Wikipedia’s news sources show a moderate liberal leaning

A recent study suggests that the news media sources cited across the English version of Wikipedia have a moderate but consistent liberal bias. This pattern appears to persist even when the factual reliability of the sources is taken into account. The research was published in 2024 the journal Online Information Review.

Wikipedia is a vast, collaboratively edited online encyclopedia that has become a primary source of information for millions of people worldwide. The platform operates on three core content policies: maintaining a neutral point of view, ensuring information is verifiable through reliable sources, and prohibiting original research. These principles are intended to make its articles balanced and accurate.

Because of these policies, the quality and neutrality of Wikipedia depend heavily on the external sources its volunteer editors cite. News media outlets are a significant source of these citations. The researchers behind this study sought to investigate whether the selection of these news sources introduced a political leaning into the encyclopedia, potentially affecting its commitment to a neutral point of view.

To conduct their analysis, the researchers began with a large public dataset called Wikipedia Citations, which contains over 29 million citations from more than 6 million articles in the English Wikipedia. They extracted the domain name, such as nytimes.com or foxnews.com, from each cited link. This process allowed them to identify the specific news outlets being referenced.

Next, the team enriched this data using two external rating systems. For political leaning, they used data from the Media Bias Monitor, a system that calculates a political polarization score for a news outlet based on the self-reported political leanings of its audience on Facebook. This score ranges from -2 (very liberal) to +2 (very conservative), with 0 representing a moderate or balanced audience.

For factual reliability, they turned to ratings from Media Bias Fact Check, an organization that assesses the accuracy of news sources. This service rates outlets on a scale from “VERY HIGH” for sources that are consistently factual to “VERY LOW” for those that rarely use credible information. By matching the domains from Wikipedia to these two databases, the researchers could assign both a political polarization score and a reliability rating to millions of individual citations.

The analysis of political polarization provided evidence of a consistent lean. The average score for all news sources cited in Wikipedia was -0.51, which falls on the liberal side of the spectrum. The distribution of scores showed that the majority of news citations came from outlets with polarization scores between -1 (liberal) and 0 (moderate).

This tendency was not confined to articles on political topics. The liberal-leaning pattern was observed across broad subject categories, including Culture, Geography, and STEM (science, technology, engineering, and math). It also appeared in articles associated with various editor communities, known as WikiProjects, from Politics and India to Biography and Military History. This suggests the effect is widespread across the encyclopedia.

The researchers then explored if this political leaning was connected to the factual reliability of the sources. One might speculate that editors favor sources with a certain political leaning because they perceive them as more factually reliable. To examine this relationship, they used a statistical technique known as multiple linear regression, which can help determine how different factors, like reliability and article topic, are associated with an outcome, in this case, the political polarization score.

The model indicated a complex relationship rather than a simple one. For instance, sources rated as “High” in reliability tended to lean liberal, while those rated “Very High” tended to lean conservative. At the other end of the spectrum, sources with “Mixed” reliability were associated with a liberal leaning, while sources with “Low” and “Very Low” reliability were associated with a conservative leaning.

This outcome suggests there is no simple, direct line connecting higher reliability with one particular political viewpoint in Wikipedia’s sources. The finding that sources with a liberal bent are chosen seems independent of a straightforward preference for the most reliable outlets available. The moderate liberal bias in sourcing appears to be a distinct phenomenon.

The authors acknowledge several limitations to their work. The study’s conclusions depend on the specific methodologies of the external rating services used to measure political leanings and reliability. Additionally, the analysis focused on a one-dimensional political spectrum from liberal to conservative, which does not capture the full complexity of political viewpoints.

Another limitation is that the analysis was conducted at the domain level, meaning it assessed the general leaning of an entire news outlet, not the content of a specific article cited. A news article from a generally liberal-leaning outlet could itself be perfectly neutral. The study also focused exclusively on the English version of Wikipedia, and the patterns may differ in other languages.

Future research could expand on this study by analyzing other language versions of Wikipedia to see if similar patterns exist. A more granular analysis that examines the content of individual news articles and how they are used to support claims within Wikipedia would also offer deeper insights. Understanding the underlying reasons for this sourcing bias, whether it stems from the media landscape or the demographics of Wikipedia’s editors, remains an open area for investigation.

The study, “Polarization and reliability of news sources in Wikipedia,” was authored by Puyu Yang and Giovanni Colavizza.