Absolutely, though I also believe it is not entirely the statisticians' fault. There have been major changes in communications technology which are probably demographically biased in some unexamined direction (switching to cell-only or even VOIP for voice communication), much of it since the last Presidential election, and on top of that people are less and less inclined to either talk to them or respond truthfully if they do. There may or may not be a bias in which group's partisans are more likely to be tapped even without those issues, for which of course the pollsters make the big bucks by supposedly taking into account in their models, but that all comes down to the validity of a legacy error-correction model with grafted-on and unproved tweaks for the current cycle at crunch-time.