View

Close Banner

Polls apart: are the political soothsayers about to get it wrong again?

Saturday 2 April 2016

The Independent

 

News /
Slug: 
Politics/ Inside Westminster

Polls apart: are the political soothsayers about to get it wrong again?

Subheadline: 

A year since predictions of a hung parliament were proved to be way off, all eyes are on the EU referendum

Where were you when the exit poll came out? Eleven months on, the dramatic moment on general election night when the Conservatives were given a clear lead over Labour, is still talked about by politicians, their advisers and journalists, most of whom expected a hung parliament. Myself included.

It was also a heart-stopping moment for the opinion pollsters, who had put the two main parties neck and neck.

For them, election night got even worse, as the Tories’ exit poll lead grew into an overall Commons majority. It was the pollsters’ nightmare scenario. They don’t make much money or deploy many staff on political polling but it is their shop window for wooing the commercial clients from whom they make their millions. So they had every interest in getting it right last May.

The pollsters’ consistent message during a long election campaign shaped the election and undoubtedly helped David Cameron to his unexpected victory. The widespread expectation of a hung parliament allowed the Tories, desperately seeking a focus for their attacks on Labour, to raise the spectre of a minority Labour government in the pockets of the Scottish National Party. It hugely damaged Labour. It also hurt the Liberal Democrats, who decided not to rule out joining a Lab-SNP deal to “lock out” the Tories.

I declare an interest: as a journalist, I am addicted to polls because they generate interesting stories.

I have been setting questions and writing up the results since 1988, working with six different pollsters. But I think their performance last year raises serious questions for both the polling and media industries.

An entire election was fought on what turned out to be a false premise. The Labour-SNP factor dominated the campaign, crowding out other issues such as how the Tories would find their £12bn of welfare cuts. If the polls had been accurate, and shown the Tories seven points ahead, the result might have been different.

Most pollsters admitted something went badly wrong and were not in denial, as they were the last time they messed up in 1992. An independent inquiry set up by the British Polling Council, which represents the companies, reported on Thursday, saying the election saw “some of the most inaccurate” election polling since it began in 1945.

It judged the main cause was unrepresentative samples, with too many Labour and not enough Tory supporters. Pollsters admit they had too many politically engaged young people (more inclined to Labour) and not enough over 70s (more likely to vote Tory).

Although there may have been a very modest “late swing” to the Tories after the final surveys, the inquiry found this would have added at most one percentage point to the Tory lead. It detected some evidence of “herding” as pollsters adjusted their final surveys in a way that made them converge, but believes this was unconscious rather than deliberate collusion.

The analysis seems sound, but the remedies less convincing. The inquiry recommended some “tweaks and modifications” to existing methods but admitted it would be up to individual firms to decide how to take them forward.

After the election, plenty of politicians and journalists vowed never to trust the polls again. But we seem to have short memories. There is no shortage of surveys about the EU referendum, and they get a lot of media coverage.

We can’t expect the “tweaks” to iron out last year’s mistakes. Referendums are different to general elections —a one-off, irreversible choice that cannot be changed next time. Yet there is a lot riding on the referendum for the polling industry. Their commercial clients will be watching closely, so they cannot afford another disaster.

They now have a new headache. At the election, there was no difference between the accuracy of telephone and online surveys. But there is when it comes to the referendum. Telephone polls show a lead of around 10 points for the Remain camp, while online ones show the result too close to call. They can’t both be right.

A possible explanation emerged this week in a report by Number Cruncher Politics and pollster Populus, which suggested that phone polls are likely to be more accurate. It argued that their samples are more representative of the population’s social attitudes (such as on national identity); that harder to reach people are likely to be more liberal than those who respond quickly online and that most internet polls allow people to reply “don’t know”, which is not an option on the ballot paper. Online pollsters naturally dispute this.

Despite what happened in the election, it would not surprise me if the polls shape the referendum campaign too. Of course, both sides will “spin” them to suit their own purposes as they will want to show momentum. The Leave camp points out that an average of all the surveys shows that a six-point lead for Remain in January is now down to two. The polls could change behaviour: if one side is 10 points ahead, people might be less inclined to vote. A close race could encourage a higher turnout. Both sides agree on one thing: turnout will be crucial.

Some politicians would ban publication of polls just before elections and referendums to stop them influencing the outcome, as they are restricted in France, Italy and Spain.

That would be a step too far. Surveys would still be conducted, and the public would then be less informed than the insiders. The findings would probably leak out or be published on websites abroad. Love them or hate them, polls are part of the democratic process. But the public, and the media, should learn to take them with a pinch of salt.