Survey USA - Ohio - Obama +3

I'm sure that's how things will play out on election day, seeing as he hasn't managed to pull beyond 5% of that vote in any poll of any value.

Couldn't be more than, say, a couple of hundred thousand or so, right?

Oh, wait...

http://www.dispatch.com/content/pages/data/community/2010-ohio-population-by-race/2010-ohio-population-by-race.html?appSession=652330012803537

I wouldn't expect for that sample to produce a very accurate estimate. The margin of error must be over 20%.

Romney isn't going to get 22% or 30% or 45% of the African American vote in Ohio. Some polls have him receiving 0% of the African American vote! Most have him at 1%-8%, with 8% being a high-range figure. I'm excluding, of course, out-and-out con artists like Gravis. So we know with as much demographic certainty as one could probably have that Obama will receive between 90% and 95% of the African American vote.

So, junior high school math problem: What would the numbers for this poll be if Obama received 95% of the African American vote rather than 70%? You purport to be a numbers cruncher and poll junky. You do the math.

Honestly, i've been keeping count and i have NEVER seen a single person of color (no black, asian, native american, etc).

I am surprised i see women in his crowds and so logically, if he is pulling 40% of the woman vote then based on the # of women i am seeing in his crowds i should look to see half that size as a representation for people of color. Where are they?

Survey USA really needs to do some better # crunching because if they are claiming 22% they are going to be way, WAY off on election night.

Obama up 58 to 39 on people already voted.

Which means Obama went down 1 point from their last poll, but Romney didn't go up.

Could just be statistical noise and it's a good sign that Romney still can't break into a lead.

Rasmussen is probably nothing. 1 point or a tie is simple statistical variance.

I would say that Obama is up about 2 to 3 points in OH. Rasmussen and PPP just don't have the track record in accuracy that Survey USA and Quinnipiac have. Survey USA also is a robocaller but it seems to get accurate polls. In 2008 and 2010 it was very good.

And if the AA vote is much higher then in this poll (which it has to be, this was too small a sample) then Obama looks pretty good in OH.

I don't think CNN has had an Ohio poll out since after the first debate.

These polls are irresponsible.

Has there ever been a republican candidate in the last 15 years that has gotten 22% of the black vote in Ohio?

The answer is simple, NO.

So this poll showing Romney with 22% of the black vote skews their overall horse-race numbers THUS showing a tighter race than it is.

Many people will claim that it was a small sample of black voters and so that means a larger MOE, YET I cannot imagine the MOE to be as big as 17% because we know for a fact that Romney will not exceed 8% of the black vote.

It should be when a pollster finds that their raw numbers indicate such unbelievable trends in a certain demographic (one that shows black voters pulling to Romney in numbers that have never seen before even this season), THEN that pollster should either weight his poll to reflect numbers that are more believable OR should try to reach more black voters to get a better assesment of their voting preferences.

It seems all polling companies are having a certain problem with each poll they produce,

whether it be a poll that undersamples Hispanics,

or a poll that shows way too much support for Romney among blacks,

or a poll that heavily samples older voters,

or a poll that has really tight "likely voter" model that eliminates many democratic voters, etc.