But out of the people who claimed to be conservatives, a quarter of them said they would vote for a Democrat.
Its an old one, but its true...
There are three kinds of lies: Lies, damned lies, and statistics.
Pollsters are true and accurate, but only within the parameters of what they have been asked to poll. Its the interpretation of those results that causes the problems. 798 is far too small a poll to be statistically significant. Theres no indication of who was polled, where from, what time, what day, what questions were asked, what order...all of which can alter things. Then there's this assertion that a quarter of all "conservatives" would vote democrat - how are they defining "conservative"? I am sure folk on this board would have very definite opinions - but certainly it is a definition that is interpretable.
I would reccomend the fine book "How to lie with statistics" by Darrel Huff. In modern society it is absolutely essential that everyone knows how to talk back to a statistic.
In the meantime...fear not! The only poll which really matters is the election.
Which I don't believe. Do you?