Chapter 6.8

Political violence and democratic norms

In the 2024 ANES, 17.2% of adult citizens said violence to pursue political goals is justified at least a little, with no party gap that the data separate from zero, while Republicans were more likely than Democrats to expect an inaccurate vote count and to call Trump the legitimate winner of 2020.

Author: Vishal Singh
Affiliation: NYU Stern
Published: 3 October 2026

In the 2024 American National Election Studies (ANES), 17.2% of adult citizens said it is justified at least a little for people to use violence to pursue their political goals in this country, and 4.6% said it is justified a lot or a great deal. The data show no gap between Democrats and Republicans that is distinguishable from zero: the Democratic share minus the Republican share was 2.2 (95% CI −0.9 to 5.3) points, so a Democratic excess of up to 5.3 points over the Republican share of 15.0% is not ruled out. Beliefs about elections divided the parties. Before the election, 35.1% of Republicans and 8.2% of Democrats expected the November votes to be counted not at all or only a little accurately. After it, 55.5% of Republicans and 4.5% of Democrats said Donald Trump was the legitimate winner of the 2020 election, and 14.7% of Democrats and 2.3% of Republicans said Kamala Harris was the legitimate winner of the 2024 election.

These are answers about approval, expectation and belief. The ANES records no violent act, threat or harassment by a respondent, so the chapter cannot say how many people have acted or would act, and the survey does not say whom the violence would be against or for what goal. The estimates describe adult U.S. citizens at residential addresses. Each interval is a 95% confidence interval that follows the sampling design: it uses the strata and primary sampling units in the file with a Taylor-series variance estimate, and it does not treat respondents as independent draws.The 2024 file combines a fresh sample interviewed mainly in person, a fresh sample interviewed mainly on the web and a panel of 2,171 respondents who were also interviewed in 2016 and 2020. The ANES asks analysts to use the variance strata and primary sampling units in the file [1]. The intervals here use the t distribution with the design degrees of freedom and describe sampling variation only. Question wording, interview mode and nonresponse are other sources of error, and a later section takes up mode.

Party means party identification, with leaners counted with the party they lean toward, unless the text says without leaners. Adults who name neither party and do not lean toward one are called neither party. Without leaners, Democrats and Republicans are the strong and not very strong identifiers, and the remaining adults are independents and leaners. The party question comes from the pre-election interview, so post-election answers are grouped by the party the respondent reported before the election.

What each question asks

Figure 1 quotes every question the chapter uses, with its answer scale, what it measures and what it is about.

The violence question asks, "How much do you feel it is justified for people to use violence to pursue their political goals in this country?" Its target is unspecified: the violence is by people in general, for any political goal, against anyone, and the wording does not say what act counts as violence. It measures approval in the abstract and asks nothing about what the respondent has done or would do, so it measures hypothetical approval and not action. Westwood and co-authors report from four survey experiments that attitudes toward political violence measured this way are biased upward by respondent disengagement and by wording that allows more than one interpretation [2]. Kalmoe and Mason study partisan violence with questions about one's own party and the other party [3], targets that the ANES item leaves open. A follow-up question asks whether violence used to pursue political goals in the United States has increased, decreased or stayed the same over a four-year comparison period (the table gives the wording), which records a perception and not approval.

The vote-count questions differ in their targets. Before the election the ANES asked how accurately the respondent expected the votes in the November 2024 election to be counted, an expectation about one election given before its result was known. After it, the ANES asked how often votes are counted fairly in this country's elections, a belief about elections in general. A further pre-election question asks how much the respondent trusts the officials who oversee elections where the respondent lives, which concerns local officials and not the count.

Three paired statements ask which of two claims is most likely to be true: who was the legitimate winner of the 2020 election, who planned and carried out the events of January 2021 that the item names by date (pro-Trump protesters, or FBI agents posing as protesters, which this chapter calls the FBI account), and who was the legitimate winner of the 2024 election. Each offers two answers, so a respondent who is unsure still chooses one, and a follow-up asks how confident the respondent is. Legitimacy is a judgment about an election and not a count of votes, and the question does not define the term.

Six further questions ask about norms for officeholders and the press. Four ask how important each of four practices is to the United States maintaining a strong democracy. The other two ask whether it would be helpful if presidents could work on the country's problems without paying attention to what Congress and the courts say, and whether the respondent favors elected officials restricting journalists' access to information about government decision-making. They ask about officials in general and name no president or party.

Figure 1

The ANES questions measure approval, expectation, perception and belief, and none records an action

Each question used in this chapter, with its variable, exact wording, answer scale, what it measures and what it is about; 2024 pre-election and post-election interviews

Question	Variable	Wording	Answers	What it measures	What it is about	Asked	Respondents
Violence justified	V241579	On another topic: How much do you feel it is justified for people to use violence to pursue their political goals in this country?	Not at all; A little; A moderate amount; A lot; A great deal	Hypothetical approval, in the abstract	People in general, any political goal, no act or victim named	Pre-election; not on paper	5,192
Violence increased or decreased	V241580	Compared to four years ago, do you think violence used to pursue political goals in the United States has:	Increased; Decreased; Stayed the same (a follow-up asks how much)	Perception of change	Violence to pursue political goals in the United States, by anyone	Pre-election; not on paper	5,181
Votes counted accurately	V241314	In the November 2024 general election, how accurately do you think the votes will be counted?	Not at all accurately; A little accurately; Moderately accurately; Very accurately; Completely accurately	Expectation, before the result was known	Votes in the 2024 general election, anywhere in the country	Pre-election; all modes	5,478
Trust in election officials	V241315	How much do you trust the officials who oversee elections where you live?	Not at all; A little; A moderate amount; A lot; A great deal	Trust	Officials who run elections where the respondent lives	Pre-election; not on paper	5,243
Votes counted fairly	V242207	In your view, how often do the following things occur in this country's elections: Votes are counted fairly?	All of the time; Most of the time; About half of the time; Some of the time; Never	Belief about elections in general, after the result	Elections in this country in general	Post-election; all modes	4,926
Legitimate winner of 2020	V242567	Which of these two statements do you think is most likely to be true? Donald Trump was the legitimate winner of the 2020 presidential election / Joe Biden was the legitimate winner of the 2020 presidential election	One of the two statements; a follow-up (V242568) asks how confident, from Not at all to Extremely	Belief about a past result	The 2020 presidential election	Post-election; not on paper	4,612
Who carried out January 6	V242571	Which of these two statements do you think is most likely to be true? The events on January 6, 2021 were planned and carried out by pro-Trump protesters / The events on January 6, 2021 were planned and carried out by FBI agents posing as pro-Trump protesters	One of the two statements; a follow-up (V242572) asks how confident, from Not at all to Extremely	Belief about who acted	Who planned and carried out the events of that day	Post-election; not on paper	4,551
Legitimate winner of 2024	V242575	Which of these two statements do you think is most likely to be true? Donald Trump was the legitimate winner of the 2024 presidential election / Kamala Harris was the legitimate winner of the 2024 presidential election	One of the two statements; a follow-up (V242576) asks how confident, from Not at all to Extremely	Belief about a result just decided	The 2024 presidential election	Post-election; not on paper	4,633
Free news organizations	V241324	For the next few items, we would like to know how important you think each one is to the United States maintaining a strong democracy. First, how important is it that news organizations are free to criticize political leaders?	Not important at all to Extremely important (five answers)	Stated importance of a named practice	News organizations and political leaders in general	Pre-election; not on paper	5,241
Branches check one another	V241325	How important is it that the executive, legislative, and judicial branches of government keep one another from having too much power?	Not important at all to Extremely important (five answers)	Stated importance of a named practice	The three branches in general	Pre-election; not on paper	5,242
Consequences for misconduct	V241326	How important is it that elected officials face serious consequences if they engage in misconduct?	Not important at all to Extremely important (five answers)	Stated importance of a named practice	Elected officials in general	Pre-election; not on paper	5,250
Agreement on basic facts	V241327	How important is it that people agree on basic facts even if they disagree politically?	Not important at all to Extremely important (five answers)	Stated importance of a named practice	People in general	Pre-election; not on paper	5,233
Presidents and Congress and the courts	V241328	Would it be helpful, harmful, or neither helpful nor harmful if U.S. presidents could work on the country's problems without paying attention to what Congress and the courts say?	Helpful; Harmful; Neither helpful nor harmful (a follow-up asks how much)	Stated view of a hypothetical change	U.S. presidents in general, no president named	Pre-election; not on paper	5,210
Journalists' access	V241331	Do you favor, oppose, or neither favor nor oppose elected officials restricting journalists' access to information about government decision-making?	Favor; Oppose; Neither favor nor oppose (a follow-up asks how much)	Stated view of a hypothetical change	Elected officials in general	Pre-election; not on paper	5,219

