It's not every day that Nate Silver, a liberal blogger and one of the most respected minds in political statistics, decides that a vast number of polls out there are inaccurate. The basis for his rejection of so many polls is simple: they don't call cell phones.
"The percentage of people who have replaced their landlines with cellphones has climbed at a remarkably steady rate," begins Silver. When pollsters call only landlines to ask questions, they wind up excluding cellphone-only households. This oversight has a big effect because "cellphone-only households ... tend to be younger, poorer, more urban, less white, and more Internet-savvy. All of these characteristics are correlated with political viewpoints and voting behavior."
Pollsters try to combat this problem by "weight[ing] their polls by demographics," Silver explains. This weighting is "something which they need to do anyway, since polls are subject to many forms of non-response bias (for instance, it's harder to get men on the phone then women)." But this solution isn't good enough to resolve the cell phone problem, he argues. Pollsters often don't weight polls "by characteristics like urban/rural location or martial status, which are predictive of both cellphone usage and political beliefs," for starters. Silver also points out that "one may encounter problems when upweighting from very small subsamples." Pollsters are not going to get very many young adults on the phone at all--can they extrapolate from the few that they do manage to get a hold of?
Calling cell phones is quite expensive, acknowledges Silver, but it just might be necessary to produce accurate polls.