Can't Control for Systemic Failure

You can control for a systemic error - i.e., if you know that "while cell-only voters were more supportive of John Kerry than voters overall," you can do the math to adjust a landline poll correctly.

But I certainly don't have your confidence about correcting for systemic failure. The landline poll is the pollster's principal means - other than exit polling, see below - of getting data. But a failure of landline polling - which I don't think has happened yet but will very soon - means that the pollsters won' t have the necessary numbers about relative biases among the cell-phone crowd to correct their data. Not only that, but as I said earlier it's not just about cell-phone only users. It's about people like me who routinely hang up on all pollsters, too.

And exit polling is very troubled in other ways. The reason is that people walk out of a polling booth in very different moods. Working voters are impatient, frustrated at the old volunteers who assisted them, and need to get back to work. It's the old, or college students, who'll have the time to approach the exit pollsters. And moreover they'll approach CERTAIN exit pollsters - it's been shown that college men will approach attractive young female exit pollsters in predicatably huge numbers while they'll blow past the earnest middle-aged collectors from a different agency. Those two pollsters will generally be asking different questions etc etc etc.

Bottom line being that polling has become a huge part of our governing process, and maybe it shouldn't be. It doesn't need to be at all; we have highly accurate polls every two years and we have mechanisms for people to lobby and contact their representatives. Why not ban public polling? I mean seriously? 1st-amendment concerns? HA! If people cared about the first amendment they'd have never allowed the monstrous McCain-Feingold law to be passed.