Show data

Data for The ANES questions measure approval, expectation, perception and belief, and none records an action
answers	asked	measures	n	question	target	variable	wording
Not at all; A little; A moderate amount; A lot; A great deal	Pre-election; not on paper	Hypothetical approval, in the abstract	5,192	Violence justified	People in general, any political goal, no act or victim named	V241579	On another topic: How much do you feel it is justified for people to use violence to pursue their political goals in this country?
Increased; Decreased; Stayed the same (a follow-up asks how much)	Pre-election; not on paper	Perception of change	5,181	Violence increased or decreased	Violence to pursue political goals in the United States, by anyone	V241580	Compared to four years ago, do you think violence used to pursue political goals in the United States has:
Not at all accurately; A little accurately; Moderately accurately; Very accurately; Completely accurately	Pre-election; all modes	Expectation, before the result was known	5,478	Votes counted accurately	Votes in the 2024 general election, anywhere in the country	V241314	In the November 2024 general election, how accurately do you think the votes will be counted?
Not at all; A little; A moderate amount; A lot; A great deal	Pre-election; not on paper	Trust	5,243	Trust in election officials	Officials who run elections where the respondent lives	V241315	How much do you trust the officials who oversee elections where you live?
All of the time; Most of the time; About half of the time; Some of the time; Never	Post-election; all modes	Belief about elections in general, after the result	4,926	Votes counted fairly	Elections in this country in general	V242207	In your view, how often do the following things occur in this country's elections: Votes are counted fairly?
One of the two statements; a follow-up (V242568) asks how confident, from Not at all to Extremely	Post-election; not on paper	Belief about a past result	4,612	Legitimate winner of 2020	The 2020 presidential election	V242567	Which of these two statements do you think is most likely to be true? Donald Trump was the legitimate winner of the 2020 presidential election / Joe Biden was the legitimate winner of the 2020 presidential election
One of the two statements; a follow-up (V242572) asks how confident, from Not at all to Extremely	Post-election; not on paper	Belief about who acted	4,551	Who carried out January 6	Who planned and carried out the events of that day	V242571	Which of these two statements do you think is most likely to be true? The events on January 6, 2021 were planned and carried out by pro-Trump protesters / The events on January 6, 2021 were planned and carried out by FBI agents posing as pro-Trump protesters
One of the two statements; a follow-up (V242576) asks how confident, from Not at all to Extremely	Post-election; not on paper	Belief about a result just decided	4,633	Legitimate winner of 2024	The 2024 presidential election	V242575	Which of these two statements do you think is most likely to be true? Donald Trump was the legitimate winner of the 2024 presidential election / Kamala Harris was the legitimate winner of the 2024 presidential election
Not important at all to Extremely important (five answers)	Pre-election; not on paper	Stated importance of a named practice	5,241	Free news organizations	News organizations and political leaders in general	V241324	For the next few items, we would like to know how important you think each one is to the United States maintaining a strong democracy. First, how important is it that news organizations are free to criticize political leaders?
Not important at all to Extremely important (five answers)	Pre-election; not on paper	Stated importance of a named practice	5,242	Branches check one another	The three branches in general	V241325	How important is it that the executive, legislative, and judicial branches of government keep one another from having too much power?
Not important at all to Extremely important (five answers)	Pre-election; not on paper	Stated importance of a named practice	5,250	Consequences for misconduct	Elected officials in general	V241326	How important is it that elected officials face serious consequences if they engage in misconduct?
Not important at all to Extremely important (five answers)	Pre-election; not on paper	Stated importance of a named practice	5,233	Agreement on basic facts	People in general	V241327	How important is it that people agree on basic facts even if they disagree politically?
Helpful; Harmful; Neither helpful nor harmful (a follow-up asks how much)	Pre-election; not on paper	Stated view of a hypothetical change	5,210	Presidents and Congress and the courts	U.S. presidents in general, no president named	V241328	Would it be helpful, harmful, or neither helpful nor harmful if U.S. presidents could work on the country's problems without paying attention to what Congress and the courts say?
Favor; Oppose; Neither favor nor oppose (a follow-up asks how much)	Pre-election; not on paper	Stated view of a hypothetical change	5,219	Journalists' access	Elected officials in general	V241331	Do you favor, oppose, or neither favor nor oppose elected officials restricting journalists' access to information about government decision-making?

Read the measures and target columns before comparing shares: the questions differ in whether they ask about approval, an expectation, a perception or a belief, and in whom or what they concern. The table gives each question as worded in the ANES 2024 codebook, with the number of respondents who gave a usable answer.Note: Variable names are those of the ANES 2024 Time Series file. The paper questionnaire omitted every question marked not on paper.Note: The file also holds a version of the violence question for the spouses and partners of respondents. That version is answered by a different set of people and is not used.Source: ANES 2024 Time Series, pre-election and post-election questionnaires, 2024; counts are unweighted numbers of respondents

Pre-election interviews ran from August to Election Day in November 2024, and post-election interviews from the week after the election to February 2025. Pre-election questions use the pre-election weight and post-election questions the post-election weight. A short paper questionnaire, completed by 245 pre-election respondents and 200 post-election respondents, leaves out most of these questions, so those items describe adults who answered on the web or in person. Of the 5,521 respondents in the file, 5,192 gave a usable answer to the violence question. A further 84 refused or broke off before it, and they are left out of its shares.

No party gap in approval of violence is distinguishable from zero

Most adult citizens, 82.8%, said violence to pursue political goals is not at all justified. At the three cut points of the five-point answer scale, 17.2% (95% CI 15.6% to 18.9%) said it is justified at least a little, 10.8% (95% CI 9.5% to 12.1%) a moderate amount or more and 4.6% (95% CI 3.7% to 5.4%) a lot or a great deal (Figure 2). Among Democrats the shares at the first and last cut points were 17.3% (95% CI 14.9% to 19.7%) and 4.1% (95% CI 3.0% to 5.3%), and among Republicans 15.0% (95% CI 13.0% to 17.1%) and 4.8% (95% CI 3.5% to 6.2%). The Democratic-minus-Republican gap was 2.2 (95% CI −0.9 to 5.3) points at the first cut point, 0.1 (95% CI −2.4 to 2.6) at the second and −0.7 (95% CI −2.5 to 1.1) at the third, and each interval includes zero.

Figure 2

Most adult citizens said violence for political goals is not at all justified, and the data show no party gap distinguishable from zero

Share of adult citizens saying violence to pursue political goals is justified at three cut points of the five-point answer scale, by party identification; pre-election 2024

At least a little justified

A moderate amount or more

A lot or a great deal

Share of the group (%)

Show data

Data for Most adult citizens said violence for political goals is not at all justified, and the data show no party gap distinguishable from zero
Grouping	Cut point	Group	Share (%)	Lower	Upper	Respondents
with leaners	At least a little justified	All adult citizens	17.2	15.6	18.9	5,192
with leaners	At least a little justified	Democrats	17.3	14.9	19.7	2,532
with leaners	At least a little justified	Neither party	30.5	23.4	37.7	346
with leaners	At least a little justified	Republicans	15.0	13.0	17.1	2,290
with leaners	A moderate amount or more	All adult citizens	10.8	9.5	12.1	5,192
with leaners	A moderate amount or more	Democrats	10.0	8.4	11.6	2,532
with leaners	A moderate amount or more	Neither party	21.7	14.8	28.7	346
with leaners	A moderate amount or more	Republicans	9.9	8.0	11.9	2,290
with leaners	A lot or a great deal	All adult citizens	4.6	3.7	5.4	5,192
with leaners	A lot or a great deal	Democrats	4.1	3.0	5.3	2,532
with leaners	A lot or a great deal	Neither party	6.3	2.5	10.1	346
with leaners	A lot or a great deal	Republicans	4.8	3.5	6.2	2,290
without leaners	At least a little justified	All adult citizens	17.2	15.6	18.9	5,192
without leaners	At least a little justified	Democrats	14.8	12.5	17.2	1,848
without leaners	At least a little justified	Independents and leaners	21.3	18.0	24.6	1,695
without leaners	At least a little justified	Republicans	15.1	12.4	17.8	1,625
without leaners	A moderate amount or more	All adult citizens	10.8	9.5	12.1	5,192
without leaners	A moderate amount or more	Democrats	8.9	7.0	10.8	1,848
without leaners	A moderate amount or more	Independents and leaners	13.1	10.7	15.6	1,695
without leaners	A moderate amount or more	Republicans	10.2	7.8	12.5	1,625
without leaners	A lot or a great deal	All adult citizens	4.6	3.7	5.4	5,192
without leaners	A lot or a great deal	Democrats	4.1	2.8	5.3	1,848
without leaners	A lot or a great deal	Independents and leaners	4.8	3.4	6.2	1,695
without leaners	A lot or a great deal	Republicans	4.9	3.3	6.4	1,625

