One week to the General Elections: some Twitter analysis

We started collecting tweets related to the UK General Elections at the end of January, exactly 100 days before election day. We are now one week away from the elections on the 7th of May and we can already do some analysis. We have collected tweets mentioning one of the following users:

  1. David_Cameron or Conservatives
  2. ED_Miliband or UKLabour
  3. Nick_Clegg or LibDems
  4. Nigel_Farage or UKIP
  5. Natalieben or TheGreenParty

NOTICE: this collection may not include tweets written by the candidates. The analysis below applies to tweets in which one of the 10 users listed above is mentioned explicitly. The technical details of the analysis are described in this other blog post: http://www.rmnd.net/some-experiments-with-postgresql-and-simple-twitter-analysis/

At the moment we have approximately 2.3 million tweets. This is the daily relative popularity of the five groups above (click for a larger version):

dailyOn average, Cameron (or Conservatives) and Miliband (or UKLabour) seem to be the most mentioned users (around 30% each), followed by Farage (20%), followed by Nick Clegg and Natalie Bennett (10% each).  Some exceptions are on the 24th of February for Natalie Bennet (40%) and on 22nd of March for Nigel Farage (63%). These two days correspond, respectively, to Natalie Bennett’s interview on LBC and to Nigel Farage’s interview with Andrew Marr.

We have  computed the sentiment of each tweet and then computed the average sentiment. This is a number between -1 (completely negative sentiment) and +1 (completely positive sentiment). We have employed the vaderSentiment tool and the result is that, overall, the sentiment is 0.057 (slightly positive).  We have also computed the average sentiment for tweets mentioning only one of the candidates+party. The results are:

  1. David_Cameron or Conservatives: 0.019
  2. ED_Miliband or UKLabour: 0.053
  3. Nick_Clegg or LibDems: 0.091
  4. Nigel_Farage or UKIP: 0.036
  5. Natalieben or TheGreenParty: 0.163

We computed the most used words, excluding so-called “stop words” such as “the”, “is”, etc., and considering only the root of words. The overall top 40 words are (sorted from most to least used):

  • cameron, david, ed, miliband, uklabour, ukip, nigel, farag, conserv, vote, thegreenparti, labour, libdem, peopl, natalieben, clegg, nick, say, get, tori, parti, like, want, would, uk, tax, go, ge2015, one, nhs, need, think, support, work, debat, time, make, elect, good, leader

We repeated the same computation for tweets mentioning each candidate+party, and the results are:

  1. David_Cameron or Conservatives: cameron, david, conserv, ed, miliband, peopl, vote, uklabour, debat, get, tori, uk, say, ukip, nhs, work, want, tax, go, labour, like, let, ge2015, nick, one, clegg, time, plan, give, would, parti, elect, need, pleas, think, make, libdem, us, back, pm, help
  2. ED_Miliband or UKLabour: ed, miliband, uklabour, labour, cameron, david, vote, tori, peopl, say, tax, get, parti, conserv, debat, would, want, like, ukip, ge2015, time, one, nhs, work, go, think, need, make, let, support, uk, plan, good, back, elect, give, today, us, know, better, well
  3. Nick_Clegg or LibDems: libdem, nick, clegg, cameron, david, conserv, ed, vote, miliband, uklabour, ge2015, ukip, parti, tori, say, peopl, get, would, labour, nhs, go, time, support, thelastleg, want, debat, like, uk, year, one, tax, elect, govern, think, need, polici, coalit, see, good, cut
  4. Nigel_Farage or UKIP: ukip, nigel, farag, vote, say, peopl, parti, get, cameron, leader, like, eu, labour, uklabour, uk, immigr, david, want, think, one, conserv, would, ed, support, miliband, suzanneevans1, polici, go, racist, polit, tori, need, ge2015, make, british, bbc, time,
    well, right, see, good
  5. Natalieben or TheGreenParty: thegreenparti, rt, natalieben, green, vote, parti, carolineluca, nicolasturgeon, leadersdeb, leannewood, polici, votegreen, polit, support, manifesto, need, peopl, leader, say, want, elect, would, ge2015, greensurg, get, like, ukip, pleas, make, uk, good, labour, well, plaid, thesnp, help, chang, join, think, cymru, tonight

These are the most retweeted tweets so far:

  • “.@David_Cameron I believe my plan can give this country a better future than yours. Disagree? Prove it – debate me & let the people decide.”, by Ed_Miliband, retweeted 4469 times.
  • “.@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.”, by Ed_Miliband, retweeted 2299 times.
  • “@David_Cameron can you please explain this!!! It’s utterly disgusting! http://t.co/R26fu3C11T”, by RoxsanRitchie, retweeted 2212 times.
  • “@Nigel_Farage @NICKIMINAJ \n\nhttp://t.co/biRsVZw1q4 http://t.co/AFYlDioq0c”, by SPhizzle, retweeted 2002 times.
  • “@nick_clegg please sign this and retweet https://t.co/ITdM52qg8v”, by GavinLilley, retweeted 1927 times
  • “I’ll debate with you @Ed_Miliband, even if @David_Cameron won’t. Any time, any place, anywhere. #BBCDebate http://t.co/GWc6JGBVYk”, by nick_clegg, retweeted 1533 times.
  • “>@PrivateEyeNews confirms Murdoch’s The Sun literally faked news to attack @Ed_Miliband #GE2015 http://t.co/oGYblEy2J3”, by TheMurdochTimes, retweeted 1487 times.
  • “.@David_Cameron Here you go, you stupid cunt. You’re welcome, you stupid cunt. http://t.co/PL2P99ZyBw”, by DaftLimmy, retweeted 1267 times.
  • “The strongest human beings are the most forgiving human beings @ParisHilton @KimKardashian @ladygaga @FLOTUS @HillaryClinton @David_Cameron”, by drissawaked, retweeted 1172 times.
  • “RT if you would like to politely ask @Ed_Miliband to clarify why he used the loaded word “Islamophobia” & exactly WHAT he plans to outlaw.”, by RichardDawkins, retweeted 1146 times.

Approximately 1.1M tweets are retweets (47.5 %), and approximately 712K tweets are “reply-to” (31 %). These are interesting figures when compared with other national elections. For instance, in the case of Twitter traffic associated to the candidates for the Nigerian elections, retweets constitute 56.5% of the overall traffic, while reply-to are only approximately 5%.

Please contact Balbir Barn or Giuseppe Primiero or Franco Raimondi if you have queries about this!