I think it is a huge mistake to equate "polls that tend to be biased towards Candidate A" with "this institution favors Candidate A's policies".
The former can occur simply because of the assumptions that your model makes. Bias in "statistical bias" does not mean the same thing as "political bias".
The former can occur simply because of the assumptions that your model makes. Bias in "statistical bias" does not mean the same thing as "political bias".