Examining Partisan Asymmetries in Fake News Sharing and the Efficacy of Accuracy Prompt Interventions

Contributions

• Tests partisan differences in fake news sharing in a controlled, balanced information environment with equal exposure to left- and right-leaning true and false headlines.
• Finds Republicans share 1.16x more false ideologically agreeable content than Democrats, but 33% more discerning for agreeable headlines overall.
• Demonstrates accuracy prompt interventions work equally well for Democrats and Republicans, ruling out differential responsiveness to accuracy cues.
• Provides evidence that most observed partisan asymmetries in social media sharing stem from asymmetric exposure rather than underlying propensity differences.

Method

Survey experiment with 2,015 U.S. adult Facebook users from YouGov, conducted August–September 2020. Respondents were balanced by party, demographics, and past presidential vote. Each respondent was randomly assigned to a control condition or one of three accuracy prompt interventions (Evaluation, Importance + Norms, or a 30-second PSA video).

All respondents rated their likelihood of sharing 20 news headlines sampled from a balanced set of 59 headlines: 5 true left-leaning, 5 false left-leaning, 5 true right-leaning, 5 false right-leaning. Headlines were evaluated on a 6-point Likert scale from "extremely unlikely" to "extremely likely" to share.

The paper measured multiplicative discernment (ratio of true to false sharing intentions) separately for ideologically agreeable and disagreeable headlines. A parallel sample of 880 Mechanical Turk respondents rated the accuracy (1–6 scale) of a larger set of 216 headlines to validate the accuracy of the news items.

Results

Partisan differences in discernment: For ideologically agreeable headlines, Democrats were 1.54x more likely to share true vs. false news, whereas Republicans were only 1.16x more likely (p = 0.042). This represents a 33% partisan asymmetry in discernment. For ideologically disagreeable headlines, both Democrats and Republicans showed similar levels of discernment (1.39x and 1.53x respectively, p = 0.57).

Accuracy prompt effectiveness: All three accuracy interventions increased sharing discernment for both Democrats and Republicans. The effect sizes were similar across parties: Democrats and Republicans in treatment conditions were 1.19x (p < 0.001) and 1.18x (p = .01) more discerning respectively compared to control. This lack of partisan asymmetry in treatment effects suggests Republicans' lower discernment for agreeable content is not due to caring less about accuracy.

Accuracy discernment capability: In the pre-test sample, Democrats and Republicans showed no significant difference in their ability to identify accurate content for agreeable headlines (ratio of discernment: 1.15, p = 0.28). For disagreeable headlines, Democrats were more discerning (1.46x ratio, p = 0.018), but this accuracy discernment difference was small and did not explain the larger asymmetry in sharing behavior.

Magnitude of partisan asymmetry: The observed partisan gap (1.54x for Republicans sharing 1.54 times more false agreeable headlines than Democrats) is much smaller than asymmetries previously reported in social media studies (2.5–8.5x from prior work), suggesting exposure to asymmetric supplies of right-leaning fake news accounts for the vast majority of observed differences in social media.

Notes

This paper makes a critical methodological contribution by using a balanced information environment to disentangle exposure from propensity. The finding that accuracy interventions work equally well for both parties contradicts some prior meta-analyses suggesting reduced efficacy for Republicans, though it is consistent with other work using more representative samples. The much smaller partisan asymmetry in the experimental setting (1.16x) compared to social media observations (2.5–8.5x) is striking and suggests that content supply and algorithmic curation—not underlying partisan differences in discriminative ability or accuracy motivation—drive most observed asymmetries. The paper is part of a broader empirical pushback against deterministic narratives of partisan susceptibility to misinformation, though the authors acknowledge that asymmetric exposure remains a substantive problem for information quality.