OSMUNDSEN, M., BOR, A., VAHLSTRUP, P. B., BECHMANN, A., & PETERSEN, M. B. (2021). Partisan polarization is the primary psychological motivation behind political fake news sharing on twitter. American Political Science Review, 115, 999–1015. https://doi.org/10.1017/s0003055421000290
Summary
This paper adjudicates among three competing psychological accounts of why people share political fake news on Twitter: ignorance (people share falsehoods because they lack the reflective, factual, or digital literacy to spot them), disruption (people share to sow chaos or express cynicism), and partisan polarization (people share to advance partisan goals and attack opponents). By linking rich survey-based psychological profiles of over 2,300 American Twitter users to their actual news-sharing behavior and to sentiment analyses of hundreds of thousands of headlines, the authors find that partisan polarization—especially out-party hatred—is the dominant driver. Their central argument is that fake news sharing is not a distinct pathology but rather the extreme tail of an ordinary, one-dimensional partisan news continuum, governed by the same goal-oriented motivations that drive real partisan news sharing.
Key Contributions
- One of the first behavioral tests linking detailed individual-level psychological measures to observed Twitter news-sharing for a large panel.
- Adjudicates three theories (ignorance, disruption, polarization) using both individual-difference and content-based evidence.
- First test distinguishing in-party love vs. out-party hatred as drivers of fake news sharing—finding out-party animus roughly twice as strong.
- Large-scale sentiment analysis of shared and front-page headlines demonstrating a partisan news-source continuum and an asymmetric supply of out-party-derogating content.
- Argues that accuracy-focused interventions (e.g., fact-checking) are unlikely to suffice, since sharing is motivated by political usefulness rather than belief in accuracy.
Methods
The authors commissioned a YouGov survey (Dec 2018–Jan 2019) of U.S. Twitter users, linking confidential responses to scraped public Twitter data (~2.7M tweets), yielding a final linked sample of N=2,337 after high, non-random attrition. Fake news sharing was operationalized at the publisher level by matching tweeted URLs against a list of 608 fake news domains, with real news matched against 260 AllSides-rated publishers. Two headline datasets (75,560 shared headlines; ~500,000 front-page headlines via Archive.org) were subjected to sentiment analysis using sentimentR and custom elite-name dictionaries to measure negativity toward each party. Predictors were measured via validated survey batteries: CRT-2 and political knowledge (ignorance), trolling and cynicism scales (disruption), and partisanship plus in-party/out-party affect scales (polarization). The main analysis used logistic regression with average marginal effects, plus multiple robustness checks.
Findings
- News sharing is rare: ~3% of tweets contained news links, ~4% of those from fake sources; sharing was highly concentrated (1% of panelists shared ~75% of fake news; only 11% shared any).
- Ignorance measures did not predict fake news sharing—cognitive reflection was unrelated, and higher political knowledge and digital literacy actually predicted more fake news sharing.
- Disruption received mixed support: apolitical trolling predicted less news sharing, while political cynicism predicted more sharing of both fake and real news.
- Partisanship and especially negative out-party affect strongly predicted fake news sharing; out-party-hatred coefficients were roughly twice the in-party-love coefficients.
- Sharing of ideologically biased real news correlated strongly with same-slant fake news sharing; partisans essentially never shared incongruent fake news.
- A partisan asymmetry emerged: Republicans shared more fake news, matched by the fact that only pro-Republican fake news consistently portrayed Democrats more negatively—suggesting a supply-side rather than psychological explanation.
- The motivations underlying fake news and real news sharing were psychologically nearly identical.
Connections
This paper’s emphasis on partisan motivation over ignorance directly complements accuracy-nudge and inattention accounts of misinformation, offering a corrective to the view that sharing stems primarily from failures of reflection (cf. Pennycook2021-jq). Its behavioral, linked-data approach and finding that fake news is a small, concentrated share of activity align with population-level exposure studies such as Grinberg2019-ua, Allen2020-nj, Guess2019-ym, and Guess2021-ym, while its framing of fake news as continuous with ordinary partisan behavior speaks to the affective-polarization literature (Iyengar2019-jj, Bail2018-fk) and to virality dynamics documented in Vosoughi2018-at.