Article Summary
Introduction
Levels of political violence support are helpful to monitor for pragmatic, diagnostic reasons. For example, if a large section of a democratic society favors the use of force and coercion to achieve their political aims, the polity may be liable to fracture and dissolve into conflict, turbulence, and animosity. Just as all manner of physical ailments are signified by common signs and symptoms—fevers, headaches, fatigue—so too are diseases of the body politic made manifest in sectarianism, intergroup animus, and political violence. In other words, studying how many people support political violence is a bit like taking our societal temperature. Troublingly, prior to this article’s publication, a widely publicized series of studies found that roughly 40% of Americans support politically motivated violence, which if correct, would be a serious signal of an unhealthy democracy. The contribution of this paper is to scrutinize and ultimately disconfirm this worrying claim.
There are few clear-cut, easily interpretable metrics for societal attitudes. How social scientists ask and frame their questions is known to influence people’s answers. Consequently, measuring attitudes—and, especially, attitudes about sensitive topics, like violence—is an error-prone endeavor that requires multiple points of observation to support dependable conclusions. This is especially important when attempting to measure a rare phenomenon, since most sources of measurement error will bias your estimate of that rare phenomenon’s prevalence upwards rather than merely making it less precise. Accounting for these considerations, the authors seek to provide a more accurate assessment of political violence support in Americans. The questions they answer in this paper are:
- How do ambiguous survey questions cause overestimates of support for violence?
- How do disengaged survey respondents cause upward bias in reported support for violence?
- How do these sources of error affect estimates of the correlates and consequences of support for violence?
- How does support for violence vary by the severity and specificity of the violent act?
Analytical Approach
Four large survey experiments (n = 4,904) were used to clarify the influence of question ambiguity and respondent disengagement on levels of measured violence support. Data were collected throughout 2021, with the first round of collection beginning two days after the violence of January 6th at the U.S. capitol. Quota sampling was used, meaning that respondents were chosen in a proportion to match the age, sex, and race/ethnicity characteristics of the U.S. public. Other data were collected via YouGov.
The authors made the following design choices to reduce the likelihood of mismeasurement:
- To use detailed vignettes that describe specific acts of violence and their perpetrators and victims, rather than vague questions that leave the definition of violence to respondents.
- To vary the level of violence (e.g., property damage versus assault versus murder) and the perpetrator’s partisan motivation (included or omitted) across vignettes.
- To measure respondents’ engagement with the survey by testing their retention of information from the vignettes and to differentially examine violence support estimates for engaged and unengaged respondents.
- To use different outcome questions that capture respondents’ justification, support, and preference for criminal charges for the violent acts, using different scales (e.g., Likert or yes/no).
- To test how their measures relate to other predictors of support for violence, such as partisan identity, aggression, and political engagement.
The following table shows all randomized conditions for each study:
| Study | Vignette | Partisan details | Sample size |
|---|---|---|---|
| Study 1 | A Democrat driving a car into a group of Republicans registering voters in Florida or a Republican driving a car into a group of Democratic protesters in Oregon. | Included or omitted | 1,002 |
| Study 2 | A man threatening a group of [Republicans/Democrats/locals] who were meeting in a [Democratic/Republican/quiet] part of town, ultimately shooting a woman while calling her a “[Democratic/Republican/] maniac.” | Included or omitted | 1,000 |
| Study 3 | Replication of Study 2. | Included | 1,863 |
| Study 4 | A man convicted of a crime (protesting without a permit, vandalism, petty assault, arson, assault with a deadly weapon, or murder) that was partisan and committed against a member of the opposing party. | Included | 1,009 |
Main Findings
In contrast to some of the existing research on support for political violence, this study found that respondents overwhelmingly reject both political and nonpolitical violence, regardless of partisan details. Compared to previously existing survey evidence, the present study provides a median violence prevalence estimate nearly 6 times smaller than the previous median estimate (18.5% versus 2.9%).
In addition, disengaged respondents show much higher measured support for political violence than engaged respondents. This is because disengaged respondents tend to respond randomly or choose extreme options on dichotomous scales. Respondent disengagement also inflates the relation between support for violence and previously identified correlates by 4x.
