Could A Social Media Poll Ever Prove To Be Dangerous?

Posted on May 2nd 2012

Could A Social Media Poll Ever Prove To Be Dangerous?

With the race for the Presidency fast approaching, I have no doubt we’ll see even more agencies releasing ‘opinion polls’ derived from social media data.

With more people using social media, there is, as Tweetminster has shown in the UK at the nationwide level, a growing validity for looking at using this data in a predictive capacity.

However, I don’t believe social media analysis is in a position to replace more traditional political polling and, unless we quickly see a far more politically representative sample of the wider population using social networks (and also willing to share their political opinions), I’m unconvinced it will be anytime soon.

However, with an increased willingness amongst the media to use this data, I think it’s important to understand what effect this could be having on elections.

Building on German political scientist Elisabeth Noelle-Neumann’s “spiral of silence” theory, in 1984, Lang and Lang suggested that poll results can set the agenda by reinforcing majority opinion.

Four years earlier, Public Opinion Quarterly made the link to agenda-setting explicit by stating that “over time, people’s political beliefs and behavior have been affected by evidence of polls presented by the press—a special case of the larger claim of the mass media’s agenda-setting functions”.

If the reporting of social media polls can have an impact on voting (either through a bandwagon or ‘backing the underdog’ effect), I think we need to better educate the media as to how we’re arriving at these results and be transparent about some of the current limitations of using this data.

As we’ve seen in recent predictions in the GOP Primaries, fringe candidates, such as Ron Paul, can be vastly overrepresented in social media conversations and their predicted vote consequently becomes overinflated.

As individuals begin to game tools like Klout and exert influence by merely creating an illusion of influence, there’s a very real danger that extremist parties will see social media as an opportunity to manipulate these newer forms of polling and attract more media coverage.

In 2006, the misreporting of a JRRT report in the UK, which stated that a quarter of Londoners would consider voting for the right-wing British National Party (wrongly, and widely, reported as revealing that 25% of the UK population were considering voting for them), was followed by a YouGov poll placing the party on 7%.

Despite the fact Populus had the party at less than 1%, the YouGov poll resulted in another bout of media coverage for the British National Party reinforcing their poll boost.

The coverage in the press helped legitimise the false idea that the British National Party was an ordinary political party and they went on to increase their votes in local elections from just 3,000 in 2000 to 230,000 in 2006.

It’s not too much of stretch to envisage a situation in which a rushed, ‘real-time’ poll, based merely on a volume of mentions for a candidate or party within social media, could be manipulated by a small number of individuals (or even misrepresented by researchers not taking the time to understand why they’re being mentioned in the first place) and lead to an extremist political party being reported as leading in the polls.

If it’s for no reason other than this, I think it’s important that we all start being more open about, not only the limitations of social media polls but, in particular, the methodology behind how we’ve arrived at the results.

Using a social media monitoring tool to quickly grab a share of mentions without understanding what’s being said is not a poll and it’s not necessarily without consequences either.

Gareth Price

Gareth Price

Head of Research at The Social Partners 

See Full Profile >