What do you mean by 'model' ~ ? Please define that for me. That term showed up in the discussion of 'skewing' and didn't make any more sense then that it does now. Statistically valid samples are organized in a number of ways ~ but the whole scheme depends on random sampling.
As you've heard people say, polling is as much an art as it is a science. It's really both things, and they are getting better at it even as reaching people is increasingly challenging. Many of these polling firms have a model for what they believe the area they are polling looks like. This is what they use to weight polling. This is the reason they can poll so few people and still be quite accurate. The information for this model comes from many sources including additional polling over time, previous exit polling, who is a likely voter, census information, etc.
This is just an example. I'm just making these numbers up so it makes sense. So lets say a pollster reaches 200 people in Detroit to find out how folks are voting for the next mayor, but 100 of those 200 respondents were white and 100% of those whites said they were voting Republican. The pollster would know this sample is not representative of the city based on their demographic models. Blacks are about 75% of Detroit, not 50%. And whites in this example are 25% of the population, not 50% So with this information they re-weight the poll to more accurately reflect who is in the city so the result is more accurate. From the raw numbers they would know that blacks were voting 100% Democrat and whites were voting 100% Republican. The raw numbers in this example would lead you to believe it was going to be a 50-50, but the pollster uses their model of the real electorate to re-weight the raw numbers and yields a poll result of 75% of the vote for the Dem candidate, and 25% for the white candidate. Now just think about that on a bigger scale. They can re-weight for all manner of things. The goal is to get a natural sample from the raw caller survey, but they re-weight to the extent they feel they have to in order to produce an accurate poll.
This last election cycle pollsters were finding a significant D+ electorate. I am sure you remember all the stories claiming the electorate was not going to be anything like 2008 which was D+7. People that did not believe the polls were claiming it would be more like D+3 or even R+1. Well, it was D+6 - which was very similar to the D/R turnout in 2008. Conservatives didn't think so many blacks would turn out, or Hispanics would turn out, or young people would turn out, single women would turn out, etc. But they did. They turned out in sufficient numbers to elect Obama by 5 million votes. Now remember D+6 doesn't mean every Democrat voted for their party, and it doesn't include the independent numbers - but it gives you a sense of what the electorate that turned out looked like. The closer a pollster was to D+6 model, the more accurate they were likely to be.
The 'skewering' crowd believed in 'models' that would enable a pollster to keep sampling until he had the number of responses he wanted ~ and then, presumably, they'd show their customer the responses they received (albeit only those that gave the result they preferred to push).
I have no idea what you're talking about. This sounds like some kind of conspiracy theory. When we talk about the "skewing" crowd most folks are referring to the people who believed polls could be "unskewed" (such as Dean Chambers).
Now the "unskewed polls" nonsense and folks who believed in that bit of silliness thought they could take existing polling and re-weight it reflect a different electorate. They rejected polling that showed a D+ turnout model and tried to rework the polls to show an R+ turnout. They were wrong. The bulk of the media polling turned out to be correct. Their "models" of what the electorate looked like were mostly right on target.
The last week I've been called for a dozen polls ~ which isn't so strange ~ I'm their token real Conservative. They'll get my response and toss it out. In the meantime I've got their phone numbers and can trace a commercial call back to its source.