Article Summary
Introduction
Partisans increasingly “follow the leader” on political issues; if their party leaders take a position on a policy, their supporters are likely to do the same. This relationship, of course, is conditional on partisans receiving the stance-taking signal from the party leaders. While most researchers consider this signal reception as binary (either partisans receive or don’t receive the signal), the reality is that messages vary in content and in the responses they elicit from certain types of partisans.
To that end, Kalmoe, Gubler, and Wood examine whether the usage of violent metaphors increases political polarization in aggressive partisans. Such violent metaphors appear (anecdotally at least) more common in modern political communication, with appeals to “fighting” for constituents, amassing “armies” of support, or going on the “attack” in campaigns. These messages may serve to polarize citizens susceptible to such messaging, as violence may cue defensiveness of their in-group and/or denigration of an out-group and result in more confrontational political stances. Less aggressive partisans do not hold similar mental “scripts” that would lead to such confrontational stance-taking, so the authors expect to only see violent metaphors polarize those with more aggressive personalities.
Analytical Approach
The authors field two survey experiments with the same general form. In each survey, respondents (pre-treatment) complete 12 items of the Bussy-Perry Aggression index in order to measure their trait aggression (0-1 scale). Then, the respondents read a vignette advocating for some position (depending on block, see Table 1). Certain phrases within each vignette are randomized to either evoke violent or non-violent metaphors, according to treatment condition. In Study 1 (n = 787, fielded on MTurk), respondents see two independently randomized blocks, and in study 2 (n = 1,118, fielded through Qualtrics) respondents only see one. One third of respondents are assigned to a pure control condition, seeing no vignettes.
Table 1: Study Details
| Study | Block | Violent/Non-Violent phrases |
|---|---|---|
| 1 | Spending | Fighting/struggling; shot/chance; need to fight for/owe it to; fight/work |
| Debt | Fighting/facing; battle/challenge; fight/work | |
| 2 | America’s Future | Fight(ing)/work(ing); fight for/help |
After each treatment block, respondents are asked a series of policy position items (n = 15 or 10 for studies 1 and 2, respectively) relevant to the block topic. These were reduced to individual-level indicators of policy conservatism/liberalism. The effect of violent metaphors on policy conservatism was estimated via OLS, including a triple interaction between violent metaphor usage, trait aggression, and partisanship (7 point scale) and all component terms/interactions. The authors also measure the marginal effect of violent metaphors on policy polarization at various levels of trait aggression using predicted values from the fitted model.
One design choice of note; because the violent metaphor condition was independently randomized between blocks in study 1, it is possible for respondents to have one block with violent phrasing preceding a block with non-violent phrasing, potentially leading to holdover effects. The authors attempt to address this by coding their main independent variable in study 1 as either 0, .5, or 1 for 0, 1, or 2 violent messages, respectively.
Main Findings
Exposure to violent metaphors appears to increase policy polarization in aggressive partisans in some policy areas. Focusing on the spending block in study 1, a trait aggression level of 0.67 (0-1 scale) is associated with the average partisan distance moving across about 1/3 of the issue scale. This result is similar in study 2, although the substantive size of the effect is half of that in study 1. There is no significant effect of trait aggression on the marginal effect of violent metaphors on polarization in the debt block of study 1.
Interestingly, there appears to be a negative effect of trait aggression on polarization at the lowest levels of trait aggression across both blocks in study 1. That is, exposure to violent metaphors decreases policy polarization among the least aggressive partisans. The authors take this as evidence of violent rhetoric being off-putting to low-aggression respondents, who may moderate their views in order to avoid conflict or confrontation.
In the appendix, the authors consider potential mechanism tests. They find some preliminary evidence that violent metaphors increase perceptions of conflicts and raise emotional valence, but do so primarily among Republicans.
Implications
Overall, authors provide some evidence of the pernicious effects of violent rhetoric in American politics. There are receptive audiences for violent metaphors in the American electorate, and they seemingly become more polarized when politicians oblige their appetites for such communication. This raises the specter of strategic communication on the part of political elites. If politicians can identify a receptive audience, their need to moderate language may become less pressing.
The authors also suggest their estimates understate the effects of violent rhetoric, since the primes used in the experimental treatments are relatively innocuous. This makes the possibility of more explicit calls to violence (such as Trump’s actions on January 6th) to be even more damaging to American politics.
In a more positive implication not explicitly considered by the authors, the results suggest a general distaste for violence among American voters. Insofar as the distribution of Americans is heavily right-skewed on trait violence (i.e. many more non-violent voters than violent ones), the net effect of politicians engaging in violent rhetoric is actually one of backlash; more partisans moderate their positions than double down. While this is certainly not an endorsement of more use of violent metaphors (as the normalization of such violent speech may dull the backlash), it does not appear the American electorate is particularly susceptible to such appeals.
Questions left unanswered
One question always left unanswered in these types of vignette experiments is the durability of the effects. The authors admit that most studies of this sort have effects that last several days at most, but also contend violent metaphors are recurrent enough in American politics to make the effects more durable. Further testing is needed to validate this claim.
More generally, while the authors focus on policy polarization in this particular paper, there is a question as to whether the observed effect on policy polarization is just an indirect effect of increasing affective polarization. That is, partisans may simply be becoming colder toward out-partisans and warmer toward in-partisans, and that change in warmth is reflected indirectly in policy positions. Disentangling the different mechanisms behind the observed results is research worth pursuing in future work.
Methods and Analysis
Was the study and its analyses pre-registered?: No
Did the study rely on proxy variables to measure polarization?: N/A
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
There are two major statistical concerns that require attention in this paper. First, the operationalization of the violent metaphors variable in study 1 as 0, 0.5, and 1 leads to artificially inflated estimates of the marginal effect of violent metaphors on polarization, since by definition the effect is a measure of a 1-unit increase in violent metaphors. By embedding a full dose of treatment in half a unit of a variable on a continuous scale, you functionally double the perceived effect. It is no surprise, then, that the standard operationalization in study 2 halves the overall effect, and in the appendix the triple interaction effect decreases by 21% after coding the violent metaphor variable as 0 or 1 only.
Second, because so much of the survey population is at the lower end of the trait violence distribution, the lower end of the estimated effect of trait violence on the marginal effect of violent metaphors on polarization is heavily influential when estimating a purely linear relationship. It is already clear that at the higher ends of the distribution, the effect is much more uncertain, so I would be interested in fitting a more flexible model, as the observed upper-tail effect may be a statistical artifact of the lower-tail results.
In general, more should be done to diagnose the exact content of the trait aggression measure and its correlates, as this appears to be a measure that is highly collinear with other demographic traits or even more pernicious survey-taking behaviors (such as inattentiveness).
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
Kalmoe, N. P., Gubler, J. R., & Wood, D. A. (2017). Toward conflict or compromise? how violent metaphors polarize partisan issue attitudes. Political Communication, 35(3), 333–352. https://doi.org/10.1080/10584609.2017.1341965
Bibtex
1
2
3
4
5
6
7
8
9
10
11
12
13
@article{doi:10.1080/10584609.2017.1341965,
author = {Nathan P. Kalmoe, Joshua R. Gubler and David A. Wood},
title = {Toward Conflict or Compromise? How Violent Metaphors Polarize Partisan Issue Attitudes},
journal = {Political Communication},
volume = {35},
number = {3},
pages = {333-352},
year = {2018},
publisher = {Routledge},
doi = {10.1080/10584609.2017.1341965},
URL = { https://doi.org/10.1080/10584609.2017.1341965},
eprint = { https://doi.org/10.1080/10584609.2017.1341965}
}