Jacob Kramer-Duffield thinks

As Nate Silver reported the other week:

Pew Research issued a study suggesting that the failure to include cellphones in a survey sample — and most pollsters don’t include them — may bias the results against Democrats. Pew has addressed this subject a number of times before, and in their view, the problem seems to be worsening. Indeed, this is about what you might expect, since the fraction of voters who rely on cellphones is steadily increasing: about 25 percent of the adult population now has no landline phone installed at all.

Clearly, this is a major problem in survey research — and one that, sooner or later, every polling firm is going to have to wrestle with. What isn’t as clear is how much of a problem it is right now.

He goes on to cover several of the key issues that are specific to this case and time. but I’ll focus for a minute on the larger-scale issues. I’ve talked about some of these ideas before, and indeed we were talking about cell-phone undercounting on the Dean campaign in 2003 and Kerry in 2004 (not, as it turned out, the biggest problem in either of those cases). But as Nate says: this is a major problem that sooner or later everyone is going to have to deal with, it’s just a question of when.

Will that be this year? Hopes of Democrats aside, probably not – or at least, not provably, given the substantial problems in constructing likely voter screens this cycle. But when the dust settles and post-election analyses are done, all the pollsters are going to have to take a good, long look at their numbers and at results, and through the lens of Pew’s results, begin to (or further) adjust their approaches. Because by 2012, an even larger share of the voting-age population will be living in cell-phone-only households, due both to continued abandonment of landlines by older demographics and the maturation of millions more who’ve never had a landline (and mostly never will).

This isn’t an impossible problem, but it’s also not solvable with a silver bullet. Polling, like any sort of research, is going to need to become more multi-modal, faster-thinking and -responding, in order to reflect anything like a generalizable sample of the population. This means working harder, thinking more and understanding better the ways in which all different sorts of people use different kinds of communications technologies.