Mapping Online Publics

By any measure, it’s been an extraordinary weekend here in Queensland. The same electoral tsunami that brought the LNP to power in March 2012 has washed the government away again in January 2015. Premier Campbell Newman has lost his own seat of Ashgrove, and Labor leader Annastacia Palaszczuk looks poised to form government – although whether Labor can govern in its own right or will need to depend on the support of the crossbenchers still remains to be determined as late counting continues.

This post is the final step in our analysis of social media activities surrounding the Queensland election, following on from previous updates here and here. A reminder about our methodology: the dataset for this analysis includes any tweets which contain the key hashtags #qldvotes and #qldpol, their variations, and related keywords, as well as mentions of the Premier and Opposition Leader and of any of the parties by name; further, we’ve identified the Twitter accounts of over 150 candidates and are capturing any public tweets directed at them, as well as any of their own tweets that include any election-related hashtags and keywords.

For this update, let us begin with a closer look at how election day itself unfolded. An overview of the major hashtags used in election-related tweets over the course of the day is useful for this purpose (as always, the totals here will add up to more than 100% because some tweets contained multiple hashtags). Starting our analysis from 8 a.m. on Saturday, we see the #putlnplast hashtag make much of the early running, as part of Labor’s (and some of the minor parties’) last-ditch campaign to ensure that no preferences are lost to the count. There is also significant use of the #ashgrove and #brisbane hashtags, pointing to the obvious importance of the urban Brisbane electorates to the eventual outcome of the vote.

By 5 p.m., the Nine News / Galaxy exit poll predicting a substantially worse than expected result for the LNP government has captured the majority of the attention: both #9news and #breaking begin to trend, and #breaking continues over the following hours as the discussion moves on from Nine’s exit poll to first booth results but the surprise about the unexpected election trend persists. Finally, as the evening progresses we see a very notable shift in focus: beyond the shock about the Queensland government’s demise, discussion now turns increasingly to its implications across the nation. #nswpol and #wapol come to trend (as the two states next in line for state elections), alongside #vicpol, and the broader discussion about the future direction of the #lnp also gathers steam.

Longer-term patterns in the data show the usual election day spike in tweeting activity, which sets in especially after polling booths close and television broadcasts begin to present first exit polls, predictions, and increasingly robust voting trends. On Saturday, in fact, this happened slightly earlier than usual, as Channel Nine – going against established media conventions – reported its exit poll even while polling booths were still open, thus potentially influencing the choices made by late voters.

Interestingly, even as an unexpectedly strong election performance by the ALP became more and more of a certainty over the course of the evening, the divergent patterns in Twitter activity that we’ve noted throughout the election remained stable: a substantially greater number of tweets mentioned terms related to the LNP than to the ALP in their content (as shown in the first graph above), but somewhat more tweets directly @mentioned or retweeted ALP candidates’ accounts than those of LNP candidates (as is evident from the second graph above) – with the accounts of independent or minor party candidates mentioned far less still.

My reading of this continues to be that Twitter users were significantly more happy to talk about the LNP than they were to directly engage with their candidates, and conversely far more happy to @mention and retweet ALP candidates than to talk about the ALP as such. Of course this also seems to be borne out by the eventual election results – while Twitter is inevitably not directly representative of overall popular opinion, here it appears to be reasonably closely aligned with it.

It is useful to compare this again with the candidates’ own tweeting activities, too. The graphs below clearly show that ALP candidates were substantially more active and more popular than their LNP and minor party counterparts, even in spite of the electorate’s overall focus on discussing the LNP government.

The first of the three graphs indicates the total number of tweets referring to the various parties – and here we see more than 2 ½ times more tweets about the LNP than about the ALP. The balance is reversed in the second graph: tweets @mentioning or retweeting ALP candidates outperformed those to LNP candidates by a factor of 1.25 – and while @mentions came out roughly even in the end, ALP candidates received more than seven times the number of retweets that LNP candidates did – this points to a substantially greater willingness to endorse candidates’ statements.

