SORRY THE SITE IS SLOW TODAY – WE’RE DOING OUR BEST TO FIX IT BUT WE ARE BEING OVERWHELMED WITH TRAFFIC
During the robust exchanges on our election prediction formula we were sympathetic to this response from a professional pollster, Graham, when asked if his industry was going to overstate Labour again.
They (we) probably will do it again. When you ask people their opinions and they tell you, as a researcher, your job is to report what they have said. If you try and put too many fixes and adjustments in it, you are just as likely to get it wrong as if you don’t.
The intriguing feature about the almost universal over-statement of Labour in the polls is that the same pattern occurs whatever methodology is used.
Face to face polls such as MORI over-stated Labour in 2001
Telephone polls such as ICM and NOP over-stated Labour in 2001
Internet polls such as YouGov have over-stated Labour
The automated pollster, Rasmussen, over-stated Labour in their only experience of working in the UK at the 2001 election.
We have no doubt that the figures we have been given over the years are honest ones based on what people have told them.
Could the problem lie with those who tell pollsters they are voting Labour and not the firms themselves? Declared Labour supporters, it would appear,are less likely to actually vote than those of other parties.
It is here where we disagree with Graham and applaud the actions of Bob Worcester of Mori. His response to the 2001 overstatement was to include in the Mori Monitor figures only those saying they are 100% certain to vote. The effect has been to raise Tory shares on Mori and depress Labour ones.
Whatever the past form of those polled reinforces our view that a correction has to be applied to those declaring for Labour compared with LDs and Tories. This is all we are doing with our formula and as we get more information we’ll refine the equation.
If and when Labour starts achieving more than its poll shares then we will be calling for a premium to be applied.