Pollsters had a good election, our survey shows
Thursday’s election was not just a test for the politicians – it was also an examination for pollsters. The industry badly needed to get it right this time. In 2015, all the final polls suggested Conservatives and Labour were neck and neck. In the event, the Conservatives enjoyed a seven-point lead. In 2017, on average the industry posted an eight-point Conservative lead; in practice Theresa May’s party was only two and a half points ahead – though this time a few polls were reasonably close to the final outcome.
The polls got it wrong in 2015 in part because they overestimated the likelihood of younger voters – who mostly vote Labour – going to the polls. In 2017, they assumed many young voters would fail to vote – but overegged the pudding. Lessons were learnt. The assumption about young voters was dropped, while attempts were made to get more people with little interest in politics to answer polls. Trouble is, nobody could be sure before Thursday whether the lessons had been learnt well enough.
In the event, the industry put in a much better performance. As many as 10 companies put their reputation on the line by publishing polls in which interviewing did not stop until the Tuesday or Wednesday before polling.
On average, these suggested the Conservatives were nine points ahead, with Boris Johnson’s party on 43 per cent, Labour on 34, Liberal Democrats on 12 and both the Brexit Party and Greens on 3 per cent. These figures mean that collectively the polls underestimated the Conservatives (who won 45 per cent) by two points and overestimated Labour (on 33) by one. They were spot on in their estimates for the Liberal Democrats and the Greens, though they overestimated the Brexit Party (2 per cent) by a point.
However, many companies were either spot on or very close. In its final poll, Opinium anticipated every tally correctly apart from being a point out on the Brexit Party. Kantar also posted a 12-point lead, albeit with both the Conservatives and Labour underestimated by a point. Survation and Ipsos MORI said the Conservatives would enjoy an 11-point lead, while Deltapoll and Number Cruncher Politics reported a 10-point advantage.
Given the difficulties of undertaking polls – including the crucial role luck plays in securing a sample of voters that proves to be representative of the country – such accuracy is remarkable.
Though it seemingly did surprise many, the scale of Johnson’s victory should not have surprised anyone. The polls consistently reported that on average the Conservatives were 10 or more points ahead. Should it emerge from the ballot boxes, such a lead was bound to translate into a very healthy parliamentary majority.
Perhaps one reason for the surprise was frequent commentary in the latter half of the campaign that polls were narrowing, putting a Conservative majority at risk. This tended to ignore that the lead had increased in the first half – and that there was no guarantee any narrowing would continue through to polling day. That said, there will be lessons for pollsters to take away from this election. If their “errors” were purely the result of luck and chance then one or two companies at least should have overestimated the Conservatives. None, however, did.
The tendency to underestimate the Conservative performance slightly and
‘The scale of Johnson’s victory should not have surprised anyone. The polls consistently reported the Tories were 10 or more points ahead – a very healthy majority’
overestimate Labour’s is on a par with the industry’s record in many recent elections, 2017 excepted.
This suggests there is still further work to do. However it should be remembered that polls provide more than headline figures. They also provide intelligence on how voters make up their minds. One clear, consistent message was that the Conservatives were gaining ground among those who voted Leave in 2016, while support among Remain voters was lower than in 2017. Polls regularly demonstrated too that Labour was struggling among Leave voters.
Leave voters are disproportionately in working-class jobs. So such a pattern of voting meant the “red wall” of traditional Leave-voting Labour seats was clearly at risk of crashing down – as it duly did.
Sir John Curtice is Professor of Politics, Strathclyde University and senior research fellow, NatCen Social Research and The UK in a Changing Europe. He is also president of the British Polling Council, but is writing here in a personal capacity