Who's talking about a Mayor for Birmingham - and Who's listening?

On Thursday 3rd of May 2012, Birmingham (UK) has a referendum on whether to have an elected mayor. We analysed twitter discussions about the subject, to identify people who might influence the debate. Full details of the data & techniques used are given below, but note that the inclusion criteria were "one tweet about the Birmingham mayor in the last week". Approximately 2/3rds of the users shown have only tweeted once about the mayor in our sample time period.

Who Mentioned the Mayor? - Influence within the debators

Who Tweeted about the Mayor - Size is number of followers in this group, colour indicates users with similar groups of followers

Figure 1 - Influence within the group of debators
Size reflects number of  followers within this group of people
Colour indicates users with similar groups of followers
(download pdf file for full zoom)

We allocated each person a colour based on who follows them. From background knowledge of the individuals involved, we attempted to roughly characterise the groups:

There's much more detail to be explored here, and feedback on the groupings (to @AndyPryke or Andy@The-Data-Mine.co.uk) is very welcome

Who Mentioned the Mayor - Influence in the Twiterverse

Figure 2 - Twitterverse influence of people who tweeted about a Mayor
Size is total number of twitter followers
Colour indicates users with similar groups of followers
(download pdf for full zoom)

Who Heard Your Voice? - Number of Tweet Views

Size is Number of impacts (tweet views)
Figure 3 - Number of people seeing tweets about a mayor from particular users
Size is total number of twitter followers * number of tweets about a Mayor
Colour indicates users with similar groups of followers

This final network visualisation indicates the total number of potential "impacts" or views of tweets from each user. The clear "winner" is @politicsinbrum, in fact they had such a high number of potential impacts, around 16,000, that we scaled it down to 3000 so that other users could be seen on the same diagram! This is, of course, a function of both their high number of tweets (121) and their large number of followers. However, many of @politicsinbrum's tweets are re-tweets of other users in the debate - so don't despair if you user doesn't appear here!

What was said?

Three Examples

Lets take a look at a three contrasting examples of what was actually said. Notice that in figure 1, @SalmaYaqoob is well linked, but positioned on the left of the diagram, rather than the centre. This is partly due to a large number of followers who are not well linked in to the main debate, but whose tweets seemed to be about the Birmingham Mayor. This is to be expected due to her large number of total followers (see figure 2) - the more followers you have, the more likely some will be external to the main debate. Deeper analysis of tweets by Salma's followers might give greater insight here.
  • "Londoners - pls do not return a Tory Mayor to lead our capital.It will devastate all those affected by Tory cuts agenda - Cameron will crow."
  • "My views on Brum Mayor vote next Thursday: 'Why I'm saying Yes to a Brum Mayor who says NO to Austerity' http://t.co/waxB
When we examine SalmaYaqoob's tweets, we find they are mainly (9 out of a total of 10) about the London mayoral election, not the Birmingham referendum. They are turning up in our dataset because Salma is tweeting from Birmingham. However, it is still likely that Salma is influencial locally due to her high number of followers4swP #brummayor"

In contrast to Salma is @politicsinbrum who tweeted 121 times on the of a Birmingham Mayor, including a large number of re-tweets and conversations.
  • "Sir Albert Böre says Cllrs will B able 2 challenge an elected mayor. Anyone told Liam Byrne he'll be called 2 account by cllrs  ? #brummayor"
  • "RT @neilelkes: Are voters shifting to anti-mayor vote? Blog post http://t.co/WKSxOlVP   #electedmayor #brummayor"
  • "@web_bod @Lesreidpolitics @politicalhackuk @Covyes. I simply RT mayoral stuff from time to time"

@HouseofTwits has a large following in the general twitterverse but has only a single mayoral tweet, which is a retweet from @gabysslave:
  • "RT @gabysslave one week to go to polling day- don't forget to vote for LE and #brummayor referendum #yes2brummayor"

There's a lot more depth and detail which can be uncovered here, in terms of how twitter is used, the numbers of tweets on Mayoral issues, conversations and re-tweets.

Wordcloud of Tweets

Word cloud of Mayoral Tweets

Figure 3 - What Was Tweeted about a Mayor?
Full size / printable on Wordle Website)

Nice overview of what we'd expect to see (e.g. "debate", "birmingham"), some hashtags and usernames of some key tweeters. It's also interesting to note the "pollution" of the dataset by tweets about the London mayoral election, sent by people in Birmingham or who mentioned Birmingham in the tweets.

Comments, Suggestions & Further Work

Any comments, suggestions or ideas for further work are very welcome:  @AndyPryke or Andy@The-Data-Mine.co.uk

We can also perform data visualisation, data analysis and data mining on your own data - please contact us for more details.

Technical Background & Acknowledements

Data Selection Criteria

Twitter accounts were included if they made tweets matching the following searches between about 5pm on Tues 22nd April & 5pm on Mon 28th April.

  • "mayor OR mayoral -foursquare near:Birmingham,UK within:10mi"
  • "#YesToBrumMayor OR #NoToBrumMayor OR #BrumMayor OR ((#Birmingham OR #Brum) AND (#Mayor OR #ElectedMayor))"

Analysis Tools

R was used to extract & process the data. In particular the twitteR package to interact with Twitter and the igraph package to represent the network of followers. Gephi provided network visualisation, with inspiration from a great tutorial by Tony Hirst.

Want to see some more?

Checkout our analysis of the impact of TedxBrum on social networks in Twitter.

andy pryke,
1 May 2012, 06:07