ICM poll is better for Yes on the unrounded numbers

It's uncanny how often this has happened with ICM - to be fair it's just coincidence/bad luck, but the rounding has yet again flattered No in the published results of the new poll for Scotland on Sunday.  On the headline figures that take account of Don't Knows, the true No lead before rounding is much closer to 10% than to the 11% we saw in the published numbers.  And when Don't Knows are excluded, Yes are at 43.44%, and are therefore just the tiniest smidgeon away from being rounded up to 44% rather than down to 43%.  Here are what the poll results look like when rounded only to one decimal place -

Should Scotland be an independent country?

Don't Knows excluded :

Yes 43.4% (-1.8)
No 56.6% (+1.8)

Don't Knows not excluded :

Yes 34.3% (-1.5)
No 44.6% (+1.2)

So the changes from last month (which the barking mad Huffington Post described as a Yes "nosedive") are even smaller than they appeared to be in the published results, which increases the overwhelming likelihood that they're just meaningless margin of error 'noise'.

You might also remember that after revealing the results of the ICM poll from two months ago, Kenny Farquharson made a big issue of the fact that if the firm's methodology hadn't just changed to take account of differential turnout, then Yes would have slipped to 40% (actually 40.4% on the unrounded numbers) rather than to 42.4%.  Well, If you scour your memory of the last 36 hours or so, you'll probably realise that no-one has made that point this time, and for a very good reason - in this case the Yes vote before turnout weighting is applied is the same (43%) as in the published results.  So whichever way you look at it, the No lead is clearly lower in this poll than it was two months ago - and by quite a substantial amount on one measure.

I'm slightly baffled as to how the Scotland on Sunday and John Curtice felt able to make a song and dance about how this poll supposedly shows that people are starting to trust the anti-independence parties on the delivery of new powers to the Scottish Parliament in the event of a No vote.  In reality, the number of people who think no new powers will be delivered or that existing powers will be taken away is, at 45%, higher than the number of people who expect new powers to arrive (40%).  It's possible that the electorate may eventually have the wool pulled over their eyes by the No campaign, but it certainly hasn't happened yet.

Now let's have a look at what John Curtice didn't tell us about the question asking people to rate themselves on a 1-10 scale of support for independence.  The first thing to say is that this only turns up deep, deep into the question sequence, which is less than satisfactory, because by that point many respondents will be self-consciously trying to keep their responses consistent with what they've already said.  Nevertheless, the results are much better for Yes than we've been led to believe.  As a number of people confirmed on the previous thread, it's standard during canvassing to regard a '5' as being exactly in the middle (ie. a complete Don't Know), meaning that anything between 1 and 4 is on the anti-independence end of the scale, and anything between 6 and 10 is on the pro-independence end.  Pedants and statisticians may bristle at that notion, because it means there are slightly more pro-independence numbers available, but nevertheless that's how it works - the human brain is programmed to see 5 as the mid-point in a scale that goes up to 10.  When looked at that way, these are the results -

Pro-independence (6-10) : 43.5%
Anti-independence (1-4) : 48.9%
Neutral (5) : 7.6%

(The above numbers exclude the small number of respondents who refused to place themselves on the scale.)

Now are you starting to see why the Yes campaign are so bullish about the numbers that are being produced by that approach?  The other point Stephen Noon made the other day (and unfortunately the ICM poll can't assist us with this one way or the other) is that, over time, people have been gradually moving up the scale towards independence, and that was happening even before any movement towards Yes was detectable in the published polls.  If that's true, it could be hugely significant.

Of course it also means that the people who currently rate themselves as a 5 are particularly important, as they literally hold the balance.  So what do they look like?  Well, 40% of them didn't even vote in the 2011 election, so that limits the numbers of clues available.  22% voted Labour, and 12.5% for the SNP.  Most say they are undecided on how they will vote in the referendum, but a significant minority of 24% currently say they will vote No, which goes a long way towards explaining the disparity with the headline numbers - these are the soft Nos we keep hearing about.

There's also a weird amount of "cross-voting" going on - supposed No voters who place themselves on the pro-independence end of the scale, and supposed Yes voters who place themselves on the anti-independence end of the scale.  Were these people just confused by the question?  It's likely that at least some of them were, given that 1% of No voters rate themselves as a 10, ie. "completely for an independent Scotland"!  But the 5% of No voters who rate themselves as a 6, 7, 8 or 9 are potentially more significant - those could well be soft Nos (very, very soft Nos).  By contrast, just 1% of Yes voters rate themselves as a 2, 3 or 4.

Annoyingly, ICM haven't asked a country of birth question this time (or if they have there's no sign of the results).  But on past form it seems reasonable to assume that English-born people are once again heavily over-represented in the sample, which almost certainly means that the headline Yes vote is a bit too low, and the headline No vote is a bit too high.  However, women aren't quite as over-represented as they were in the previous two polls, which is a touch irritating given that (very unusually) more women than men in this poll say they will vote Yes!