The paper
Derek E. Holliday, Justin Grimmer, Yphtach Lelkes, and Sean J. Westwood (2025). Who Are the Election Skeptics? Evidence from the 2022 Midterm Elections. Election Law Journal: Rules, Politics, and Policy 24(1), 1–17.
doi:10.1089/elj.2024.0010 (opens in a new tab) · Replication data: doi:10.7910/DVN/1ZAI9W (opens in a new tab)
Main text
p. 1, Abstract
Published
Over half of skeptics claim they are skeptical because of how elections are run
Correction
Nearly half of skeptics (45%) claim they are skeptical because of how elections are run
Basis Among the 844 coded answers to the open-ended evidence question from respondents who said fraud occurred, 45.2% (weighted) fall into the four election-administration categories of Table 2, the 45% the text reports on p. 10. The share exceeds half only among Republicans (54.1%; see the correction to p. 11).
pp. 1–2, Introduction
Published
Most recently, the U.S. Representative Katie Porter claimed that the 2022 California primary for U.S. Senate was rigged against her.
Correction
Most recently, the U.S. Representative Katie Porter claimed that the 2024 California primary for U.S. Senate was rigged against her.
Basis Representative Porter's claim concerned the California Senate primary held on March 5, 2024.
p. 2, Introduction
Published
With closed-ended survey measures we find that overall only 15.4% of all skeptics claim that all elections have fraud, and only 0.7% of our sample believes that fraud determined the outcome in all 50 states and DC.
Correction
In open-ended responses we find that overall only 15.4% of all skeptics claim that all elections have fraud (Figure 5), and with closed-ended survey measures we find that only 0.7% of our sample believes that fraud determined the outcome in all 50 states and DC.
Basis 15.4% is the weighted share of answers to the open-ended evidence question coded “Logical deduction” (no evidence offered; fraud is assumed to happen), the category the text reports as 15% on p. 11. The 0.7% comes from the closed-ended question asking in which states fraud decided the outcome.
p. 4
Published
(Trotta, Harte and Trotta, 2022)
Correction
(Trotta and Harte, 2022)
Basis See the correction to the reference list (p. 17).
p. 6
Published
Our findings reveal no significant or meaningful links between individual-level potential confounders (such as age, gender, education, and political identity) and the survey date.
Correction
Our findings reveal no significant links between survey date and gender or political identity. Race and education do predict survey date: holding the other characteristics constant, White respondents were interviewed on average 1.5 days earlier than other respondents, and respondents with a four-year degree 1.3 days earlier than those without one (p < .001 with or without survey weights). With survey weights, older respondents were also interviewed slightly earlier (0.2 days per decade of age, p = .005). These differences are small, and controlling for age, gender, education and race leaves the partisan divergence in Table 1B intact: the week 1–4 effects become −0.25, −0.26, −0.27, −0.28 for Republicans and 0.11, 0.13, 0.11, 0.14 for Democrats.
Basis The test described in the text, run on the public data: days from Election Day regressed on age, gender, a four-year degree, White race and party (N = 5,244), with HC2 standard errors. Party and state are unrelated to survey date. The published Table 1B week effects are −0.24, −0.25, −0.27 and −0.29 for Republicans and 0.09, 0.11, 0.08 and 0.08 for Democrats.
p. 7
Published
Eventually, however, partisans largely returned to their premidterm baseline perceptions of election accuracy. [...] Republicans, after a year-long period spent (on average) being more likely to disagree than agree the 2022 election results were accurate, eventually returned to pre-election perceptions of accuracy. This return to baseline for Republicans is confirmed in an analysis of n = 14,449 panel respondents embedded in our broader survey (see Supplementary Appendix S1, section 1.2.2). [...] While skepticism may be genuine (Graham and Yair, 2023), any increase in skepticism resulting postelection dissipates within one electoral cycle.
