Finding new ways to track voters’ moods, beyond polls and punters

Coal seam gas has proved a hot topic ahead of the NSW election, not only on the campaign trail but also on social media, where a ‘social mood reader’ can reveal the depth of community feeling. NewZulu/Tony Markham/AAP

The recent Queensland election result surprised everyone – including the professional pollsters and punters. Sportsbet declared the result and paid out for a win to the Liberal National government one day before the January 31 poll.

We were so confident yesterday that we decided to pay out early on the Liberals as it looks a foregone conclusion. The punters obviously agree with us as they have moved to $1.01.

In that case, the punters got it wrong. Yet that hasn’t stopped the betting agency doing the same thing again ahead of the March 28 polling day in NSW.

The most recent Fairfax/Ipsos poll shows the Baird government on track to victory, with a 54% to 46% two-party preferred lead over Labor. But as many people paying close attention to the polls have warned – including NSW Premier Mike Baird and ABC election analyst Antony Green – the election could be tighter than the polls show.

So beyond traditional polls and betting markets, how else could we try to gauge how people feel ahead of future elections? Social media is a goldmine of real-time information on public sentiment – and there are new ways to tap into how people really feel, including with a “social mood reader”.

Polling is getting harder to do well

Relying too heavily on traditional political polling or even the usually reliable betting market is a risky strategy for political parties and pundits.

Audience measurement for ratings and polling began in the 1930s. The two men behind those measures – Archibald Crossley and George Gallup – were close colleagues.

Audience ratings and political polling were created to provide accurate samples and also to stop “hypoing”, an industry term that means distortion by vested interests wanting to control what public opinion looked like.

Ratings and polling rely on robust samples, which properly represent the defined populations from which they are drawn. Modern political and audience ratings pollsters have always tried to draw an accurate statistical sample for their surveys.

But experts such as the former head of the American Association for Public Opinion Research, Peter Miller, have found that maintaining the quality of traditional and web polls around the world is getting harder by the day.

Some of the reasons include the shift away from landlines to mobile phones; the challenge of web identity [transparency]((http://transparency.aapor.org/index.php/transparency); the unwillingness of many people, especially young people, to participate in surveys; and, not least, increased privacy concerns.

Tapping into the public mood on social media

What people post and share on social media – such as on Twitter, Instagram and Facebook – can show political moods in real time and it provides qualitatively rich data.

In surveys and focus groups, people are asked questions about their views or behaviour. In social media networks, people’s views are on display as they are expressed. Those views, in turn, are often re-posted elsewhere and opinions are built up on particular issues.

As QUT’s Social Media Research team has covered in more detail in The Conversation, among the most popular topics on Twitter in the #nswvotes campaign have been #nswnotforsale and #csg (coal seam gas).

Similar themes dominated the Queensland election, particularly privatisation, which was often discussed with the hashtags #assetssales or #Not4Sale. Those topics’ popularity were driven partly by widespread public concern but also by a well-organised Queensland union-led Not4Sale campaign, working closely with the Labor Party. (Interestingly, if you try to visit follow the Queensland not4sale.org.au link, it now automatically redirects you to stoptheselloff.org.au – the NSW unions’ anti-privatisation website.)

What people said and shared about “asset sales” in the Queensland election clearly indicated their voting intentions.

How a ‘social mood reader’ works

Depending on whether you only follow conversations you agree with or actively seek out different views, the risk with social media is that it can sound like an echo chamber.

So how can a pollster, or anyone, be confident that a negative or a positive “mood” in social media networks will translate into voting behaviour? New methods are emerging that try to capture social mood from what people say in their social media networks.

Dr Brett Adams’s work at Curtin University is one example, with his development of a “social mood reader”, originally developed for autistic groups, to identify quickly the mood of different networks.

Here’s how it works. After getting permission from an individual, the social media reader can access not only what they say and share publicly, but also their private information: who they follow and like (such as if they follow Mike Baird on Facebook, or Stop the Sell Off on Twitter), where they get their information from (for instance, liking The Conversation or ABC News on Facebook), who they are friends with, and what they are seeing in their feeds. It also goes beyond just social media, including other online services the person uses, such as email.

There are some key differences between traditional polls or surveys (which take people time to participate in and which are often done on behalf of political parties or commercial interests) and the social mood reader, including:

  • it takes a participant no time to be involved with the social mood reader, as they are simply giving access to what they are already saying and doing on social media
  • people like the chance to have their say on matters they care about, especially when they know it is for independent research, rather than for commercial purposes.

An individual participant who agrees to share their information with the social mood reader gets to nominate some of the key circles they belong to: for instance, family, friends, work, community groups, political groups etc. You can see an example of how that looks below.

When there is crossover between those various groups, you then see a line or multiple lines criss-crossing between those groups.

What the social mood reader platform currently looks like, with researcher Mark Balnaves at the centre and his various social connections all around him, on an issue where the community was largely fairly happy (or green). Mark Balnaves

Red for angry, green for happy

I used Dr Adams’s social mood reader in a 2011 trial with City of Geraldton-Greenough in Western Australia, to see how the local community felt about a number of issues: from the rise of fly-in, fly-out mining and what that meant for the community, right down to more local issues such as how people felt about bike safety and the future of a playground roundabout.

Different moods are represented by different colours, based on the Affective Norms for English Words (ANEW). For example, a very deep red, for example, represents “enraged”, while deep green would indicate a mood of great “happiness”.

The colours attributed to various moods in the social mood reader. Brett Adams

Nearly 5,000 people participated in the Geraldton-Greenough council study. What the social mood reader revealed was that the single most controversial issue proved to be fly-in, fly-out mining, with the overall community mood being red, or “angry”, at what it meant for their community.

The Geraldton locals were also angry when asked about the prospect of losing their playground roundabout; when the council saw that result, they kept it.

We also used Brian Sullivan’s CivicEvolution platform in the study, which is an easy way for people to pass on their ideas. So as well as testing for how people felt about issues, we gave participants in our study the chance to propose their ideas directly to the council.

On bike safety, for instance, the council not only found out it was a widespread community concern via the social mood reader, but also got some constructive suggestions about what could be done to improve safety.

Giving people easier ways to have their say

As far as I know, Dr Adams’ social mood reader has not been used in any Australian elections yet. But how could it help politicians if it were used in an election?

A good MP should always have a general sense of how her or his electorate feels. But rather than relying just on polls, focus groups, door knocking and gut instinct, a social mood reader could also be used to reveal how a community really feels.

The Geraldton-Greenough council was able to see how red and angry locals were about the prospect of losing their playground roundabout before it happened – and that changed the council’s mind, keeping the community happier.

We know that Australian politics is already highly poll-driven, with everyone from the prime minister down closely watching opinion polls and focus groups. So do we really want one more method of telling politicians their reforms are unpopular, when sometimes those policies might be the right thing to do?

There is no easy answer to that. But the upside of technology like the social mood reader is that it does give the community an immediate opportunity to have their say, participate in the democratic process and feel as if they are really being heard.


Read more of The Conversation’s coverage of the 2015 NSW election.