Five Million Tweets

Politics and social media

On the front page of the Australian Prime Ministers site we chart the last 100 days’ activity on Twitter, tracking public tweets which mention current and recent prime ministers (both full names and twitter handles) as well as the hashtags #auspol and #qanda. We harvest around 40 000 tweets every day: we’re at over 7 million total at the time of writing. For this analysis, we’re using exactly 5 million tweets from 8 December 2015 to 22 May 2016. 

It’s a huge amount of data. Printed out, 5 million tweets would fill 100 000 pages, a stack of A4 paper 10 metres tall, weighing almost half a tonne.

Incumbency has its advantages

The current Prime Minister has significantly more tweets directed at or written about them than the previous PMs, which is unsurprising given the media attention they enjoy (or, more often, endure).

Chart showing twitter mentions for the current and most recent PMs: Malcolm Turnbull 995,846: Tony Abbott 254,771: Kevin Rudd 31,042, Julia Gillard 24,556.

Hashtags

Twitterers can use hashtags to categorise or identify one or more topics for their tweet. The hashtags arise organically from community usage: some hashtags such as #auspol or #ausvotes, are used to connect to the broad conversation; some relate to State politics; others such as #panamapapers are more targeted, and reflect concerns of the day. 67 hashtags appear in more than 10 000 tweets.

Chart showing the hashtags with more than 10,000, tweets in the dataset: auspol 3,417,184, ausvotes 265,569, qanda 254,871, lnp 131,289, abc 83,923, qt 58,355, insiders 56,468, nbn 55,682, election 53,990, climate 52,152, nswpol 50,168, budget2016 47,983, election2016 47,443, australia 47,061, letthemstay 42,412, qldpol 41,936, shenhua 38,631, csg 38,352, turnbull 37,829, lnpfail 37,099, abc730 35,626, gonski 34,468, refugee 33,509, ausvotes2016 30,366, turc 29,955, tax 29,285, refugees 29,012, nauru 28,897, safeschools 26,098, springst 23,142, medicare 22,071, news 21,778, thedrum 21,629, negativegearing 21,096, ausunions 20,475, climatechange 20,163, coal 19,833, panamapapers 19,444, politas 18,611, manus 18,205, afp 17,247, abcnews24,16,944, banksrc 16,854, wapol 16,632, penaltyrates 16,354, alp 16,137,4corners 16,100, asylumseekers 15,809, abcc 15,148, abbott 14,742, senate 14,657, pmlive 13,600, icac 13,528, australian 13,459, lateline 13,427, dutton 13,308, greens 13,093, marriageequality 12,406, npc 11,930, labor 11,878, tpp 11,856, csiro 11,807, afpraids 11,593, malcolm 11,029, taxrort 10,952, gst 10,861, tony 10,605

Who’s tweeting?

There are 258 129 individual user accounts represented in the dataset, but over half have tweeted just once (some are “sock puppet” accounts, computer-generated for marketing or other purposes: others are very real). At the other end of the spectrum, the 915 users with more than 1 000 tweets each have contributed 48% of all tweets. This power law distribution is common in online communities.

Chart showing percentage of tweets contributed by number of tweets per user. Over 10,000 tweets: 0.1% of users, 7% of tweets; 1000 to 9999 tweets: 0.35% of users, 41% of tweets; 100 to 999 tweets: 2% of users, 30% of tweets; 10 to 99 tweets: 9% of users, 14% of tweets; 2 to 10 tweets: 31% of users, 6% of tweets; 1 tweet: 57% of users, 3% of tweets.

When do they tweet?

Our dataset doesn’t contain tweet locations, but most Australian political tweeters will be on the eastern seaboard. Tweet volumes are relatively consistent during daylight hours, with slight bumps at breakfast, lunch and late evening. The late-evening results are skewed by the #qanda hashtag, which has a very significant uptick during the hour on Monday evening when its TV show is aired. 

Chart showing average tweet volumes by hour of day, peaking at around 1800 per hour at breakfast, lunch and after dinner.

Michael Honey is the founder of Icelab, a design and development studio working with MoAD to build Australian Prime Ministers.