| gergroy said: are polls usually all over the place like this? I can't remember having polls that varied this dramatically in the last presidential race. For example, marist came out with a poll of iowa showing the president up by 9 points yesterday, and today PPP has a poll out showing Romney up by one point, that is a 10 point spread between the two. Now, since these polls supposedly have a margin of error of +-4 points, it is actually impossible for both of these polls to even be that accurate. So, how do these polls help if they can't even be accurate within their margin of error? |
It's all in the voting models.
Some polls sample registered voters. While other polls sample likely voters.
Around 75%+ of eligable voters are registered by only 50-60% vote each year. So each polling company has their own formula and questions they use to decide what makes a registered voter "likely" to vote.
As such, the likely voter samples can vary widely depending on how it's done.
Likely registered voter polls favor Obama, while likely voter polls lean more Romney.
Which is why Romney got such a huge boost in some polls. Essentially after he clobbered Obama in the first debate... and got a lot of apatehtic people who were seen as unlikely to vote, excited and ready to vote.
Hence for example why Galup seemingly has Mitt Romney way ahead of Obama way more then most polls. Gallup has Romney up by 7 points among likely voters, but only 1% up on registered ones. (Or maybe Obama's ahead, i'm not sure.)
There are a LOT of enthusiastic pro-romney voters now. (That didn't exist previously).
While most Obama votes fall in the category of "Eh, at least he's not Romney."
