News outlets less biased, more similar than perceived, study suggests

Many viewers shun one news organization or another for a perceived conservative or liberal slant, but a recent study suggests these outlets have more in common than generally believed.

The study, led by UMSI assistant professor Ceren Budak, looked at how 15 major news outlets in the US select and frame political issues. Budak and her co-authors, Stanford University assistant professor Sharad Goel and Microsoft Research senior researcher Justin M. Rao, combined machine learning and crowdsourcing to enable investigation at a larger scale than previous studies on the topic. 

The researchers first used supervised learning algorithms, which employ a dataset of classifications gathered from human judges online, to identify tens of thousands of articles related to political events. They then went online again to recruit hundreds of human judges, who classified a random subset of the articles based on topic and ideology.

The resulting ideological ordering of news outlets was consistent with previous research—and popular perception. Organizations like Breitbart News Network, Fox News and the Wall Street Journal fell on the conservative end of the scale, while those such as Daily Kos, the New York Times and the Huffington Post leaned more liberal.

However, the team found that the news organizations are actually much more similar than conventional wisdom would have it. Aside from political scandals, there was little difference in the presentation of topics at the various outlets, with neither Democrats nor Republicans coming off particularly favorably or unfavorably. Furthermore, again aside from political scandals, each major news outlet selected similar stories to cover over a broad range of topics generally unrelated to its ideological position.

The team also found that when there is ideological bias in coverage, it generally comes not from advocacy for a preferred party bus from criticism of the other party.

The article, “Fair and Balanced? Quantifying Media Bias through Crowdsourced Content Analysis,” appeared in the Public Opinion Quarterly’s 2016 special issue on party polarization.

Posted May 2, 2016