You can’t draw those big conclusions from recent election polling
In a recent post the Indy on Sunday’s John Rentoul had this to say about the EP14 and Newark polling and its implications for GE15.
“…By-elections in which a seat fails to change hands may not seem the most exciting of events, but they are important because they allow us to assess the accuracy of opinion polls. Two were carried out in Newark in the final two weeks, one by Lord Ashcroft and the other by Survation. Both predicted a comfortable Conservative win over Ukip. But both had the Tory share of the vote too low by three points and Labour’s share too high by two and four points respectively. The same thing happened in the six final polls before last month’s European Parliament election: on average the pollsters had the Tory vote 1.7 points lower than the result and Labour’s vote 1.5 points higher…”
Tightly fought by-elections where the main contenders are throwing in an enormous amount of resource can be subject to late swings and the polling showed that by two days before things were moving even further to the Tories. Survation in its first poll had CON on 36%. Four days later that was up to 41% – clearly that was where the momentum was.
The polling highlighted the likely trend over the final period.
With the EP14 polling Rentoul writes about the “average” – but that was distorted, I’d suggest, by one or two firms not doing very well. One of them, the ComRes online survey for Rentoul’s own paper, the IoS, came bottom of the polling accuracy table overstating UKIP by a huge 5.5%, LAB by 1.4% while understating CON by 3.9%.
Thus, as the chart above shows, ComRes online had UKIP 13% ahead of CON when the actual gap was 3.6%.
Survation got the Tory total to within 0.9% and both ICM polls, phone and online, overstated the CON share. All the firms bar ICM got LAB to within the margin of error.
It’s hard from all of this coming to any conclusion for GE15.