Some reminders of the latest polling “form”
The last time that an NOP poll was tested against real results was just six months ago at the by-election in Hartlepool – the town where they elected a “monkey” (above) as Mayor.
When their poll came out showing a 33% Labour lead I dismissed it as “wildly over-stated” and came under a lot of criticism here. Come polling day the actual Labour lead was 7% so the survey was a massive 26% out.
For Hartlepool was the most recent of a very long line of Labour over-statements by the pollsters but at least in that by-election NOP got the top two parties in the right order.
This was not the case for another pollster, Populus, fifteen weeks earlier at the Euro Elections when their final survey got the position of all four of the leading parties wrong. Their final week figures were CON 24: LAB 25/26: UKIP 13: LD 16-18. The actual result was CON 26.7: LAB 22.6: UKIP 16.1: LD 14.9.
So instead of the Labour lead of upto 2% reported by Populus there was a Tory lead of more than four per cent. Thus the pollster’s overstatement of Labour’s position was the equivalent of more than a quarter of those who actually voted for the party.
Hartlepool and the Euro Elections are the most recent examples of UK pollsters being tested against real results.
I refrain from betting on horse-racing but I am advised by those who do to check the form of the runners.