Polling errors
Back in 2012 when Mike Smithson first asked me to edit PB one thing he told me pay close attention to was the thoughts of his friend Lord Hayward, the man who has an excellent record in spotting polling failures, going back to the shy Tories of 1992.
Today’s Guardian has the following report
Some of the polls before the general election may be overstating Labour’s huge lead over the Conservatives, a Tory election expert has said.
According to Robert Hayward, a peer and former MP who first identified the phenomenon of “shy Tories” before the 1992 election, his analysis of local election results suggests that the Conservatives are getting more support from voters who say they are undecided than is showing up in the polls.
“About 33 years on, I am yet again convinced that a statistical bias exists in the polls,” he said.
Polling in the run-up to the election has shown commanding leads for Labour of more than 20 points in some surveys, such as YouGov, while others, such as Opinium, show a gap of about 14.
Some constituency-level polling by YouGov suggests the Conservatives are heading for near wipeout. Its April projection points to more than 400 seats out of 650 going to Labour.
Most experts, MPs and political strategists across the parties agree that Labour is almost certainly heading for victory, but the size of its lead will determine its majority in parliament.
Hayward said it was “not shy Tories” all over again, and that Labour still undoubtedly had a large lead over the Conservatives, but he was concerned the polls were “getting it wrong”.
“I think the biggest problem for me is how you allocate the Conservative 2019 voters who are now ‘don’t knows’. I do believe the ‘don’t knows’ will vote in fairly large numbers, and the indications from local council results is that they are beginning to break towards the Conservatives.
“Because of the lack of commitment displayed so far, I’m not sure where things are going to end up, because events during this election could have more of an effect than in many elections because there are so many undecided.”
One polling industry expert also agreed there was a risk the polls could contain methodological or structural errors. They said: “I think there are lots of reasons to be concerned about a polling error this time around. There are serious data quality issues across the industry, where it’s getting tougher to ensure the people taking part in surveys are real, taking the surveys seriously, and representative of the wider population. Pollsters are working harder to combat it, but that task is getting tougher, and there is a risk that it affects the accuracy of the polls this election.
Some of the polls before the general election may be overstating Labour’s huge lead over the Conservatives, a Tory election expert has said.
According to Robert Hayward, a peer and former MP who first identified the phenomenon of “shy Tories” before the 1992 election, his analysis of local election results suggests that the Conservatives are getting more support from voters who say they are undecided than is showing up in the polls.
“About 33 years on, I am yet again convinced that a statistical bias exists in the polls,” he said.
Polling in the run-up to the election has shown commanding leads for Labour of more than 20 points in some surveys, such as YouGov, while others, such as Opinium, show a gap of about 14.
Some constituency-level polling by YouGov suggests the Conservatives are heading for near wipeout. Its April projection points to more than 400 seats out of 650 going to Labour.
Most experts, MPs and political strategists across the parties agree that Labour is almost certainly heading for victory, but the size of its lead will determine its majority in parliament.
Hayward said it was “not shy Tories” all over again, and that Labour still undoubtedly had a large lead over the Conservatives, but he was concerned the polls were “getting it wrong”.
“I think the biggest problem for me is how you allocate the Conservative 2019 voters who are now ‘don’t knows’. I do believe the ‘don’t knows’ will vote in fairly large numbers, and the indications from local council results is that they are beginning to break towards the Conservatives.
“Because of the lack of commitment displayed so far, I’m not sure where things are going to end up, because events during this election could have more of an effect than in many elections because there are so many undecided.”
One polling industry expert also agreed there was a risk the polls could contain methodological or structural errors. They said: “I think there are lots of reasons to be concerned about a polling error this time around. There are serious data quality issues across the industry, where it’s getting tougher to ensure the people taking part in surveys are real, taking the surveys seriously, and representative of the wider population. Pollsters are working harder to combat it, but that task is getting tougher, and there is a risk that it affects the accuracy of the polls this election.
“The other issue is undecided voters. There’s three times more people who voted Conservative and are undecided than voted Labour and are undecided, and where these voters end up going will make a big difference. The pollsters make different assumptions about this group … The range of this is probably like Labour 1997 to Canada 1993 – it’s a Conservative party historic defeat with one methodology, to complete wipeout with another methodology.”
For gamblers, particularly spread bettors, it is a very concerning that the Tory seat range could be between 2 and 165. If Lord Hayward thinks we could have a potential polling failure on this level, it confirms my hunch we could see a 1997 style result not a Canada 1993 or 1931 in reverse results, I am betting accordingly.
TSE