Shares fall from the first panel to the third, and the Democratic and Republican dots overlap in every panel, with or without leaners. In the default view the neither party row is farther right in the first two panels and has the widest intervals. Each dot is the weighted share of adult citizens who answered at or above the cut point named above the panel, and each line is a 95% interval that follows the sampling design.Note: The first view counts leaners with the party they lean toward and shows adults who name neither party separately. The second view counts only strong and not very strong identifiers as Democrats or Republicans and groups the remaining adults as independents and leaners.Note: Intervals are 95% intervals from a Taylor-series variance that uses the survey strata and primary sampling units, with the t distribution and the design degrees of freedom. A cell needs at least 100 respondents and an effective sample of 100; no plotted cell fell below that floor.Note: Refusals, break-offs and respondents not asked are left out of each share. The paper questionnaire omitted most of these questions.Source: ANES 2024 Time Series, pre-election interviews, 2024; weighted (V240107a); 95% design-based intervals

Without leaners the Democratic share at the first cut point is 14.8% (95% CI 12.5% to 17.2%) and the Republican share 15.1% (95% CI 12.4% to 17.8%), a gap of −0.3 (95% CI −3.9 to 3.4) points, and the independents and leaners who remain are at 21.3% (95% CI 18.0% to 24.6%). The sign of the gap at the first cut point differs between the two groupings, and neither interval excludes zero.

The highest share is among adults who name neither party and do not lean: 30.5% (95% CI 23.4% to 37.7%) of the 346 who answered said violence is justified at least a little. They are 6.7% (95% CI 5.8% to 7.7%) of adult citizens. They are also the group least likely to say they follow government and politics always or most of the time (27.8% against 56.5% of Democrats and 52.8% of Republicans), and adults who follow politics less often than that said violence is justified at least a little more often (20.9% (95% CI 18.3% to 23.5%) against 13.9% (95% CI 12.1% to 15.7%)). This pattern fits the finding that respondent disengagement inflates estimates of this kind [2], and it fits other differences between these groups. The attention question cannot tell the explanations apart.

On the perception question, 62.7% (95% CI 60.6% to 64.7%) of adult citizens said political violence had increased over the four-year comparison period, as did 62.3% (95% CI 59.3% to 65.4%) of Democrats and 64.4% (95% CI 61.8% to 67.1%) of Republicans, and the two party intervals overlap.

Younger and less educated adults approve of violence more often

Figure 3 splits the share saying violence is justified at least a little by age and by education. In the youngest of the four age groups 31.3% (95% CI 26.1% to 36.4%) gave that answer, against 11.0% (95% CI 8.9% to 13.0%) in the oldest, a difference of 20.3 (95% CI 14.7 to 26.0) points. Among adults with a high school education or less 22.8% (95% CI 19.7% to 25.8%) gave it, against 11.4% (95% CI 9.3% to 13.6%) among adults with a bachelor's degree or more, a difference of 11.4 (95% CI 7.9 to 14.8) points. Both intervals lie above the upper end of the party gap in the previous section.

Figure 3

Younger and less educated adults were more likely to say violence is justified than older and college-educated adults, in both parties

Share saying violence to pursue political goals is justified at least a little, by age group and education, for all adult citizens and for Democrats and Republicans with leaners; pre-election 2024

All adult citizens
Democrats
Republicans

Age

Education

Share of the group (%)

Show data

Data for Younger and less educated adults were more likely to say violence is justified than older and college-educated adults, in both parties
Dimension	Category	Group	Share (%)	Lower	Upper	Respondents
Age	18 to 29	All adult citizens	31.3	26.1	36.4	513
Age	18 to 29	Democrats	32.1	25.2	39.1	275
Age	18 to 29	Republicans	27.1	19.3	35.0	188
Age	30 to 44	All adult citizens	18.0	15.1	20.9	1,268
Age	30 to 44	Democrats	14.1	10.4	17.7	610
Age	30 to 44	Republicans	19.5	15.0	23.9	541
Age	45 to 59	All adult citizens	14.1	11.4	16.9	1,155
Age	45 to 59	Democrats	14.0	10.0	17.9	558
Age	45 to 59	Republicans	12.1	8.2	16.0	508
Age	60 or older	All adult citizens	11.0	8.9	13.0	2,027
Age	60 or older	Democrats	13.0	9.6	16.3	1,004
Age	60 or older	Republicans	8.5	5.7	11.2	935
Education	High school or less	All adult citizens	22.8	19.7	25.8	1,148
Education	High school or less	Democrats	23.6	18.6	28.6	441
Education	High school or less	Republicans	18.7	14.9	22.5	581
Education	Some college	All adult citizens	17.7	14.8	20.6	1,632
Education	Some college	Democrats	18.1	13.7	22.4	703
Education	Some college	Republicans	16.0	11.5	20.5	796
Education	Bachelor's degree or more	All adult citizens	11.4	9.3	13.6	2,370
Education	Bachelor's degree or more	Democrats	12.8	9.8	15.7	1,371
Education	Bachelor's degree or more	Republicans	9.0	6.2	11.8	893

The share falls with age and with education among all adults and within each party, and the Democratic and Republican dots overlap in every row. Each dot is the weighted share saying violence is justified at least a little, and each line is a 95% interval that follows the sampling design. Age and education overlap and are not adjusted for one another.Note: Age is age on Election Day in four groups; education is the five-category ANES summary collapsed to high school or less, some college, and a bachelor's degree or more. Respondents with no age or education answer are left out of those panels.Note: Democrats and Republicans include leaners. Intervals are 95% intervals from a Taylor-series variance that uses the survey strata and primary sampling units, with the t distribution and the design degrees of freedom. A cell needs at least 100 respondents and an effective sample of 100; no plotted cell fell below that floor.Source: ANES 2024 Time Series, pre-election interviews, 2024; weighted (V240107a); 95% design-based intervals

The age pattern appears in both parties. In the youngest group the share is 32.1% (95% CI 25.2% to 39.1%) among Democrats and 27.1% (95% CI 19.3% to 35.0%) among Republicans, and in the oldest group 13.0% (95% CI 9.6% to 16.3%) and 8.5% (95% CI 5.7% to 11.2%). Within every age group and every education group the Democratic and Republican intervals overlap. Age, education and attention to politics overlap with one another, and the figure does not separate them.

Republicans doubt the count and the 2020 result more often than Democrats do

Figure 4 shows six measures by party. Before the election 35.1% (95% CI 32.3% to 37.9%) of Republicans expected the votes to be counted not at all or only a little accurately, against 8.2% (95% CI 6.7% to 9.7%) of Democrats, a gap of 26.9 (95% CI 23.9 to 30.0) points with Republicans higher. After the election 33.2% (95% CI 30.5% to 35.8%) of Republicans and 19.9% (95% CI 17.5% to 22.2%) of Democrats said votes in this country's elections are counted fairly less often than most of the time, a gap of 13.3 (95% CI 9.7 to 16.8) points. The two gaps come from different questions with different targets, so the smaller post-election gap is not a measured change. Little or no trust in local election officials was 33.0% (95% CI 30.2% to 35.8%) among Republicans and 11.9% (95% CI 10.0% to 13.8%) among Democrats.

