Thoughts on polls and "accuracy"

[Spike89]

Saw an interview with a sheet-wearing self-avowed racist the
other day where he claimed to be undecided and it got me
thinking about the polls in general. Taking the dude at his
word, I'm willing to believe he hasn't decided to vote for
McCain--I can't believe he's really considering voting for
Obama. He'll almost certainly vote McCain, not vote, or choose
a "protest" vote. In other words, he isn't a true
swing voter. This got me thinking more about the polls and the
polls I've participated in.
  I've already mailed my Oregon ballot in for Obama, so I'm
obviously not undecided, in fact, I'm committed. However, my
Mom is still holding onto her ballot, intending to vote for
Obama. In her 70s, she doesn't like McCain or republicans, but
is not exactly 100% comfortable with Obama either. In almost
any poll, my Mom and I would both count for Obama, especially
if I'd been polled a couple days before sending my ballot. We
are not the same--Obama has my vote sewn up, my Mom will vote
for him, unless something changes her mind in the next week or
so.
  The point is that no poll lead is truly safe without
drilling deep into the statistics. Theoretically, if you had a
state that polls 50.1 for Obama, 49.9 for McCain, but that was
very polarized, i.e., everyone was like me and unshakable,
that would truly be a dark blue state despite almost perfectly
fitting the definition of a tossup. Likewise, a seemingly huge
advantage like we have in Pennsylvania, could be much less
than it appears if all (or even many) the Obama poll answers
come from folks like my Mom.
  As far as I know, none of the EV sites out there really
account for this. They talk about "states leaning"
red or blue, but it isn't the states that vote, it is the
people in them. 
  At this point, how many voters are like me, and how many are
like my Mom?

[Muttocracy]

1. some polls ask people whether they're decided, or if there's a chance they might change their vote

usually several percent, so I think Barack is beyond that factor in many swing states now.

[Spike89]

3. True, everyone I've taken has, but...

I don't see that information being factored into the various sites tracking EV. I really got to wondering why McCain might be leaving Colorado to pursue Pennsylvania and figured he must have a source saying that although his deficit is 10+ points, it isn't hard support. I can't imagine why else he thinks he has a chance there.

[MJkcj]

2. I wonder too

I live in Maryland which is solid blue but I have been canvasing in Virginia and I find it very mixed between Mccain and Obama. It feels very close. It makes me super nervous. I talked to a lot of people who say they are for Obama but unless they go to the polls and vote it counts for nothing. I think the GOTV effort is going to be CRUCIAL. I am hopeful, but take nothing for granted.

[FloridaGrl]

4. don't you believe there

can be a reverse of the scenarios you just gave? Where people don't disclose the fact they are voting for Obama for various reasons. There is a margin of error in polling, no poll claims to be 100%.

[Spike89]

5. Sure, I actually think Obama is further ahead

than the polls reflect--I think reverse Bradley effect, cell phone voters, and a huge GOTV effort will have a huge effect and aren't adequately being reflected in the polls. The reality is that most of the polls are based on this being a typical election, but it isn't. Typically, partisans (like us and trolls at FR) can be counted on and a fairly large number of "low info voters" that aren't really into politics, but vote for various reasons. Obama has energized a huge population that aren't traditional voters and there is no way pollsters can really anticipate turnout. In most years, about 50% of Democrats vote and although there are fewer Repugs, they tend to get slightly higher turnout resulting in the very close elections of 2000 and 2004. The pollsters try and get about half and half likely dems and repugs in their samples, but that isn't accurate if 60% or more of the Dems decide to vote this election.
Anyway, back to your question...margin of error is a statistical effect and unrelated to the issue I was talking about. I'm concerned that volatility isn't being discussed...I don't know if most people are "locked" into their candidate or not. If they are, we've already won big. If not, no lead is safe and we really need to work to preserve the lead.