Reply #12: it's a bit sketchy so far

[OnTheOtherHand]

Averaging the pollsters together the way he has is seriously apples-and-oranges, especially when the number of polls per pollster ranges from 3 to 184. There's no good way to solve this problem; we just don't know enough.

Pew has generally concluded that properly weighted landline surveys have provided similar results on political issues as dual-frame surveys that include cell phone users. That may no longer be the case, but I haven't seen strong evidence one way or another.

Let me be clear: if one just does landline calls and doesn't weight them at all, the results will be biased.