Berlinski, N., Doyle, M., Guess, A. M., Levy, G., Lyons, B., Montgomery, J. M., Nyhan, B., & Reifler, J. (2023). The effects of unsubstantiated claims of voter fraud on confidence in elections. Journal of Experimental Political Science, 10, 34–49. https://doi.org/10.1017/xps.2021.18
Summary
This paper provides direct experimental evidence that unsubstantiated elite claims of voter fraud erode public confidence in elections. Using a survey experiment fielded to a nationally representative U.S. sample shortly after the 2018 midterms, the authors expose respondents to real tweets from prominent Republicans alleging fraud and find that this exposure significantly reduces confidence in electoral integrity. The damage is concentrated among partisans for whom the claims are congenial—Republicans and Trump approvers—and, critically, corrective fact-checks from mainstream sources fail to measurably offset the harm. Notably, the effects are specific to election confidence and do not extend to broader support for democracy as a system of government. Written against the backdrop of Trump’s 2020 fraud narrative and January 6, the study frames unfounded fraud allegations as a distinct threat to perceived electoral legitimacy.
Key Contributions
- Direct experimental (rather than mixed observational/panel) evidence that unfounded elite fraud claims causally reduce election confidence.
- Enhances external validity by using actual tweets from real political leaders in their original format, going beyond prior hypothetical-vignette designs.
- Shows that mainstream fact-checking is insufficient to counter congenial misinformation—an important corrective to optimistic assumptions about correction efficacy.
- Empirically distinguishes erosion of election confidence from erosion of support for democracy as a regime type.
- Connects belief-echoes, elite-cue, and motivated-reasoning literatures to the fragility of electoral legitimacy in advanced democracies.
Methods
An online survey experiment (YouGov, N = 4,283) fielded in December 2018/January 2019. Respondents were randomly assigned to four conditions: a placebo (non-political tweets), a low dose (four fraud tweets), a high dose (eight fraud tweets), and a low dose plus fact-check (four fraud tweets with four corrections from AP, PBS NewsHour, and NYT Politics). Election confidence was measured with a seven-item composite validated by factor analysis, and a separate five-item battery measured support for democracy. Analysis used OLS with robust standard errors, including models interacting treatment with partisanship and Trump approval as congeniality moderators. The design was preregistered with documented deviations.
Findings
- Low-dose exposure reduced composite election confidence by 0.147 SD; high-dose by 0.168 SD (both p < 0.005).
- No reliable difference between low and high doses—higher dosage did not amplify the effect.
- The fact-check condition still significantly reduced confidence relative to placebo (−0.092 SD, p < 0.05); corrections did not offset the harm.
- High-dose effects were significant only among Republicans (~0.27 SD) and Trump approvers (~0.34 SD).
- Fact-check effects did not vary by party or approval; negative effects persisted even with corrections present.
- Treatments had overwhelmingly null effects on broader support for democratic governance.
- Beliefs that ballots are secure from tampering were unaffected—consistent with that specific claim not being questioned in the stimuli.
Connections
This paper’s finding that mainstream corrections fail to neutralize congenial misinformation speaks directly to work on the limits and mechanisms of fact-checking and debunking, such as Lewandowsky2026-ob and Pennycook2021-jq. Its emphasis on elite cues and motivated reasoning shaping who accepts fraud narratives connects to research on partisan sharing and susceptibility to falsehoods like Osmundsen2021-et and Guess2020-rr, while its focus on the delegitimation of elections resonates with studies of election-related disinformation and rumoring such as Starbird2025-jj.