Figure 4

Republicans doubted the count and the 2020 result more often than Democrats did, and Democrats doubted the 2024 result more often

Share giving the doubting answer on six measures, by party identification; two measures from the pre-election interview and four from the post-election interview, 2024

Expected count inaccurate (before the election)

Votes counted fairly less than most of the time (after)

Little or no trust in local election officials (before)

Trump the legitimate 2020 winner (after)

FBI account of the January 2021 events (after)

Harris the legitimate 2024 winner (after)

Share of the group (%)

Show data

Data for Republicans doubted the count and the 2020 result more often than Democrats did, and Democrats doubted the 2024 result more often
Grouping	Measure	Group	Share (%)	Lower	Upper	Respondents
with leaners	Expected count inaccurate (before the election)	All adult citizens	21.8	20.2	23.4	5,478
with leaners	Expected count inaccurate (before the election)	Democrats	8.2	6.7	9.7	2,638
with leaners	Expected count inaccurate (before the election)	Neither party	29.9	23.1	36.7	373
with leaners	Expected count inaccurate (before the election)	Republicans	35.1	32.3	37.9	2,444
with leaners	Votes counted fairly less than most of the time (after)	All adult citizens	27.7	26.0	29.4	4,926
with leaners	Votes counted fairly less than most of the time (after)	Democrats	19.9	17.5	22.2	2,407
with leaners	Votes counted fairly less than most of the time (after)	Neither party	44.8	36.4	53.1	320
with leaners	Votes counted fairly less than most of the time (after)	Republicans	33.2	30.5	35.8	2,177
with leaners	Little or no trust in local election officials (before)	All adult citizens	22.5	20.9	24.2	5,243
with leaners	Little or no trust in local election officials (before)	Democrats	11.9	10.0	13.8	2,551
with leaners	Little or no trust in local election officials (before)	Neither party	30.9	23.3	38.4	349
with leaners	Little or no trust in local election officials (before)	Republicans	33.0	30.2	35.8	2,323
with leaners	Trump the legitimate 2020 winner (after)	All adult citizens	29.2	27.3	31.0	4,612
with leaners	Trump the legitimate 2020 winner (after)	Democrats	4.5	3.3	5.7	2,297
with leaners	Trump the legitimate 2020 winner (after)	Neither party	32.5	24.5	40.5	289
with leaners	Trump the legitimate 2020 winner (after)	Republicans	55.5	52.6	58.5	2,009
with leaners	FBI account of the January 2021 events (after)	All adult citizens	24.5	22.6	26.4	4,551
with leaners	FBI account of the January 2021 events (after)	Democrats	4.7	3.5	6.0	2,291
with leaners	FBI account of the January 2021 events (after)	Neither party	28.1	20.1	36.0	284
with leaners	FBI account of the January 2021 events (after)	Republicans	46.4	43.0	49.7	1,965
with leaners	Harris the legitimate 2024 winner (after)	All adult citizens	9.3	8.0	10.6	4,633
with leaners	Harris the legitimate 2024 winner (after)	Democrats	14.7	12.4	16.9	2,283
with leaners	Harris the legitimate 2024 winner (after)	Neither party	16.4	9.5	23.2	293
with leaners	Harris the legitimate 2024 winner (after)	Republicans	2.3	1.2	3.4	2,042
without leaners	Expected count inaccurate (before the election)	All adult citizens	21.8	20.2	23.4	5,478
without leaners	Expected count inaccurate (before the election)	Democrats	7.2	5.4	9.0	1,926
without leaners	Expected count inaccurate (before the election)	Independents and leaners	24.0	20.8	27.1	1,795
without leaners	Expected count inaccurate (before the election)	Republicans	35.7	32.5	38.8	1,734
without leaners	Votes counted fairly less than most of the time (after)	All adult citizens	27.7	26.0	29.4	4,926
without leaners	Votes counted fairly less than most of the time (after)	Democrats	18.7	15.9	21.5	1,765
without leaners	Votes counted fairly less than most of the time (after)	Independents and leaners	31.0	27.2	34.7	1,597
without leaners	Votes counted fairly less than most of the time (after)	Republicans	33.9	30.4	37.3	1,542
without leaners	Little or no trust in local election officials (before)	All adult citizens	22.5	20.9	24.2	5,243
without leaners	Little or no trust in local election officials (before)	Democrats	9.9	7.7	12.1	1,863
without leaners	Little or no trust in local election officials (before)	Independents and leaners	26.5	23.1	29.8	1,715
without leaners	Little or no trust in local election officials (before)	Republicans	32.7	29.4	36.1	1,645
without leaners	Trump the legitimate 2020 winner (after)	All adult citizens	29.2	27.3	31.0	4,612
without leaners	Trump the legitimate 2020 winner (after)	Democrats	3.9	2.5	5.3	1,684
without leaners	Trump the legitimate 2020 winner (after)	Independents and leaners	27.4	23.7	31.0	1,487
without leaners	Trump the legitimate 2020 winner (after)	Republicans	59.9	56.2	63.6	1,424
without leaners	FBI account of the January 2021 events (after)	All adult citizens	24.5	22.6	26.4	4,551
without leaners	FBI account of the January 2021 events (after)	Democrats	5.0	3.3	6.6	1,681
without leaners	FBI account of the January 2021 events (after)	Independents and leaners	25.8	22.2	29.5	1,459
without leaners	FBI account of the January 2021 events (after)	Republicans	46.1	42.3	50.0	1,400
without leaners	Harris the legitimate 2024 winner (after)	All adult citizens	9.3	8.0	10.6	4,633
without leaners	Harris the legitimate 2024 winner (after)	Democrats	15.6	13.0	18.2	1,675
without leaners	Harris the legitimate 2024 winner (after)	Independents and leaners	8.8	6.7	10.9	1,497
without leaners	Harris the legitimate 2024 winner (after)	Republicans	2.5	1.1	3.8	1,446

Republican dots are to the right of Democratic dots in five panels and to the left in the panel on the 2024 winner. The neither party dot lies between the two parties in four of the six panels. Each dot is the weighted share giving the answer named above the panel among adult citizens who answered, and each line is a 95% interval that follows the sampling design.Note: The first view counts leaners with the party they lean toward and shows adults who name neither party separately. The second view counts only strong and not very strong identifiers as Democrats or Republicans and groups the remaining adults as independents and leaners.Note: Two measures come from the pre-election interview and use the pre-election weight; the other four come from the post-election interview and use the post-election weight. Party identification comes from the pre-election interview in all six.Note: Intervals are 95% intervals from a Taylor-series variance that uses the survey strata and primary sampling units, with the t distribution and the design degrees of freedom. A cell needs at least 100 respondents and an effective sample of 100; no plotted cell fell below that floor.Note: Refusals, break-offs and respondents not asked are left out of each share. The paper questionnaire omitted most of these questions.Source: ANES 2024 Time Series, pre-election interviews (August to November 2024) and post-election interviews (November 2024 to February 2025); weighted (V240107a and V240107b); 95% design-based intervals

On the 2020 election, 55.5% (95% CI 52.6% to 58.5%) of Republicans said Trump was the legitimate winner, against 4.5% (95% CI 3.3% to 5.7%) of Democrats, a gap of 51.0 (95% CI 47.9 to 54.1) points. Without leaners the Republican share is 59.9% (95% CI 56.2% to 63.6%). Of the Republicans who said Trump was the legitimate winner, 52.5% (95% CI 48.2% to 56.8%) were very or extremely confident in that answer, which is 29.1% (95% CI 26.2% to 32.0%) of all Republicans. On the events of January 2021, 46.4% (95% CI 43.0% to 49.7%) of Republicans chose the FBI account, against 4.7% (95% CI 3.5% to 6.0%) of Democrats, and 19.0% (95% CI 16.4% to 21.5%) of Republicans chose it and were very or extremely confident.

The 2024 question points the other way. Among Democrats, 14.7% (95% CI 12.4% to 16.9%) said Harris was the legitimate winner of the election that Trump won, against 2.3% (95% CI 1.2% to 3.4%) of Republicans, a Republican-minus-Democratic gap of −12.3 (95% CI −14.8 to −9.9) points. The gap on the 2024 question is smaller in absolute value than the gap on the 2020 question, and 7.2% (95% CI 5.6% to 8.8%) of Democrats chose Harris and were very or extremely confident.

Adults who name neither party have estimates between those of the two parties on the pre-election count, trust in local officials, the 2020 result and the FBI account, and above both parties on how often votes are counted fairly: 44.8% (95% CI 36.4% to 53.1%) chose the less-than-most-of-the-time answers. Their distance from the midpoint of the two parties' estimates, positive when they are closer to the Republican answer, is 8.2 (95% CI 1.4 to 15.1) points on the count and 8.4 (95% CI 0.6 to 16.2) on trust in local officials, so on those two they are closer to the Republican answer. On the 2020 result and the FBI account the distances are 2.5 (95% CI −5.8 to 10.7) and 2.5 (95% CI −5.8 to 10.8) points, intervals that include zero, so the data do not place the group nearer either party there. The group is small and its intervals are wide.

