I woke up this morning to people fretting about a 'bad' Survation poll in the Daily Record, although in fact it's not as bad as you'd think if you inhaled the Record's reporting. For some unknown reason (well, we can probably guess), they've used percentage changes from the Holyrood election of 2016 rather than from the last comparable poll - giving the misleading impression of a very sharp and sudden drop in SNP support. They don't actually make clear whether this is an online or telephone poll, but assuming it's online (the vast majority of Record-commissioned polls are), here are the correct figures with percentage changes measured from the last online Survation poll that was published around three weeks ago -
Scottish Parliament constituency ballot:
SNP 38% (-5)
Conservatives 26% (+2)
Labour 25% (+2)
Liberal Democrats 9% (n/c)
Scottish Parliament regional list ballot:
SNP 32% (n/c)
Labour 23% (n/c)
Conservatives 23% (+2)
Greens 9% (-1)
Liberal Democrats 9% (+1)
SNP 36% (-5)
Conservatives 27% (+1)
Labour 26% (+2)
Should Scotland be an independent country?
Yes 45% (-2)
No 55% (+2)
Although the changes aren't as dramatic as the Record are making out, the SNP are significantly down on two of three ballots (strangely they're not down at all on the Holyrood list). It's hard to make much sense of that, because three polls were published at the time of the SNP conference (two Survation, one Panelbase), which all agreed that the SNP had come through the reporting of the Alex Salmond story unscathed. So if there has been a setback, it must have occurred over the course of October, and I can't think of any obvious recent trigger for a 5-point drop. My own instinct is that this is much more likely to be random sampling variation - ie. within the standard margin of error, the sample Survation interviewed this time may have been a bit SNP/Yes-light. We'll have to wait for the next poll or two to find out for sure.
There's an absolutely nonsensical section of the Record article which observes that a small swing from SNP to Labour since last year's Westminster election would see Labour gain eight seats, pushing the Tories back into third place. That's true as far as it goes, but the rather more salient point is that the Record's own poll shows that no such swing has occurred - the SNP's lead over Labour stands at ten points, exactly the same as June 2017. On a uniform swing, the SNP would lose no seats at all to Labour, and would gain one seat (Stirling) from the Tories. And if the Record are so keen to go off on a tangent and talk about hypothetical swings that were not detected by the poll, it's mysterious that they neglected to mention that a small swing from Labour to SNP could see the SNP gain six seats, practically wiping Labour out once again.
As ever, the SNP and Green showings on the Holyrood list have to be regarded as extremely suspect. A relatively low SNP list vote, and a relatively high Green list vote, seem to be standard 'house effects' of Survation's online polling - perhaps because the question characterises the list vote as the "second" vote, thus giving some respondents the false impression they are being asked for a second preference vote. Nevertheless, given the direction of travel elsewhere in the poll, it's perhaps surprising (and reassuring) that the SNP haven't slipped down to the 20s on the list.
John Curtice is quoted in the Record piece as saying that the poll shows the "fragility" of the SNP's position. But in fact, the Holyrood seats projection puts the pro-independence parties just four seats short of retaining their majority. It's pretty remarkable that they could be so close to doing that when the SNP are in the 30s on both ballots - which arguably illustrates the in-built strength of the SNP's position, providing they can retain a handsome lead over the second-placed party on the constituency vote. So you can look at the situation in more than one way.
UPDATE: Survation have belatedly tweeted about the poll, and have done so in this annoyingly ambiguous way -
New Scotland polling with fieldwork 18th-21st October in today's @Daily_Record. (Changes vs 3rd-5th Oct*)
Westminster voting intention;
CON 27% (nc)
LAB 26% (nc)
LD 7% (+1)
SNP 36% (-1)
AP 3% (nc)
Does the placing of the asterisk indicate that they are making a potentially misleading comparison between an online poll (this one) and a telephone poll from earlier in the month? Or does it mean that the new poll was conducted by telephone as well? I hope it's not the latter, because if so it would mean that half of the above blogpost is based on a false premise, but we'll find out when the datasets are published...