Election polls failed to predict the results in 2015 and 2017 – can they be trusted now? – The Sun

POLITICAL polls have been disastrously wrong in the past – but what do they mean for the 2019 election and should Brits trust them?

The only poll to predict Theresa May's hung parliament in 2017 has forecast that Boris Johnson will get 339 seats tomorrow, comfortably returning him to No10.

But they also said the margin of error means they can't rule out the possibility of a hung Parliament.

If Boris fails to make gains in key areas, Britain could end up with months more political uncertainty, with Brexit and the future of the country all up in the air.

Throughout the campaign the polls have fluctuated – but all of them have predicted a Tory government with various majorities. Are any of them right?

Parliament dissolved on November 6 and politicians went full-pelt into the election campaign.

Polls have consistently had Boris Johnson in the lead with anywhere between 17 and 6 points over Jeremy Corbyn.

Top pollster and Professor of Politics, Sir John Curtice, said a majority government this week is far from a sure thing.

"It's perfectly clear that the Prime Minister does not have it in the bag, the variation in the polls is too great," Sir Curtice told The Sun Online.

"The Tories are favourites. They are in pole position. They are expected to win," he said.

"But its not a certainty and there is sufficient uncertainty about the exact position in the polls."

POLLS AT TIPPING POINT

In a final election poll in 2017, Theresa May was cast as much as a 96 seat majority – yet she ended up with a hung parliament just a few hours later.

With a forecast much slimmer than that, a Tory win is far from guaranteed.

David Cameron's majority win was famously missed by pollsters in 2015.

Dozens of polls wrongly predicted a Remain win in the 2016 Referendum too.

So why are these different and can they be trusted?

Political researcher and polling expert with YouGov Chris Curtis defended the use of polls, saying methodology in the industry has had a huge overhaul to deal with their past mistakes.

WEIGHTING 

Polling results are usually weighted against the composition of the population to try and weed out any bias in a sample size, which, according to the British Polling Council, should generally be upwards of 1000 people.

Pollsters will weight their results by factors such as age, education, and past voting record.

For Ipsos Mori, another one of the polling giants, ethnicity has been added as a weight – but whether or not there is a child in the household has been scrapped since 2017.

Mr Curtis said YouGov also weight for political attention.

He said: "Those that don't pay attention to politics, that aren't watching Newsnight everyday, will vote differently so its an important one to do."

TURNOUT

How many people actually turn out to vote on election day is one of the hardest things to predict for pollsters, according to Mr Curtis.

"All we can do is look at the turnout patterns that we did before and hope we can capture voting intention this time around, but (there) is not a very good record of (predicting) turnout and any big changes in the dynamics of turnout can change the results," he told The Sun.

A final Ipsos Mori poll in 2017 had the two parties neck and neck.

When the poll it was adjusted for turnout, it gave the Conservatives an eight-point lead.

Voter turnout was one of the main concerns in having an election in the middle of winter – the first of its kind for decades.

Some polls, such as those done by ICM, used to be weighted by a pre-determined level of turnout, and no longer are this election season.

Others such as Savanta ComRes, have ditched their voter turnout models used in previous elections.

Much has been made of the MRP survey by YouGov, which was the only poll to pick Theresa May’s defeat in 2017, but the rest of the polls failed to predict the result because of the assumption young Labour voters would not turn up to vote.

Whether there is another 'youthquake' in this election remains to be seen, and could have a huge effect on polling day.

VOTING PATTERNS

Voting patterns have changed dramatically since the 2015 election, and tactical voting has become a commonly-used phrase among some.

Both of these factors make it difficult for pollsters.

According to one set of figures by BMG Research, 24 per cent of people plan to use their vote tactically.

Over the weekend, there were warnings as many as three million people would change their minds on who they would vote for with just days to go until polling day.

Assumptions made on the basis of a “uniform national swing”, where changes in national vote share are applied to each area make the results unpredictable.

In a tight general election, where the two parties are so close together, small errors can make a big difference.

METHODOLOGY

How respondents for a poll are chosen and how the poll is conducted (by phone or online) can also impact the overall result by influencing the spread of voters in a sample.

Voters should also be particularly wary of polls paid for by pressure groups, as the question asked can be skewed one way or the other.

Mr Curtis, however, said methodological faults which saw polls so wrong in 2015 have been adjusted.

In 2015, too many young Labour voters were counted in the polls, and not enough older voters, thought to be simply because many were done online, rather than by phone.

"In 2015 we were not surveying enough people who weren't university educated, and we weren't surveying enough elderly people," Mr Curtis said.

"We now make sure we weight for over 65s instead of over 60s, as age is increasingly the biggest dividing line on how people will vote," he said.

"We also weight by educational qualification, so we have enough people who did go to university and enough people who did not."

CAN THEY BE TRUSTED?

Mr Curtis stands by the polls, saying despite their flaws, they can predict the mood of the country more accurately than any other measure we've got.

"The type of polls we do at organisations like YouGov is the best indication we have of what the public currently think," he said.

"We constantly hear about vox pops and response on the doorstop, but fundamentally that is not a good measure of what the public are thinking at the general election."

So, take it with a pinch of salt, make sure you vote yourself, and wait for the results.

Recent Polls

Savanta ComRes (December 7): 8 point lead to the Conservatives (41%), Labour (33%)
YouGov (December 6): 10 point lead to the Conservatives (43%) to Labour (33%)
PanelBase (December 6): 9 point lead, Conservatives (43%) to Labour (34%)
Ipsos Mori (December 6): 12 pont lead to Conservatives (44%) to Labour (32%)
Opinium (December 8) 15 point lead to Conservatives (46%) to Labour (31%)
YouGov MRP survey (December 10): 9 point lead to Conservatives (43%) to Labour (34%)

Source: Read Full Article