How sensitive are the poll figures?

How sensitive are the poll figures?

Let’s take a step back from the day-to-day swings and look at the overall picture, focused on the likelihood of a Conservative overall majority (I’m not going to look here at changes between the opposition parties). The result at the last election was 

Con 42 Lab 40 LD 7 UKIP 2 Green 2 others 7

The current polling average is 

Con 37 (-5) Lab 25 (-15) LD 17 (+10) BXP 11 (+9 over UKIP) Green 4 (+2)

I’m taking here the 3 latest polls as I write (see here)

Start with uniform swing (UNS). If exactly these changes happened everywhere, with a 5% Lab-Con swing and a 7.5% Con-LD swing, the Conservatives would gain 47 seats from Labour while the LibDems would gain 11 from the Conservatives. This would give the Tories 353 seats, a working majority of about 60. (Source: Here.) 

How sensitive are these leads? Assume there is some tactical voting. I know that lots of us would not consider it, but it’s an objective fact that there are folk out there who do, and they mostly are either BXP->Con, Green->Lab or LD, or Lab->LD.

Polls suggest that about half the Opposition voters are up for this. There are roughly twice as many LDs/Greens and Lab/Greens than BXP, so there should be more tactical voting to Labour and LibDems than from BXP to Con. Let’s call that differential tactical voting. 

If, to take a modest assumption, the Lab->Con swing is in practice 2.5% less as a result, the outcome is 24 Tory gains. If in addition the Con->LD swing is 2.5 more for the same reason, we see 17 LibDem gains from the Conservatives. The overall result is 324 Tory seats, which is roughly a tie. If the differential effect is larger than 2.5%, the Conservative majority disappears.

There are lots of qualifications in this. Obviously sentiment will shift during the campaign – though 2017 was unusual in seeing a large shift. There will be constituency variation, but this tends to even out – some surprise gains, some near misses.

Quite a bit of tactical voting happened last time already, so what we’re looking at here is net new tactical voting resulting from people seeing the current position (for instance, first-time incumbency bonus for some Labour MPs). We know voters (like our kind host) who voted tactically last time and won’t this time, but on the whole voters seem more polarised than 2017 and more likely to vote tactically rather than less.

What is striking to me is that the numbers involved are relatively small. Many people are currently assuming a Tory landslide – well, 60 with no tactical voting at all is a decent majority, but it’s not 1983/1997 scale. And a 2.5% differential tactical vote shift seems to me perfectly plausible. Moreover, the position is robust against Lab-LD shifts.

If Labour recovers some ex-LD votes during the campaign, they do a bit better and LDs make fewer gains, but the overall picture is broadly the same; likewise if the LibDems pick up a bit more from Labour. The game-changer would be if lots of Tory voters switch to Labour or vice versa, but recent polling data shows 90% of both blocs of voters wouldn’t even consider that, much less do it.

If I had to guess, I’d now predict a Tory majority of 15-30 overall, but anything from 60 to minus 30 and a Labour minority government looks perfectly credible. Do your own research, but despite the feverish atmosphere I’d bet against larger changes either way.

Nick Palmer

 

Comments are closed.