How did the pre-election polls get it so wrong? By misjudging the turnout
We all know by now that the polls got the election result wrong last month. After months of expecting a very tight result the final published pre-election polls bore this out: they ranged from 31 per cent to 34 per cent for Labour and 31 per cent to 36 per cent for the Conservatives. In the event Labour received just over 30 per cent of the vote and the Conservatives 37 per cent — worse for Labour and better for the Conservatives than any of the polls. Of the 1,000-plus polls published since the “omnishambles” Budget of 2012 only one put Conservative support higher than their actual result.
After the debacle of polling during the 1992 election campaign, the pollsters did much to adjust their methodologies to take account of “shy Tories” and over-reporting for Labour. With the exception of YouGov, most pollsters continued to understate the Tories and overstate Labour but by 2010 they were congratulating themselves on how close they collectively got to the final result. Yet things clearly went wrong this year.
What has not been widely noted is that most of the polls in 2015 actually overestimated the number of votes that the Tories would obtain. Ipsos Mori’s final poll predicted the Tories would receive roughly 12.5 million votes — 1.2 million more votes than their actual result of just over 11.3 million. Its figures for Labour were even more askew, predicting 12.2 million votes against the 9.35 million obtained, an overestimate of 2.85 million. The poll supposed a turnout of 82 per cent, something last seen in the UK in a general election in 1951. The pollsters themselves thought this was rather high and stated in their analysis that they were actually expecting a turnout of between 72 and 74 per cent. In the event, the turnout was 66 per cent, an increase of 1 percentage point from 2010 and much more in line with what one might have expected.
If the turnout expectations of the pollsters had been more widely publicised during the campaign their surveys would in all likelihood have been treated with considerably more scepticism than they were. It also gives credence to the idea that much of the error in the polls was due to “lazy Labour” voters — those who when pressed stated they would vote Labour but in the event did not bother to turn out.
The extraordinary results in Scotland can also partly be explained by differential turnout. Fewer people voted for the SNP (1.45 million) than had voted last year for Scottish independence (1.6 million). But the turnout on the pro-Union side slumped much more decisively. The combined vote of the unionist parties was 1.4 million while the vote against independence was 2 million.
The other explanations for the errors in the polling — “shy Tories” and a late swing — do in all likelihood also play their part. In hardly any UK elections have the polls over-predicted Conservative support. The one notable exception was Boris Johnson’s second election victory as Mayor of London in 2012 where all the polls expected a higher vote share for him, and even then it did not affect the result.
A general election eve telephone poll by Survation indicated a clear late move towards the Conservatives. It was conducted on the afternoon and evening of May 6, the day before the election, got the Conservative result exactly right and was within 1 per cent of the Labour figure. Survation made the mistake of not publishing this poll: it thought it was too far out of kilter with the others. The private pollsters of the Labour and Conservative parties have now stated that they were much more accurate in predicting the final result than the public polls were. But as Stephan Shakespeare, founder and chief executive of YouGov, cynically noted, “It’s amazing how the secret polls were all bang-on right.”
The results mean that Labour will have a real struggle to win the next election. Its vote went up from 2010 by 1.5 percentage points and it gained 740,000 more votes — but it gained them in the wrong places. There was a swing to Labour in the seats it already held — and the safer the seat, the bigger the swing. In seats Labour hoped to gain there was actually a swing to the Conservatives.
The electoral system has worked to the advantage of the Labour party for many years: it has needed to gain fewer votes to obtain the same number of seats as the Conservatives. This has been reversed. Anthony Wells of UK Polling Report estimates that if Labour and the Tories obtained the same number of votes, 34.5 per cent, the Conservatives would have almost 50 more seats. The Tories could obtain an overall majority with a lead of about 6 points; Labour would need a lead of nearly 13. This is before the proposed boundary changes, leading to more equal-sized constituencies, are implemented, which would further skew things in the Tories’ favour.
If Labour cannot overturn the SNP landslide, matters seem bleak for them. But talk of this being the end of Labour are surely overblown. After all, in 1992 John Major won the election with 14 million votes — the highest number of votes any party has ever won in the UK — and we all know what happened five years later.