Making sense of the polls

Making sense of the polls

Tories win UK election outright majority as polls completely fail

In a huge shock, the Tories have won an outright majority of the UK House of Commons seats. With all 650 seats officially declared, the Tories have won 331 seats, a gain of 22 on the 2010 results. Labour has 232, down 26, the Scottish National Party (SNP) has 56 seats, up 50, and the Liberal Democrats have been shattered, winning only eight seats, down 49. The Tories thus have over 326 seats, the requirement for an overall one party majority.

The vote shares were 36.9% for the Tories, up 0.8%, 30.4% for Labour, up 1.5%, 12.6% for the UK Independence Party (UKIP), up 9.6%, 7.9% for the Lib Dems, down 15.2%, 4.7% for the SNP, up 3.1% and 3.8% for the Greens, up 2.8%. Despite winning far more votes than the Northern Ireland parties, the Greens and UKIP won only one seat each. Vote shares excluding Northern Ireland, which the major UK parties do not contest, were 37.7% for the Tories, 31.1% for Labour, 12.9% for UKIP, 8.0% for the Lib Dems and 3.8% for the Greens - a Tory win over Labour of 6.6%.

Compared with pre-election polling, England was a catastrophe for Labour. Polls had at least a 4% swing from the Tories to Labour in England, which would have resulted in at least 30 Labour gains from the Tories. Instead, Labour’s English vote was up by just 3.6%, with the Tories up 1.4%, for a net swing of 1.1% to Labour. As expected, a “sophomore surge” for Tories elected in 2010 meant that there were only five Tory net losses to Labour in England. Both parties gained as the Lib Dems were massacred, and this helped the Tories win an overall majority.

In Scotland, the SNP won a massive landslide, winning 56 of Scotland’s 59 seats, to one seat each for the Tories, Labour and the Lib Dems. This represented a gain of 50 seats for the SNP, a 40 seat loss for Labour and a 10 seat loss for the Lib Dems, with the Tories unchanged. The vote shares were 50.0% for the SNP (up 30.0%!), 24.3% for Labour (down 17.7%), 14.9% for the Tories (down 1.8%) and 7.5% for the Lib Dems (down 11.3%).

There were many high profile casualties of this election. The Scottish Labour leader, Jim Murphy, was defeated by the SNP, and shadow Chancellor Ed Balls lost to the Tories. UKIP leader Nigel Farage failed to win his target seat from the Tories. While Lib Dem leader Nick Clegg held his seat, he has now resigned as leader following the Lib Dem annihilation. Labour leader Ed Miliband has also resigned as leader.

The election results showed a massive failure by the UK pollsters. You can see the final polls in this Wikipedia table, and only SurveyMonkey came close to the final Tory vs Labour margin of 6.6%; this pollster had not conducted any previous UK election polls. All other pollsters had final polls that were a tie +/- 1%. On Nate Silver’s US site, there is speculation that the polls were “herding” - that is, becoming close to each other in an artificial way that gave the blatantly wrong result. Note that the Survation phone poll that had the Tories winning by 6% was not released until the day after the election, as Survation “chickened out” of releasing it on election day - this is clear evidence of herding.

The UK polls must do a better job of identifying likely voters. Turnout for this election was only 66%, up 1% on 2010. In contrast, most pre-election polls had well over 70% “absolutely certain to vote”. Demographics that vote Conservative are more reliable voters than Labour voting demographics, so the weaker than expected turnout adversely affected Labour’s performance.

Only one serious forecaster predicted that the polls could be so wrong. Number Cruncher Politics used three models of past elections, and averaged them to get a Tory lead of 8.4% in the popular vote. At the time this looked very brave, but this analysis has been proved right.

Some of the pre-election forecasts did better than the polls on the Tory vs Labour seat gap, as they assumed that the polls were somewhat biased to Labour. However, these models grossly overestimated the Lib Dem seat count. In a previous article, I had expressed scepticism with the Ashcroft polls that were used for the forecasts, and had predicted that the Lib Dems would do even worse than they would under uniform swing assumptions.