Liberal media bias in American news: fact or fiction?

An infographic from the 4th Estate

The commonly held view that mainstream American news has a ‘liberal bias’ is a myth, according to new research.

A study by, released on August 9, found that in coverage of the 2012 US presidential election, a greater percentage of negative comments were directed at President Obama and the Democratic party than their Republican rivals.

The study took into account news reports from publications ordinarily considered to be Democrat-friendly, including the New York Times, the Boston Globe, the Washington Post, CNN, NPR and MSNBC, as well as Rupert Murdoch’s Fox News, which is regarded as being Republican in its political leanings.

Between May 1 and July 15 this year, Republicans were quoted more frequently Democrats, and, according to the research, negative coverage for Barack Obama was 17% higher than for Republican candidate Mitt Romney.

In the study, CBS News was found to have been negative towards Democrats 63% of the time compared to only 33% for Conservatives. The Washington Post was slightly more balanced with 61.8% negative reporting of the Democratic Party. USA Today was found to be negative 59.6%, the New York Times 57.1%, and CNN 56.7% of the time. The only news sources that were balanced towards positive reporting of Democrat or ‘liberal’ views were NPR which was positive 51% of the time and MSNBC 63.4% of the time.

Michael Howe, chief technology officer for 4th Estate said “Media bias is certainly the perception, but it’s based on a lot of anecdotal evidence and people talking about ‘What I think …’ We hope to change the nature of the debate.”

The 4th Estate project describes its mission as replacing “anecdote and speculation about the 2012 election coverage with visual intelligence based on statistical analysis.” Further information can be found at

This entry was posted in News and features and tagged . Post a comment or leave a trackback: Trackback URL.
  • The Daily Satire

    The idea that a machine can understand the nuances of meaning enough to decide whether or not a story is biased is absolutely laughable. People are so naive when it comes to technology – its not magic and it does have limitations. Computers are great at maths, but not so great at literary interpretation.

  • 4th Estate Project

    Two points. RE: The Daily Satire – our technology is an AI-human hybrid, so we do use human feedback to ensure data quality and integrity. RE: The primary story – it would be wrong to assert that we are claiming the ‘Liberal Media Bias” is a complete myth. We believe it is a much more complicated story than it is presented in the culture – hence, the constant battling and heckling over the topic between liberals and conservatives. Our goal is to be able to put some quantitative research behind the issue, so as a society, we can have more enlightened and nuanced discussions on the topic.

  • The Daily Satire

    Thanks for the reply, its good to know that you do have human input as well. I still wouldn’t trust it though. For example, your methodology says: “Explicit statements of praise or criticism toward a candidate are marked as Positive or Negative to that candidate, respectively.” Which seems to suggest that a communist criticising Obama for not being left wing enough would be treated as a negative story about Obama in the same way that an Obama supporter criticising Mitt Romney would be marked as negative about him. So they would have 1 ‘point’ each, but there would still be a liberal bias.

    Also I would be interested to hear how your system attempts to deal with satire.

    • 4th Estate Project

      Let me address your criticism and your question with the same ‘stone’. We think of our ‘sentiment’ much more along the lines of perspective than sentiment. Meaning all of our sentiment assignments are a function of a particular point of view – not some difficult to pin down, agree on, or assess notion of universal negative or positive sentiment.

      Now if you are going to think along the lines of perspective, then the source of the content (who is speaking) becomes a critical feature. In fact, we use this feature for a wide assortment of tasks including sentiment analysis. In this way we handle satire, sarcasm, and irony by assessing the source both for repeated behavioral patterns (Steve Colbert) and for one time inconsistencies with previous patterns both for the individual and for the group characteristics that that individual represents. If a feminist woman referred to herself as “one tough bitch”, this would most likely be inconsistent with multiple groups she might be associated with (social advocacy, women) and would be treated as sarcasm and positive toward a feminist perspective whereas if Rush Limbaugh were to say the same thing, his group characteristics (agent provacateur, men) would probably reinforce this as sarcasm and negative toward the feminist perspective.

      I am not sure I am entirely following your argument in regards to the communist and Obama critics.The notion of a Liberal Media Bias is so much more complicated than it is usually portrayed in the periodic arguments over it, and I am not overly interested in it per se. I am more interested in having deeper, data informed discussions about the patterns we find in coverage – both in Election coverage and across the other important topics of our times. So my response to your example would have nothing to do with media bias. Instead, I would say that our platform captures rich detailed data about news media amplification patterns (who is criticizing whom, as well as countless other data insights); that these patterns shift significantly over time; and that it is important to society to understand these patterns. Our goal is to create an authoritative platform that can start to build consensus on a host of topics such as media bias, etc.

  • Pingback: בוטוקס ללא כאבים

  • Pingback: Yes, the media are massively biased towards Obama. But the real problem is the obsession with opinion polls – Telegraph Blogs

  • Facebook

  • Twitter

  • How to build a local website

  • Twitter