Quinnipiac last polls for OH, VA, and FL tomorrow morning (7 AM)

[ajb]

wow 'Seriously' you are a bit thick aren't you?

No, actually pollsters call numbers randomly and amass a pool of adult responses.  They then reweight the poll to census data. So if they get 60% women responding they adjust the weight.   Bear in mind that over 90% of respondents do not agree to take the survey. This issue of response rate has become a big problem. There is some evidence that there is some response bias after certain events like conventions where those of a particular political persuasion are more enthused and more likely to agree to be polled. But again the pollster does not map anything out before they begin calling except to make sure their calls are balanced geographically.

The fundamental misunderstanding with the new attack on polling is a belief that pollsters predetermine what the party ID makeup will be. The only pollsters who do that are Rasmussen and the Battleground poll.

I am not talking about party identification. Most pollsters do not touch party ID. However, do adjust adjust polling for race, gender, age, etc. If they use solely census data for that, their methodology is likely flawed.

Turnout does not equal census data as all races, ages and even gender do not vote equally. Women vote more than men. Whites vote more than Hispanics. Old people vote more than young people.

At the end of the day, no matter how many phone calls you make, the data is never perfect.

There has to be some sort of judgment call made and the data has to be raked to mirror some reasonable expectation of what the electorate will be.

They use census data to weight their sample of registered voters according to race, age and sex, typically. Then they ask a series of questions to determine how likely a particular voter is to vote -- questions ranging from "how excited are you about voting?" to "do you know where the polling place is in your neighborhood?" to "did you vote in 2008/2010?"
The likely voter numbers then come from the subset of registered voters who made it through this likely voter screen. You'll not that age, race and sex do not play a role here. So if some particular demographic is unlikely to vote, the typical pollster doesn't weight them less because of that, but individual voters from that demographic who fail the likely voter screen won't show up in the likely voter numbers.

[ajb]

There is no real reason for undecideds to go towards Romney. I think a lot could depend on if more Romney leaning undecideds show up than Obama leaning undecideds. Now there is a possibility of that occurring.

Turnout will be crucial, of course, but I'm not sure that's the end of it. Don't you think late undecideds could heavily shift given the painfully lingering economic problems and general perception that Obama was a disappointment in delivering the sweeping changes he promised (even embodied)? I understand there are contrary Democratic arguments to be made (e.g. economy finally now showing tangible signs of growth, obstructionist GOP fillabustering, Romney would be worse, etc), but considering the typical late undecided at this point was likely an 08 Obama voter who remains soured out of disappointment with the economy and percieved lack of sweeping change, couldn't they realistically largely abandon Obama?

Nothing's off the table till all the votes are counted. But every day and every poll (and every early voter) makes it just that little bit less likely.