THE BLOG

Can We Ever Trust The Polls Again?

01/08/2017 17:13
Ed Sykes / Reuters

Following a string of surprise political results confounding pollsters and the commentariat alike - including two General Elections, the small matter of Donald Trump and an EU referendum - you can't blame those questioning whether we can ever trust polls again.

In an answer that will surprise no one (as the Group CMO of a polling company), I firmly believe the answer is 'yes'. It will, however, require work on both sides.

Why does this matter?

The short answer is, democracy flourishes when pollsters get it right. When they get it wrong, it skews political reaction to what the public want.

Would GE2017 have happened if it hadn't been for the polls? Would David Cameron have gambled on Brexit if he'd had reliable numbers? From our impending EU exit to the DUP's crowning as Government kingmakers, it's not unfair to suggest that dodgy poll results have led to fundamental political changes.

Polls also change the way people vote. Take the tale of the stay-at-home-Tory who, expecting a landslide, decided not to venture out on a rainy Thursday. Meanwhile, the undecided Labour voter, fretting about their party's hopes under Jeremy Corbyn and simultaneously thoroughly uninspired by a lacklustre Conservative campaign, committed to following their vote for their Party - anticipating a leadership change on June 8 because, well, the polls.

The last General Election campaign showed more change in voting intention and the biggest swing from beginning to end than any in living memory. It also presented an interesting challenge to the industry: the role of 'models' versus 'polls'. This may sound like pollster geekery but the difference is key. Polls show voter share and reflect voting intentions of a representative sample, with weighting applied to certain factors (e.g. a voter's likelihood to actually vote.)

Models are different. Models predict seat share and call election results in a first-past-the-post world. They don't poll constituency-by-constituency, but rather survey a huge number of people across the UK and apply their findings seat by seat. This is done by looking at the demographics of each constituency and allying this with previous voting behaviour in both the EU referendum and General Election.

It sounds all lovely and technical but the fact remains: when it comes to a published poll or model, it is a human-being making the final call on how to interpret and weight raw data. Will young people turnout? Will the Labour vote be suppressed? Many questions must be answered in the most public of forums. The industry can devise new models that allow for corrections and assumptions all day long, but it is still a living, breathing person who has to make a decision on that data.

Blue-tick harbouring social media commentators will readily point to the data-heavy Twitters, Facebooks and Instagrams of the world for the future of political predictions. That will never work. The filter bubble and our socially-influenced online behaviours are simply too far removed from the realities of how we actually think, and the political complexities of a country.

For many outside our blog-reading bubble, the only political engagement is on voting day. Until then, many remain invisible to the Westminster commentariat and pollsters alike.

Going forward, pollsters will have to factor in the results from those who cannot be reached and the increasing unreliability of voters in our volatile world. The industry must explore how they collect data: how do we have a genuine, non-intrusive conversation with a representative sample of voters? Technology points at solutions, but perhaps the answer relies in more traditional (and more expensive) forms of contact. Ultimately, polling brands must embark on a trust-building exercise: this means public-facing campaigns to lift the hood on their data and transparency in their methods.

There is some solace. Polls will never be perfect but they're the best benchmark to measure public opinion. They're also the best form of accountability: pollsters live and die by their polls and it's in their interest to make them as accurate as possible. Companies like ICM spend a lot of time, money and resources to get them right. Our democracy benefits from such commitment.

Polls are politics' data-processing sausage factories - most people don't care how they work, they just want them to be right. They want to trust to them. The willingness is there, evidenced by an onslaught of surveys following the last election. This appetite, married with a commitment to transparency and investment from the pollsters, is good news. Indeed, anything that provides informed objectivity outside our social media filter bubble must be championed.

Comments

CONVERSATIONS