Trump voters disputed the 2020 result more often than Harris voters disputed the 2024 result

A respondent's view of an election differs with whether the respondent's candidate won, a pattern that has been described as a winner effect in voter confidence [4] and, across countries, as losers' consent [5]. Figure 5 puts the two elections on the same footing by showing the share of voters who said the candidate who lost was the legitimate winner. In 2020, 60.6% (95% CI 57.4% to 63.9%) of respondents who recalled voting for Trump said so, against 4.2% (95% CI 3.0% to 5.4%) of those who recalled voting for Biden. In 2024, 12.7% (95% CI 10.3% to 15.0%) of Harris voters said Harris was the legitimate winner, against 1.4% (95% CI 0.4% to 2.4%) of Trump voters. Among the losing side's voters the difference between the two elections is 48.0 (95% CI 44.0 to 52.0) points, and the very or extremely confident answers show the same order (33.3% (95% CI 29.9% to 36.8%) for Trump voters in 2020 and 6.8% (95% CI 5.1% to 8.5%) for Harris voters in 2024).

Figure 5

Trump voters called Trump the legitimate 2020 winner more often (60.6%) than Harris voters called Harris the legitimate 2024 winner (12.7%)

Share saying the candidate who lost was the legitimate winner, by the voter's choice in each election; post-election interviews, November 2024 to February 2025

Show data

Data for Trump voters called Trump the legitimate 2020 winner more often (60.6%) than Harris voters called Harris the legitimate 2024 winner (12.7%)
Election and voters	Share (%)	Lower	Upper	Respondents
2020 election: Trump voters	60.6	57.4	63.9	1,495
2020 election: Biden voters	4.2	3.0	5.4	2,135
2024 election: Harris voters	12.7	10.3	15.0	1,976
2024 election: Trump voters	1.4	0.4	2.4	1,547

In each election the losing side's voters are to the right of the winning side's voters, and the 2020 losing side is farther right than the 2024 losing side. Each dot is the weighted share of voters who said the candidate who lost was the legitimate winner, Trump in 2020 and Harris in 2024, and each line is a 95% interval that follows the sampling design.Note: The 2020 vote is the vote recalled in the 2024 pre-election interview (V241106x), restricted to Biden and Trump voters. The 2024 vote is the post-election report of the presidential vote (V242067), restricted to Harris and Trump voters.Note: Intervals are 95% intervals from a Taylor-series variance that uses the survey strata and primary sampling units, with the t distribution and the design degrees of freedom. A cell needs at least 100 respondents and an effective sample of 100; no plotted cell fell below that floor.Note: Refusals, break-offs and respondents not asked are left out of each share. The paper questionnaire omitted most of these questions.Source: ANES 2024 Time Series, post-election interviews, November 2024 to February 2025; weighted (V240107b); 95% design-based intervals

The comparison does not isolate a difference between the parties. The 2020 question was asked years after that election and the 2024 question in the months after it, and the survey does not measure what candidates and officials said about either result in between. Winning does not remove all doubt about counts, either: after the 2024 election 31.6% (95% CI 28.3% to 34.9%) of Trump voters said votes are counted fairly less often than most of the time, against 13.6% (95% CI 11.3% to 15.8%) of Harris voters. On that question the difference between the sides was in place after the result, although the data cannot show how much it changed.

Strong Republicans differ from Republican leaners on the 2020 result only

Figure 6 uses the seven categories of the party identification question: strong, not very strong and leaning for each party, and neither party. Approval of violence at the first cut point is not higher among strong identifiers than among the other identifiers of their party: strong Democrats minus other Democrats is −4.7 (95% CI −9.2 to −0.2) points and strong Republicans minus other Republicans is −1.2 (95% CI −5.7 to 3.3). Among Democrats the leaners are above the strong identifiers, at 23.4% (95% CI 17.2% to 29.6%), a difference of 8.6 (95% CI 1.9 to 15.3) points, and adults who name neither party have the highest estimate, 30.5% (95% CI 23.4% to 37.7%). Approval of violence does not rise with strength of identification in either party.

Figure 6

Strong Republicans stood apart from Republican leaners on the 2020 result only, and approval of violence did not rise with strength of identification

Share giving the answer named above each panel, by the seven categories of party identification; pre-election and post-election interviews, 2024

Violence justified at least a little

Expected count inaccurate

Trump the legitimate 2020 winner

FBI account of the January 2021 events

Share of the group (%)

Show data

Data for Strong Republicans stood apart from Republican leaners on the 2020 result only, and approval of violence did not rise with strength of identification
Measure	Party identification	Share (%)	Lower	Upper	Respondents
Violence justified at least a little	Strong Democrat	14.8	11.8	17.8	1,252
Violence justified at least a little	Not very strong Democrat	15.0	10.9	19.0	596
Violence justified at least a little	Leans Democratic	23.4	17.2	29.6	684
Violence justified at least a little	Neither party	30.5	23.4	37.7	346
Violence justified at least a little	Leans Republican	14.9	11.6	18.2	665
Violence justified at least a little	Not very strong Republican	16.5	11.4	21.5	539
Violence justified at least a little	Strong Republican	14.4	11.1	17.6	1,086
Expected count inaccurate	Strong Democrat	6.9	4.7	9.1	1,311
Expected count inaccurate	Not very strong Democrat	7.8	4.8	10.8	615
Expected count inaccurate	Leans Democratic	10.7	7.4	14.1	712
Expected count inaccurate	Neither party	29.9	23.1	36.7	373
Expected count inaccurate	Leans Republican	33.9	28.7	39.2	710
Expected count inaccurate	Not very strong Republican	27.6	22.5	32.7	577
Expected count inaccurate	Strong Republican	39.9	35.9	44.0	1,157
Trump the legitimate 2020 winner	Strong Democrat	2.9	1.7	4.0	1,154
Trump the legitimate 2020 winner	Not very strong Democrat	5.8	2.7	9.0	530
Trump the legitimate 2020 winner	Leans Democratic	6.2	3.9	8.5	613
Trump the legitimate 2020 winner	Neither party	32.5	24.5	40.5	289
Trump the legitimate 2020 winner	Leans Republican	45.7	39.6	51.9	585
Trump the legitimate 2020 winner	Not very strong Republican	45.0	38.0	52.0	466
Trump the legitimate 2020 winner	Strong Republican	67.7	63.4	71.9	958
FBI account of the January 2021 events	Strong Democrat	5.0	3.0	7.0	1,152
FBI account of the January 2021 events	Not very strong Democrat	4.8	2.2	7.5	529
FBI account of the January 2021 events	Leans Democratic	4.1	2.5	5.8	610
FBI account of the January 2021 events	Neither party	28.1	20.1	36.0	284
FBI account of the January 2021 events	Leans Republican	46.9	40.8	53.0	565
FBI account of the January 2021 events	Not very strong Republican	35.5	28.1	42.9	466
FBI account of the January 2021 events	Strong Republican	51.8	47.3	56.3	934

In each election panel the Democratic categories sit at the left and strong Republicans have the highest estimate, but their lead over Republican leaners is distinguishable from zero only on the 2020 winner. In the violence panel the dots do not rise toward either end, and the highest estimates are for Democratic leaners and adults naming neither party. Each dot is a weighted share among adult citizens who answered, and each line is a 95% interval that follows the sampling design.Note: The count and violence measures come from the pre-election interview and the 2020 and FBI measures from the post-election interview. Each uses the weight for its interview. Intervals are 95% intervals from a Taylor-series variance that uses the survey strata and primary sampling units, with the t distribution and the design degrees of freedom. A cell needs at least 100 respondents and an effective sample of 100; no plotted cell fell below that floor.Note: Refusals, break-offs and respondents not asked are left out of each share. The paper questionnaire omitted most of these questions.Source: ANES 2024 Time Series, pre-election interviews (August to November 2024) and post-election interviews (November 2024 to February 2025); weighted (V240107a and V240107b); 95% design-based intervals

The three election measures are low and close together across the three Democratic categories and higher among Republicans, where strong identifiers differ from leaners on only one measure. Strong Republicans said Trump was the legitimate winner of 2020 at 67.7% (95% CI 63.4% to 71.9%), against 45.7% (95% CI 39.6% to 51.9%) among Republican leaners, a difference of 21.9 (95% CI 14.6 to 29.2) points. On the expected count the two categories are at 39.9% (95% CI 35.9% to 44.0%) and 33.9% (95% CI 28.7% to 39.2%), a difference of 6.0 (95% CI −0.7 to 12.7) points, and on the FBI account at 51.8% (95% CI 47.3% to 56.3%) and 46.9% (95% CI 40.8% to 53.0%), a difference of 4.9 (95% CI −2.6 to 12.4) points, and both of those intervals include zero. Strong Republicans have the highest estimate of the seven categories on all three measures, but the data distinguish them from Republican leaners only on the 2020 result. The categories do not form a ladder either: not very strong Republicans chose the FBI account less often than Republican leaners, by 11.4 (95% CI 2.1 to 20.6) points.

