Reply to post:

From landslide to buried alive: Why 2017 election forecasts weren't wrong

I ain't Spartacus Gold badge

Yes. That was a point I was going to make, but got sidetracked. The polls in the US weren't too bad either. Although some of the state level polling wasn't so hot. The Dutch polls recently were about right as well.

To be fair, it's always hard to know whether the polls are accurately following the voters changing their minds. For example, as polls tend to be sampled over 2 or 3 days, and you can't publish them on election day after 7am, polls are always crap at picking up late changes in voting intention.

They're also not good at showing tactical voting when an election isn't due. As although they ask "who do you intend to vote for at the next election", most voters seem to interpret that as what party do you currently support. Some pollsters ask twice, once to capture that, and then a second time concentrating on the voter's constituency, to try and pick up tactical voting trends.

But for some reason UK polling is rather off at the moment.

Partly it's a problem because the pollsters still do joint Welsh, English and Scottish polling, rather than large enough samples in each area to have a decent margin of error. And voting patterns are totally different.

Partly it's because our turn-out is changing quite a lot at the moment, with GE turnout going up, and the two recent referenda being very well attended. The AV one, not so much.

But in general, I think the pollsters learned the wrong lesson from 2015. They tried to correct their samples statistically, even though they know their samples are bad. Because it was too hard to get new, more accurate, samples. And to be fair, the post-mortem final report was only issued a couple of months ago, as they thought they had time until the next election.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon