Just about everyone else has been short on facts, eg bongsaway just called polling crazy.
This is my view on polling at this point. I had statistics in college. My professor said there are lies, damn lies, and statistics.
Polls are probably one of those most inaccurate statistic out there. why? you can make a poll say anything you want.
In order to determine the accuracy of a poll there are several things you have to look at.
1. Is always always sample size. How many people were polled or asked. Let us take state polling. If I have a state with 400k people and I poll 1000 of them. My sample size is bunk.
It is simply not enough people out of the entire pool to conduct a fair poll.
2. Who was polled. This goes back to Dewey/Truman and Clinton/Trump.
If you look at most polls you will find that out of 1000 people they poll 435 democrats 325 republicans and maybe 240 independents.
So right off the bat we have a bias response in the polling data. then there are those people that don't respond or reveal who they are voting
for or they lie about who they are voting for.
My take on polling is thus. If you ask 1000 people what they think then those 1000 people speak for themselves. you do not get to project those 1000 people
onto the other 400k people that live in a state.
The other issue with polling is this as well. We do not have a national election so any poll referencing a national poll is pure bunk.
we have 50 individual elections that are mutually exclusive of each other.
so to really see who is going to win the election you have to break the polls down by state. ensure that your sample size is more than just 1000 people out of a state
that has 1m people or so you then have to ensure that you are polling the largest block of voters which are independents. then you might you might come close
to a 4% margin of error.