Finally, ALP candidates were also quite simply a great deal more active than those of the LNP or any other party: overall, they posted more than 7,400 tweets over the course of the campaign, compared to just over 3,000 tweets from everyone else (of which just under half were posted by LNP candidates). Clearly this very active use of Twitter gave ordinary users a greater opportunity to endorse ALP candidates’ statements by retweeting them, and thus to increase their visibility on Twitter and beyond (since tweets are often also cross-posted to Facebook and other social media platforms).

As with any election, it would be foolish to say that its social media activity (or any other single factor, for that matter) won Labor the election, of course – the ALP was also very active on Twitter during the previous Queensland state election in March 2012, for example, and during the Australian federal election in September 2013, but it seems certain that no amount of social media action could have prevented the landslides it suffered in those cases.

In a campaign as close as this year’s, however, surely the ability to spread additional election messages through social media won’t have hurt Labor’s candidates. It’s a truism in marketing that messages delivered by trusted friends through word of mouth are considerably more effective than generic advertisements, and this is true also for the particular form of marketing that elections represent: trusted Twitter or Facebook contacts sharing on a Labor candidate’s call to number every box and put the LNP last, for example, are likely to be a great deal more effective than the various party banners plastered all over the polling booths.

In the 2015 Queensland state election, then, it is similarly likely that no amount of social media activity from LNP candidates would have prevented a very significant correction of the previous election’s extraordinary result. However, by remaining largely inactive – in comparison to their major challengers – the LNP and its candidates essentially ceded the social media space to the ALP and the other opposition parties altogether; even LNP supporters on Twitter had very little party content that they could have used to counter the many retweets received by ALP candidates’ tweets. Giving up the social media contest in this way, right from the start of the campaign, must be seen as another failure of the LNP campaign – and in an election as close as this, every last mistake matters.

This year’s Queensland state election campaign may be very brief, but it’s certainly been action-packed. In my previous post I provided an overview of what were roughly the first two weeks of the campaign; mid-way through the final week, here’s a further update of how the social media campaign has unfolded. As a reminder about our methodology, the dataset for this analysis includes any tweets which contain the key hashtags #qldvotes and #qldpol, their variations, and related keywords, as well as mentions of the Premier and Opposition Leader and of any of the parties by name; further, we’ve identified the Twitter accounts of over 150 candidates and are capturing any public tweets directed at them, as well as any of their own tweets that include any election-related hashtags and keywords.

In this post I’m covering the timeframe of 8-27 January. And in addition to this retrospective analysis, remember that there is also our Queensland Election Social Index (QESI), which takes in Twitter and Instagram data and generates a live overview of the current balance of attention and interaction around the various parties contesting the election; you can see QESI in action over on the QUT Social Media Research Group site.

Reviewing the overall patterns we’ve seen over recent days, then, the general observation I made in the previous post still stands: while the two major parties remain relatively closely matched in terms of @mentions of candidates’ Twitter handles, with a slight lead for the ALP (as seen in the second graph below), there is still considerably more discussion of the LNP and its leader Campbell Newman than there is of the ALP and Annastacia Palaszczuk (as evident from the first graph), with all other parties considerably less prominent.

Over the past few days, and excepting the comparative lull in election-related tweeting during Australia Day, we’ve observed a gradual increase especially in the number of posts per day which discuss the parties (less so for the number of @mentions of candidates), and here again especially where the focus of discussion is on the LNP government; this is a common trend in social media activities during election campaigns, whose volume almost always increases as we come closer to election day.

Amidst that overall increase are a number of key moments: these include the LNP campaign ‘launch’ on Sunday 18 January, with spikes both in general discussion and in @mentions of Campbell Newman and other LNP leaders; the ALP campaign ‘launch’ on Tuesday 20 January, with an even greater spike both in ALP-related tweets and in @mentions of Annastacia Palaszczuk and her team; and a substantial spike in in-text mentions as well as @mentions of both leaders during the televised Queensland election debate on Friday 23 January.

But over the weekend preceding Australia Day itself, there is also sustained discussion especially about the LNP and its leadership, and more so in the text of tweets themselves than in tweets which @mention Newman or other LNP candidates – indeed, at more than 14,000 tweets per day, 24 and 25 January record the greatest number of tweets discussing the LNP that we have seen so far. This is unusual as weekend days are typically relatively slow days for political discussion on Twitter, and long weekends in summer doubly so.