The degree of violence measured also had an impact. Support for violence varies substantially depending on the severity and specificity of the violent act, with minimal support for assault and murder and higher support for low-level property crimes.
Finally, providing partisan details about who committed or suffered from the violent act does not substantially affect support for violence relative to identical but apolitical acts. Aggression is a relatively strong predictor of support for political violence, but the magnitude of this effect, too, is inflated by disengaged respondents. By contrast, support for violence is weakly related to other relevant individual characteristics, such as partisanship, affective polarization, and political engagement.
To zoom out, these findings sustain several notable takeaways. First, support for political violence is not a mass phenomenon in the United States, but rather a niche preference among a small segment of the population who often also support violence in general. This runs counter to the idea that soaring support for political violence is symptomatic of larger societal ailments. Second, methodological flaws in prior surveys have created misleading estimates of support for political violence that overstate its prevalence and intensity among Americans. This finding illustrates the importance of careful, rigorous measurement of social phenomena. It is not enough to administer a handful of loosely related questions about a sensitive issue; measurement is difficult and requires devoted research attention. Third, support for political violence is contingent on the context and consequences of the violent act, rather than on abstract moral judgments or partisan motivations. This finding certainly does not hold true for everyone and merits further examination, yet most people appear to be relatively pragmatic, rather than deontological or ideological, in their assessments of political violence.
Questions left unanswered
This paper sought to accurately estimate support for political violence in the mass public and found that it is quite low. Understanding the motives of those few who do support violence is an avenue for future research. In this paper, the authors did not emphasize larger, systematic modalities of political violence visible in certain American and international political events (e.g., riots, coordinated acts of terrorism, attempted coups). Further, they did not contextualize or seek to explicate the motives and ideology of the violent offenders in their vignettes, only denoting them as partisans. Assessing violence in response to specific ideological ends (e.g., assassinating a Senator who plans to vote on a consequential policy; seeking to jumpstart a communist or fascist revolution; bombing a black church; shooting anti-trans educators) might provoke different patterns of results than assessing partisan violence in the (relative) ideological abstract. Supporters of political violence (however rare they are) may be motivated by partisanship, but many perpetrators of political violence appear to be captured by specific, radicalizing belief systems. These considerations may also extend to other contextual factors, such as the socio-cultural characteristics of the violent perpetrator. We still don’t know whether embedding these cultural factors in the violence-related vignettes will shape measured support for political violence. Although the context of partisanship doesn’t appear to unduly influence the results, perhaps people have a contextual line across which they will, indeed, endorse political violence (e.g., to combat Nazis in 1930s Germany). Identifying a way of assessing this line, its variation across individuals, and the individual differences that push the line in one direction or another (if one exists) seems to be a fruitful avenue for further investigation.
Methods and Analysis
Was the study and its analyses pre-registered?: Yes
Did the study rely on proxy variables to measure polarization?: No
Were standard p-value thresholds used (p<.05 or 95% Confidence Intervals that don’t overlap zero)?: Yes
- Largest p-value presented as significant: 0.05
Were correlational results interpreted with causal language?: No
Limitations / Weaknesses
The paper draws upon survey experiments that may not accurately encapsulate the real-world motivations and emotions that catalyze support for political violence. In addition, the paper uses vignettes, either based on authentic or contrived events, which could fall short in portraying the intricate diversity of political violence scenarios and could potentially introduce confounding factors or biases that inadvertently skew respondents’ judgments. Relatedly, it is unclear whether responses to these hypothetical survey items reflect any sort of “true” attitude.
Open Data & Analyses
Does the article make the replication data publicly available?: Yes
Does the article make the replication analysis scripts publicly available?: Yes
Article Citation
Westwood, S. J., Grimmer, J., Tyler, M., & Nall, C. (2022). Current research overstates American support for political violence. Proceedings of the National Academy of Sciences, 119, e2116870119.
Bibtex
1
2
3
4
5
6
7
8
9
10
11
@article{westwood2022current,
title={Current research overstates American support for political violence},
author={Westwood, Sean J and Grimmer, Justin and Tyler, Matthew and Nall, Clayton},
journal={Proceedings of the National Academy of Sciences},
volume={119},
number={12},
pages={e2116870119},
year={2022},
publisher={National Acad Sciences}
}