Putting that ICM poll in context
The ICM poll last night has produced various comments from people dismissing it or defending it. Let's put it in some context. Firstly there is no large swing here. All the parties are within 2 points of ICM's last poll, so there is no reason to think it is obviously a rogue. Indeed, the small shift from Labour to the Conservatives is entirely in line with the only other post-Budget poll from YouGov.
Both companies have the Conservatives at 37%. The difference is basically the Liberal Democrats - YouGov have them at 11%, ICM have them at 16%. In YouGov polls a large chunk of former Liberal Democrats now say they'll vote Labour, so as a consequence of their different Lib Dem scores ICM also end up showing a lower Labour score than most other companies... hence the 1 point Tory lead yesterday.
This is not down to margin of error or we would see ICM giving the Lib Dems a lower score than YouGov half the time. ICM are consistently showing higher Lib Dem scores than other companies, suggesting it is a systemic methodological difference.
Part of the difference we have a straight explanation for. Most companies ignore people who say don't know, ICM re-allocate 50% of them to the party they claim they voted for in 2010, on the basis of evidence from past elections that this is what people saying don't know tend to end up doing. Over recent months this has typically increased the level of Lib Dem support ICM finds by 2%.
However, this is not the whole explanation - without it we would still have seen recent ICM polls showing the Lib Dems in the mid teens, when every other company has them between 9% and 11% (equally, Populus also reallocate don't knows, albeit using a formula slightly less favourable to the Lib Dems and they have the Lib Dems at 11%).
There is no obvious methodological reason for the rest of the difference between ICM's Lib Dem score and that other companies produce. In his post-mortem of the election result Martin Boon of ICM suggested they may be weighting the Liberal Democrats somewhat too highly... but looking at the current weighting targets ICM weight the Liberal Democrats to pretty much the same figure as other companies using past vote weighting, so that cannot be the reason for the difference.
Another suggestion is the wording of ICM's question, which mentions "in your area". I think this is probably too subtle to make such a large difference to the results and, besides, Angus Reid use the same sort of wording and have the Liberal Democrats at 10% (The example normally used of prompting by constituency making a difference to voting intention questions is a poll I did of marginal seats for PoliticsHome back in 2009. That, however, asked constituency voting intention straight after national voting intention and used very gung-ho prompting to shove people towards considering their own area. It wasn't a tiny subtle word change like this).
Hence in terms of saying whether ICM are right on the Liberal Democrats (and therefore Labour), or whether the other companies are, there is no easy answer since we don't know what is causing it. I expect, in practice, most people will tend to believe the results they want to.
Besides, while it may seem important this week whether the polls are showing Labour or the Conservatives head, remember that the Conservatives got a boost of a couple of points after the June budget too. It lasted all of a week. My guess is that we'll be soon back into the more solid Labour leads we've become accustomed to.
