(By the way, the second part of the title has nothing to do with potential further independence referenda…)
Ipsos-MORI today published a poll showing the Scottish National Party with an astonishing 29 point lead in voting intention for Scotland’s Westminster seats. Toplines, together with my estimates updated with this week’s data, are as follows:
That’s a 26 point swing from LAB to SNP, which would be impressive in a by-election, but in this context, really is off the swingometer. It also shows the SNP in a much stronger position than my (updated) estimate. What accounts for the difference?
You can see from the chart’s rightmost tab that the SNP are 10 points higher than my estimate, now with more than half of the poll. Who are those votes are coming from? Labour are 1.5 points lower, within the margin of error, as are the changes for most of the parties.
But the Conservatives are all the way down on 10%, 7.6 points lower than my estimate. That is possible, but it would be at odds with all other available data, including the Panelbase/SNP poll, which showed the Scots Tory vote moving sideways post-referendum.
It’s worth noting that the last time (that I’m aware of) that Ipsos-MORI did Scottish polls (in the months following the election in 2010), they tended to show the Tories 3 or 4 points lower than YouGov and the SNP higher by a similar amount. We can’t be sure if that’s still the case now until YouGov (on whose data my number crunching is based) do their own poll, but this picture is at least consistent with consistency of house effects.
Certainly the two pollsters’ methodologies are completely different. YouGov poll online and prompt for the main parties (including for voters living in Scotland, the SNP) or “Another party” and include all voters besides DKs, non-voters and refusals, with weighting by party ID. Ipsos-MORI poll by phone, do not prompt for any party and include only those “certain to vote”, with no political weighting. From the tables, the turnout filter doesn’t seem significant. Not using political weighting, on the other hand, can lead to exaggerated volatility. This too could explain part of the difference – time will tell.
What we can be sure about is that there has been a very substantial swing. I’m not going to bother with a Westminster seats diagram, but it would be virtually all yellow, with one red and one gold seat.
This has a huge impact on the electoral skew, which I’ll be writing more about soon. Take these kind of seat numbers with a bucket of salt, but IF they were an indication, then with GB vote shares of CON 33, LAB 33, LIB 8, a UNS gives Labour 38 more seats than the Conservatives. This poll suggests they would lose 40 seats and the Tories their sole Scottish seat. Obviously there are many other moving parts (some of which offset each other) but with this sort of Caledonian collapse, you can see how the electoral map might no longer favour Labour. They would lose 19% of total Scottish votes, but 40 (68%) of Scotland’s 59 seats. In technical terms, the distributional efficiency of their vote would be far lower.
This would massively increase the chance of a hung parliament. There is almost almost no change to the Tories’ chances of winning an overall majority, as they would simply lose their one Scottish seat. Labour, on the other hand, would lose 40 seats and would need to find them somewhere else. Getting even a bare majority without those 40 Scottish seats would be like winning an 80 seat majority with them.
I stress that the above is hypothetical – this magnitude of move cannot be uniform (UNS implies that LAB vote shares would be negative in a number of seats!) and I’d need to see more polls showing the SNP in the 50s before I’m convinced they’re really up there. But it does illustrate how important this arithmetic is, south of the border as well as north of it.
Update (22:10): Turns out we didn’t have to wait long for a YouGov poll, they’ve published one for the Times an hour after I posted this! Toplines are SNP 43 LAB 27 CON 15 LIB 4 – so it seems that the MORI vs YouGov house effect is very little changed from 2010.
I also means that the much-derided crosstabs (and the ‘weighted’ method) were both within the margin of error for every party:
For the latest on Scottish polling, please click here. You can also: