(no subject)

[beccaelizabeth]

I just read a thing in the Guardian about polls
and found out they're still doing them by phoning people?
they reckon the main problem is persuading people they're jot PPI calls so they don't get hung up on.

but who answers the phone to strange numbers anymore?

no wonder the polls are going more and more wrong.



also it said if they can't get enough answers they do 'weighting'
which means counting each answer they did get as for instance 1.2 answers to make up for it.
... if we'd done that in stats class at school we'd have been epic failed.


the other set of people they have do online polls for small money.

again, that's a very specific and self selecting group of people.

I don't have any suggestions for getting a more representative sample, but I don't see how those methods can do it.

Comments

[elf]

Phone polls are increasingly subject to errors, but we don't have anything better to use for a replacement. Self-selected online polls are worse.

Right now, they're working on history and inertia: Answers from this area skew that way, so we can adjust the results in that direction. And that gets mostly-accurate answers. Over time, that'll be increasingly less useful, as small shifts in opinions move away from the historic norms, and as various sub-issues break away from the associated party platforms.

We may need to replace them with boots-on-the-ground in-person polling, which is substantially more expensive AND brings in the whole issue of "how do you get pollsters that people will talk to?"