Consider this:
You get a call on your cell during the weekday. It's from an unknown number OR a 1800 number. What do you do?
You going to ignore it. If it's someone important they'll leave a message.
Say you answer it, and the stranger on the other side says "Hi, I'd like to ask you some questions"
Again, your answer is likely going to be "end call". You're not wasting your minutes on some random caller, or worse, a telemarketer.
But the fact is, this is how polls are done. They operate the exact same way as telemarketers and the response they get is no different.
Basically all they get are answers from landlines, which are usually older people. So unless your poll is "Who would win the election if only seniors could vote?", your poll is going to have a flawed sample and the results are meaningless.
Even worse, if you weren't neutral and WANTED to show a particular result, all you have to do is just call a red/blue area; it's no secret where they live, that's how gerrymandering works. Notice how polls rarely tell you where they got their sample from? That's one of the reasons. CNN got caught doing this, that why they have to say on every poll "Our sample is slightly skewed towards liberal voters"
Internet polls are worse for obvious reasons like how easy it is to manipulate the results by voting multiple times.
So it makes me laugh when people talk about poll results. The only survey results I would trust are those which you take legitimately random people and compensate them for their time. But no election poll can do that, so we're stuck with either "neutral" polls with flawed samples or "tainted" polls that only exist to paint your favored candidate winning.
