What’s Keeping Britain’s Pollsters Awake Ahead of the Election?
What’s Keeping Britain’s Pollsters Awake Ahead of the Election?
Voter Volatility
Almost half of voters supported different parties in the elections between 2010 and 2017, according to research published by academics at the British Election Study. At 49%, it’s the highest level of switching ever, driven by political shocks including the credit crunch and the Brexit referendum.
Put simply, people’s political allegiances are weaker, and their behavior harder to predict. “This is the most volatile and uncertain election I have ever worked on,” said Andrew Hawkins, chairman of ComRes Ltd.
Pollster Volatility
The difficulty is reflected in the results. While some polls have the Conservatives on 40%, they have sunk as low as 32% in others. Labour have ranged from 22% to 29%.
Turning Votes Into Seats
For Ben Page, chief executive of Ipsos MORI, the “overarching” worry is translating the national vote share that polls produce into the thing that matters: seats in the House of Commons. “It becomes more difficult as we now have four sizable parties,” he said.
A party could win “a handsome majority” with a vote share below 35%, according to ComRes’s Hawkins.
Individual seats often behave in counterintuitive ways. Take Wimbledon, in South West London. In 2017, the Tories held it with 47% of the vote, Labour were second with 36% and the Liberal Democrats third with 15%. But the Lib Dems view it as a target seat this time, aiming to leapfrog Labour. “By looking at a national poll, you could never call it,” said Anthony Wells, of YouGov Plc.
The Brexit Factor
Why do the Lib Dems fancy their chances in Wimbledon? Brexit. Barely a third of voters in the district backed leaving the European Union, and, as the party with the strongest anti-Brexit position, the Liberal Democrats think they can win in such places.
Meanwhile, the Conservatives, who have held the district for all but 13 of the last 100 years, are worried about it. Under Boris Johnson’s leadership, they are a pro-Brexit party, something that could hurt them in some of the wealthy suburban seats where they’ve traditionally been strong.
They’re aiming to make up that difference in places they’ve never won before. “The question is how that plays out,” said Page. “There is no guarantee it will happen.”
The Campaign
One of the problems for pollsters in 2017 was that the campaign itself saw voters change their minds. In particular, people developed a more favorable view of Labour leader Jeremy Corbyn once they’d seen him on the stump.
The polls picked up a shift toward Labour, but, for reasons we’ll come to, still underestimated the party’s success. But this time, with both Corbyn and Johnson already well-known, “the campaign is unlikely to have as much effect as last time,” according to Joe Twyman of Deltapoll.
New Methods
One polling success of 2017 was an experiment by YouGov called Multilevel Regression and Post-stratification, or MRP. It used a much larger sample, 50,000 voters over the course of a week compared to the usual 1,000 over a couple of days, to produce a seat-by-seat estimate.
It wasn’t perfect, underestimating Tory support in Scotland for example, but it did predict that Theresa May would lose her majority, something that seemed unthinkable at the start of the campaign.
Expect to see more pollsters experimenting. But be careful. “It’s not a panacea,” said YouGov’s Wells. “You still have to have good data, and you still have to make the right decisions in modeling.”
Old Problems
Voters have become harder to contact. Some groups, such as younger people, no longer have fixed-line telephones, while internet-based panels miss voters who aren’t online. And, even when they can be contacted, people aren’t always good predictors of their own behavior.
“The big challenge, as always, is when we speak to 1,000 people, how many of these will actually vote?” said Page. In 2015, pollsters overestimated Labour’s vote share and, to fix the problem, they used turnout models in 2017 -- working on the basis that not everyone who said they would vote Labour was actually going to do so. The result was an underestimate of the party’s vote.
According to Wells, the industry has responded by dumping the tweaks it adopted. “If I have a worry for the industry, it’s that we go back to the problems of 2015,” he said. “At the last election, the polls overestimated the Conservative performance, but that’s very unusual. Historically, polls have overstated Labour.”
Still, Damian Lyons Lowe, of Survation, is looking forward to the race. “It if wasn’t challenging, it wouldn’t be interesting or useful,” he said.
To contact the reporter on this story: Robert Hutton in London at rhutton1@bloomberg.net
To contact the editors responsible for this story: Tim Ross at tross54@bloomberg.net, Thomas Penny, Robert Jameson
©2019 Bloomberg L.P.