Correction
Republican perceptions of election accuracy did not return to their premidterm baseline. [...] Republicans, after a year-long period spent (on average) being more likely to disagree than agree the 2022 election results were accurate, recovered only partly. Between January 24 and February 15, 2024, 39% of Republicans agreed that the outcome accurately reflected the preferences of those who voted, close to the 37% of November 9–29, 2022, and 22 percentage points below the 61% who agreed before Election Day (p < .001). An analysis of n = 14,449 panel respondents embedded in our broader survey (see Supplementary Appendix S1, section 1.2.2) shows the same pattern: across the whole period the Republican trend is flat (−0.0005 per week on the five-point scale, p = .381), and from January 12, 2023 onward it is 0.0027 per week (p < .001). [...] While skepticism may be genuine (Graham and Yair, 2023), the postelection increase in Republican skepticism had not dissipated by early 2024.
Basis Weighted shares of Republicans who agree or strongly agree, from the long-term survey file. The pre-election period is November 4–8, 2022 (the item was first asked on November 4; Election Day counts as pre-election, as in Table 1); the later windows are the paper's own post-election and current periods (pp. 8–9). Republican agreement reached 33% in the paper's height-of-skepticism window (January 12–February 3, 2023). The public table_s1.R reproduces the Table S1 Republican coefficient, −0.0005, which shows no return to baseline; the post-January 2023 slope is the same model fitted to interviews from the start of the height window.
pp. 9–10
Published
Specifically, we asked respondents, who answered “Yes” to “Do you think fraud occurred in the 2022 midterms?” [...], two open-ended questions. They were asked, “What evidence have you seen that leads you to believe that voter fraud occurred in the 2022 midterms?” In addition, they were asked, “To change an election outcome massive fraud is necessary [...] Why do you think those who commit fraud on this scale are able to get away with it?”
Correction
Specifically, we asked respondents, who answered “Yes” to “Do you think fraud occurred in the 2022 midterms?” [...], an open-ended question: “What evidence have you seen that leads you to believe that voter fraud occurred in the 2022 midterms?” A subset of 713 of them, defined by their answer to a closed-ended follow-up question, was also asked, “To change an election outcome massive fraud is necessary [...] Why do you think those who commit fraud on this scale are able to get away with it?”
Basis In the replication data, answers to the second question exist for 712 of the 1,055 respondents who said fraud occurred. All of them have the same value (1) on the closed-ended item w9_2022_1_elections4, and none of the 342 respondents with any other value has an answer.
p. 11
Published
Issues with election administration were particularly prominent among the Republicans who claimed that fraud occurred (55%)
Correction
Issues with election administration were particularly prominent among the Republicans who claimed that fraud occurred (54%)
Basis The four administration categories in the Republican column of Table 2 sum to 54.1% (14% + 20.3% + 7.6% + 12.2%).
pp. 11–12
Published
That said, only three respondents identified the movie 2000 mules as the source of their skepticism. [...] We find, for example, that only three skeptics even mention Dominion voting machines in their open-ended response
Correction
That said, only 15 respondents cited the movie 2000 mules as evidence of fraud, and 17 mentioned it in either open-ended response. [...] We find, for example, that only 7 skeptics even mention Dominion voting machines in their open-ended responses (5 of them in answers to the evidence question)
Basis Counts of respondents who said fraud occurred whose open-ended answers mention “mules” or “Dominion”. The printed three and three are the counts in answers to the second question (why fraud goes unpunished) alone; the Smartmatic count of one is correct. The film and the company remain rare among the 844 coded answers.
p. 13, Figure 6 note
Published
N=844; Weighted values and 95% CIs
Correction
N = 579 (Democrats 46, Independents 98, Republicans 435); Weighted values and 95% CIs
Basis figure_6.R merges the coded answers to the second question with the survey file; 579 responses carry a code (see the correction to pp. 9–10). The plotted percentages are correct for these responses.
p. 17, References
Published
Trotta, Daniel, Julia Harte, and Daniel Trotta. 2022. “Democrats win governors’ races in three crucial ‘blue wall’ states.” Reuters.
Correction
Trotta, Daniel, and Julia Harte. 2022. “Democrats win governors’ races in three crucial ‘blue wall’ states.” Reuters.
Basis The Reuters article is bylined Daniel Trotta and Julia Harte.