There’s a really useful rule in psephology which my colleague Stephen Bush likes to remind us of. It’s called Twyman’s law, and it goes: “if a poll is interesting, it’s probably wrong“.
As it becomes increasingly clear that Scotland has voted No, it’s worth bearing Twyman’s law in mind. Because there was a really interesting poll a couple of weeks ago – the one which showed that Yes was in front. And, lo and behold, it looks like it was wrong.
(Still plenty of time for this to go awry, of course, and the result in Dundee has brought the running totals a bit closer, but right now it looks like the good guys have won.)
M’colleague Dan Hodges has been complaining at length that David Cameron has torn up the constitution of Britain at the sight of that one, rogue poll. And he’s right. He says that what must come out of this is a new settlement in which England gets the same set of new powers that Scotland does. I think he’s probably right about that, too.
But more importantly than that, the thing that really has to come out of this is: we’ve got to stop paying attention to outlier polls. Yes, even if they might make a good headline or tell us what we want to hear or anything else. There are dozens of polls in each election campaign. If they give the right result to within a 95 per cent confidence interval, you’d expect one poll in 20 to put the wrong side ahead. So when one poll out the 20 or so that have been taken shows one result, and the other 19 show the other, you can pretty much guarantee that the one is wrong. (I’m not just saying this after the fact, by the way.)
This is obvious. But somehow it got the whole of the UK gibbering with existential fear. Still, I’m sure we’ll have learnt our lesson for next time.