Social media and society

#spill: How Twitter Reacted to the Labor Leadership Challenge

It’s that time of the electoral cycle again where the Australian Labor Party changes leaders in response to its flagging opinion polling. As with the 2010 leadership spill, which we touched on here, there was a great deal of activity on Twitter last night: the #spill hashtag, in particular, served as one forum through which rumours, information, commentary, and snark were shared in some quantity. Here’s a (very) quick analysis of how the spill played out on Twitter last night.

First, there is a pronounced rise in #spill activity from around 16:15 onwards, as Prime Minister Julia Gillard announces the leadership ballot in a live interview with Sky News; this builds gradually into an 1,100-tweets-per-minute frenzy just before 20:00 as the results of the vote are announced. Other spikes in activity occur around 18:40, as factional leader Bill Shorten announces that he is switching his support to Kevin Rudd; at 21:20, as Julia Gillard makes her farewell speech as Prime Minister; and 22:50, as PM-elect Kevin Rudd and deputy Anthony Albanese hold their own press conference. Opposition Leader Tony Abbott’s response to the events of the day, in a presser from 23:05, doesn’t quite register as much any more – and he is probably not helped by the fact that only a handful of television networks covered it live.

#spill tweets per minute.

The graph above also shows the distinction between various types of tweets: here, original tweets (i.e. tweets which are neither retweets not @replies) are fairly evenly matched with retweets; the latter dominate especially in the later stages of the event, as the volume of participation in #spill, but also amount of actual, confirmed information worth sharing increases.

Genuine @replies (as distinct from retweets) constitute only a small percentage of the total #spill volume, most likely not because Twitter users weren’t replying to one another, but because those replies only rarely carried the #spill hashtag and therefore do not form part of our dataset. Generally, there will be a great deal of follow-on conversation, and of non-hashtagged discussions, in addition to what we are able to analyse here – the #spill dataset itself forms only the tip of the iceberg of last night’s Twitter activity.

Overall, more than 31,000 unique Twitter users posted or retweeted #spill messages last night. If we divide that userbase following the well-known 1/9/90 rule (ordering them by the volume of tweets they contributed, and distinguishing them into a top 1% of lead users; the next 9% of highly active users; and the remaining 90% of least active users), we find a few notable differences between these groups. It’s the handful of lead users who contributed disproportionately many original tweets, while the less active users participated more frequently by retweeting other users’ tweets. However, the lead users did not dominate the #spill hashtag – the top users punched above their weight by contributing some 13% of all tweets (so they were lead users), but others participated strongly as well.

#spill user contributions.

Much of the retweeting and @replying activity within the #spill hashtag centred around a handful of keyparticipants. These include some of the politicians involved – chiefly the Prime Minister and her challenger – as well as the news organisations and journalists covering the event. The following graph provides some insight into the distribution of Twitter users’ attention, and shows whether these accounts were predominantly referenced in retweets or genuine @replies:

#spill key @mention targets.

Unsurprisingly, Gillard and Rudd are mentioned rather than retweeted; neither of them found the time during the leadership spill to do any tweeting of their own. Beyond this, the public broadcaster figures strongly, and mainly as a source of information: #spill tweets by both @abcnews and @ABCNews24 are widely shared by other users. The smaller component of @replies to these accounts mainly comment on the quality of the ABC’s coverage (including some occasional vision switching issues at ABC News 24, and a rather presumptive graphic behind 7.30 host Leigh Sales just as she said the ABC would wait for official confirmation of the ballot result rather than share rumours). Most of the @mentions of the @leighsales account also relate to this graphic, or comment on her overall performance as host; for obvious reasons, she did not tweet during the broadcast.

Other prominent users are present mainly because one or two of their tweets managed to capture the spur of the moment and were widely retweeted. These include

  RT @lucethoughts: Somewhere, Tony Abbott is sitting in a darkened room,
  slowly stroking a Persian cat. #spill

– which, remarkably, was the only #spill tweet from that account, but was still widely retweeted even during the following morning – as well as a somewhat clumsy attempt by job site Seek to cash in on the #spill story:

Hey @JuliaGillard, we can help you out! ;) #spill #auspol

(a great number of retweets here as well, but also quite a few critical comments about the lack of respect shown for the outgoing Prime Minister).

Here’s how these mentions distribute across the overall #spill timeframe:

#spill @mentions over time.

This illustrates quite clearly the long-term impact of the @lucethoughts tweet as well as the much shorter-lived reaction to @seekjobs. And more broadly, of course, we’re seeing the focus of the #spill community shift between mentions of @JuliaGillard and @KRuddMP, as well as the changing attention to (and/or resonance of) the different news organisations.

Finally, for what it’s worth, a quick indication of the global attention paid to this event. This is based only on those #spill tweets which contained their sender’s GPS details – accounting for little more than 1% of all #spill tweets. So, the map below is indicative of interest only, and each dot on this map may represent many more Twitter users in those places who chose not to reveal their current location.

#spill geographical distribution.