I have heard this as well sometime around the last election cycle. I think there was a study done showing the discrepancy between polls done in person on the street and those done over the telephone. I can't find anything about it right now and I can't remember the outcome.

The Man likes it this way. With the polls slanting Republican it is much easier to steal elections by a much wider margin, and will become easier as time goes on. (e.g. 2016 general population 65% Dem vs. over 60 year old dinosaurs with landlines 49% Dem)

/tinfoil hat off

It was reported about last time but didn't turn about to be much of a factor because the youth, who are most of the cell-only users don't bother to vote much. This time it'll probably be a bigger factor for at least two reasons: it's four years later and that many more people are cell only users; Obama has a huge use following and is planning a massive GOTV effort.

Disclaimer: I'm volunteering for the campaign and doing data entry for the Get Out The Vote system.

Luke Russert talked about this during his first report at the DNC. (This may be up on msnbc or on youtube?) It seems to be a much bigger issue this time - especially considering how many young folks & students seem to be supporting Obama/Biden.

There were huge numbers of young folk roaming the streets of Denver during the DNC. And many of them were either registering voters, protesting a cause or actually doing something. These weren't just kids sitting on their asses. It was a good thing to witness in person.

I've been really impressed with Obama's team and the amount of internet and phone texting they've used and encouraged.

I don't think it's that big a deal. The pollers are well aware of the problems of unrepresentative samples, and they weight their results accordingly.

So for instance if they have only half as many young people in a sample, they might double the weighting of those votes to make the sample more representative.

It's the more subtle effects that are harder to compensate for. For instance, people who have a generally conservative attitude might be more likely both to vote for conservative candidates and to stick with their landlines.

So, it's best to be generally cautious of polls. The "margin of error" they generally quote (usually 3%) is more of a minimum measure, just down to statistical noise. There are also systematic errors, so they're less accurate than they say.

I think it's definitely plausible. Although, younger voters could be split like regular voters. There may be more young liberal voters but younger voters aren't as likely to show up at the polls.

TheophilleEscargot said the same thing I'm trying to say, but better.

I know a considerable number of people in my generation / age bracket (mid 30s-mid 40s) with no landline. In fact, I'd say it's a small but statistically significant chunk of the white, educated middle class, middle aged, middle management batch of colleagues I work with on a daily basis. I myself haven't had a land line for the past ten years. So, uh... I dunno what my point is. Just saying.

Believe me, this is just ONE factor affecting the "reliability" of polls. As delicious as they are to munch on for any political junkie, they individually have quite a few issues. Every pollster, for example, has its own model for who is a "likely voter". If there's any reason for Obama supporters to be nervous, it's the historically unreliable "youth vote" (18-30s). Assuming you can reach them on the phone, they (for many, many years) say they're going to vote more often than they actually do. So it could be that these factors balance out.

Most important, though, is that the figures you are seeing are ALREADY massaged according to the pollsters' models to account for what they professionally (if they're not a politically-associated pollster) believe is the appropriate weight.