Corrections to published work
- corrections
- 507
- papers, 2012–2026
- 34
I checked 34 of my published papers, from 2012 to 2026, against everything an outside reader can see: the published articles and supplements, the public replication archives, the preregistrations and the works they cite. Where the public data allowed it, I recomputed the reported numbers, tables and figures. I did this because readers rely on what a paper reports, and they should be able to see where my published work and its evidence disagree. These pages list the errors the check found and correct each one.
Each paper has its own page. A short paragraph at the top says what the corrections change. The corrections follow in page order, with the supplementary information in a separate section. Each entry gives the location, the text or element as published and the correction. Where a correction changes a number, table or figure, the corrected value comes from code that downloads the public replication data and recomputes it.
Most of these papers have coauthors, who are named in each citation. The corrections here are mine.
Papers
Newest first. The number at the left counts the paper’s corrections.
PNAS Nexus (2026) · 28 corrections
The paper's conclusions stand. District competitiveness (PVI) is defined only for House districts, so the models that control for it include only the 12 senators from states with a single at-large House seat. The committee-assignment result holds for House members; senators show no difference.
Levels of Office and Voter Accountability for Democratic Norm Violations
Public Opinion Quarterly (2026) · 15 corrections
Study 1 shows small but significant differences by level of office, which the paper reported as absent: defection from norm-violating copartisans is slightly higher in local races than in state and national races, the opposite of the hypothesis. Study 2 does not show this pattern. Republicans defect less than Democrats at every level of office. The Study 2 priming hypothesis was preregistered as primary and is not supported.
Public responses to atrocities depend on partisanship and context, not legal labels
Proceedings of the National Academy of Sciences (2026) · 15 corrections
The paper's main finding stands: legal labels such as genocide do not measurably change public responses to atrocities, while partisanship and the country involved do. With the robust standard errors the paper describes, no label effect is significant. The confidence intervals bound label effects to within ±0.21 to ±0.37 standard deviations, so effects of that size cannot be ruled out. Sudan elicits more sympathy than China, the reverse of the published sentence.
Proceedings of the National Academy of Sciences (2026) · 1 correction
The correction does not change the letter's conclusions.
Systemic electioneering from the evangelical pulpit: Evidence from a computational analysis
Proceedings of the National Academy of Sciences (2026) · 18 corrections
The observed rates of political advocacy in sermons are unchanged. The paper presented them as bounds on true prevalence, and the independent coder's check does not support that: against that coder, the final labels have a precision of 73.8% and a recall of 31.8%. Adjusted for both, the rates are about 2.32 times the observed rates, and advocacy that the language-model stages missed would raise them further.
No news is good news? The declining information value of broadcast news in America
PLOS One (2025) · 12 corrections
None of the corrections changes the paper's conclusions.
The potential existential threat of large language models to online survey research
Proceedings of the National Academy of Sciences (2025) · 19 corrections
The paper's conclusions stand. Figure 8 compared shares of respondents with scale means. On a common metric the comparisons with Mummolo and Peterson change, and the synthetic respondent's demand effect remains far larger than the effect among human respondents.
Who Are the Election Skeptics? Evidence from the 2022 Midterm Elections
Election Law Journal: Rules, Politics, and Policy (2025) · 11 corrections
Two claims change. 45% of skeptics cite how elections are run; the abstract says over half. Republicans' confidence in the 2022 results recovered only partly: in early 2024 it was close to its immediate post-election level and 22 points below its pre-election level.
Why depolarization is hard: Evaluating attempts to decrease partisan animosity in America
Proceedings of the National Academy of Sciences (2025) · 16 corrections
The paper's conclusion stands: these interventions have small effects that do not grow with repetition. The preregistered primary outcome was affective polarization, and the paper reported out-party warmth without disclosing the switch. On the preregistered outcome, no dose in Study 1 has a significant effect, and one of six comparisons with control in Study 2 is significant.
Affective polarization is uniformly distributed across American States
PNAS Nexus (2024) · 11 corrections
None of the corrections changes the paper's conclusions.
Correcting misperceptions of partisan opponents is not effective at treating democratic ills
PNAS Nexus (2024) · 18 corrections
None of the corrections changes the paper's conclusions.
Persistent polarization: The unexpected durability of political animosity around US elections
Science Advances (2024) · 14 corrections
The paper's conclusion stands: partisan animosity was largely stable around the 2022 election. The data show small, statistically fragile shifts around Election Day, which the paper described as absent. The design had 22% to 26% power to detect the smallest effect the paper said it could detect (Cohen's d = 0.04).
Proceedings of the National Academy of Sciences (2024) · 8 corrections
The decline in Republican support for partisan murder after the attempt stands as a description of the data. The robustness claim changes: the published placebo comparison ("nearly 50 times smaller") was a unit error. A decrease as large as the observed one occurs in 3.2% of placebo windows, so it lies in the tail of the placebo distribution without being unique to the period after the attempt.
Uncommon and nonpartisan: Antidemocratic attitudes in the American public
Proceedings of the National Academy of Sciences (2024) · 10 corrections
None of the corrections changes the paper's conclusions.