Cold feelings toward the other party go with count doubts among Republicans, not with approval of violence

The feeling thermometer rates a party from the coldest to the warmest answer. Chapter 2.1, Feelings toward the other party follows those ratings over time and owns that trend. Here the out-party rating is a Democrat's rating of the Republican Party or a Republican's rating of the Democratic Party, taken in the same pre-election interview as the violence and count questions. Any pattern below is an association measured at one time, and it does not say which came first or whether one produced the other. Chapter 2.2, Negative partisanship asks that question for voting.

The coldest bin, defined in Figure 7, holds 37.6% (95% CI 35.0% to 40.1%) of Democrats and 40.9% (95% CI 38.5% to 43.3%) of Republicans. Among Republicans in the coldest bin, 47.9% (95% CI 43.6% to 52.2%) expected the votes to be counted not at all or only a little accurately, against 32.4% (95% CI 26.8% to 38.0%) in the middle bin and 17.5% (95% CI 13.0% to 22.0%) in the least cold bin. Each ten-point step colder goes with 5.6 (95% CI 4.4 to 6.8) points more Republican doubt about the count, and with 5.9 (95% CI 4.7 to 7.2) points after holding age, education, sex, race and ethnicity and strength of party identification fixed in a linear probability model. Among Democrats the slope is −0.6 (95% CI −1.5 to 0.2) points, an interval that includes zero.

Figure 7

Colder ratings of the other party went with more doubt about the count among Republicans, and with no more approval of violence in either party

Share giving each answer by the party member's rating of the other party on the 0 to 100 feeling thermometer, Democrats and Republicans with leaners; pre-election 2024

Democrats
Republicans

Violence justified at least a little

Expected count inaccurate

Share of the group (%)

Show data

Data for Colder ratings of the other party went with more doubt about the count among Republicans, and with no more approval of violence in either party
Measure	Out-party rating	Party	Share (%)	Lower	Upper	Respondents
Violence justified at least a little	Out-party rating 0 to 10 (coldest)	Democrats	15.5	11.8	19.2	1,024
Violence justified at least a little	Out-party rating 0 to 10 (coldest)	Republicans	14.3	11.1	17.5	1,013
Violence justified at least a little	Out-party rating 11 to 30	Democrats	14.2	10.8	17.6	933
Violence justified at least a little	Out-party rating 11 to 30	Republicans	14.2	10.2	18.2	737
Violence justified at least a little	Out-party rating 31 to 100 (least cold)	Democrats	23.1	17.0	29.2	565
Violence justified at least a little	Out-party rating 31 to 100 (least cold)	Republicans	17.6	12.5	22.6	532
Expected count inaccurate	Out-party rating 0 to 10 (coldest)	Democrats	7.2	4.7	9.8	1,077
Expected count inaccurate	Out-party rating 0 to 10 (coldest)	Republicans	47.9	43.6	52.2	1,095
Expected count inaccurate	Out-party rating 11 to 30	Democrats	6.5	4.0	9.1	960
Expected count inaccurate	Out-party rating 11 to 30	Republicans	32.4	26.8	38.0	774
Expected count inaccurate	Out-party rating 31 to 100 (least cold)	Democrats	10.7	7.4	14.0	590
Expected count inaccurate	Out-party rating 31 to 100 (least cold)	Republicans	17.5	13.0	22.0	567

In the count panel the Republican dots fall from the coldest bin to the least cold bin while the Democratic dots stay low. In the violence panel neither party's dots rise toward the coldest bin. Each dot is a weighted share among adult citizens in the bin, and each line is a 95% interval that follows the sampling design. Ratings and answers come from the same interview, so this is an association measured at one time.Note: The out-party rating is the Democratic Party rating (V241166) for Republicans and the Republican Party rating (V241167) for Democrats. The top bin runs from 31 to 100, and so includes ratings at the scale midpoint of 50 and above, because the narrower top bins had too few respondents to plot.Note: Intervals are 95% intervals from a Taylor-series variance that uses the survey strata and primary sampling units, with the t distribution and the design degrees of freedom. A cell needs at least 100 respondents and an effective sample of 100; no plotted cell fell below that floor.Source: ANES 2024 Time Series, pre-election interviews, 2024; weighted (V240107a); 95% design-based intervals

For approval of violence at the first cut point the slope per ten-point step colder is −0.4 (95% CI −1.5 to 0.6) points among Republicans and −1.8 (95% CI −3.2 to −0.4) among Democrats, and 0.3 (95% CI −0.9 to 1.5) and −0.9 (95% CI −2.3 to 0.5) after the same adjustment. The Democratic estimate without adjustment has an interval below zero, so each step colder goes with fewer Democrats saying violence is justified, by the margin just given and the opposite of what an account in which hostility produces approval would predict, and the adjusted interval includes zero. No interval for violence reaches past 3.2 points per step in either direction. The link between cold ratings and count doubts among Republicans is not evidence that hostility produces doubt. Republicans who rate the Democratic Party coldest also differ from other Republicans in ways the adjustment does not hold fixed, such as what they have heard about the count, which the survey does not measure.

Answers differ by sample group on some items

The 2024 ANES interviewed three groups separately: a fresh sample interviewed mainly in person, a fresh sample interviewed mainly on the web and a panel interviewed on the web. Each has its own weight and design variables. The user guide cautions against attributing differences among interview modes to the mode itself, because respondents within a sample group were not assigned a mode at random [1]. Figure 8 repeats four estimates for each group.

Figure 8

Republican doubt about the count and Democratic doubt about the 2024 result were lower in the in-person sample than in the web sample, and the party gap on the 2020 result did not differ

Share giving the answer named above each panel, by ANES sample group, Democrats and Republicans with leaners; 2024 pre-election and post-election interviews

Democrats
Republicans

Violence justified at least a little

Expected count inaccurate

Trump the legitimate 2020 winner

Harris the legitimate 2024 winner

Share of the group (%)

Show data

Data for Republican doubt about the count and Democratic doubt about the 2024 result were lower in the in-person sample than in the web sample, and the party gap on the 2020 result did not differ
Measure	Sample group	Party	Share (%)	Lower	Upper	Respondents
Violence justified at least a little	Fresh in-person sample	Democrats	17.5	12.8	22.3	522
Violence justified at least a little	Fresh in-person sample	Republicans	14.7	11.0	18.3	436
Violence justified at least a little	Fresh web sample	Democrats	19.0	15.7	22.4	956
Violence justified at least a little	Fresh web sample	Republicans	15.2	12.0	18.5	862
Violence justified at least a little	Panel (web)	Democrats	12.7	9.2	16.3	1,054
Violence justified at least a little	Panel (web)	Republicans	12.9	9.6	16.3	992
Expected count inaccurate	Fresh in-person sample	Democrats	6.9	3.9	9.8	524
Expected count inaccurate	Fresh in-person sample	Republicans	26.2	20.3	32.1	437
Expected count inaccurate	Fresh web sample	Democrats	7.7	5.4	10.1	969
Expected count inaccurate	Fresh web sample	Republicans	40.1	35.9	44.4	882
Expected count inaccurate	Panel (web)	Democrats	8.7	6.1	11.3	1,058
Expected count inaccurate	Panel (web)	Republicans	40.0	35.9	44.1	999
Trump the legitimate 2020 winner	Fresh in-person sample	Democrats	3.8	1.7	5.9	465
Trump the legitimate 2020 winner	Fresh in-person sample	Republicans	55.1	49.7	60.5	381
Trump the legitimate 2020 winner	Fresh web sample	Democrats	4.9	3.2	6.6	831
Trump the legitimate 2020 winner	Fresh web sample	Republicans	55.2	50.4	60.0	712
Trump the legitimate 2020 winner	Panel (web)	Democrats	5.1	2.6	7.6	1,001
Trump the legitimate 2020 winner	Panel (web)	Republicans	58.7	53.4	64.0	916
Harris the legitimate 2024 winner	Fresh in-person sample	Democrats	9.3	5.2	13.4	457
Harris the legitimate 2024 winner	Fresh in-person sample	Republicans	0.3	0.0	0.7	389
Harris the legitimate 2024 winner	Fresh web sample	Democrats	20.4	16.2	24.5	826
Harris the legitimate 2024 winner	Fresh web sample	Republicans	3.5	1.7	5.3	722
Harris the legitimate 2024 winner	Panel (web)	Democrats	14.3	10.7	18.0	1,000
Harris the legitimate 2024 winner	Panel (web)	Republicans	1.6	0.3	2.9	931

