ICM/Guardian - 38/34/19

There is a new ICM poll in this morning's Guardian. The topline figures, with changes from the previous ICM poll in late June, are CON 38%(-3), LAB 34%(+3), LDEM 19%(+3). That's a bouncing back for the Lib Dems from 16% in the previous poll and the first survey to show the Conservatives below 40 since Populus after the budget.

On other questions, people are pessimistic about the economy, with 51% thinking it is likely to fall back into recession compared to 43% who think it will not. On the defecit reduction plans so far, 38% think they go too far, 39% think they are about right and 16% think they do not go far enough.

YouGov's overnight voting intentions were CON 42%, LAB 35%, LDEM 15%. Net government approval was just plus 3, the lowest YouGov have shown it so far in their daily government approval figures.

I expect there will be a lot of attention paid to the difference between the Lib Dem figure in ICM and YouGov. The rather unexciting truth is that it's probably premature to read much into it at all. The previous ICM poll in June was pretty much in line with YouGov's, with ICM showing 16% and YouGov 17%. All in all, there have been 5 ICM polls since the election, in 2 cases they showed an identical Lib Dem figure to YouGov, in one case they had the Lib Dems lower than YouGov and in two cases higher than YouGov.

The sheer volume of YouGov's daily polling means we know their Lib Dem score is averaging about 15%, and if you get a 13% or a 17%, it's likely just noise. In more traditional polling with just one or two polls a month, you can't really be sure if something is an outlier or not - there may well be no difference here at all (or perhaps a smaller difference than these particular figures imply).

If it persists over time, then I'll look at it properly, since it's certainly plausible that there's a difference. ICM's "spiral of silence adjustment" consists of reallocating people who say don't know to the party they voted for at the previous election, so if a lot of former Lib Dem voters are now saying don't know, this adjustment will help the Liberal Democrats. In Martin Boon and John Curtice's article on the 2010 election polls they also said they thought they might have been weighting the Lib Dems too highly, and would be looking at it in the future. So I can think of some theoretical reasons why ICM might show a higher Lib Dem score than YouGov, but on the evidence we've got at the moment, I'm not certain they consistently are.

We've also got a MORI poll due in the next day or two.

UPDATE: Full tables for ICM are here. For what it's worth, 2010 Lib Dem voters were slightly more likely to say don't know to voting intention than former Con or Lab voters... but it wasn't enough to increase their topline figure. In fact, the effect of the topline adjustment was to decrease Labour's support by 1 point. There do not, as yet, appear to be any changes to ICM's methodology from pre-election (aside, of course, from the fact that ICM will be weighting to 2010 recalled vote instead of 2005 recalled vote).