A few thoughts updating those from early Monday morning. Since penning this in the small hours while waiting for Tower Hamlets to report their data, YouGov’s Stephan Shakespeare pointed out to us that they had factored in support for AIFE in their voting prompt (but AIFE results were still likely higher than their polling so we’ve made some changes to factor that in).

Analysing The Pollsters – Getting UKIP Right

• 4 of 7 pollsters in their final polls had higher UKIP figures than transpired. We think that factoring in the vote share loss to UKIP for Mike Nattrass’ clear spoiler party “An Independence From Europe” (1.49%) makes sense . His party name and description being so similar to UKIP and at the top of the ballot suggests to us removing their % from pollster’s UKIP final headline figures is a fair approach to measuring accuracy (although we all should have made an adjustment ahead of time).

• Considering this clear attempt to confuse voters, which UKIP are taking up with the Electoral Commission we believe UKIP would have achieved 29% of the vote, not the 27.5% the party ended up with.

• YouGov to their credit presented respondents with the actual ballot paper choices by region and found AIFE’s “support” at something around 1% – although we suspect this may have been their own respondents believing they were indicating a hurried vote for UKIP – Outside of mistaken votes intended to be cast for UKIP – AIFE’s support in our view is likely to have been effectively zero in reality.

• For measuring UKIP support then, Survation, TNS, Opinium and YouGov all had UKIP’s vote share within 3% on this measure. YouGov were the closest to our (AIFE adjusted) target, just 1% out from UKIP’s final %. Having YouGov’s resources and funding however to conduct a 6000 person poll on the eve of polling would certainly reduce margin of error timing that captured what appeared to be a softening UKIP lead in the campaign’s final days.

• Predicting the winner however caused some some important divergence. Of the 7 pollsters with “final” polls, YouGov, showed headline figures on this measure indicating a tighter UKIP victory (just half of 1%) than transpired – a 2.1% UKIP lead or 3.6% had Mike Nattrass not launched his cunning plan sucking votes from unwitting UKIP supporters. Even without this AIFE adjustment, ICM for example had UKIP 4 points behind their stated polling winner – Labour – and actually had UKIP in third place behind the Tories.

• Aside from Opinium, most firms overstated Labour’s support to some extent. Survation were so concerned about lacklustre turnout by Labour supporters we introduced an additional step discounting intention to vote if the voter was not aware of election day, which helped our final Labour figure. This turned out to be very important as Populus and ICM for example had large Labour leads over UKIP which of course did not transpire.

Some caveats are Important. Populus were the least accurate across all parties and had UKIP in third – with Labour winning the election but did not to our knowledge make a “final” poll – their last work being 2 weeks from polling day.

Damian Lyons Lowe.