Wild recent polling produces wild results
The columnists had a lovely job this week: the Johnson government in unprecedented meltdown (seemingly owned by the opposition, in possession of a minus-43 majority, a heated debate about what kind of prison food the former PM might expect, etc etc) and yet a swarm of polls suggesting things are not too bad. A correspondent asked what they mean for the seats outcomes according to my machine; feeling all relaxed after a nice run* and with the model newly re-written to strip out ancient bugs, I decided to oblige.
First, Opinium, which shocked us with a 10 point Tory lead, and the LibDems down at 17. Even with a moderate degree of “Libdemmyness”, as I have christened my skew on the LibDem voting patterns, you – obviously – get a handsome Tory majority.
Then came YouGov, even more shocking – 14 point lead!
Even fewer surprises there. No one would quibble at this being a deserved victory for a No Deal Brexit – though the result is “unfair” in that the Conservatives would win 11 seats for every percentage point share, Labour just 7, the combined BXP-CON vote is pretty compelling.
But then we got ComRes, and a very different story : Conservative lead of just 3 (for a situation where Brexit is not delivered as of 31 October)
Here the Conservatives’ governing majority is wiped out, the strategy has failed. There’s an intriguing multi-coloured government somewhere in there. Corbyn has lost seats though; is he under pressure? Esher and Walton falls (Uxbridge does not).
And here, worse (or better: let’s stay neutral) – the Labour lead envisaged if a Brexit Extension is imposed
Total disarray for the Conservatives, a small victory for the Labour party, a big one for the LibDems.
Finally, Delta and ComRes did a couple of similar ones showing the Tories with a small lead, like this
Bottom line: well, it is all obvious. And as for which of these top line numbers feels right, you tell me. The people I mix with are appalled at what Johnson is doing; the vox pops done in the Observer and BBC appear to cheer him on.
The bulk of the fights are about Conservatives, and that surely matters
Final observation. I was noticing that no matter how much I messed around with my model, the seat switches to the LibDems were nearly all from the Conservatives – even if on most polls the Labour vote is down as much as the Conservatives’, and if we assume Conservatives are losing votes to BXP, the LibDems must be getting more from Labour.
For example, in that last Delta-ComRes result, there were 48 Con seats falling to LibDems, 12 to the SNP, and the Conservatives gaining back 28 from Labour. Only 6 Labour ones fall to the LibDems. Why?
It appears to be because of the 2017 results, where only 7 of Labour’s seats are held against the LibDems in 2nd, while 29 of the Conservatives’ are. And after the swing above, even more are set up that way. We would move into a situation where a quarter of the House’s seats are Conservative-LibDem fights, but only a tiny percentage are Labour-LibDem fights.
If I have got this right, it feels significant, for tactical voting. At the headline poll level, it looks like the fight is all about who gets to be the Anti No Deal party; at a seat level, the tactical sorting may be a lot easier than you think. It is a fight against Tories in most places. Next polling model post should be about how on earth to model that ….
*I apologise. It motivates me.