In the in-person sample the Republican dot on the count is to the left of the web and panel dots, and the Democratic dot on the 2024 winner is to the left of the web dot. On the 2020 winner the dots are close across groups. Each dot is a weighted share using the sample group's own weight, and each line is a 95% interval from that group's strata and primary sampling units. Respondents were not assigned a mode at random, so a difference may reflect mode, sample or both.Note: Each sample group uses its own weights and variance variables: V240101 for the fresh in-person sample, V240102 for the fresh web sample (web interviews, without the paper cases) and V240106 for the panel. The in-person sample excludes Alaska and Hawaii. Intervals use the t distribution with each group's own design degrees of freedom, which are smallest for the in-person sample and the panel.Note: Refusals, break-offs and respondents not asked are left out of each share. The paper questionnaire omitted most of these questions.Source: ANES 2024 Time Series, pre-election interviews (August to November 2024) and post-election interviews (November 2024 to February 2025); weighted (V240107a and V240107b); 95% design-based intervals

Republican doubt about the count differs with the sample group: 26.2% (95% CI 20.3% to 32.1%) expected an inaccurate count in the in-person sample, against 40.1% (95% CI 35.9% to 44.4%) in the web sample and 40.0% (95% CI 35.9% to 44.1%) in the panel, and the web sample minus the in-person sample is 13.9 (95% CI 6.7 to 21.1) points. The Democratic answer that Harris was the legitimate winner of 2024 is lower in the in-person sample than in the web sample, at 9.3% (95% CI 5.2% to 13.4%) against 20.4% (95% CI 16.2% to 24.5%), a difference of 11.1 (95% CI 5.3 to 16.8) points, and the panel lies between them at 14.3% (95% CI 10.7% to 18.0%). The Republican-minus-Democratic gap on the 2020 result is similar across the groups, at 51.3 (95% CI 46.1 to 56.5), 50.3 (95% CI 45.2 to 55.4) and 53.6 (95% CI 47.7 to 59.6) points in the in-person, web and panel groups. The gap on the expected count is smaller in the in-person group, at 19.3 (95% CI 13.4 to 25.3) points, than in the web and panel groups, at 32.4 (95% CI 27.6 to 37.3) and 31.3 (95% CI 26.3 to 36.2). Approval of violence at the first cut point is lower in the panel, at 13.4% (95% CI 10.8% to 15.9%), than in the web sample, at 18.5% (95% CI 16.2% to 20.8%).

Interviewer presence can change answers to sensitive questions [6], and the panel members differ from fresh respondents in having been interviewed in 2016 and 2020, so the data cannot say whether these differences come from the interview mode or from the people in each group. The estimates in the rest of the chapter pool the groups with the full-sample weights, so a pooled share on an item that differs by group averages over that spread.

Both parties call the norms important, with gaps of different sizes

Figure 9 gives the share who said each of four practices is very or extremely important to the United States maintaining a strong democracy, and the shares answering the questions about presidents and the press. For news organizations being free to criticize political leaders the Democratic share is 65.4% (95% CI 62.7% to 68.1%) and the Republican share 50.8% (95% CI 47.4% to 54.2%), a Republican-minus-Democratic difference of −14.6 (95% CI −18.8 to −10.3) points. For the other three practices the Republican shares run from 69.6% to 86.1% and the Democratic shares from 78.2% to 91.7%, with Republicans lower on each.

Figure 9

Republicans were less likely than Democrats to rate four democratic practices important and more likely to favor presidential freedom from Congress and the courts or limits on journalists

Share giving the stated answer on six norm questions, by party identification with leaners; pre-election 2024

Democrats
Neither party
Republicans

Show data

Data for Republicans were less likely than Democrats to rate four democratic practices important and more likely to favor presidential freedom from Congress and the courts or limits on journalists
Question and answer	Group	Share (%)	Lower	Upper	Respondents
News organizations free to criticize leaders: very or extremely important	Democrats	65.4	62.7	68.1	2,548
News organizations free to criticize leaders: very or extremely important	Neither party	42.9	35.0	50.8	347
News organizations free to criticize leaders: very or extremely important	Republicans	50.8	47.4	54.2	2,320
Branches check one another: very or extremely important	Democrats	83.9	81.9	86.0	2,545
Branches check one another: very or extremely important	Neither party	57.7	50.4	65.0	351
Branches check one another: very or extremely important	Republicans	78.5	75.6	81.4	2,325
Consequences for misconduct: very or extremely important	Democrats	91.7	90.0	93.4	2,549
Consequences for misconduct: very or extremely important	Neither party	74.6	68.1	81.0	352
Consequences for misconduct: very or extremely important	Republicans	86.1	83.7	88.5	2,325
Agreement on basic facts: very or extremely important	Democrats	78.2	75.7	80.6	2,540
Agreement on basic facts: very or extremely important	Neither party	53.4	45.2	61.6	352
Agreement on basic facts: very or extremely important	Republicans	69.6	66.5	72.8	2,321
Helpful if presidents could ignore Congress and the courts	Democrats	16.0	13.7	18.4	2,531
Helpful if presidents could ignore Congress and the courts	Neither party	17.9	11.8	24.1	347
Helpful if presidents could ignore Congress and the courts	Republicans	20.0	17.7	22.3	2,315
Favor officials restricting journalists' access	Democrats	11.1	9.1	13.2	2,537
Favor officials restricting journalists' access	Neither party	13.2	8.0	18.4	349
Favor officials restricting journalists' access	Republicans	15.8	13.6	18.0	2,314

On the four importance questions the Republican dot is to the left of the Democratic dot, and on the two questions about presidents and journalists it is to the right, at lower levels in both parties. Each dot is the weighted share giving the answer named in the row, and each line is a 95% interval that follows the sampling design. Adults naming neither party are shown for reference.Note: Democrats and Republicans include leaners. The four importance questions ask how important each practice is to the United States maintaining a strong democracy; very or extremely important are the top two of five answers. Intervals are 95% intervals from a Taylor-series variance that uses the survey strata and primary sampling units, with the t distribution and the design degrees of freedom. A cell needs at least 100 respondents and an effective sample of 100; no plotted cell fell below that floor.Note: Refusals, break-offs and respondents not asked are left out of each share. The paper questionnaire omitted most of these questions.Source: ANES 2024 Time Series, pre-election interviews, 2024; weighted (V240107a); 95% design-based intervals

Among Democrats 16.0% (95% CI 13.7% to 18.4%) and among Republicans 20.0% (95% CI 17.7% to 22.3%) said it would be helpful if presidents could work without paying attention to what Congress and the courts say, a difference of 4.0 (95% CI 0.9 to 7.0) points. Among Democrats 11.1% (95% CI 9.1% to 13.2%) and among Republicans 15.8% (95% CI 13.6% to 18.0%) favored restricting journalists' access to information, a difference of 4.7 (95% CI 1.8 to 7.6) points. The presidential question is related to the strong-leader questions asked in other surveys, and Chapter 6.7, Support for democracy takes up those questions across countries. Chapter 6.6, Tolerance of speech covers willingness to let speakers speak.

These are stated commitments to a practice in general. A respondent can call a practice important and still accept a violation by a favored official, and the survey has no question that pairs a norm with a violation by the respondent's own side.

What these data cannot show

One cross-section cannot show change. The cumulative ANES file used elsewhere in this book has no item on political violence, on votes counted accurately or on the legitimacy of an election, so no earlier ANES study on disk is comparable. The 2024 file carries identifiers that link panel respondents to the 2016 and 2020 studies, but those studies are not in the data used here. Whether approval of violence or doubt about counts rose or fell needs surveys that repeat the questions over time. Projects that field repeated survey questions on political violence, democratic norms and the legitimacy of elections include the Polarization Research Lab (America's Political Pulse) and Bright Line Watch. This chapter does not use them and reports nothing from them.

The file records no action. It has no question about violence, threats or harassment by the respondent, or about taking part in the events of January 2021. The only actions it records are ordinary participation such as voting and joining a protest march, rally or demonstration. Approval in the abstract is not evidence that a respondent would act, and an estimate of how many have acted would need administrative records, such as law-enforcement case files, that the survey cannot supply.

