Are the polls skewed?

Something's not adding up.

McCain is slightly ahead of Obama in the polls? I couldn't wrap my mind around it until I began to question poll methodology.

In normal times, Gallup's polling methods, as explained here, should work fairly well. However, we do not live in normal times. We are in an economic downturn, to put it mildly.

I heard on the news that one in four hundred and sixteen homeowners will be facing foreclosure this month. I have heard that seven thousand homes foreclose each month. I have heard two statistics on the total number of homeowners that either faced foreclosure or who are behind on mortgage payments, but the one I can source is 6.82%.

Do you think the people behind on their mortgage payments or who are in foreclosure are answering their phones? How likely are they to vote for a guy who can't remember how many homes he owns?

Then, I read another statistic today: one in five households is either behind on credit card payments or over the limit on at least one card. That's twenty percent of American households. Do you think those people are answering their phones?

Gallup says that they randomly generate phone numbers--cell phones included, according to their FAQ--but if people aren't answering their phones, which is common with people who are behind on their bills and have autodialers calling them nine times a day or so, the statistics have to be skewed, do they not?

How likely do you think that the twenty percent of households where autodialers are constantly calling are answering their phones for Gallup? The polling company says they call back numbers where they haven't reached anyone, but I have trouble believing that they reach a significant enough number of these households to keep their polls from straying well outside of the margin of error.

The document from Gallup was written in 1997, and said that 95% of homes had a telephone (15% of homes use cell phones only, according to Gallup). How many homes have phones today? How many wrong or disconnected numbers do they get before getting enough good ones in their random sample? Are ten percent of the homes where bill collectors are also calling (inscessantly!) answering Gallup's polls? Fifteen? Five?

I think the methodology is no longer sound, because the probability of selection is not equal for all Americans anymore, and it is this probability of selection being equal that Gallup relies upon for statistical accuracy--as do all scientific polling organizations.