Most if not all of the polls conducted lately are telephone polls. I have a cell phone, and most cell phone services don't publish phone numbers of subscribers in databases accessible by third-parties. Moreover, I'm part of a growing demographic, the members of which have ditched their land lines entirely in favor of cell connectivity exclusively.
If my own experience is any indication, much of this demographic includes younger individuals who are comfortable with technological evolutions and paradigm shifts, or at least that of not having a "home" phone. Although I don't have any statistical evidence, I'd hazard a guess that the majority of this demographic is probably more liberal than conservative.
But by virtue of the fact that cell phone customers' numbers aren't public, this demographic is not represented in the polls.
Although there are several other factors that prevent such polls from reaching individuals who comprise a fairly representative sample of the electorate, what I'm curious about is whether the collective political affiliation of this demographic is a substantial enough factor to really throw the results into doubt ... and whether this cell-only factor is offset or exacerbated by other factors.