|
|
 Originally Posted by Poopadoop
My understanding is that they try to, but given those adjustments involve previous elections that aren't under identical conditions, then how successfully they can possibly do so is hard to guage until the event in question has already happened.
Even then these modelling methods have an issue with being unfalsifiable, at least until they predict "X has 100% chance of winnng election". In every other case where the favourite loses, they can say they didn't say it was impossible, just unlikely.
The farthest I would be willing to go as one of these guys would be to say "If the polls accurately capture voter behaviour, based on our simulations X has Y % chance of winning." But then no-one would visit my site because I admit I'm not actually omniscient and that my assumptions are potentially fragile, and so I wouldn't make any money as an honest election modeller.
Ah, ok, yeah, I mean I think we essentially agree-- I just can't not say something when it seems like someone is claiming "the weather man is always wrong." The biggest issues seems to be that poll respondents are self selecting.
|
Reply With Quote
