some other answers suggest political bias is not objective, probably because they are confounding "preference" with "observation of preference (bias)". Ambiguity in natural language obscures the issue, but objectivity is possible. Preference is subjective, but measurement of preference (i.e. observed bias) is objective. Consider
supremacy of the color red (a subjective qualial preference), in contrast to
current preference (i.e. bias) for red amongst infants (difficult to measure, but still objective). Likewise,
favorability of party X is subjective, but
observed favor for party X, though hard to measure, is still objective. Also, objectivity is not confined to non-stochastic phenomena. In cases where objective phenomena are difficult to measure, clever scientists will search for plausible proxy metrics that are easier to measure, as is the case with the following:
Please see this publication from UCLA, which estimates ADA scores for major media outlets. According to the introduction:
...we count the times that a media outlet cites various think tanks.
We compare this with the times that members of Congress cite the same think tanks in
their speeches on the floor of the House and Senate. By comparing the citation patterns
we can construct an ADA score ... These findings refer strictly to the news stories of the outlets [and not] editorials, book reviews, and letters to the editor from our sample
The publication also contains a section Previous Studies of Media Bias
It's worth considering possible shortcomings of the article's described methods. Firstly, you have to be willing to accept that think-tank citation ratios that coincide with those of congressional members are an acceptable proxy for media outlet bias. This requires that we accept other assumptions prima facie, like a consistent manner of think-tank citation behavior across the political spectrum, etc.
Still, this is, to date, the most serious analysis I've seen. As regarding objectivity, the described methods seem sufficiently detailed so as to be repeated, and hence the results independently measurable and verifiable. I've intentionally omitted their conclusions - if you want to know, read the article ;)
Here are two other studies I found:
- WHAT DRIVES MEDIA SLANT? EVIDENCE FROM U.S. DAILY NEWSPAPERS
- Fair and Balanced? Quantifying Media Bias through Crowdsourced Content Analysis
The first tries to compare "unique phrase usage" analysis of media outlets to members of Congress. I'm skeptical of the second, which tries to recognize bias with machine learning techniques, trained with data procured by a set of volunteers. They might have sufficed with just generalizing the volunteers' results since the AI can only ever (likely) perform as well and no better than it's training set provided by flawed humans.
And here is a data source of news chyron archives used in some more serious analyses.
Finally, an article probing the meta-concerns of determining media bias.