Driven by data; ridden with liberty.
After Prime Minister Theresa May called a general election to seek greater certainty and stability over negotiations (i.e. a larger majority), the Conservatives are now the largest party in a hung parliament.
Unlike in 2015, the final pre-election voting intention polls did not show very similar results. For members of the British Polling Council, there were point estimates of the Conservative lead ranging from one point (Survation) to 13 points (BMG Research).
The Conservative vote share has been accurately predicted, with an unweighted polling average of 43.7%, compared to the GB result of 43.5%.
Similarly, the Liberal Democrats and Greens were accurately predicted.
Where pre-election polls diverged from the final result was in the Labour and UKIP shares, with every pre-election poll underestimating Labour’s 41% share of the GB vote. Note: this is the vote share in England, Scotland and Wales, and Labour’s share of the UK vote is 40.0%. Since these polls are of adults in Great Britain, the proper comparison is with the GB vote shares.
The average error by company on Labour’s vote share was 5.2 points. The error on UKIP’s share was 2.5 points, which is a major error on a party that received only 1.9% of the GB vote.
For market research companies, political opinion polling is a small part of their operation, which receives a large amount of media and public attention.
Whilst there were more companies conducting polls throughout the campaign, there were nine polling organisations that conducted a ‘final poll’, which included either Tuesday 6th June or Wednesday 7th June in their fieldwork range.
The two polling companies with the smallest two party mean average error were Survation (conducted between 6th and 7th June) and Kantar Public (conducted from the 1st June to 7th June), both at 1.8 points.
Unsurprisingly, these two companies performed the best at estimating the Conservative lead, with Survation underestimating the lead by two points, and Kantar Public overestimating by two points.
Whilst their final poll showed a seven point lead, YouGov had been previously showing leads of three to four points.
Noticeably, there does not appear to be a large issue of sampling, as in 2015.
The British Polling Council expects companies to present what lessons they have learned from this election, rather than participate in an inquiry.
Commissioned by the broadcasters (BBC, ITV and Sky), the exit poll measures changes in party shares at over 100 polling stations across the country, asking respondents to replicate their vote in a mock ballot box outside the station.
At 10pm, the exit poll was published, and showed that the Conservatives were the largest party, but the central estimate was they would achieve 314 seats, short of an overall majority.
With all the results, the exit poll was accurate, with the Conservatives getting 318 seats (compared to 314 in the exit poll), Labour had 262 (against the projection of 266). The SNP has 35 (slightly above the projection of 34), and the Liberal Democrats came to 12 seats (below the exit poll projection of 14).
The Greens have one MP and UKIP have no MPs, just like the exit poll said.