A brief qualitative examination of the tenor of those tweets will bring no joy to LNP supporters: the majority of these tweets address issues such as the Premier’s threats that local election promises will only be kept in electorates which vote for LNP candidates; his claims that illegal bikie gangs were donating to the ALP campaign (and subsequent suggestion that journalists should Google for evidence); and his absence from the Ashgrove candidates’ debate. Newman is strongly criticised on each of these points, and the fact that users posted such criticism even on the long weekend may be an indication just how badly those issues sat with the general electorate.

Over the course of the past ten days, the gap in the number of @mentions and retweets received by candidates of the two major parties has closed somewhat, as seen in the second graph below, although the stark differences specifically in the retweets of candidates’ messages persist: few Twitter users have chosen to retweet LNP candidates’ posts in the campaign to date. This is party due to the fact that LNP candidates are also tweeting a great deal less than ALP candidates, of course (as the first graph below shows), so there is a smaller number of messages that could potentially be retweeted – but even so, the differences are significant.

Amidst all of this activity around ALP and LNP, incidentally, candidates for the minor parties are receiving very few @mentions or retweets – but since we last reviewed their numbers, independent candidates have pulled ahead of Greens candidates in the total number of @mentions and retweets received. Long-serving independent Nicklin MP Peter Wellington is well in the lead amongst this group, especially following his stated intention to complain to the Queensland Police and Electoral Commission over the Premier’s statements on local election promises.

Independent candidates and the Greens have been quite active in tweeting from their Twitter accounts, however, together coming close even to matching the Twitter activities of LNP candidates (who continue to remain comparatively quiet in their uses of social media). Well ahead of the rest of the pack remains the ALP, whose candidates across the state have been sending twice as many tweets as all other candidates combined.

Such activity is not evenly distributed, however: at well over 400 tweets each (since 8 January), candidates Gail Hislop (Burleigh), Penny Toland (Broadwater), Leanne Donaldson (Bundaberg), and Mark Bailey (Yeerongpilly) have been substantially more active than their colleagues, who have typically remained below 100 tweets. (Indeed, there is an argument to be made that lower volumes of tweeting activity may be more effective, since they allow for key messages to stand out better and avoid overwhelming an account’s more casual followers with constant updates.)

Finally, the past ten days have also seen the emergence of a range of additional hashtags accompanying the campaign, in addition to the obvious #qldvotes and #qldpol hashtags and a handful of other generic variations, as the graph below shows (as before, percentages above 100% are due to the use of multiple hashtags in the same tweet). While the #imwithstupid controversy of the first few days of the campaign has gradually disappeared and even the discussion of the federal government’s aborted initiative to change Medicare rebates by now seems like a distant memory, the various campaign ‘launches’ and other events have resulted in a handful of short-lived hashtags: #strongchoices for the LNP ‘launch’, and #qldforum as well as #pplsforum for the TV debate.

The ALP ‘launch’ on 20 January, by contrast, did not see the emergence of a major hashtag in itself (our Queensland Election Social Index briefly showed both #alpqldlaunch and #qldalplaunch as trending hashtags, so perhaps the confusion over which term to use kept it from lasting prominence); the trending of the #abbott hashtag on the same day may be merely coincidental. This is not necessarily bad news for the ALP, however: a substantial portion of the #strongchoices tweets accompanying the LNP event were strongly critical of the LNP’s policies, after all.

More recently, #ashgrove-related discussion has trended again, especially in relation to Campbell Newman’s no-show at the local candidates’ debate, and pro-ALP hashtag #putlnplast has also grown in prominence. A new entry is the World Wildlife Fund-promoted #fightforthereef hashtag, which emerged on 21 January and has persisted until today.

