ICM Marginals poll	Voted CON 2005	Voted LAB 2005	Voted LD 2005
A	“I am certain which party I will vote for at the General Election”	67%	46%	41%
B	“I think I know who I’ll vote for but could change my mind”	18%	25%	31%
C	“I’m a floating voter and I will not decide who to vote for until later in the election campaign”	14%	28%	24%

Can we trust their assumptions about how they’ll vote?

Above are some findings from the weekend’s marginals poll from ICM which I highlight to show the scale of the problem created by the “don’t knows/won’t says”.

For the firm, in common with all the other phone pollsters bar MORI, makes an assumption about the intentions of those who say they will vote but at this stage don’t know which way. As can be seen above they represent a big segment of the electorate in the marginals particularly amongst 2005 Labour voters.

Rather than excluding them from the poll ICM has a formula that allocates 50% of them in accordance with the party that they voted for five years ago.

This can have a huge impact on the headline figures as the detailed data from the marginals poll shows. This moved a Labour deficit of two points before the adjustment to a Labour lead of one point.

Without this the swing in these marginal seats would have been at 8.3% – its actually reported at 6.3%. That’s the difference between being in “hung parliament territory” and a sizeable Tory majority.

When faced with a “don’t know” response the one exception amongst the phone pollsters MORI seeks to tease out what this group will do by asking what party interviewees are “leaning to”. This gets quite a good response and to my mind is much more satisfactory than the approach of ICM. At least it’s based on real responses rather than assumptions.

My concern is that the headline figures from this poll might be disguising the challenge that Labour is facing in these key seats.

I am inviting the boss of ICM, Martin Boon, to take part in a live Q&A on PB like we have done with other pollsters.

Mike Smithson