The violence question does not say who would be hurt, for what goal or by what act, so the shares cannot be sorted into approval of attacks on officials, on the other party or on no one in particular. The answer "a little" accounts for 6.4% (95% CI 5.3% to 7.6%) of adult citizens, about as many as the 6.2% (95% CI 5.1% to 7.3%) who chose "a moderate amount", and it is the answer whose meaning the wording leaves most open. The data cannot separate respondents who read the question carefully from respondents who did not, and the attention pattern above is a description and not a correction.

The thermometer, strength and age results are associations and make no claim about cause. Neither the party gap nor the gaps by age and education are adjusted for one another, except where the text says so. The 2020 and 2024 legitimacy questions were asked at different distances from their elections and carry different histories of statements by candidates and officials, which the survey did not measure, so a difference between them mixes timing, who won and party.

The population is adult citizens at residential addresses. It leaves out noncitizens, people in institutions and group quarters, and, for the in-person sample, residents of Alaska and Hawaii. Respondents who refused a question, and paper respondents who were not asked, are left out of its shares, and nonresponse to the survey itself is handled only through the weights. The intervals describe sampling variation. They do not include the differences among sample groups shown above or the effect of question wording.

Data sources

ANES 2024 Time Series2024 Time Series full release (May 19, 2026): political violence, vote counts, election legitimacy, the events of January 2021 and democratic norms, with party identification, age, education, race, sex and party thermometersYears: 2024 (pre-election and post-election interviews) · Weights: V240107a (pre-election) and V240107b (post-election) with PSU V240107c and stratum V240107d; sample-group weights V240101, V240102 and V240106 · License: Free download after accepting the ANES terms of use; cite the release.

All data sources

Methods

Sample. ANES 2024 Time Series, full release of May 19, 2026: 5,521 respondents in the pre-election study, 4,964 of whom completed the post-election interview. The target population is U.S. citizens aged 18 or older at residential addresses (48 contiguous states and the District of Columbia for the fresh in-person sample, 50 states and the District for the fresh web sample and the panel). The file has a fresh in-person sample (1,042 respondents), a fresh web sample (2,308, including paper cases) and a panel of 2016 and 2020 respondents (2,171). Paper-and-pencil respondents (245 before the election, 200 after) were not asked most of the questions used here. Weights and design. Pre-election items use V240107a and post-election items V240107b, with variance PSU V240107c and variance stratum V240107d, as the user guide directs for the full release. The sample-group figure uses each sample group's own weight, PSU and stratum (V240101a, V240101b, V240101c and V240101d for the fresh in-person sample; V240102 for the fresh web sample, which excludes the paper cases; V240106 for the panel). Estimates and intervals are not computed with the model-based Kish formula used in other chapters. Estimator and intervals. A share is the weighted ratio of an indicator within a domain. The variance is a Taylor-series linearization with one-stage stratified cluster sampling and no finite-population correction. Domains stay inside the design (a subpopulation), so a stratum is never left with one PSU by dropping rows. A difference of two shares in one sample linearizes the difference, which keeps the covariance that shared PSUs create. Intervals are 95% intervals from the t distribution with the design degrees of freedom: PSU-stratum units minus strata, 2,339 for the pre-election design (2,473 units, 134 strata) and 2,000 for the post-election design (2,134 units, 134 strata). A difference between two sample groups combines two independent variances and uses Welch-Satterthwaite degrees of freedom. An interval is cut at 0 and 100 for a share; one cell, Republicans in the fresh in-person sample on the 2024 winner, is cut at 0. For the 161 shares plotted in the figures, the design effect (the design-based variance divided by the model-based variance with the same weights) runs from 0.5 to 1.7 with a median of 1.1, so design-based intervals are narrower than model-based intervals for some cells and wider for others. Items and cut points. Violence justified (V241579): at least a little is answers 2 to 5, a moderate amount or more is 3 to 5, a lot or a great deal is 4 and 5. Expected count (V241314): not at all or a little accurately is answers 1 and 2. Votes counted fairly (V242207): less often than most of the time is answers 3 to 5. Trust in local election officials (V241315): not at all or a little is 1 and 2. 2020 winner (V242567): answer 1 is Trump. Events of January 2021 (V242571): answer 2 is the FBI account. 2024 winner (V242575): answer 2 is Harris. Very or extremely confident is answers 4 and 5 of the follow-up (V242568, V242572, V242576). Norm items: very or extremely important is answers 4 and 5 (V241324 to V241327); helpful is answer 1 of V241328; favor is answer 1 of V241331. Refusals, don't-know answers, break-offs and respondents not asked are missing (the loader sets negative codes to missing) and are excluded from the denominator of each share. Groups. Party is the pre-election seven-point summary (V241227x): with leaners, answers 1 to 3 are Democrats, 5 to 7 Republicans and 4 neither party; without leaners, answers 1 and 2 are Democrats, 6 and 7 Republicans and 3 to 5 independents and leaners. Age is age on Election Day (V241458x, with the paper age V241457 where only that was given) in four groups; education is V241465x collapsed to high school or less, some college, and a bachelor's degree or more. The 2020 vote is the recalled vote (V241106x), the 2024 vote the post-election report (V242067). Attention to politics is V241004 (always or most of the time against about half the time or less). Thermometer and regressions. The out-party rating is V241166 for Republicans and V241167 for Democrats, grouped as 0 to 10, 11 to 30 and 31 to 100 (the top bin of 51 to 100 had an effective sample under 100 for some cells and was merged). Slopes come from weighted linear probability models fit separately within each party, with the Taylor-series sandwich variance on the same strata and PSUs. The adjusted model adds age group, education, sex (V241550, which was also asked on paper), race and ethnicity (V241501x) and strength of party identification (strong, not very strong, leaner). Suppression. A cell needs at least 100 respondents and a Kish effective sample of at least 100; no plotted cell fell below the floor. Checks. A set of 66 estimates and intervals (headline shares, party, age and education gaps, losing-side shares, strength-of-identification contrasts, neither-party distances from the party midpoint, sample-group cells and five regression slopes) was reproduced with the survey package in R (svydesign with nest = TRUE, subset domains, svycontrast and svyglm); every estimate and interval end agrees to the digits printed, and verify_independent.py in the chapter folder repeats the comparison. Within a sample group the interview mode was not assigned at random, so mode effects are not separated from sample differences.

References

American National Election Studies (2026). ANES 2024 Time Series Study Full Release [dataset and documentation], May 19, 2026 version. Stanford University and the University of Michigan. electionstudies.org/data-center/
Westwood, Sean J.; Grimmer, Justin; Tyler, Matthew; Nall, Clayton (2022). Current research overstates American support for political violence. Proceedings of the National Academy of Sciences 119(12): e2116870119. doi:10.1073/pnas.2116870119
Kalmoe, Nathan P.; Mason, Lilliana (2022). Radical American Partisanship: Mapping Violent Hostility, Its Causes, and the Consequences for Democracy. University of Chicago Press.
Sinclair, Betsy; Smith, Steven S.; Tucker, Patrick D. (2018). "It's Largely a Rigged System": Voter Confidence and the Winner Effect in 2016. Political Research Quarterly 71(4): 854-868. doi:10.1177/1065912918768006
Anderson, Christopher J.; Blais, Andre; Bowler, Shaun; Donovan, Todd; Listhaug, Ola (2005). Losers' Consent: Elections and Democratic Legitimacy. Oxford University Press. doi:10.1093/0199276382.001.0001
Tourangeau, Roger; Yan, Ting (2007). Sensitive Questions in Surveys. Psychological Bulletin 133(5): 859-883. doi:10.1037/0033-2909.133.5.859

How to cite

@misc{singh2026politicalviolenceanddemocraticnorms,
  author = {Singh, Vishal},
  title = {Political violence and democratic norms},
  year = {2026},
  url = {https://vishalsingh.org/politics/political-violence-and-democratic-norms},
  note = {In The Public Mind. Published 3 October 2026.}
}

What each question asks

No party gap in approval of violence is distinguishable from zero

Younger and less educated adults approve of violence more often

Republicans doubt the count and the 2020 result more often than Democrats do

Trump voters disputed the 2020 result more often than Harris voters disputed the 2024 result

Strong Republicans differ from Republican leaners on the 2020 result only

Cold feelings toward the other party go with count doubts among Republicans, not with approval of violence

Answers differ by sample group on some items

Both parties call the norms important, with gaps of different sizes

What these data cannot show

Data sources

Methods

References

How to cite

Reading paths