Does Affective Polarization Undermine Democratic Norms or Accountability? Maybe Not
American Journal of Political Science (2023) · 16 corrections
The paper's conclusions stand. One test changes: respondents assigned to lower affective polarization rewarded bipartisan behavior significantly less than those assigned to higher affective polarization (p = 0.040), the opposite of the literature's prediction. The paper reported this difference as not significant.
Effects of a US Supreme Court ruling to restrict abortion rights
Nature Human Behaviour (2023) · 21 corrections
The pre-registration called for dropping respondents unaware of the leak or the ruling, and the article neither applied nor reported that exclusion. The corrections disclose the deviation and keep the full-sample estimates as primary: they are intent-to-treat estimates of the shift in public opinion, and awareness was measured after each event. Under the exclusion, several estimates lose statistical significance, including the effects on perceived fetal viability and on support for the Dobbs holding after the leak; both sets of estimates are reported. Figure 5c counted engagements, and the corrected figure counts tweets.
Which Police Departments Want Reform? Barriers to Evidence-Based Policymaking
Journal of Experimental Political Science (2023) · 16 corrections
The paper's conclusion stands: agencies' willingness to collaborate shows no selection on performance. A coding error reversed the population result. Smaller jurisdictions agree less often than larger ones, and the difference is not significant after correction for multiple comparisons.
Current research overstates American support for political violence
Proceedings of the National Academy of Sciences (2022) · 17 corrections
The paper's main finding stands: surveys overstate American support for political violence. Two claims change. With the measure held fixed, disengaged respondents inflate the relationship between support for violence and its correlates by a factor of 1.2 to 2.1; the abstract says 4. Engaged respondents also show small but statistically significant leniency toward in-party perpetrators.
American Journal of Political Science (2022) · 3 corrections
None of the corrections changes the paper's conclusions.
The Inseparability of Race and Partisanship in the United States
Political Behavior (2022) · 20 corrections
The paper's conclusions stand. Study 3 included respondents who are neither white nor Black, with their out-race rating set to the scale midpoint. Restricted to white and Black respondents, as in Study 1, Study 3 supports the same conclusions, and the spillover from partisanship to racial affect is 90% of the reverse effect; the paper reports 77%.
Political Behavior (2022) · 23 corrections
Two results change. The increase in speeches engaging bipartisanship after a primary is small and not significant once standard errors are clustered by legislator (b = 0.34, p = .079). Swapped labels in the SI tables hid a partisan difference in the experiments: the bipartisan treatment raises support among Democrats and liberals, while among Republicans and conservatives it does not raise support for overtly partisan bills (Study 1) or for the policy (Study 2). The paper's other results are unchanged.
Beyond Attitudes: Incorporating Measures of Behavior in Survey Experiments
Advances in Experimental Political Science (2021) · 12 corrections
None of the corrections changes the chapter's conclusions.
The Limits of Partisan Loyalty
Political Behavior (2021) · 18 corrections
None of the corrections changes the paper's conclusions.
Projecting Confidence: How the Probabilistic Horse Race Confuses and Demobilizes the Public
The Journal of Politics (2020) · 22 corrections
None of the corrections changes the paper's conclusions.
Are there Still Limits on Partisan Prejudice?
Public Opinion Quarterly (2019) · 13 corrections
None of the corrections changes the paper's conclusions.
The Origins and Consequences of Affective Polarization in the United States
Annual Review of Political Science (2019) · 6 corrections
None of the corrections changes the review's conclusions.
An exploration of Donald Trump's allegations of massive voter fraud in the 2016 General Election
Electoral Studies (2018) · 15 corrections
None of the corrections changes the paper's conclusions.
The tie that divides: Cross-national evidence of the primacy of partyism
European Journal of Political Research (2018) · 12 corrections
The paper's cross-national findings stand. One passage about Belgium changes: the ideological pattern there comes from the left-wing parties, and right-wing supporters do not favor fellow right-wing supporters across the regional divide.
Political Analysis (2017) · 21 corrections
The ensemble method and its applications are unaffected. Two simulation results change: Find It's poor performance in Table 2 came from the prediction code used to evaluate it, and the archived Monte Carlo 2 results come from a dense design, which the paper describes as sparse.
The Limits of Partisan Prejudice
The Journal of Politics (2017) · 18 corrections
Pooled across parties, the paper's findings stand. Mislabeled columns in Table A20 hid one partisan difference: among Republicans in Study 3.1, more affectively polarized participants agreed more with police action against Democratic protesters and proposed larger fines.
Fear and Loathing across Party Lines: New Evidence on Group Polarization
American Journal of Political Science (2015) · 26 corrections
One claim changes: in Study 4 the penalty for opposing partisans and the bonus for copartisans do not differ significantly, so the data do not show that outgroup animosity outweighs ingroup favoritism. The paper's other findings are unchanged.
The Role of Persuasion in Deliberative Opinion Change
Political Communication (2015) · 21 corrections
The paper's conclusion stands: persuasive argumentation drives opinion change after deliberation. Its effects are larger than those of knowledge change and the most consistent of any predictor; on free trade and protectionism, distance from the group's initial attitude has the larger coefficient.
Selective Exposure in the Age of Social Media
Communication Research (2014) · 7 corrections
None of the corrections changes the paper's conclusions.
American Political Science Review (2012) · 4 corrections
None of the corrections changes the paper's conclusions.