Could the Tories End Up On <100 Seats?
When statistical models break down
I'm a coward. If I get into car (necessarily as a passenger) I think about a chance that I'll be killed in a crash. Generally speaking, that is very low - a very small % of car journeys end fatally. That is true even if you’re sitting in the car and it’s driving at speed down a road with a row of houses at the end - a straight-line projection says you’ll plow into the houses and die but a statistical model says cars rarely plow into houses.
But what if the driver is blind? Or foaming at the mouth? Or cheerfully informing you that they’ve Done The Research and it proves that cars must have special anti-crash systems as they grin and jam their foot on the accelerator?
That’s the point at which you should revisit the implicit assumptions in the statistical model. Sure, cars rarely plow into houses - but then blind/foaming/insane people are rarely behind the wheel.
The great thing about simple statistical models is that they avoid all the confusion and complexity of a mechanical, causal model. But it is important not to then read that as if it were necessarily some sort of automatic mechanism that operates regardless of the actions of any specific agent.
Specifically, the statistical model I have in mind is the Polls Almost Always Narrow From Midterm - laid out very clearly here by Rob Ford (Midterm Limbo Dance). There’s a tendency for people to read this as a contradicting straight-line projections (the Tories are currently projected to win 45-134 seats *if an election were held tomorrow and limited[1]/no tactical voting occurred*). But really they are just two different models with different implicit/explicit assumptions (“parties generally swerve”/”what if they didn’t/couldn’t?”)
Some (reasonable!) people have a *lot* of confidence in the “parties generally swerve” model. Should they? Seems to me that confidence should be proportionate to how representative the historical post-WWII electoral dataset is of the context of the next GE. Some of that context is broadly[2] ‘external’ to the Conservative Party (world events/domestic economics/Labour Party dynamics) but the party I’m focusing on here is the ‘internal Conservative Party component’.
I don’t have a conclusive/objective answer to that - hard to think about how you’d go about measuring that other than looking at the “subsequent incumbent GE upswings from worst mid-term result” as Rob Ford. Of course, if you do that you notice that the only time an incumbent had a *negative* upswing was in 1997 which is the last time the Conservative Party were headed towards the end of governing megaterm with euroskeptics as the dominant disruptive faction in the coalition.
A common model is to say that the 1997 result/negative Tory “upswing” was due to specific properties of Tony Blair. But an alternate model would be to say that it was due to the historically unusual development of disruptive Tory radicalism. If current straight-line projections look rough for the Tories now, how much worse do they look if take 4.3% off them?
This is just speculation. In general, I like simplest models with the simplest, crudest assumptions and I’m dubious about This Time It’s Different claims. I just also think it’s reasonable to crank down the level of confidence about that in line with the amount of evidence that Maybe This Time (^And The One Before^) It Is Different.
It’s hard[3] to know how much upswings are caused(!) by the corrective actions of political parties and how much just an automatic mechanism operating within voters - but the part that does depend on party action obviously turns on the capacity and willingness of the Conservative Party to act.
And there are at best question marks about the institutional capacity of the party to react - I’m particularly struck by the sense in which people in the party seem to be acting like the driver who just assumes the car must contain an anti-crashing mechanism rather than considering that *they* might be the anti-crashing mechanism.
ERROR: Misread Rob Ford’s chart - the only ‘negative incumbent upswing’ is not Major govt in 92-97 but Blair 97-01! Cons don’t do well 92-97 (3rd lowest upswing at 7%) but they do worse in 1974feb and 2019.
[1] ElectoralCalculus.co.uk doesn’t include specific constituency tactical voting explicitly, but might catch some of the effect implicitly.
[2] Obviously Truss managed to have an immediate and lasting impact on the UK economy so we’re not talking about distinct and non-interacting things, this is just a convenience distinction
[3] For me, a person following political science publications at *at least* one remove!