Contributions
- First large-scale study linking detailed psychological profiles (party identification, partisan affect, cognitive reflection, political awareness, digital media literacy) to actual Twitter news-sharing activity
- Empirical evidence that partisan polarization, not cognitive deficits or deliberate disruption, explains fake news sharing
- Demonstration that fake news follows a partisan continuum: both Democrats and Republicans share partisan-slanted content, but with different source bases and asymmetric effects
- Evidence that content matters: headline sentiment and news source ideology significantly predict sharing decisions independent of source origin
- Analysis of 42 fake news domains and 20 real news publishers, coded for partisan slant
Method
The authors surveyed 2,337 US Twitter users via YouGov (December 2018–January 2019) and extracted 2,709,052 tweets posted by those participants. Psychological measures included:
- Partisanship (7-point scale: "Strong Democrat" to "Strong Republican")
- Partisan affect (positive feelings toward own party, negative feelings toward opposing party)
- Cognitive reflection (Cognitive Reflection Test)
- Political awareness and digital media literacy (knowledge questions, media source judgments)
- Political cynicism and trolling behavior (Likert scales)
News domains were classified by AllSides as "Fake," "Leaning Republican," "Center," "Leaning Democrat," or "Fake." The authors analyzed which news stories users shared, whether from fake or credible sources, and the negativity of headline sentiment using sentiment analysis.
Logistic regression models estimated the association between psychological predictors and the probability of sharing at least one story from each type of news source, controlling for demographics (age, gender, income, education, race) and political interest.
Results
- Among 2.7M tweets, 3% linked to fake news; 4% to real news. Nearly 1% were responsible for all fake news shares, with 11% of links responsible for all fake news sharing.
- Partisan polarization strongly predicts sharing both fake and real news. Strong Republicans were 15 percentage points more likely to share pro-Republican fake news; strong Democrats showed no comparable effect.
- Partisan asymmetry: Republicans are significantly more likely to share from sources with a history of publishing false claims.
- Ignorance (low cognitive reflection, low factual knowledge, low digital media literacy) does not substantially predict fake news sharing when controlling for real news sharing.
- Political cynicism shows modest association with fake news sharing.
- The strongest predictor of fake news sharing is partisan affect—particularly negative feelings toward the opposing party.
- Headline sentiment: Participants select news based on whether stories attack the opposing party; pro-Republican fake news sources are more negative toward Democrats; pro-Democratic fake news sources follow similar patterns.
- Fake news shares concentrate at the partisan extremes of the news source continuum: strongly partisan fake sources attract the most polarized participants.
Notes
Strength: Large sample size with actual Twitter behavior, not self-reported sharing, and detailed psychological measures validated in prior work. Comprehensive control for competing theories (ignorance, disruption) tested simultaneously. Honest about limitations: self-selected Twitter sample, though demographics closely match national survey benchmarks.
Limitations: Study period (2018–2019) predates 2020 election intensification and COVID-era polarization shifts; results may reflect baseline polarization rather than crisis-driven dynamics. Relies on AllSides classifications for source ideology, which has known biases. Analysis is observational; cannot establish causality or rule out unmeasured confounders (e.g., algorithmic amplification as a mediator).
The paper's main insight—that partisan goals, not accuracy concerns, drive fake news sharing—has significant implications for interventions. Fact-checking alone may fail if people prioritize political usefulness over accuracy.