Quote:
Originally Posted by MysticCat
Plus, as been noted elsewhere at GC, polls still just call landlines. They're missing the growing number of people (mostly younger people I would guess), who have ditched landlines in favor of cellphones alone.
|
This is becoming less of an issue, as far as I know (and this is only through reading industry mags and articles), although it's being replaced with a new problem - more and more polling is happening online, through "banks" of volunteer phone numbers including cell phones, or using on-site polling.
This simply replaces one form of selection bias with another, but since we really can't determine exactly what effect that bias will have, it's safe to take all polls with a grain of salt this early.
I guess my point is that if the polling is done via landline alone, I think it's fairly safe to say Obama's numbers are slightly underreported (even accounting for the fact that a huge number of the college kids on his side simply won't show up). Other polling, however, is just going to miss period, and completely ignore the error bars because they're making assumptions from 1983.
The two major political polling companies are not exactly known for being thorough or scientifically sound - they are, however, known for charging an assload for information, taking large payments from political committees, and firing out new polls as fast as possible.