Notable in their absence, finally, are any hashtags associated with the widely criticised decision by the Prime Minister to recommend a knighthood for Prince Philip, such as #sirprincephilip or #knightmare, although on both 26 and 27 January these hashtags certainly emerged to substantial prominence in overall Twitter discussion in Australia. While (as the prominence of the #medicare and #abbott hashtags clearly indicates) federal political topics are clearly relevant to the Queensland election debate on Twitter, it looks like this particular issue at least has not been connected with the state campaign to date, then – due possibly also to Campbell Newman’s relatively swift negative response to the idea.

Somewhat earlier than expected, Queensland finds itself in the throes of a very tightly contested state election campaign, which was called on 6 January and concludes with election day on 31 January. Just like candidates and the media, we in the QUT Social Media Research Group too have been scrambling to put in place the infrastructure to track and analyse the social media elements of the campaign in close detail, even more so because we’ve hoped to roll out several new technologies for this campaign, compared to our coverage of the 2012 state election. Some of these will now have to wait until the New South Wales state election instead.

One innovation we have already launched is our Queensland Election Social Index (QESI), which takes in Twitter and Instagram data and generates a live overview of the current balance of attention and interaction around the various parties contesting the election; you can see QESI in action over on the QUT Social Media Research Group site.

Here, however, I’m focussing on a closer analysis of the Twitter-specific activity on the election trail to date. The dataset for this analysis includes any tweets which contain the key hashtags #qldvotes and #qldpol, their variations, and related keywords, as well as mentions of the Premier and Opposition Leader and of any of the parties by name; further, we’ve identified the Twitter accounts of over 150 candidates and are capturing any public tweets directed at them, as well as any of their own tweets that include any election-related hashtags and keywords. In this post I’m covering the timeframe of 8-18 January.

What’s immediately obvious is that there is a great deal more discussion on Twitter about the incumbent LNP government than about the ALP opposition; over the course of this timeframe, we’ve picked up almost 82,000 tweets referring in some way to the LNP in the text of the tweet, compared to fewer than 29,000 tweets for the ALP. To date, Twitter discussion of the ALP has surpassed that of the LNP in numbers only for a handful of hours on 16 January, during Labor’s release of its economic policy.

Beyond this, there is only a smattering of discussion about any of the other parties, with the Palmer United Party (PUP) rising briefly especially during its campaign ‘launch’ on 18 January, as well as on 14 January when federal leader Clive Palmer and state leader John Bjelke-Petersen simultaneously retweeted @PalmerUtdParty’s announcements of all state candidate names. (I suspect from this that Palmer’s and Bjelke-Petersen’s accounts are operated by the same media team.)

Here is an overview of activity patterns to date (click to enlarge):

This focus of overall discussion on the LNP, in the first of the graphs above, is perhaps unsurprising given the speculations about a strong swing against the LNP, and the possibility of Premier Newman losing his own seat – under the circumstances, much more of the discussion in the media, too, has been about why the LNP’s massive majority from 2012 has been pared back in under three years, rather than about the electoral agenda of the ALP. Additionally, however, the difficulties in spelling Labor Opposition Leader Annastacia Palaszczuk’s name might also have meant that our tracker did not capture all of the tweets referring to her, and that our figures for in-text mentions of Labor may be slightly too low.

But the second of the graphs above, which focusses only on @mentions and retweets of candidate accounts, tells a markedly different story. Here, the advantage is with the ALP, whose candidates were @mentioned almost 25,000 times over the same timeframe, compared to only 18,000 @mentions for LNP candidates and a far lower level of attention to all the other parties. The LNP records a major spike only on the day of its campaign ‘launch’ (18 January), with the ALP’s largest spike again coming as it released its economic policy. (As I write this on 20 January, the ALP ‘launch’ is taking place, and we’ll explore its resonance on Twitter in a future update.) We are currently tracking 57 ALP and 60 LNP candidate accounts, incidentally – so the differences in activity around the candidates are not driven by cohort size.

Such @mentions are not necessarily in themselves indications of endorsement, of course; indeed, many of the @mentions of LNP candidates during the early days of the campaign are critical tweets directed at LNP MP Verity Barton, following revelations of her traffic infringements. Similarly, both Newman and Palaszczuk come in for a substantial amount of criticism (and vitriol) alongside more reasoned debate and support.

