Should swing be worked out from these – not the result?
The 2005 final polls:
C33% L38% LD23% – MORI
C33% L38% LD22% – HARRIS
C32% L37% LD25% – YouGov
C32% L38% LD22% – ICM
C32% L38% LD21% – Populus
C33% L36% LD23% – NOP
C33% L37% LD21% – BPIX
C31% L39% LD23% – COMRES
The 2005 result:
C33.2% L36.2% LD22.7%
Would this factor in likely Labour polling over-statement?
In the past week while we’ve been putting the focus on what lead would produce a majority couple of PBers have emailed me with the same idea – rather than calculating swing from the the 2005 votes why not do it when each poll comes out from the the final 2005 surveys of the firm involved?
I’ve listed the closing polls above which was by historical standard pretty good. Also by standards of the last five elections it followed the trend that we have got used to – Labour being over-stated.
So when we get a new poll in this crucial period should we make a calculation of the swing from the firm’s final 2005 survey if only to remind ourselves that polls are not perfect?
A problem is that in 2010 all but two of the current regular polls from 2005 are operating in a different manner. YouGov radically changes their party ID weighting; Ipsos-MORI abandoned face-to-face polling and switched to the phone while ComRes have adopted past vote recall as a means of trying to achieve a politically balanced sample.
But having said that it’s not a bad idea to highlight every now and then that the pollsters do have a long history of Labour over-statement. When the latest Populus came out on Monday night one commenter observed that the same poll at exactly the same point before the 2005 election had Labour in the 40s with a 9% lead.