Here is a poll prediction from 2004, a year in which PUMAs and the Bradley effect weren’t potential considerations in the polls:
Final Pre-election Harris Polls 11-02-04: Too Close to Call but Kerry Makes Modest Gains
The Harris Poll ^ | 11-02-04 | harris interactive
Posted on Sunday, November 02, 2008 9:10:40 AM by grandpa jones
The final Harris Poll from Nov 2, 2004...
Another piece of evidence pointing to a likely Kerry victory is that online Harris Polls in these large, key states, which may well determine the Electoral College result, all show modest Kerry leads. However, all these leads are within the possible sampling error for these surveys. Assuming the forecast is correct, Kerry is likely to win all three large states, and almost certainly the White House along with it. The sample sizes were well over 1,000 likely voters in Florida (1,433), Pennsylvania (1,204), and Ohio (1,218).
(Excerpt) Read more at harrisinteractive.com ...
http://www.freerepublic.com/focus/f-news/2122618/posts
And you have to remember the dems stand for nothing so they will always bend with whatever poll there is not matter what it says. They can always lie and change their minds later.