🌅 Subscribe to the Waugh Zone and we’ll send you the key takeaways from election night bright and early on Friday morning.
The exit poll is out and it’s predicted a monumental loss for Jeremy Corbyn’s Labour Party.
If it’s correct, the Conservatives will stomp home to victory tonight with a majority of 86.
In recent years, the exit poll has been a very good predictor of the outcome but given some infamous poll failings lately (think Brexit and Trump), why is this one any different?
The exit poll is different from other opinion polls because instead of asking people how they intend to vote, it asks people how they voted.
How accurate is it?
Pretty bloody accurate.
David Firth, Professor of Statistics at the University of Warwick, helped develop the model that has been in use since 2001. He told HuffPost UK that he and his colleagues “put some statistical thinking into it” – which had a profound effect.
After some notable failures in the 90s, the current system of calculating the exit poll has a very impressive track record of forecasting the result – including being bang on in 2005.
In 2017, the exit poll predicted the Tories would end up with 314 seats, just four short of the 318 the party actually won.
Labour was forecast to win 266 (they finished on 262), the Lib Dems 14 (12), the SNP 34 (35), Plaid Cymru 3 (4), Ukip 0 (0), the Greens 1 (1) and others 18 (18).
How is it so accurate?
The crucial difference is simply when it is conducted.
“The biggest distinguishing feature of the exit poll is you know everyone who took part in it has voted because you’ve seen them come out of the polling station,” says Firth.
The fieldwork takes place throughout election day in a selection of constituencies around the country.
In 2017, 144 seats were chosen – 141 of which had also been used at the 2015 election, PA Media reports, and a similar number is expected to be used for this election.
The seats are dotted across Great Britain and are places where the result is expected to be close.
In these seats, interviewers will approach voters as they leave polling stations and ask them one question: who have they voted for?
“They are asked to complete a duplicate ballot to the one they’ve just completed in the polling station and they fold it and put it in a box,” says Firth
These ballot boxes are then passed to the academic researchers so they can begin analysing the contents.
What’s the point though?
If you’ve ever wondered why someone goes to all the trouble of trying to predict the election result, late at night and just hours before the actual result is called, you’re in good company.
“Back in the 1970s my friend Phil Brown at Imperial College used to do the exit polling with the Royal Statistical Society. One evening they met the Queen and Prince Phillip and his question was exactly this.”
The answer is perhaps a little more mundane than you’d imagine – it’s essentially to fill a bit of time on TV.
“If you look at the current affairs audience, the general election is the biggest current affairs audience broadcasters get,” Prof Firth says.
“But between 10pm and midnight they’ve got no actual results to talk about so they need something to discuss in their primetime.
“The exit poll is payed for by the broadcasters and I think that’s why they pay for it.”
This brings us nicely to our next point.
Who pays for and conducts the poll?
As in previous years, the exit poll has been commissioned jointly by the BBC, ITN and Sky News, and its contents will be revealed live on television by each of the main broadcasters.
The first stage, the fieldwork, is carried out by a polling company – in the case of this election, Ipsos Mori.
The second stage, the analysis, is undertaken by a team of academic researchers.
They will begin by comparing the responses with the results from exit polls in previous elections, to allow them to estimate the change in party support.
They will then produce statistical models of how the change in party support is expected to vary according to the social and political characteristics of each constituency, before applying these models to all seats in Great Britain (no exit polling is carried out in Northern Ireland).
Having come up with predicted levels of support for this election, the analysts calculate the probability of who will win each seat, before finally producing their forecast of how many seats have been won by each party.
Is there anything can mess it up?
Of course, with any type of prediction of future events there is a margin of error and a degree of luck involved.
Firth says: “When you take a random sample, you might just get a really unlucky sample who have all voted the same way for example.”