Regular readers of this blog will know that we’ve now examined Twitter activity around a number of recent events in some detail – from the Labor leadership spill in Australian politics in June 2010 through to the subsequent election, to the recent floods in Queensland and beyond. On that basis, we now also in a position to make some comparisons between these events: in the first place, to examine how they unfolded, and how much of the wider Twitter userbase they’ve been able to mobilise.
So, building on the work we’ve already done, and adding a few more case studies into the mix, here’s an overview of activity within selected Twitter #hashtags – in each case, over the course of their most active day. The process is similar in each case: retrieve a full #hashtag archive from Twapperkeeper, run our explodetime.awk Gawk script over the data to identify daily and hourly activity, then pick the 24 hours during which the volume of tweets in the #hashtag was most significant.
The #hashtags selected for this exercise are the following:
- #spill: Rudd/Gillard Labor leadership challenge speculation on 23 June 2010
- #ausvotes: election day in the Australian federal election on 21 Aug. 2010
- #qldfloods: the height of the south-east Queensland floods on 12 Jan. 2011
- #vicfloods: the height (?) of the Victorian floods on 15 Jan. 2011 (but note that some 20% of tweets in #vicfloods on 15 Jan. also contained the #qldfloods hashtag – so there may have been some overall discussion about floods and other natural disasters)
- #wikileaks: the arrest of Julian Assange on 7 Dec. 2010
- #egypt: 8 Feb. 2011, dubbed the “Day of Egypt’s Love” and apparently marking the largest protest gatherings to date (but also see my note below)
Because the last two cases concern international events (and took place in European/African timezones), we’ll be looking at a period from 00:00 to 23:59 GMT for #wikileaks (since Assange was arrested in London) and from 00:00 to 23:59 GMT+2 for #egypt (Egypt’s timezone); for the rest, we’re using Australian Eastern Standard Time (GMT+10).
(I also need to note that Twapperkeeper has had a significant outage over the past days, which means that #egypt data between 4 Feb. and 8 Feb. is currently missing from the archives – so take the reliability of the #egypt data with a grain of salt for now. In fact, the tweet count for 00:00 to 00:59 on 8 Feb. is only 20, compared to over 2200 during each of the following three hours, which seems unlikely, too.)
The results put these events into perspective – and unsurprisingly, it’s an international event which ends up showing the most overall activity (as shown in the totals on the left side of the graph – note that these bars are cut off at 20,000 tweets to aid visualisation). (Click on the graph for a full-size image.)
Update: Argh – the original version of this post contained a pretty major error: for #spill, I’d worked off a dataset that included only @replies, rather than all the tweets using the hashtag. I’ve updated the graph below to using the full dataset – which brings the #spill peak almost up to the level of #ausvotes now (a maximum of 13,415 tph for #spill, 14,809 tph for #ausvotes). Oops.
Additionally, the international focus of #wikileaks and #egypt also means that diurnal patterns of activity are far less pronounced, of course – it’s always daytime somewhere around the globe, and so the #wikileaks discussion, for example, keeps ticking over even when Europe, America, or Asia and Australia are sleeping. The same goes for #egypt, but here we do see some indication of a ramping-up of tweeting volume during the Egyptian day; at the same time, there is also a pronounced drop during the middle of the day, and it would be interesting to explore whether this was caused by local factors (the unfolding of events during the day, or even another Internet clampdown by the government) or is an artefact of unrelated factors (such as continuing problems with Twapperkeeper).
For #wikileaks, we also see a substantial ramping up of activities from 10:00 (GMT) onwards, which is when first reports of Assange’s arrest surfaced; this lifts the #wikileaks tweeting volume from an average of almost 1500 tweets per hour before ten to nearly 7500 tweets per hour for the rest of the day. A similar ramping-up occurs in two of the Australian cases: both on election day and on the day of the first #spill speculation, things heat up considerably towards the end of the day. This is easily explained by the choreography of events during each day, of course: activity on #ausvotes during election day is naturally focussed on the evening, as the counting of votes begins and first electorate results are announced; in the case of the 2010 election, the uncertain result of the vote meant that high volumes of tweeting activity continued well into the early hours of the following day (not shown here). Similarly, speculation about a leadership challenge in the Labor party first appeared during TV primetime, and was soon confirmed – and as such confirmation was received and the next morning nominated for the party room vote, we also see a slowdown of activity towards the end of the day (with another spike in the morning, as we already know). By contrast, the (comparatively very limited) #spill activity earlier on 23 June refers to the BP oil spill in the Gulf of Mexico, and not to Australian politics.
Finally, the two flood events result in comparatively far less Twitter activity, but that should not be seen to negate the argument that social media like Twitter and Facebook did play an important role during the flood response. Rather, it simply indicates the relatively more limited userbase participating in these #hashtags: for the most part, only local residents could have sensibly engaged in #qldfloods and #vicfloods discussions (and indeed, outsiders tweeting into those conversations with generic messages of support or other comments were at times told to stay out and keep from clogging the channel with non-essential tweets). Given that the immediate participant base for #qldfloods measured only some 2-3 million south-east Queensland residents, of whom only a fraction would have been on Twitter, while a national event such as the leadership #spill affected the entire national population of over 22 million, that #qldfloods clocked up over 11,500 tweets on its peak day (compared to #spill’s 14,000) should be seen as a substantial achievement, in fact. Similarly, with the Victorian floods affecting mostly relatively rural areas of the state, its 1,100 tweets are not insignificant.
However, at the same time it seems obvious that some themes simply exercise the current Twitter community much more than others. Political topics would seem to be one category; political topics which directly connect with Internet themes (as in #wikileaks) doubly so. It’s still far too early to develop a comprehensive perspective on this, of course – one of the aims for our project, therefore, is to examine a wide range of the themes which Australian users tweet about, and to track how the associated #hashtags perform in comparison to one another as well as in comparison to the long-term average within the Australian twittersphere. And of course international comparisons will also be of great interest – just how Twitter-active are Australian Internet users, in comparison with other nations?
One interesting pointer towards the answer to that question, at any rate, already seems to be contained in the graph above: the way #ausvotes outperformed all the other #hashtags collected here – even the international cause célèbre of Wikileaks / Julian Assange on the day of his arrest (with up to 14,500 tweets per hour in #ausvotes, as compared to just over 10,000 for #wikileaks) – is frankly astonishing. Do Australians tweet a lot, then? Clearly, yes. More so, per capita, than Twitter users from other countries? Time will tell…
Just as I published this, the Twitter blog released some figures on tweeting activity during the recent Superbowl (which I’m told is some kind of American football thingy ;-), using a very similar approach to our research here. The headline result: #superbowl activity peaked at just over 4,000 tweets per second as the game concluded – and the most-mentioned names were those of the half-time break entertainers. By comparison, #ausvotes peaked at 18 tweets per second (at 21:10:37 on election night) – and most of those tweets commented on the election of Wyatt Roy as the youngest-ever member of parliament.