Yes, I know there is a main story about this, and I started this as a comment to it, but it became too verbose.
There is a story here about bias in Rasmussen polls, yet Rasmussen was found to be the most accurate Presidential Election poll. I would like to explain how it is that both statements are correct. To understand this, we must understand the key difference between an election poll and a non-election poll.
The election polls, polls that predict the election outcome, have a gold standard that they can be measured against. The election itself. A pollster risks losing credibility biasing this poll.
Then there are the polls that have no gold standard to be measured against. THE VAST MAJORITY OF POLLS FALL INTO THIS CATEGORY.
The most notorious is the Presidential Approval poll - there is no nationwide referendum on whether you approve of the president or not. Remember, in the presidential election, the ballot does not say "Do you approve of Barack Obama's performance as President". A vote for him does not necessarily mean approval (you could find him the lesser of two evils), and a vote for someone else does not necessarily mean you don't approve (you might prefer another candidate).
Another example, the generic congressional ballot. We don't vote generic congressional ballots, we vote for specific candidates in specific districts. The TOTAL number of people that vote Democratic or Republican for Congress is irrelevant and rarely reported. It's the number of people that vote Democratic or republican in a DISTRICT. It is entirely possible that more people can vote for one party, yet the other party wins more races. In fact, that's how gerrymandering works.
Another example is issue polls. We don't have nationwide referendums.
You cannot be called for bias in these polls because there is no gold standard to measure against.
Rasmussen's non-election polls are HEAVILY biased. He leverages his record with the election polls that are conducted a few days before the election (that can be measured against a gold standard) to give credibility to his non-election polls whose accuracy cannot be measured.