But our analysis also reveals some very substantial differences across the parties in the types of responses to candidates’ accounts: as the second of the graphs below clearly indicates, more than 42% of the @mentions received by ALP candidates turn out to be retweets, while only just over 7% of the tweets directed at LNP candidates are retweets. This points to a very different pattern of engagement with candidates’ accounts across the two parties – and as retweeting frequently indicates a certain degree of endorsement, it appears that a much greater number of Twitter users are prepared to publicly endorse ALP candidates’ than LNP candidates’ tweets at this stage of the election campaign.

As the first of the graphs above shows, LNP candidates have also been sending substantially fewer election-relevant tweets than ALP candidates; this would explain to some extent the difference in the total numbers of retweets each party’s candidates have received, but not the vast divergence in the ratio of @mentions and retweets we have seen here. It is important again to stress that the number of tweets sent which we’re presenting here is not a total number of all tweets originating from candidates’ accounts, but counts only those tweets which also contain election-related content (hashtags, keywords, @mentions of other candidates’ accounts, etc.). It is possible, therefore, that the LNP numbers underestimate candidates’ full activities if those candidates fail to include hashtags like #qldvotes or #qldpol in their campaign tweets – but if so, those non-hashtagged tweets will also be more difficult to find for regular Twitter users, not just for our tracking infrastructure.

On the other hand, if ALP candidates are indeed substantially more active on Twitter than their LNP counterparts, that also points to a clear divergence in the campaign strategies. The early election date and short campaign timeframe proposed by the Queensland Premier already indicate a desire to run a relatively streamlined campaign, probably operated comparatively tightly by head office rather than encouraging individual candidates (most of them sitting MPs) to run their own, individual local campaigns. By contrast, the ALP’s lack of representation in the current parliament necessarily means a much stronger focus on local electoral races, as its candidates seek to win (or win back) their seats, and this must necessarily mean a much greater engagement by local candidates.

I would not be surprised if the ALP had looked closely at independent Cathy McGowan’s campaign in Indi during the last federal election, which combined dedicated on-the-ground effort with strong social media efforts to unseat a sitting MP against the overall national swing – and if the high level of activity from ALP candidates we’re seeing in our data was an expression of such a local-centric campaigning style.

Finally for this first analysis, an update on the major hashtags we’ve seen trending in our dataset to date – these are hashtags which have been used alongside generic hashtags such as #qldvotes or in tweets about and at the candidates. The graph below shows the ten most prominent of these hashtags, which offer a reasonable picture of themes in the campaign so far (multiple hashtags in the same tweet mean that the totals add up to above 100% at times).

Early on, the #imwithstupid hashtag dominated, reflecting an incident in which an anti-LNP protester and Twitter parody account operator wearing said slogan was charged with public nuisance for standing in the vicinity of an LNP stall, causing some degree of controversy). Federal politics made its way into the campaign with a flare-up of debate under the #medicare hashtag over the federal government’s proposed, then quickly withdrawn $20 reduction of Medicare rebates, while the absence of Prime Minister Tony Abbott from the LNP campaign is likely to be a driver of the #abbott hashtag which has been a persistent undercurrent of the Twitter debate. A similarly persistent debate continues over coal-seam gas policy, under the #CSG hashtag.

The prominent hashtags also bear out the overall focus of general discussion on the LNP rather than ALP: hashtags such as #lnp  and #lnpfail are prominent throughout the Twitter stream so far, while campaign slogan #strongchoices makes an appearance only on the Sunday of the LNP campaign ‘launch’ (and then as often as not with critical rather than supportive connotations). The significant presence of #ashgrove also points to persistent discussion of Campbell Newman’s chances of retaining his own seat.

Finally, the significant visibility of #nswpol in the Queensland election debate points to the fact that not all tweets we are seeing here are necessarily coming from Queensland-based political observers; the implications of the Queensland election outcomes for the coming New South Wales poll as well as for federal politics are already being hotly debated. And of the news organisations, Nine News has been most successful in attaching its hashtag to the Twitter coverage, with its journalists religiously attaching the #9news hashtag to their updates, and many users amplifying its visibility through retweeting.