Could it even show Labour to be level-pegging?
Thanks to the Lib Dem poll watcher, Mark Senior, on a previous thread for highlighting the very high proportion of 2005 Labour supporters in the sample for the latest Guardian ICM poll.
From the table it can be seen that one in two of those who said they had voted at the last election told the pollster that they had opted for Labour. Contrast that with the 36.1% GB share that Blair’s Labour actually achieved on that day.
Generally pollsters find the proportion answering Labour to the past vote question is bigger than the election figure – but I cannot recall seeing anything on this scale before.
Was there something special, I wonder, about last weekend that caused a disproportionate number of 2005 Labour supporters to be home ready to answer ICM’s questions when the pollster called?
And if there was something special was that also picked by Ipsos-MORI which had its December survey in the field at precisely the same time?
It is to deal with abnormal samples that most pollsters make a corrective adjustment so that the make-up of those who take part is more politically balanced. Just look at the right hand column above to see the big changes that ICM had to make.
They don’t weight to the actual result but have a formula to deal with false recall that generally means that their shares are based on a six point Labour lead in 2005 rather than the 2.9% that we saw on the day. The exception is PB’s pollster, Angus Reid, which weights to the actual result – one reason why its Labour shares are usually the smallest.
So what’s going to happen with the MORI poll. For lone amongst the main firms MORI does not weight by past vote or party ID although it does have a public sector worker weighting.
Mark Senior has projected that applying the MORI methodology to the ICM data would have produced 37-34-17. It might be, of course, that MORI’s sample was absolutely normal and this effect was confined to ICM.
We are probably going to have to wait until Saturday evening to get details of MORI. My guess is that like last month the poll will be in the Observer