conservatives) and Miliband (Labour) are the most mentioned users (around 30%), followed by Farage (UKIP) at 20% and then Nick Clegg (Lib Dems) and Natalie Bennett (Greens) both averaging at 10%.

These mentions roughly correlate in terms of proportions to the poll of polls available on the BBC website.

Political sentiments

We also employed the vaderSentiment tool to help capture the general sentiment in the Twitter data set. A sentiment is expressed as a number between -1 (completely negative) and +1 (completely positive).

Here the results are a little surprising and should raise some questions for party analysts.

Across all the tweets, there is a slightly positive sentiment – 0.057. At the very least, party analysts will be pleased that the long election campaign has not plunged into negative opinion.

Tweets related to Natalie Bennett and the Greens were the most positive (at 0.163). David Cameron and the Conservatives were the least positive (0.019). Ed Miliband and Labour came in at (0.053) and Nick Clegg clocked up a positive sentiment value of 0.091. Farage and UKIP managed a respectable 0.036.

What does it all mean?

How can we explain the strong positive showing of the Greens and the Lib Dems?

Inspections of the words found in the tweets of each of the parties provide some clues. For the two main parties, there are references to each other indicating that there is debate occurring (we infer a debate when words associated with the other party are found in the tweet). If there is a debate there is also likely to be criticism, hence there is lower positive sentiment score of the two parties that really matter.

In contrast, the words in tweets concerning the Greens, Liberal Democrats and other parties are mostly focussed on themselves. For example, in the case of the Greens, we can see words such as ‘manifesto’, ‘greensurge’ and also references to other parties broadly aligned with the Green Party such as the SNP and Plaid Cymru.

There is little or no reference to the other main parties. Hence, the tweets are mostly inward looking and therefore likely to be positive. The range of positive sentiments for the Lib Dems and UKIP can be similarly explained.

Stemmed words in the tweets that mention the Lib Dems do contain references to Labour and Conservative indicating some debate and therefore the potential for negative sentiment and so contributes to lower positive sentiment than that for the Greens.

So a possible unfortunate interpretation of the higher positive sentiment for these parties is that the lack of cross party debate in these tweets reflects their lower importance.

Labour should be pretty pleased with their representation in the Twittersphere. Despite some sides of the press pushing an agenda of ‘Weird Ed Miliband’, something our colleague Martin McGrath discussed in his recent blog post, Twitter has proven a relatively safe haven for the party. Indeed, the most retweeted tweets are all criticisms of the Conservatives and their reluctance to debate in a public setting.

.@David_Cameron I believe my plan can give this country a better future than yours. Disagree? Prove it – debate me & let the people decide.

— Ed Miliband (@Ed_Miliband) April 16, 2015

.@David_Cameron why are you running scared of TV debates? The British people want a head-to-head TV debate. Let’s give it to them.

— Ed Miliband (@Ed_Miliband) March 5, 2015

Different levels of engagement

Our twitter data collection has also included the recent Nigerian general elections and there are some interesting differences between those elections and the UK elections.

For the current UK election, approximately 1.1 million tweets are retweets (47.5%) and around 712,000 tweets are replies (31%). In contrast, retweets constituted 56.5% of the overall Twitter traffic during the Nigerian elections while replies were approximately only 5%.

Clearly, the UK public is significantly more interested in engaging in a public dialogue with the political parties, by directing comments to the parties and their leaders.

Can this be something that party analysts take advantage of in the last few days of the campaign?

Middlesex Applied Software Engineering research group

Share this post

 

Tags: general election, government, politics, polls, research, social media, twitter, voting