I tend to doubt - doubt - in polling. There have been MANY times in the past when polling has been a poor bellwether of the outcome, because circumstances went outside of the normal parameters assumed by pollsters.
But I also doubt they're accurate when over the same period there are *massive* differences in polling.
The news shows tend to believe the RCP average. I think that's a good start, but I would also throw out outliers which are consistently on the edge of reason.
When one poll puts Trump ahead by 4 and another puts Hillary ahead by 10, at least one of them is wrong but likely both are very wrong.
I do however see indications that there is probably more Trump support than is measured.
The problem is the focus on "Likely" voters. Pollsters don't grab names out of a hate and do guesswork, because you can't set up your sample unless you know SOMETHING about the person you're asking.
Despite what you think, polls are not random, nor SHOULD a poll be if it has any chance of being accurate.
So how do they know? Well, *past* data, for one. Voter rolls, for another. Because it's pretty much a waste of time to poll people who aren't even registered to vote.
Since a large portion of registered voters STILL don't show up at the polls, they go with "Likely Voters" which is usually very accurate. "Likely" being defined as someone who has voted recently and often
(often in the sense of, in each cycle - not illegally).
But there's a problem - "Likely" voters completely miss two groups of people - NEW voters and people who have chosen to exercise their right to vote who haven't been voting.
They're pretty much out of the sample completely. If they should be significant, the polling will be all wrong. If not, their numbers will be overwhelmed by the rest of the voters.
Some polls use samples that reflect either registration enrollment or past experience - for instance, if 32% of the voting population is registered as Democrat, that's what the sample should be.
On the other hand, if 34% of the persons who VOTED were registered Democrats, they may go with THAT. But what happens when a significant surge occurs in any demographic?
There's no accounting for it. I think pollsters KNEW that in 2008, many voters were going to show up to cast a vote for the first black President - and also accurately noticed that not nearly as many
showed up to re-elect him, although it was enough.
But no one has a clue this time around.
WILL we see a Reagan type surprise turnaround, or will the pollsters be spot on? Honestly, I think election night will have a LOT of surprises. A LOT of them.
And they may not favor YOUR candidate, whomever he/she may be.